Introduction to Graduate Fluid Mechanics: C.R.

Smith Table of Contents/Objectives

Introduction to Graduate Fluid Mechanics

(an Electronic Text)

Charles R. (Chuck) Smith

Professor Emeritus
Department of Mechanical Engineering
Lehigh University

Second Edition
Self-Published

Copyright 2018

i
Introduction to Graduate Fluid Mechanics: C.R. Smith Table of Contents/Objectives

Dedication

This book has been a long labor of writing

and development, and I dedicate this book to
two exceptional people in my life:

My wife, Christine, who put up with my

constant occupation of my retirement office
to write and complete this book. She was
often a writer's widow, but never complained
about my continued obsession.

The other is my long-time colleague,

Dr. David J. D. A. Walker, who was a
consummate fluid mechanician and an
outstanding teacher, and my research
partner for over 30 years. He passed away
too soon to collaborate on this book.

ii
Introduction to Graduate Fluid Mechanics: C.R. Smith Table of Contents/Objectives

Table of Chapters

Part I: Basic Concepts

Chapter Title (left click on a chapter title to go to that chapter) Page

1 Introduction/Classification of Forces on a Fluid ……………….. 1

2 Vector and Vector Operators: The Basics ……………………... 18

3 Kinematics and the Substantial Derivative ……………………. 38

4 Motion and Deformation of a Fluid ……………………………. 64

5 Differential Equations of Motion ………………………………. 82

6 Simple Solutions to the Navier-Stokes Equations ……………. 115

7 Equations of Motion for Inviscid Flow ……………………….. 180

8 The Stream Function and Vorticity …………………………... 211

9 Two-Dimensional Potential Flow Theory …………………….. 231

10 Changes in Circulation and Kelvin’s Theorem ………………. 295

11 The Vorticity Transport Equation ……………………………. 318

12 More Complicated Navier-Stokes Solutions …………………. 372

Part II: Applications

13 Boundary Layer Theory ………………………………………. 403

14 Approximate Solutions of the Boundary Layer Equations …. 440

15 Flow Separation and Drag …………………………………….. 474

16 Introduction to Non-Newtonian Fluid Behavior …………….. 502

17 Instability and Turbulence: An Introduction ……………….. 570

iii
Introduction to Graduate Fluid Mechanics: C.R. Smith Table of Contents/Objectives

Objective and Scope of the book

The objective of this textbook is to provide a solid grounding in the principles of incompressible
fluid mechanics for beginning graduate students. The book, and the accompanying course which
would utilize it, should provide a student with the background to continue their studies into
specialty areas (e.g. aerodynamics, boundary layers, turbulence, etc.), and parallel areas (e.g. heat
transfer and thermodynamics). The book is relevant to students pursuing graduate degrees in
mechanical, civil, chemical, aeronautical, or nuclear engineering. I have also had students with
an initial degree in physics take my course, upon which this book is based, and found that their
background allowed them to use the book successfully.

The book does not cover control volume fluid mechanics or hydrostatics, since these are topics
in which students should be well grounded from their undergraduate studies. It also does not
cover numerical methods, since I feel that doing so distracts from the understanding of the
fundamentals of fluid mechanics, and that the time is better used providing a broad theoretical
background. It is my belief that students must develop a solid theoretical understanding of fluid
mechanics before they can pragmatically employ effective numerical solution techniques.

The thrust of the book is on the processes of how a fluid is modeled, the development of the basic
equations of fluid mechanics, the simplification of these basic equations using appropriate
assumptions, the establishment of proper boundary/initial conditions, and on methods of reducing
the resulting equations through judicious parameter scaling. While a number of traditional and
non-traditional examples are solved in the book using straight-forward analytical techniques, the
solution of more complicated problems using advanced analytical (e.g. asymptotics) or numerical
techniques (e.g. finite difference or finite elements) is not considered. I feel that once a student
understands the process of modeling and appropriately reducing the governing equations, they
can then consider solution techniques, both analytical and numerical, through advanced courses
in those topics. However, as appropriate, equation-solving programs, such as MATLAB or the
website Wolfram Alpha, will be employed to solve problems not amenable to analytical closed-
form solutions.

This book provides a student with a broad scope of the cumulative areas of incompressible fluid
behavior, including the processes of modeling a fluid, the properties of importance, how they
change and modify in an Eulerian environment, and the extension to areas of practical application
(boundary layers, lift, drag, non-Newtonian flows, and the basics of turbulence). The material
covered is applicable to a one-semester, survey course, or could be extended to a second semester,
depending on the depth to be pursued.

One practice I have tried to follow throughout the book is to assure that there is sufficient
development and demonstration of key topics and examples, and not to shortcut the development
of example problems. In many texts, a classic problem will be presented with too few details,
such that the student cannot follow the complete process from problem statement to solution. I
have tried to provide that completeness, at the expense of added length.

iv
Introduction to Graduate Fluid Mechanics: C.R. Smith Table of Contents/Objectives

I have also included a series of problems with each chapter, as appropriate. These are intended
for use as homework, and to illustrate the application of the material in a chapter. In some cases,
the problems may require the use of material from previous chapters as well. If you are an
educator teaching a course using this book, you can contact me to obtain detailed solutions to all
homework problems.

In this second edition of the book, I have added a chapter on the basics of non-Newtonian fluid
mechanics. The chapter covers the basics of non-Newtonian behavior and two of the methods
used to model non-Newtonian materials: the power-law and ideal Bingham plastic models. This
chapter 16 illustrates the complex changes that such models introduce into very simple flow
configurations: Couette and Poiseuille flow between parallel and curved surfaces. I have tried to
demonstrate the details of modeling such flows, and not just focus on the end result. Additionally,
I have included a number of graphs of velocity, flowrate, and pressure gradient behavior which
clearly illustrate the complexities of even simple non-Newtonian behavior.

As I noted in the first edition, since this is an electronic book, I have attempted to provide
electronic links to as many of the terms and references as possible. Where possible, I have linked
references directly to the appropriate bibliographic source on the web, which hopefully avoids
exhausting web searches for sources that are often hard or expensive to obtain.

How to "Read" this Book

This book contains 17 Chapters within one hyperlinked Adobe PDF file. The first 12 chapters
are designated as Part I, which covers the basic concepts of graduate fluid mechanics; the last 5
chapters comprise Part II, which covers the practical application of the basic concepts to key
areas of fluid mechanics (i.e. boundary layers, drag, non-Newtonian flows, and turbulence). The
book begins with a Table of Chapters, with each chapter title hyperlinked to the respective
chapter; left clicking on a chapter title will take you to the beginning of the selected chapter.

Each chapter begins with a Contents section, listing the key topics within the chapter. Each
topic and subtopic listed in the Contents section is again hyperlinked to that topic location
within the chapter, so that if the book is read electronically, one can quickly access any of the
topics/subtopics by simply left clicking on the link. There are also links within the chapters,
which will take you to either other portions of a chapter, or to electronic references.

To return to the location of a previous link, one can simultaneously hit the "Alt" and "" keys
on the keyboard, or use a "previous view" icon in the "page navigation" tool bar in Adobe
Acrobat Reader. Note: if you want to use the previous view icon (looks like a blue circle with a
left pointing arrom), you must add it to the page navigation toolbar. To add the previous view
icon, select the "Tools" tab, the "Customize Toolbars", and then navigate to the “Page
Navigation Toolbar”and check "Previous View". The "Previous View" icon should now appear
in the page navigation toolbar. If you just want to return to the beginning of the book, just hit
the "Home" key. These Adobe Acrobat Reader tools should allow you to navigate quickly
within the electronic chapter contents.

v
Introduction to Graduate Fluid Mechanics: C.R. Smith Table of Contents/Objectives

If you choose to print each chapter, you should have the book PDF file resident on your
computer, and then open and print the pages you wish. Do not try to print chapters directly
from a website. This sometimes works, but just as often creates complications.

Use this material whatever way works best for you. And enjoy it.

If you like the book, find any errors, or have suggestions for additions to the book, please send
me an email and let me know your thoughts.

Also, this is a public domain document, so feel free to pass on copies to any and all that might
be interested in the topic. I am self-publishing this book, and it is open access at no charge.
However, if you feel you would like to provide some compensation for the use of this book, feel
free to send me a check for whatever you would like to 1209 Oxbridge Drive, Lutz, FL 33549,
USA.

Chuck Smith
November 6, 2018

vi
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 1

Chapter 1

Basic Concepts

Contents

1.1 The Concept of a Fluid ..................................................................................................... 1

1.2 A Fluid as a Continuum ................................................................................................... 2
1.3 The Concept of Viscosity ................................................................................................. 3
1.3.1 Table 1: Viscosity of Water at Standard Pressure .................................................... 6
1.3.2 Table 2: Viscosity of Air at Standard Pressure ......................................................... 6
1.4 Types of Fluids: Newtonian and Other ............................................................................. 7
1.5 The Differential Element .................................................................................................. 8
1.6 Body Forces ................................................................................................................... 10
1.7 Surface Forces .............................................................................................................. 10
1.8 Stresses ........................................................................................................................ 11

1.1 The Concept of a Fluid

We all inherently know what a fluid is. Most lay persons, when asked for an example of
a fluid, would probably cite liquids, like water, as a fluid. However, gasses also qualify
as a fluid — although most people don't as readily think of those as a fluid, since they are
generally not visible. In general, all liquids and gasses are fluids. But what properties
qualify a substance as a fluid? And how does a fluid differ from a solid?

In general, a fluid is a substance which will deform continuously under an applied

shearing stress (i.e. a force applied tangentially to a surface of the fluid). In our solid
mechanics courses we learn that a solid will deform under an applied shear stress, but it
will only deform a fixed amount proportional to the applied stress. In other words, a
solid will deform in some proportion when a shear stress is applied, but once a balance
between the stress and the amount of deformation is reached, a solid will cease to deform
and will remain in a state of static deformation (unless it undergoes creep, due to thermal
or environmental effects). On the other hand, a fluid will continue to deform as long as a
shear stress is applied.

Another way of defining a fluid is that it is a substance that cannot sustain a shearing
stress. The key here is that the application of a shear stress causes a fluid to deform
continuously. This implies that the applied shear must balance with the rate of
deformation of the fluid. We will use this concept tying fluid shear stresses to the rate of

1
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 1

deformation (generally termed the strain rate) to characterize different types of fluids, and
to relate shearing stresses on a fluid to the strain rates through the use of a proportionality
constant known as the coefficient of viscosity.

1.2 A Fluid as a Continuum

We now have a way of judge if a substance behaves as a fluid. However, to deal with a
fluid mathematically, we must be able to develop ways to model the fluid behavior; and
to model the fluid appropriately, we need to consider the collective behavior of all the
molecules that comprise the fluid. In most macroscopic (large scale) situations, we do
not perceive the individual molecular behavior. However, from a microscopic
perspective, and statistical mechanics, we recognize that individual fluid molecules move
in rather randomized paths, and interact with other molecules through linear and
rotational collisions.

During these collisions, molecules exchange both momentum and energy, which
collectively give rise to the properties we normally associate with macroscopic materials,
such as pressure, temperature, internal energy, and density. However, these collective
properties require that the mean-free paths of the molecules be very small, such that
macroscopic property changes appear continuous. This requires that the random
molecular interactions are so manifold within a given volume of the fluid that one cannot
discern a change in a macroscopic property as molecules move in and out of a given
volume. As pointed out by Frank Smith (2003), the limiting volume for a fluid to exhibit
"continuous" behavior is roughly 10-9 mm3 of air at standard conditions. Such a volume
would contain approximately 3x107 molecules, which is sufficient to establish the
appearance of continuous properties in air. Since liquids have smaller mean free paths,
liquids can also be considered continuous fluids at (and even below) this limit.

Since most engineering applications deal with volumes much larger than this lower limit,
properties within essentially all fluids (with the exception of very low pressure gases,
such as at the edge of outer space or in extreme vacuums) can be considered to behave in
a continuous manner. The assumption of a fluid as a continuum allows the application of
mathematical modeling approaches which assume continuous behavior down to
infinitesimal volumes – i.e. volumes that are very small, but still larger than the
molecular limit required for continuous behavior within a fluid.

Recently, the consideration of nano-scale structures containing nano-scale fluids has

challenged the continuum assumption, and has required the application of alternative
methods of fluid modeling. However, for the purposes of most engineering concerns, and
within this text, we will consider fluids as a continuum.

2
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 1

1.3 The Concept of Viscosity

There are a number of different types of fluids, each characterized by the way the fluid
deforms relative to an applied shear stress. To assess the type of fluid behavior, we first
must model how a shear stress applied to a fluid causes a deformation rate, i.e. strain rate,
within the fluid. The simplest way to illustrate this shear stress/strain rate relationship is
to consider a fluid element (a small region of fluid) sheared in one plane by a single shear
stress, , as shown in Fig. 1.1.

ut 

t
u

 

y u  u  0  u

u0


Figure 1.1 A fluid element deformed at a strain rate /t by an applied shear stress .

We consider the changes here small, but continually increasing with time, such that the
strain rate, /t, will be a constant value as long as the shear stress  is maintained
(which means that the shear strain angle, , will continuously increase with time). In
most common fluids, like water, air, and oil, the strain rate will be linearly proportional to
the applied shear, such that:


 (1.1)
t

as indicated in figure 1.1. If the differential in velocity due to the shear stress is u across
a distance y, then from figure 1.1 we can geometrically show that:

ut
tan    , if  is considered small. (1.2)
y

Using Eqs. 1.1 and 1.2, we can show that:

 u
  (1.3)
t y
Now, taking the limit as y becomes small in Eq. 1.3, mathematically we have:

3
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 1

u du du
  lim  or   (1.4)
y 0 y dy dy

Assuming a proportionality constant of µ in Eq. 1.4, we can write:

du
 (1.5)
dy

In Eq. 1.5, µ, is termed the coefficient of viscosity, often called the absolute or dynamic
viscosity of the fluid. The larger the value of µ, the more viscous a fluid is. The units of
µ are stress-time, with units of [FT/L2] or [M/(LT)], where F, M, L, and T represent
respectively force, mass, length, and time in the units system being employed. Viscosity
typically varies with temperature and pressure, although temperature is generally the
dominant effect, with liquids becoming less viscous with increasing temperature, whereas
gasses become more viscous (generally) with increasing temperature.

Some typical values of absolute viscosity for various substances at standard conditions of
1 atmosphere pressure and 20°C are:

Air: 1.84x10-5 kg/(m-s)

Water: 1.0x10-3 kg/(m-s)
SAE 30 oil: 0.29 kg/(m-s)

So, water is more than 50 times more viscous than air, and motor oil is 290 times more
viscous than water! The following gives comparisons for other selected fluids:

Absolute Viscosity at
Fluid Room Temperature
(kg/m-s)
Air 1.84 x 10-5
Water 1 x 10-3
Olive Oil 1 x 10-1
Glycerol 1 x 100
Liquid
1 x 101
Honey
Golden
1 x 102
Syrup
Glass
1 x 1040
(Molten)

4
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 1

Another convenient form of viscosity that is often used in fluid mechanics is termed the
kinematic viscosity, , which is defined as the ratio of the absolute viscosity to the fluid
density, i.e.:


 (1.6)


where  is the symbol for fluid density (e.g. kg/m3). So, the units of kinematic viscosity
will be in [L2/T]. For the same three fluids cited previously, the kinematic viscosities are:

Air: 1.53x10-5 m2/s

Water: 1.01x10-6 m2/s
SAE 30 oil: 3.25x10-4 m2/s

Note that the kinematic viscosities of the three fluids vary much less from each other than
the absolute viscosities. Since both viscosity and density vary with temperature, so will
kinematic viscosity. The chart below shows the variation of the kinematic viscosity of
some common fluids with temperature.

Kinematic Viscosity
- common fluids
Kinematic Viscosity (10 m /s)
2
-6

Temperature (deg C)

The following tables list the absolute and kinematic viscosity for water and air at
atmospheric pressure, over a range of temperatures.

5
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 1

Table 1: Viscosity of Water at Atmospheric Pressure

Absolute Kinematic
Temperature
Density Viscosity Viscosity
-t- -- -- --
(oC) (kg/m3) (kg/m-s) x 10-3 (m /s) x 10-6
2

0 1000.0 1.787 1.787

5 1000.0 1.519 1.519
10 1000.0 1.307 1.307
20 998.0 1.002 1.004
30 996.3 0.798 0.801
40 992.4 0.653 0.658
50 989.2 0.547 0.553
60 983.2 0.467 0.475
70 978.2 0.404 0.413
80 972.6 0.355 0.365
90 966.3 0.315 0.326
100 972.4 0.282 0.29

Table 2: Viscosity of Air at Atmospheric Pressure

Absolute Kinematic
Temperature Density Viscosity Viscosity
-t- -- -- --
o
( C) (kg/m3) (kg/m s) x 10-5 (m /s) x 10-5
2

-50 1.469 0.687 0.468

0 1.295 1.736 1.341
10 1.249 1.787 1.431
20 1.206 1.837 1.523
25 1.186 1.862 1.570
30 1.166 1.886 1.617
40 1.129 1.934 1.713
50 1.094 1.982 1.812
60 1.061 2.029 1.912
70 1.030 2.075 2.014
80 1.001 2.121 2.119
90 0.974 2.166 2.225
100 0.947 2.210 2.333
150 0.835 2.423 2.900
200 0.747 2.624 3.512
300 0.617 2.994 4.854
400 0.525 3.330 6.342

6
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 1

You can find more information on both absolute and kinematic viscosity on the internet.
Some useful sites are: Gas viscosity calculator, Viscosity-Wikipedia, The Physics Hyper
Textbook and The Engineering ToolBox

1.4 Types of Fluids: Newtonian and Other

In section 1.3 we made the assumption that shear stress in a fluid will vary linearly with
the strain rate. In the majority of fluids, such as air, water, and motor oil, this proves to
be a good assumption. This linear relationship between shear stress and strain rate was
first proposed by Newton in his classic "Principia Mathematica" in 1686 (although he
didn't use the same terms, but close enough). In honor of his observation, all fluids that
display liner shear-strain rate relationships are termed a Newtonian fluid.

However, what about fluids that do not display such a linear relationship? They are all
lumped into a category known as non-Newtonian fluids, which is then split into
subcategories, depending on the shear–strain rate behavior. Figure 1.2 shows four
possible generic behaviors of non-Newtonian fluids, compared with a Newtonian fluid.
These non-Newtonian fluids are designated as: (1) dilatant (shear-thickening), which
increase in resistance with increasing shear stress; (2) pseudoplastic (shear-thinning),
which decrease in resistance with increasing shear stress; (3) plastic (strongly shear-
thinning), which show a marked decrease in resistance with a shear stress increase; and
(4) Bingham plastic, where the material behaves initially like a solid, until it yields under
a limiting shear stress, and behaves subsequently as a plastic. An ideal Bingham plastic
(shown), would display a linear relationship after it begins to yield; however, most real
Bingham plastics will generally display a plastic-type behavior after yielding.

Shear
stress Ideal Bingham plastic

Plastic
Pseudoplastic
Newtonian
Yield
Dilatant
stress

Shear strain rate

Figure 1.2 The generic behavior of various non-Newtonian fluids relative to

Newtonian fluid behavior.

7
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 1

The modeling and assessment of non-Newtonian fluids, though of significant interest, is

quite messy. And most practical engineering applications generally are concerned with
fluids that behave in a Newtonian manner. So, in this book, we will generally deal with
Newtonian fluids. However, Chapter 16 of this text will briefly examine simple non-
Newtonian fluids in basic flow configurations, to illustrate their variation in behavior
from Newtonian fluids.

Note that there are a number of texts which address the behavior of non-Newtonian
fluids, and once you understand how to model and solve the equations for Newtonian
fluids, this will put you in a better position to address non-Newtonian fluid behavior.

You can find more information on non-Newtonian fluid on the web. Some useful sites
are: The Physics Hyper Textbook, and Non-Newtonian fluids-Wikipedia.

1.5 The Differential Element

Fluids, like all matter, are governed by Newton’s second law, where inertial
accelerations/decelerations of a body are balanced by the sum of the forces applied to the
body, no matter how small. As long as a fluid can be considered a continuum, and each
particle of the fluid can respond independently to the applied forces, we generally find it
useful and necessary to model the fluid behavior using a differential fluid element. This
differential element is assumed small enough such that property changes across the
element are minimal (but not negligible), but large enough to still be assumed a
continuum (i.e. no discontinuities of the fluid motion within the differential element).
 
This approach allows us to ultimately express Newton’s second law ( F  ma ) in terms of
a very small, differential volume (i.e. an infinitesimally small control volume). The
geometry of this differential volume can be expressed in terms of an appropriate
reference coordinate system (generally either Cartesian, cylindrical, or spherical), such
that we can describe both the differential volume and the surfaces surrounding this
differential volume in terms of the coordinate directions (e.g. x, y, z for a typical
Cartesian coordinate), as shown in figure 1.3

Vol = d = dxdydz
y
dy

dz
x dx

Figure 1.3 An infinitesimal differential fluid element

8
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 1

The volume of a differential element in a Cartesian system, as shown in figure 1.3, is

defined as d = dxdydz, or the product of the three orthogonal dimensions of the
differential element. Note that this element is not necessarily a cube, since no restriction
is placed on the magnitude of the differential dimensions, other than the assumption that
they are small enough such that second-order and higher changes (i.e. spatial derivatives)
of the fluid properties (e.g. velocity) in a selected direction are negligible, relative to first-
order, linear changes of the property in that same direction. This is important, since this
assumption allows us to develop differential equations modeling the property changes
throughout the fluid by use of a truncated Taylor series expansion (more on this in
Chapter 5).

Another important characteristic of this differential element is that it allows a systematic

representation of the various bounding surfaces, which is important for establishing
contact forces (such as pressure). For example, the area dzdy in figure 1.3 is the area (i.e
the magnitude) of the surface bounding the right and left sides of the differential element.
In order to characterize both the magnitude and the direction of orientation of a surface,
we require a systematic method for establishing the spatial orientation of each surface
relative to a selected coordinate system. Accordingly, we define the direction that a
bounding surface faces by a unit vector oriented normal to that surface. By convention,
this unit vector always points outward from the differential volume. Therefore, for the
surface bounding the right side of the differential element shown in figure 1.3, the normal
vector would point to the right, as shown in figure 1.4. However, for the same magnitude
surface bounding the left side of the differential element, the normal vector would point
in the opposite direction (i.e. to the left). As required, both of these normal vectors would
point outward from the differential element.

y
dy
Normal Vector (pointing outward)
dz
x dx

Figure 1.4 Illustration of normal vector for right face of differential element

Note that for the right-facing surface the normal vector points in the positive x-direction
(for the coordinate system shown). However, if we were considering the outward normal
for the left-facing surface of this differential element, the surface vector would point to
the left, in the negative x-direction, which is (of course) outward from the differential
element for the left surface. The direction of the outward normal is particularly important
when determining the absolute direction that a force will act when applied to a particular

9
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 1

surface of the differential element. We will discuss this issue, and its importance, further
in section 1.7 below.

Now we will illustrate how we make use of the differential element concept to describe
the respective forces that act on a fluid. To do this, we note that the forces on a fluid are
separated into two basic types: (1) Body forces, and (2) Surface forces.

1.6 Body Forces

Body forces are forces that develop within a fluid without physical contact, and are
proportional to the mass of the fluid contained within a region. Particular examples are
forces due to gravitational or magnetic fields. For a differential fluid element, as shown in
figure 1.5, the local body force due to gravitational attraction is given by
   
dFB  gd  gdxdydz  , where dFB is the differential vector body force acting on the

differential fluid element,  is the local fluid density, g is the gravitational vector field,
and d is the volume of the differential element.
Vol=d
y   
dy
g
dFB  gd  gdxdydz 
dz
x dx

Figure 1.5 Illustration of gravitational body force generated on a differential

element

In this text, we will only address gravitational effects, since it is the most common body
force encountered by a fluid. However, magnetic effects can be quite important when
dealing with the presence of a magnetic field within electrically conductive fluids, such
as mercury, or even sea water. In fact, the field of magneto hydrodynamics is primarily
concerned with the influence of magnetic fields on electrically conducting fluids.

1.7 Surface Forces

Surface forces are forces exerted by direct contact at a fluid boundary. The most
common examples of such forces are normal forces generated by pressure, or shear
stresses generated by fluid friction.

Note that surface forces, regardless of their source, are vectors that are comprised of: (a)
one normal force, which acts perpendicular (normal) to a particular surface, and (b) two
tangential forces, which act tangential to the surface, and perpendicular to each other.

10
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 1

Figure 1.6 shows a general example of a surface force acting on a planar surface, and its
directional components relative to a Cartesian coordinate system fixed on the surface
(here, the z-direction is taken as normal to the plane).

z
y
Fsy
Fsz F
(or Fn) x
Fsx

Figure 1.6 Surface forces acting on an arbitrary planar surface.


We assume that F is an arbitrary surface force acting on the plane. We can then

represent F by its respective components, where Fsz is the normal component of force
acting perpendicular to the surface (in the z-direction), and Fsx and Fsy act respectively in
the x-direction and y-direction, where x and y are oriented at right angles within the
plane.

1.8 Stresses

Although we are ultimately concerned with the total forces acting on either a body or
surface (e.g. an airplane wing or the inside surface of a pipe), for the purposes of
mathematically modeling the fluid behavior, it is more efficient to write the governing
equations for fluid mechanics in terms of very localized stresses, where stress = force/unit
area. Here we discriminate between the normal and tangential stresses as follows:

 Fn
1) normal stress ii  lim (1.7)
A  0 A
 Fs
2) shear stress ij  lim (1.8)
A  0 A

Here, i and j are direction coordinates which reflect respectively the defining surface of
action and direction of the stress, according to the reference coordinate system employed.
Note that a stress has both (a) a direction, and (b) a surface upon which it acts. For
normal stresses, the direction of action is along the normal to a respective surface.
However, for shear stresses, the direction of action will be tangential to the surface and
perpendicular to the surface normal. Thus, we need to employ a method for defining
stresses that clearly characterizes all the directional forces (i.e. one normal, and two
shear forces) that act on any given surface. We do this by using a double subscript
notation that characterizes both (1) the plane, or surface, upon which a stress acts, and (2)
the direction in which the stress acts. This creates a higher-order type of vector

11
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 1

description of a stress field known as a tensor, which is a collection of nine scalar

components that describe both the direction and the surface of action for all possible
stress components. For a given coordinate system, the double subscript notation is
defined as i j , where  is the generic stress component. Here, the first subscript (i)
indicates the surface of application (defined by the direction of the normal to the surface)
and the following subscript (j) indicates the direction of action of the stress. For a
Cartesian coordinate system, an example is:

i.e.  y x  stress (normal or shear) (1.9)

direction of action (x, for this case)

surface of application (y, for this case)

This raises the question of which direction a calculated stress will actually act on a
specified surface. One needs to recognize that a “positive” stress (as determined by an
appropriate calculation) will not necessarily act in a “positive” direction relative to a
specified coordinate system. While this seems strange at first, one can easily realize that
for a particular geometry and flow, an imposed stress will have an absolute direction in
which it will act (which is independent of the chosen coordinate system). For example,
pressure exerted by a fluid on a surface will always act “against” the surface, and
gravitational forces will always act toward the center of the earth, regardless of how we
choose our coordinate system. So a stress created by a particular flow may act in a
particular inherent direction (e.g. to the right of this page), but this may be construed as
acting in either a positive direction (coincident with our reference coordinate) or a
negative direction (in opposition to a reference coordinate), depending on how we chose
to orient our coordinate system, and the particular surface of action that is selected.

Note that a surface is defined as either a “positive” or “negative” surface depending on the
direction of the outward normal for a selected surface (relative to the coordinate system).
The directions of the stress components acting on a specified surface are then established
according to the sign of the absolute value of the stress component, which is either
assumed (for purposes of differential modeling and equation derivation) or calculated (if a
mathematical expression is already available). The convention for defining a “positive”
stress, as indicated below, depends on the product of the signs of: (a) the surface vector
(the outward normal) and (b) the direction of the stress. If the product is positive, the
absolute stress component is positive, whereas a negative product indicates a negative
absolute stress component. The following table defines all the possible signs for planes
and directions of action, and the corresponding sign of the absolute stress component.

12
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 1

Stress Component Sign Convention

Plane Direction  Absolute Stress Component

+ + +
- - +
+ - -
- + -

Product  Component Sign

As an example, consider the application direction of shear stress components acting on

various surfaces of a differential element volume, where all the absolute stress
components are all assumed to act positively. Here, we define the sign of all external
surfaces of the differential element by their outward pointing normal relative to the
Cartesian coordinate system shown. For example, the outward normal for the top surface
of the differential element shown in figure 1.7 points in the positive y-direction. Thus,
yz, a z-direction shear stress, which acts in a “positive” absolute direction on this surface,
will act in the positive z-direction, as shown, since the product of the sign of the surface
normal (+) and the shear stress directional sign (+) must yield a positive (+) absolute sign.

y
outward normal, y-surface (+)

yz xy
outward normal, x-surface (-) xy x
outward normal, x-surface (+)
z

Figure 1.7 Illustration of typical orientations of “positive” stresses acting on

selected surfaces of a differential fluid element

Correspondingly, for xy , a y-directed shear stress acting on the right surface of the
differential element (with outward normal in the positive x direction), the shear stress will
also act in a positive y-direction in order to represent a positive absolute shear
component. However, if one considers the same y-directed shear stress acting on the left
surface of the differential element (with an outward normal in the negative x direction),
the absolute direction for xy will now act in the negative y-direction, as shown in figure
1.7.

Note that this sign convention is incorporated into the governing differential equations
when they are derived, and is implicit in the equations when they are integrated (in either

13
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 1

closed form or numerically) to determine the velocity field for a particular fluid flow.
Thus, when the resulting velocity equations are subsequently differentiated to establish
quantitative (absolute) values for the stresses, the stress sign convention provides a
logical, systematic method to establish the absolute direction that a calculated stress will
act.

Example: Consider a Couette Flow (Couette, 1890), a simple viscous flow created
between two infinite flat plates, where an upper plate moves parallel to a fixed lower
plate, as shown.

y
Uo
h
x

Without going through the derivation of the velocity flow field (we do this later in
Chapter 6), we will simply note here that the velocity expression for a steady laminar
flow of a fluid between the moving and fixed surfaces is given by (for the coordinate
system shown):

 y
V  u î  vĵ  U o   î (1.10)
h

u and v represent the velocity components in the x and y directions, respectively, so by

comparison we have:

y
u y   U o   and v  0 (1.11)
h

In Eq. 1.11, u(y) is the x-direction equation for the velocity as a function of y, Uo is the
constant velocity of the upper plate, and h is the constant spacing between the plate
surfaces (i.e. the thickness of the fluid layer). The unit vector î indicates that the flow is
in the x-direction. Now, from Eq. 1.5 the local shear stress for this one-directional flow
is proportional to the local rate of deformation of the fluid, or the gradient of the velocity,
given by:

du U
Shear Stress     yx   o  0 (1.12)
dy h

Here  is the absolute fluid viscosity, which is the proportionality of the shear stress to
the velocity gradient, or the rate of fluid deformation.

So, having determined the local “absolute” magnitude of the shear stress within the fluid,
we can now assess in which direction the stress will act. What is not intuitively apparent
is that the direction of the stress depends on which surface is of interest to us. For

14
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 1

example, are we interested in the stress (and thus force) exerted on the fluid by the
moving plate, or on the plate by the fluid? Or are we concerned with the stress exerted
by the fluid on the fixed lower plate, or the moving upper plate? Just as we did when
establishing the direction of action of a specified stress component on a differential
element, we must first establish our surface of interest (and thus action), and then
determine the absolute direction (relative to our selected coordinate system) for our
calculated stress. Figure 1.8 below illustrates several different shear stress scenarios for
selected surfaces of interest.

Uo (+)
(+) yx upper plate
yx (-) fluid on
on fluid
(-) upper plate
y
(+) fluid on
x (+) yx lower plate
lower plate yx (-) on fluid
(-)
Figure 1.8 Absolute directions of the shear stress acting on selected surfaces
for a Couette flow.

Consider, for example, the direction of action of the calculated shear stress, yx, acting on
the upper plate (which is proportional to how much force would have to be exerted to
keep the plate moving). As shown on the left schematic of figure 1.8, the outward
normal for the surface of the upper plate in contact with the fluid is in the negative y-
direction. Since the absolute value of yx that we calculated in Eq. 1.12 was positive, this
means that in order for the sign product of the surface and the shear stress direction to
yield a positive orientation, the shear stress must act in a negative x-direction, as shown.
This makes physical sense when one considers the physics of the problem. As we move
the upper plate to the right, the fluid resists, and we experience a force (i.e. stress) on the
plate that acts in opposition to the plate movement (i.e. to the left, or negative x-direction
relative to our selected coordinate system).

On the other hand, if we wish to determine the direction of action of the shear stress
acting on the fluid adjacent to the upper plate, we consider the right schematic of figure
1.8, which isolates the fluid between the plates. Here the outward normal for the upper
fluid surface points in a positive y-direction (out of the fluid). Thus, since the absolute
value of yx must still have a positive orientation, the surface-direction sign product must
again be positive, which means that the shear stress on the upper fluid surface must act in
the positive x-direction. At first this result might seem confusing, since this is opposite to
the result we obtained when considering the upper plate surface. However, if we again
consider the physics of the situation, the movement of the upper plate to the right in
essence “pulls” the adjacent fluid to the right, which is felt by the fluid as a stress in the
positive x-direction. Clearly, the stress has the same absolute value at this location, but

15
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 1

the direction of action depends on the surface of consideration. As Newton pointed out,
every action causes an equal and opposite reaction.

Using the same approach and convention, the direction of action of the shear stress on the
lower, fixed surface can also be determined as shown in figure 1.8 -- and again the
“absolute” direction of application depends on whether one considers the contact surface
of the plate or the fluid.

As a further exercise, consider a change of orientation of the coordinate system, to that

shown below. Here the reference x-y coordinates point in the opposite directions from
the original coordinate system employed, and have their origin at the surface of the upper
plate, rather than the lower plate.

x Uo
h
y

While this is physically the same flow situation, using this coordinate system the velocity
equation for the fluid between the plates is now given by:

  y y 
V   U o 1   ˆi or u y   U o   1 (1.13)
 h h 

While this equation looks quite different, it describes the same exact linear velocity
distribution, only referenced to the oppositely oriented coordinate system. Accordingly,
if we calculate the corresponding shear stress we have:

du U
   yx   o  0 (1.14)
dy h

Note that this is identical to the value determined in Eq. 1.12, using a markedly different
coordinate system. However, since the coordinate system is oriented differently, we
might intuitively expect that this would have a significant influence upon the “absolute”
direction of the applied shear. To examine this, consider figure 1.9, which assesses the
shear stress application to the same surfaces considered in figure 1.8, only with the
oppositely-orientated coordinate system.

16
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 1

(-)
(-) yx upper plate
x fluid on
yx (+) on fluid
upper plate
(+)
y
(-) fluid on
(-) yx lower plate
lower plate on fluid
yx (+)
(+)

Figure 1.9 Absolute directions of the shear stress acting on selected surfaces
for a Couette flow—oppositely-oriented coordinate system.

Note that although the reference coordinate directions and origin are different, when we
apply our stress direction convention, the “absolute” direction of the applied shear
stresses is identical, both in magnitude and absolute orientation, as they should be. So,
regardless of how we define our reference coordinates, the final results of our modeling
process, and the analysis of the governing equations, will yield the same absolute result,
as long as we apply our convention for the action of a stress on a specified surface.

References

Smith, F. M. (2003). Fluid Mechanics, McGraw-Hill, Boston

Couette, M (1890). Ann. Chem. Phys., ser.6, vol. 21, pp. 433-510.

17
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 2

Chapter 2

Vectors and Vector Operators: The Basics

Contents

2.1 Coordinate Systems ...................................................................................................... 18

2.1.1 Cartesian Coordinates .......................................................................................... 19
2.1.2 Cylindrical Coordinates ........................................................................................ 20

2.2 Vector Differentiation ................................................................................................... 21

2.2.1 Derivatives of Unit Vectors ................................................................................... 22

2.3 Vector Operator  ........................................................................................................ 23

2.3.1 Gradient .............................................................................................................. 25
2.3.2 Divergence .......................................................................................................... 26
2.3.3 Cross Product or Curl ............................................................................................ 28
2.3.4 The Laplacian and Other Useful Identities ............................................................. 30

2.4 Review of Line and Surface Integral Theorems ............................................................... 32

2.4.1 The Gauss Divergence Theorem ............................................................................ 33
2.4.2 The Curl Theorem ................................................................................................. 33
2.4.3 The Gradient Theorem ......................................................................................... 34
2.4.4 Stokes' Theorem .................................................................................................. 35

A fluid can move and change its orientation and properties continuously in space and
time. Thus, to characterize fluid behavior properly we must know both the magnitude
and direction of changes to fluid behavior at any point. This requires that we represent
fluid behavior in terms of vector functions, which describe the magnitude and direction of
both fluid motion and property changes throughout the extent of the fluid. Although
there are excellent books that cover the details of vector calculus (e.g. Schey, 1997), and
a student of fluid mechanics is expected to have had prior mathematical instruction on the
basics of vector calculus, the present chapter reviews a few of the basic concepts of
vector calculus that are particularly important for the derivation and utilization of the
governing equations of fluid mechanics.

1. 2.1 Coordinate Systems

To properly establish a vector representation of a fluid property (such as velocity),

requires that we first establish an appropriate coordinate system. Note that a vector
function representing a fluid property will always reflect an absolute behavior, which is
inherent in the fluid behavior (e.g. water always flows downhill). However, how we
choose to represent that absolute behavior mathematically depends on the coordinate

18
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 2

system we employ to describe the behavior. It is to our advantage to choose a coordinate

system that most easily represents the type of behavior of interest. For example, if we are
concerned with flows over flat surfaces, or within rectangular tubes, a Cartesian or
rectangular coordinate system (x, y, z) is best suited for representation, since all property
changes will occur either along the planar boundaries or normal to them, and the
Cartesian coordinate system can be easily aligned with these boundaries. Alternatively, a
cylindrical or radial coordinate system (r,, z) can best describe, and be aligned with,
flows that pass either around, along, or through geometries that are radially symmetric,
such as cylinders and circular tubes, pipes, and ducts. And finally (at least for our
purposes), spherical coordinates (r, , ) provide a coordinate system that is well suited to
flows over or within spherically-symmetric geometries, such as geophysical flows within
the atmosphere. For the purposes of the present text, we will confine our interests to the
former two coordinate systems, Cartesian and cylindrical, which have the broadest
applications, and are admittedly simpler to employ. However, realize full well that the
same principles that we develop for Cartesian and cylindrical coordinate systems apply
equally well in spherical coordinates (or any other coordinate system of choice), albeit
yielding more complicated governing differential equations.

2.1.1 Cartesian Coordinates

A typical Cartesian coordinate system is shown in figure 2.1 below, characterized by

three orthogonal coordinate axes initiating at a common origin.

y a

ĵ î x
k̂
z

Figure 2.1 Schematic of a Cartesian coordinate system

Normally used with planar or rectangular geometries, a Cartesian coordinate system will
generally be oriented such that two of the axes lie in the primary plane of interest (e.g. on
the surface of a flat plate), with the other axis projecting perpendicular, or normal, to the

plane of interest. As with all coordinate systems, a vector property, call it a , when
described in a Cartesian system will consist of three directional components (ax, ay, az),
which indicate the magnitude of the vector contribution in each coordinate direction.
This allows the vector to be expressed in terms of the directional components as:


a  î a x  ĵ a y  k̂ a z (2.1)

19
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 2

Here î, ĵ, and k̂ are unit vectors, which are each of unity magnitude and oriented in the x,
y, and z directions respectively, as shown in figure 2.1. Note that ax, ay, and az can be
either constants or, more likely, functions of one or more of the coordinate variables x, y,
and z.

2.1.2 Cylindrical Coordinates

Similar to the Cartesian system, the cylindrical (also called radial or polar) coordinate
system employs three coordinate axes: r, , and z. However, only two of these (r and z)
originate from a common origin, as shown below.

î î r
y
r

î z x
z

Figure 2.2 Schematic of a radial coordinate system

The third coordinate is a reference angle for the r axis, and describes the angular location
of the r axis relative to a fixed reference. This coordinate system will generally be used
for axisymmetric geometries, such as rods and tubes, with the z coordinate axis located
along the axis of rotational symmetry (e.g. the centerline of a rod or tube). Here, we can

describe a vector a in terms of directional components ar, a, and az as:


a  î r a r  îθ a θ  î z a z , (2.2)

In Eq. 2.2, î r , îθ , and î z are again unit vectors of unity magnitude, and oriented
respectively in the r, , and z directions, as shown in figure 2.2.

There are two things to note about the cylindrical coordinate system that make it more
complicated than the Cartesian system (at least when performing vector calculus
operations). The first complication is that the r and z coordinates reference spatial
changes, whereas the  coordinate references an angular change. This creates some
complication when one takes the derivative of a vector expressed in cylindrical
coordinates, since derivatives with respect to z and r yield spatial property changes, while
derivatives with respect to  indicate angular property changes. Consequently, it is the
effect of these angular changes of the r coordinate that yields the corresponding spatial

20
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 2

change with respect to . In that these angular changes are a combination of changes in
both r and , the terms comprising the governing differential equations will be a bit more
complicated, and in some cases contain additional terms, relative to the Cartesian form of
the governing equations.

The second complication is that the îr and îθ unit vectors within the cylindrical system
change their spatial orientation with changes in the angular coordinate, . Thus, a
change in the angular orientation of a vector will also result in a corresponding change in
the orientation of the îr and îθ unit vectors, since these unit vectors are oriented
respectively along ( îr ) and normal to ( îθ ) the r coordinate axis, and r changes its angular
orientation whenever there is a change in . For example, if r changes its angular
orientation (i.e. rotates) by 180, the îr and îθ unit vectors will change their orientation by
180 as well. This coupling of îr and îθ with  manifests itself when one differentiates a
vector in cylindrical coordinates, which results in non-zero unit vector derivatives for
îr and îθ , as we will discuss in section 2.2.1.

2.2. Vector Differentiation

Generally, vector differentiation is performed in essentially the same manner as non-

  
vector differentiation. Going back to basics, if we let a be a vector such that a  a (t ) ,
we can define a derivative with respect to t as:
   
da a a ( t  t )  a ( t ) 
 lim  lim  da instantaneous
dt t 0 t t 0 t a (t )  rate of change
 dt
a

a (t  t ) 
a approximate
 rate of change
t

 
   
So, if at   t 2 î , then da dt  dt 2 dt î  t 2 dî dt  2t î , since we assume that the unit
vector î is not a function of t (which is the case for the Cartesian coordinate system). Of
course, the process is the same for partial differentiation as well. However, the
assumption that a unit vector is not a function of the independent coordinate variables
depends on the coordinate system employed, as shown in the following.

21
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 2

2.2.1 Derivatives of Unit Vectors

First, let’s consider the easiest situation, the Cartesian coordinate system. This is really
simple, because the Cartesian unit vectors -- î, ĵ, k̂ -- are of constant magnitude (unity) and
independent of changes in x, y, z. Thus,

  
, ,  0 for all Cartesian unit vectors.
x y z

So, when performing a vector operation, whether it is the formation of a derivative or an

integration, we treat Cartesian unit vectors the same as we do constants in any operation.

Now, let’s consider the cylindrical coordinate system. Here, î z is independent of r, z, 

-- so we can treat the î z unit vector as a constant during differentiation. However, the
unit vectors îr and î , while they are independent of changes in r and z, they are (as
pointed out in section 2.1.2 above) dependent on changes in . Thus, when we consider
the differentiation of unit vectors in the cylindrical coordinate system, we obtain the
following:

ˆir ˆi ˆiz

0 0 0
r r r
ˆir ˆ ˆi ˆiz
 i  ˆir 0 (2.3)
  
ˆir ˆi ˆiz
0 0 0
z z z

Here, differentiation of î r and î with respect to r and z yields zero values; however,
differentiation with respect to  yields two non-zero values, which are unit vectors
themselves. This is of particular importance when dealing with vector calculus
operations employing the  operator, which we will discuss in section 2.3.

Example: Using the basic concept of a vector derivative, the following demonstrates
 ˆi
how we establish the non-zero partial derivative for r . Here, we represent îr as a
θ
vector of unity magnitude, oriented along the coordinate r. Now we assume this
vector undergoes a small angular change of . Since the orientation of îr is a
function of , this means that the change of îr with respect to  is from îr () to îr
(+). Thus, we can write the derivative for îr with respect to  as:

22
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 2

 îr î (  )  îr ()

 lim r (2.4)
  0 

Now, to approximate the difference, îr     îr  in Eq. 2.4, we examine the
geometric change of these vectors, as shown below. We label this difference
 îr , and note that  îr is represented by the small arc created by the rotation of the
unit vector îr by an amount  in the î direction. Since the magnitude of îr is
unity, as indicted, then  îr = î .
note : iˆr  1

î îr
îr      
 îr  îr  î
y magnitude
r 

 îr 
x direction

Thus, inserting our value of  îr into Eq. 2.4 yields:

îr î (  )  îr ()  î  î

 lim r  lim r  lim  î  (2.5)
 0    0    0 

 î
As an exercise, use a similar approach to derive the result for by considering the

changes in the unit vector î for a small angular change, .

2.3 Vector Operator 

To develop the governing equations of fluid mechanics, we must be able to

mathematically model the changes in the various properties of a fluid, both spatially and
temporally. Sometimes we are concerned with changes from one point to another (like
with the Bernoulli equation), and sometimes with the local changes, particularly when
developing the governing differential equations of motion. These properties run the
gamut from scalars (e.g. fluid density, pressure) to vectors (e.g. velocity, vorticity). To
model these changes, we have several different mathematical methods to establish these
property changes, described variously as the divergence, the curl, and the gradient of a

23
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 2

property field. These descriptions can often get rather complicated, since we have to
consider the various possible changes that can take place with respect to all the
coordinate directions.

To help facilitate these “change” descriptions mathematically, we have a very convenient

mathematical operator that allows us to write a bit of short hand for the various changes.
This operator is designated the “del” operator, and is written as:

  
  î  ĵ  k̂ (in Cartesian coordinates) (2.6)
x y z
and
 1  
  î r  î  î z (in cylindrical coordinates) (2.7)
r r  z

Examination of the above definitions illustrates that the  operator reflects the
cumulative changes for all three respective coordinate directions. For example, the first

derivative function for the Cartesian coordinate expression in Eq. 2.6 above, î ,
x
assesses the changes in a property relative to the x-direction—the unit vector indicates
that this change is an x-direction contribution to a spatial derivative vector. Applying this
 operator to a designated property (e.g. via a dot or cross product operation), we can
establish how that property is spatially “changing” at a specified point.

However, one must exercise care when using the  operator in cylindrical coordinates,
since, as pointed out above in Section 2.2.1, the differentiation of the unit vectors îr and
î  , which change their orientation with respect to the angular coordinate , will yield
other unit vectors. This is particularly important to understand when performing
operations with the  operator. The rule is that when performing operations that involve
both vector differentiation and subsequent vector operations, one must always perform
the derivative function (of both the magnitude function and the unit vectors) first, and
then perform the prescribed vector operation. For example, if we are taking the dot

product of bî with the angular derivative of V  a ˆi , (where a and b are constants) we
r 

would perform the operation as:

Differentiation Vector Operation


ˆi 
ˆ
b ir 
V

 bˆir 


   
   
  
a ˆi  bˆir  a  ˆi      bˆir  a 1 ˆi    ˆir  ab


24
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 2

Note that if we didn’t consider the derivatives of the unit vectors in this operation, our
result would have been zero! So, the rule is: perform all derivative functions first, before
performing the vector operations.

Now, let’s consider the operations that can be performed using the  operator: first on a
scalar field, and then on a vector field. Recall that a scalar field is one for which the
property has magnitude, but no direction. Pressure, fluid density, and temperature are
examples of scalar properties: they will have a magnitude at a location, but do not have a
direction associated with the property. On the other hand, velocity, acceleration, and
vorticity (angular rotation) are properties that have both a magnitude, and a particular
direction at any given location. The application of the  operator will yield very
different types of information about a property field, depending on whether the property
is a scalar or a vector. And from the point of view of modeling of fluid behavior, it is
important to understand what information results from the use of the  operator.

2.3.1 Gradient

Consider a scalar property (like pressure or fluid density) that is given by   f ( x, y, z) .

Such a property describes what we term a scalar field, which means that the function
f(x,y,z) will describe the magnitude of the property  at any coordinate location within a
region (the property “field” encompassed within the coordinate space). Clearly, this
scalar property only has magnitude, but no direction of action. However, if the scalar
property is changing with location (e.g. a change in pressure as we walk up a mountain),
we can establish the local rate at which the property will change as we move in any
direction from a specified location within the field. Clearly, these changes will have both
a magnitude and an associated direction of that change. Taken collectively for all
coordinate directions, these changes are termed the “gradient” of the property and can be
mathematically determined by performing the  operation directly on the scalar function
, such that:

  
  î  ĵ  k̂ (2.8)
x y z

The result of this operation is a vector function,  , with magnitude and direction
indicating the spatial rate of change of  .

Example: Consider the static pressure field in a swimming pool given by

  p(z)  p o   g z , where  is a constant fluid density, g is the gravitational constant,
po is the pressure at the surface of the pool, and z is a Cartesian coordinate pointing

25
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 2

vertically downward from the pool surface. Accordingly, the gradient of the pressure in
the swimming pool is given by:

0 0

  
  î  ĵ  k̂   g k̂ (i.e. a constant change in z-direction)
x y z
g
direction  po
magnitude
z

Here, the gradient indicates that the pressure is increasing at a constant magnitude in the
positive z direction (i.e. as we move farther below the surface of the water). Other
examples of scalar fields in a fluid where a gradient can exist include fluid density,
temperature, and concentration fields. We’ll show how this concept can be utilized in
Section 2.4 below, and later in Chapter 4.

2.3.2 Divergence

Now let’s consider the changes that can take place in a vector field (again in Cartesian

coordinates), V  Vx î  Vy ĵ  Vz k̂ , which has both magnitude and direction. Clearly,
since the  operator is a vector operator, in order to operate on a vector field, one must
employ an appropriate vector operation. Our options are the dot product of two vector
functions, or the cross product (or curl) of two vector functions. Here we consider the dot
product first, with the  operator prescribed to operate (left side of the dot product) on the
vector field (right side of the dot product), with the result:

  
  V   î
  
 
V V V
 ĵ  k̂   Vx î  Vy ĵ  Vz k̂  x  y  z (2.9)
 x y z  x y z

Equation 2.9 is a scalar termed the “divergence,” which indicates the degree of total
“expansion” of a vector field at a point. Note that this expansion (of course a negative
value would indicate a contraction) is the sum of the local respective changes in all three
coordinate directions. The divergence is intimately related to the compressibility of a
fluid -- the spatial change in the local density at a point. We will show later, in Chapter

4, that for incompressible fluids (like water),   V = 0 is an expression of the

conservation of mass, if the vector V represents the velocity field of a fluid. Thus, if the
density of a fluid can’t change, the local velocity changes in all coordinate directions
must exactly balance each other (i.e. there can be no “expansion,” or “contraction” of the
volume that an incompressible fluid occupies).

26
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 2

Example: Divergence in Cartesian Coordinates


Consider a velocity field given by V  ax î (a = constant).

Here, there will be a non-zero divergence, indicating that the fluid is expanding at a
point.

  x y
  V  î  ax î  a a 
x x
expansion in x
expansion at point x


Now consider a velocity field given by V  xî  yĵ . contraction in y
y
  ˆ 
  V   î
 
 j  k̂   xî  yĵ  
x y
 
 11  0 expansion in x
 x y z  x y
x

In this latter case, the divergence is zero, indicating that the fluid is nonexpanding,
and any expansion in the x-direction (a positive change) is balanced by a contraction
(a negative change) in the y-direction. This balanced expansion/contraction is
reflected by increasing or decreasing velocity components to and from a point.

Example: Divergence in Cylindrical Coordinates

As was noted in section 2.2.1, when dealing with the differentiation of a flow field
described in cylindrical coordinates, one must be careful to perform any indicated
differentiation (of both the vector magnitude and the unit vectors) first, and then
perform the prescribed vector operation. To illustrate the importance of this process,

consider the divergence of a velocity field given by V  V ˆi , where V0 is a constant.
0 r

Here, the divergence is given by (using Eq. 2.7)

   1   
  V   îr  î  îz   V0 îr
 r r  z

 1 
 îr  V0 îr  î  V0 îr  îz  V0 îr
r r  z
î
0 0 0 0 0
 V ˆi  1  V ˆi   V ˆi 
 ˆi r   0 ˆi r  V0 r   ˆi   0 ˆi r  V0 r   ˆi z   0 ˆi r  V0 r 
 r r  r      z z 
 

1 V
 ˆi  V0 ˆi  0
r r

27
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 2

Here, the divergence of the flow field is infinite at the origin, and diminishes rapidly
as the flow moves radially outward (a bit like a radially-symmetric explosion). While
not a terribly realistic fluid flow, this example does illustrate the importance of
completing the unit vector differentiation before performing the vector operation.
Note that if we assumed that the derivative of the îr unit vector was zero, our result
would also be zero, which certainly is not the case.


Note that for a general vector field of V  Vr îr  V î  Vz îz , the divergence in
cylindrical coordinates can be shown to be:

 1
V  rVr   1 V  Vz  Vr  1 V  Vz  Vr (2.10)
r r r  z r r  z r

Here, the last term in Eq. 2.10 is the result of the change in the îr unit vector with
îr
angular displacement (i.e. a non-zero derivative of the unit vector,  i  ).


2.3.3 Cross Product or Curl

Next, we consider the cross product of a vector field, with the  operator again prescribed
to operate (left side of the cross product) on the vector field (right side of the cross

product). The result for a velocity field V (in Cartesian coordinates) is:

  
  V   ˆi  ˆj
 ˆ  

 k   Vx ˆi  Vy ˆj  Vz kˆ 
 x y z 

(2.11)
 Vz Vy ˆ  Vx Vz ˆ  Vy Vx  ˆ
   i    j    k
 y z   z x   x y 


Equation 2.11 gives a resultant vector often termed the “curl” of V . The curl indicates
the amount of angular change or rotation that occurs within a vector field due to changes
in the vector property in directions normal to each coordinate direction. If the vector
field is a velocity field, the cross product will give the angular velocity of the fluid, which
is expressed as a proportional “vorticity” in fluid mechanics. We will show later (in

Chapter 7) that when   V = 0, a velocity field is termed to be “irrotational” -- a

28
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 2

condition which allows some significant simplifications to be made to the governing

equations of fluid mechanics.

Example: Consider a two-dimensional


velocity field given by V  byî  Vx î (a simple uniformly sheared flow, like a
Couette flow, as shown). Eq. 2.11, the cross product for this two-dimensional flow,
simplifies to:
0
  Vx   Vx   (by )  (by ) y
V    ĵ    k̂  ĵ  k̂
 z   y   z y
 bk̂
x
Thus, the rotation of the fluid would be like a ball rolling left to right (or clockwise

around the z-coordinate, which points out of the page). In contrast, if V  ax î (the

flow field shown to have non-zero divergence in section 2.3.2), then   V  0 ,
indicating that although the flow is expanding, it has no rotation. This is also the case

for the second flow field we examined in section 2.3.2, V  xî  yĵ . For that case,

again   V  0 , which indicates a non-expanding field; additionally, one can show

that   V  0 , which indicates that the flow field also exhibits no rotation.

Note that in cylindrical coordinates, the curl for a velocity field given by

V  Vr îr  V î  Vz îz will be:

   1  ˆ  
  V   ˆi r  ˆi  iz  
  Vr ˆir  V ˆi  Vz ˆiz
 r r  z 
 1 Vz
 
V   V V 
   ˆi r   r  z  ˆi    
1  V 
rV  r  ˆi z
 (2.12)
 r  z   z r  r  r  
 1 Vz V  ˆ  Vr Vz  ˆ  V 1 Vr V  ˆ
    ir     i    iz
 r  z   z r   r r  r 
 
Here, the last term in Eq. 2.12 is the result of changes in the î unit vector with
î
angular displacement (i.e. a non-zero derivative of the unit vector,  i r ). Thus,

taking the cross product yields an extra z-direction component.

29
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 2

2.3.4 The Laplacian and Other Useful identities Using the  Operator

In the derivation and presentation of the various equations of fluid mechanics, it is

common, and often useful, to present portions of the equations in terms of functions of
the  operator. This provides both a short cut method for concisely representing
manifold complex terms, and also allows equations to be represented in a more generic
fashion, applicable to all coordinate systems. Among the  operator functions that prove
particularly useful is the Laplacian, 2, which evolves from equations which incorporate
the divergence of the gradient of a scalar property. Examples of such scalar fields in
fluid mechanics include pressure, temperature, or density fields. For example, if we
consider a scalar field, , in Cartesian coordinates the divergence of the gradient of  is
written:

           2  2  2
 2       î  ĵ  k̂    î  ĵ  k̂   2  2  2
 x y z   x y z  x y z

So here, the Laplacian 2 (generally read “del squared”) is:

2 2 2
2    (2.13)
x 2 y2 z 2

Equation 2.13 is a nice compact way to express the sum of the second-order derivatives
of a scalar with respect to all coordinate directions. Note that an equation of the form
 2   0 is classically termed the Laplace equation (which arises in many different areas
of engineering and science), and thus the designation of 2 as the Laplacian.

In cylindrical coordinates, the Laplacian is of course more complicated, due to the need
to perform the unit vector derivative operations with respect to , and is given by:

21     1 2 2 1  2 1 2 2
  r        (2.14)
r r  r  r 2  2 z 2 r r r 2 r 2  2 z 2

Additional term due to

unit vector differentiation

Note that the Laplacian function in cylindrical coordinates is often written in the more
“compact” manner on the left side of Eq. 2.14. However, writing the expanded function,
to the right, illustrates the additional term that arises from unit vector differentiation

30
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 2

There are a number of other -based functions that are commonly employed to
consolidate differential vector equations, and which are often confusing if one doesn’t
follow the proper order of operation (i.e. which operation you perform first). The general
rule for multiple operations is that one performs the right-most operation first, and then
 
work back to the left. For example, for     F one would perform the curl of F first

  F, then perform the curl of that result [i.e.     F ]. If for     F we
intended the operation on the left to be performed first, we would use parentheses to
indicate a variance in the order of operation. For example, if the expression was written

as     F , we would perform the del cross product first, then take the cross product

of the result with F . The catch here is that sometimes certain operations don’t work in
the right-to-left hierarchy, and thus inherently must be performed in the only logical
manner available (I’m not sure how this practice evolved, but it has). For example,
 
consider F  F . By performing the right hand operation first, this would suggest that we

should have  operate on F first. However, a vector operation (dot or cross product) is
 
not indicated for F , and this can’t be a gradient operation, since F is a vector. So one

must perform the only vector operation available, F   , first and then (since the result is a

non-vector functional operation) operate on F. So, in essence, the proper way to write
 
this operation should be: F   F .  

Another confusing function is the Laplacian operating on a vector,  2 F . Here, if we
 
follow the origin of the Laplacian above, we would write:  2 F    (F) , which again
doesn’t make sense. So, what is generally implied by this representation is:

  2 2 2  
 2 F   2  2  2 F     F (2.15)
 x y z 

A subtle point, but one that can be confusing when faced with manifold “compact”
representations of vector equations.

One of the nice aspects of using the  operator is its capability to express more
voluminous governing equations in compact form (as we shall see in Chapters 4 and 5).
Once an equation is written in this consolidated format (like the Laplace equation,
 2   0 ), it is applicable to a flow in any general coordinate system. In addition, certain
operator functions can often be expanded into sometimes more convenient identity
operator functions, which may often assist in the simplification, interpretation, or
metamorphosis of the original equations. Some of the most common identities

31
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 2

employing the  operator are listed below (Where f and g represent generic scalar
 
functions, and F and G represent generic vector functions):

(a)   f  0 (2.16)
(b) fg   f g  gf


(c)     F  0 
  
    
 

(d)     F     F  2F (i.e.     F    F )
  
   
(e)   fF  f   F  F  f
  
   
(f)   fF  f   F  f  F
         
         
(g)  F  G  G   F  F   G  F    G  G    F
         
      
(h)   F  G  G   F  F   G  F   G  G   F  
     
    
(j)   F  G  G    F  F    G 
(k)   f   2 f
  
(ℓ)   F   2 F (i.e.     F )
 
(m)    2 F   2   F  
 
(n)    2 F   2   F  
Although some of the expressions in Eq. 2.16 look a bit daunting, we will show in the
following chapters how some of these identities can be put to good use in the derivation
and evolution of the governing equations of fluid mechanics.

2.4 Review of Line and Surface Integral Theorems

There are several theorems of vector calculus that prove very useful in either developing
or simplifying the equations of fluid mechanics, many of which reflect the use of the 
operator in some form. Most of these evolve from the divergence theorem of Gauss,
which demonstrates a remarkable equivalence between the integration of a vector
property over a surface and the volume integral of the divergence of the property over the
volume encompassed by the surface. Here, we expand the functional applications of
Gauss’ theorem to include not only the divergence, but also the curl of a vector field, and
the gradient of a scalar field as well. Note that we will not develop a mathematical proof
of these theorems, but will simply state and explain their use, and the applicability to the
equations of fluid mechanics. For further background on the development, and the
application of these theorems, one is referred to the very useful book by Schey (1997) or
any good book on vector calculus.

32
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 2

2.4.1 The Gauss Divergence Theorem


If F is any continuously differentiable vector field, extending over a volume, , which is
encompassed by a surface area, A (e.g. the velocity field inside of a ball-shaped spatial
volume), then:

   

(  F )d   n̂ 
A
F dA    dA
F
A
(2.17)


(where n̂ is the outward normal unit vector to the surface, such that dA  dAn̂ )

Recalling that the divergence of a vector field gives the degree of expansion (or
contraction) at a point, the left side of Eq. 2.17 is the summation of the divergence of the
vector field within a specified volume. In fluid mechanics one construes this as the
change in the velocity field within a bounded volume (again, think of the volume
enclosed by a ball-shaped region). This expansion (or contraction) is equivalent to the
right side of Eq. 2.17, which represents the “flux” of a vector field across the surface
surrounding the volume in question. The flux is simply the component of the vector field
that crosses the bounding surface (e.g. the component of a velocity field that is normal to
the surface). For example, if a fluid is flowing in or out of our envisioned ball, and thus
crossing the ball’s surface, then there is a flux of velocity in or out of and across the
surface. On the other hand, if a fluid (or some component of the velocity vector) flows
parallel to the surface bounding the ball, there will be no flux of the velocity (or that
component) across the surface.

What this divergence theorem describes is essentially the principle of conservation of

mass in fluid mechanics. As we will show in Chapter 5, this theorem also proves quite
useful in demonstrating the coincidence between the governing integral equations of
fluid mechanics for a control volume, and the comparable point-wise differential
equations.

2.4.2 The Curl Theorem

This is a variation on the divergence theorem, in essence using the same mathematical

arguments that Gauss employed, but for the curl of F instead of the divergence. This
yields the following relationship, which is similar to Stokes theorem for a bounded
surface, which we discuss below in Section 2.4.4.

   
 
(  F)d   n̂  FdA    F  dA
A A
(2.18)

33
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 2


Equation 2.18 equates the summation of the curl of F within a fixed volume to the
negative of the surface integral of the cross product of the vector field over the bounding
surface. At first glance, Eq.2.18 seems like a rather complicated expression, with unclear
physical relevance. However, if one thinks in terms of a velocity vector field, the left
hand term of Eq. 2.18 reflects the summation of the rotational behavior within a defined
volume, which is proportional (for incompressible flows) to the angular momentum of
the fluid within that volume. This equates to the right hand term of Eq. 2.18, which is the
summation of the component of the velocity that acts parallel to the bounding surface

surrounding the fluid volume (recall that dA always points out and perpendicular to the
  
bounding surface, thus the cross product F  dA will only include that component of F
oriented parallel to the surface). Equation 2.18 proves to be a particularly useful
relationship for deriving and simplifying the equations of fluid mechanics that relate to
the vorticity and circulation of a fluid, which we discuss in Chapters 10 and 11.

2.4.3 The Gradient Theorem

This is yet a further variation on the divergence theorem, again using the same
mathematical arguments, but for the gradient of a continuously differentiable scalar
function, , giving:


 ()d   n̂dA   dA

A A
(2.19)

The left side of Eq. 2.19 is the summation of the gradient of  over a fixed volume.
Recalling that the gradient reflects the degree and direction of change of a scalar
property, the left side reflects a summation of the total spatial change of the property 
over a fixed volume. The right side of Eq. 2.19 is the summation of the scalar  acting on
the surface bounding the volume . As we will show in Chapter 7, Equation 2.19 proves
particularly useful for assessing the behavior of the scalar properties of height and
pressure in the derivation of the Bernoulli equation for an inviscid fluid.

The previous three relationships are quite remarkable in that they provide vehicles for
assessing the behavior of a vector field within a fixed volume by simply assessing certain
characteristics of that vector field at the bounding surface of the volume. One might be
tempted to envision a similar diagnostic technique that would assess the internal behavior
of the human body, or a house, simply by assessing their external characteristics.
However, we (humans), and our houses, are not continuously differentiable,
homogeneous material, but are composed of a number of different media. So such an
approach, while attractive, won’t submit to the surface integral approach. One needs an
identifiable, contiguous medium to apply these volume-to-surface theorems.

34
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 2

2.4.4 Stokes’ Theorem

We often don’t need to consider the behavior for a volume, but are only interested in the
behavior on a two-dimensional surface bounded by a continuous curve, such as a circle.
Such a situation is addressed by Stoke’s Theorem, which implies that the integral of a

continuously differentiable vector field F around a closed curve C is equal to the integral

of the normal component of the curl of F over the surface bounded by C, or:

    
 n̂  (  F)dA   (  F)  dA   F  dr
A A C
(2.20)


Here, dr is a differential line element lying along curve C, and n̂ is a unit vector normal
to the surface.

n̂
C
A


dr

  
If one thinks of the vector field, F , as a velocity field, V , then the curl of V will give the
angular rotation or vorticity. So Eq. 2.20 provides a method to relate the collective
vorticity over an area (which we will learn in Chapter 8 is the circulation, or total
rotational strength of an area of flow) to the line integral of the velocity around the
periphery of that area. Equation 2.20 is again a remarkable relationship, since it implies
that we can assess the rotational strength of a vortical flow by simply measuring the
tangential component of velocity at the edge of a region. This allows (for example) one
to assess the “strength” of a tornado or a hurricane by appropriate measurements at the
edges of those huge cyclonically rotating regions (of course, measurement of those edge
velocities is still no easy process). We will examine and make broad use of the Stokes
theorem when we derive various vorticity and circulation theorems in Chapters 8, 10, and
11.

References

Schey, H.M. (1997). div grad curl and all that, W.W. Norton & Company, New York.

35
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 2

Problems

1. If  is a scalar function, show by expansion in Cartesian coordinates that the following hold:

a. That     0
b. That  is normal to a line of constant 

2. If V is a vector function, show by expansion in Cartesian coordinates that the following hold:


a. That     V  0 
 V2 
b. That V   V    V  V   
 2 
c. That   V  V is normal to V

3. Let V  yˆi  x 2 ˆj . Show that Stoke's Theorem ( i.e.  V  ds   n̂    VdA ) holds

S A
over a square with corners at (x,y) = (1,1), (1,-1), (-1,-1), and (-1,1).
 
 V  ds   n̂    VdA ) holds

4. Let V  y 2 î  xyĵ . Show that Stoke's Theorem ( i.e.
S A
over a square with corners at (x,y) = (2,2), (2,3), (3,3), and (3,2).

5. In cylindrical coordinates
 1  
  î r  î   î z
r r  z
and
  r, , z  and V  v r ˆir  v  ˆi  v z ˆiz

a. Determine an expanded expressions for   

b. Determine an expanded expression for   V

6. If F  yˆi  x 2 ˆj and f = x + y2, by calculation of the left and right functions in the
equations, show that the following are valid:

  
   
a.     F     F  2F
  
   
b.   fF  f   F  f   F


7. If F  rˆir  rˆi and f = r + r, by calculation of the left and right functions in the equation,
show that the following is valid:

  
   
  f F  f   F  f   F

36
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 2

 
8. If F  yˆi  x 2 ˆj and G  y 2 ˆi  xyˆj , by calculation of the left and right functions in the
equation, show that the following is valid:

     
  
  F G  G   F  F  G   
 
9. If F  xyˆi  x 2 ˆj , determine   F , and then show that the following is valid:



 F  0 
 
10. If F  rˆir  r 2 ˆi , determine   F , and then show that the following is valid:



 F  0 

37
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 3

Chapter 3

Kinematics, Flow Lines, and the Substantial Derivative

Contents

3.1 Acceleration of a Fluid Particle in a Velocity Field .......................................................... 39

3.2 Flow Lines .................................................................................................................... 47

3.2.1 Flow Line Examples: Steady Flow .......................................................................... 49
3.2.1.1 Path line: Steady Flow .............................................................................. 49
3.2.1.2 Streak line: Steady Flow ........................................................................... 50
3.2.1.3 Stream line: Steady Flow .......................................................................... 51
3.2.2 Equivalence of Flow Lines in Steady Flow .............................................................. 52
3.2.3 Flow Line Examples: Unsteady Flow ..................................................................... 53
3.2.3.1 Path line: Unsteady Flow .......................................................................... 53
3.2.3.2 Streak line: Unsteady Flow ....................................................................... 54
3.2.3.3 Stream line: Unsteady Flow ...................................................................... 55
3.2.4 Non-Equivalence of Flow Lines in Unsteady Flow .................................................. 56

3.3 The Substantial (or Material) Derivative ........................................................................ 57

3.4 The Substantial Derivative: Cartesian and Cylindrical Coordinates .................................. 60

3.4.1 Substantial Derivative in Cartesian Coordinates .................................................... 60
3.4.2 Total Acceleration in Cartesian Coordinates .......................................................... 60
3.4.3 Substantial Derivative in Cylindrical Coordinates .................................................. 61
3.4.4 Total Acceleration in Cylindrical Coordinates ........................................................ 61

The assessment and prediction of the impact and consequences of fluid motion require
that we develop convenient physical descriptions that allow the effective modeling,
mathematical formulation, and analysis of fluid behavior. This is simplified by
employing the concept of a fluid as a continuum, wherein fluid properties are considered
to be distributed smoothly throughout the space of consideration. This generally implies
that although a fluid is a collection of molecules, for essentially all cases of practical
interest these molecular collections will behave as a continuous medium (a gas at
extremely low pressures is a particular exception). Thus, when examining a fluid
property (e.g. density), we assume that the property displays no discontinuous behavior
within the region of consideration. Another way of putting this is that the spatial
derivatives of any fluid property (i.e. the change in a property in any direction) will
always have a finite value. This does not mean that a property cannot undergo large
changes, but just that it cannot undergo a discontinuous change.

38
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 3

To approach the modeling and eventual analysis of fluid behavior requires that we be
able to spatially describe the appropriate properties of a fluid within a region of interest
(e.g. within a pipe or over an airplane wing). We must also be able to account for
changes with time (if the flow behaves in a time-dependent manner), such as the increase
in the velocity of a water stream during a hard rain. Since we generally focus on the
behavior of incompressible fluids in this text, the primary properties of interest will be
velocity and pressure, with velocity being the more problematic property to deal with
appropriately. Of particular interest is how we properly account for fluid acceleration,
since acceleration is a key property in Newton’s second law of motion. Our ultimate
charge is to develop a form of Newton’s second law that can describe the behavior of a
continuously distributed fluid. However, application of Newton’s second law to a fluid
creates a fundamental problem, since this requires that we can somehow collectively
identify and follow the acceleration/deceleration of all the fluid particles comprising a
fluid, and likewise the collective forces acting on those particles. This suggests that if we
employ the Lagrangian formulations utilized for assessing the physics of discrete
particles, we could end up with a particularly difficult bookkeeping process, since a
Lagrangian approach considers the change in the spatial position of particular particles
as a function of time alone. However, by taking a more global point of view, we can
exploit what we term a field concept, where we view the collective motion of all particles
as a continuum within a specified region, or field.

3.1 Acceleration of a Fluid Particle in a Velocity Field

The use of a field concept is generally termed an Eulerian description, wherein we

assume that a fluid property can be described as a function of its position in space (x, y, z
in Cartesian coordinates) and its evolution with time (t). Thus, in a Cartesian system we
  
would describe a velocity field as V  V( x, y, z, t ) , where V( x, y, z, t ) is the Eulerian
velocity, as illustrated in figure 3.1. This is somewhat akin to watching a river flow by
through a telescope focused on a single point in the river; the speed with which the water
passes across the telescope’s field of view is dependent upon the location that is viewed
by the telescope, and the specific time when we make our observation.

y

Vx, y, z, t 

Figure 3.1 An Eulerian velocity field

39
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 3

In contrast, a Lagrangian description focuses on the motion of one specific particle

within the velocity field, and tracks the positional change of that particle as a function of
time. This is like watching a small block of wood float by in that same river, and
tracking the position of that block of wood from an initial instant of time onward.

So, to summarize the difference between an Eulerian and Lagrangian description:

Eulerian – Specifies the velocity field (i.e. velocity behavior in space and time).
This approach specifies how a particle at a specified point will move
(think of it as viewing a series of particles passing by a small
observation window)

Lagrangian – Specifies how an individual particle will move through space as a

function of time (like watching a leaf floating in a river)

Now, Newton’s second law requires a balance between: (1) the forces acting on a
particle, and (2) the momentum change of the particle, which is proportional to its
acceleration. Our problem is how do we mathematically relate the local motion of a fluid
particle to the broader velocity field, of which the particle is a part, without having to
follow all of the individual particles? Let’s approach this by assuming that at any instant
of time, t, the location of a particle can be represented (here we use Cartesian

coordinates), by a position vector, r  xˆi  yˆj  zkˆ , as shown in figure 3.2. Note, that
although we perform this analysis in Cartesian coordinates, the result applies equally well
to any coordinate system.

y particle at time, t particle

 
path r  r (x, y, z)
particle at

   V( x, y, z, t )  uî  vĵ  wk̂ (Cartesian)
r r  dr time, t+dt
x

Figure 3.2 Representation of the Lagrangian change in position of a particle within

an Eulerian field.

Note that for the Cartesian coordinate system, as shown in figure 3.2, the components of
the velocity in the x, y, and z direction are generally designated u, v, and w. Now, as the
particle moves within the velocity field over an infinitesimal amount of time, t, it attains
a new position within the velocity field given by:

40
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 3

  
r  r  r  x  x ˆi  y  yˆj  z  zkˆ

where r is an infinitesimal position change that occurs over the time change t.

Assuming that this positional change takes place due to motion of the particle within the

velocity field, then the velocity of the particle, Vp , at the initial time t and the final time
t+t are given by:
 
Vp  V( x, y, z, t ) at time t,
t

and
 
Vp  V( x  x, y  y, z  z, t  t ) at time t  t .
t  dt


To establish the mathematical change in Vp over this time change, we can apply the chain

rule of differential calculus, which allows us to express Vp as:

   
   V V V V
Vp  Vp  Vp  x p  y p  z p  t
t  t t x y z t

where xp, yp, and zp are the differential changes in the relative coordinate position of
the particle over the differential time change, t. We recognize that the differential
displacement components of the particle are given by the product of the local velocity
component and the differential time change (i.e. xp = ut, yp = vt, and zp = wt),
since the particle will move at the local Eulerian velocity at its specified location within

the velocity field. Thus, we can write the expression for Vp as:

       
 V V V V  V V V V 
Vp  u t  v t  w t  t   u v w  t

x y z t  x y z t 

Now, we divide through by t and take the limit of this differential velocity change,
 
Vp / t , as t0. The result is the total acceleration for the particle, a p , moving

within the Eulerian velocity field, V( x, y, z, t ) :

    
 Vp V V V V
a p  lim  u v w
t 0  t x y z t

 
We reflect this particle acceleration by a p  DV Dt , where the derivative operator,

41
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 3

D( )/Dt, is known as the “total”, “substantial’, or “material” derivative (more on this

later). Thus, we write:
    
 DV V V V V
ap  u v w  (3.1)
Dt x y z t

Total acceleration Advective Local

of a particle acceleration acceleration

This general derivative operator, D( )/Dt, can also give the temporal changes in other

particle properties as the particle moves within an Eulerian velocity field, V( x , y, z, t ) .
For the present case, we are concerned with the changes in the particle velocity itself.
However, this total derivative operator can also be used to mathematically establish
changes in any other property associated with a material particle, such as density,
temperature, or position, as we will discuss in Section 3.3.

Note that this particle acceleration within a velocity field is comprised of two different
types of velocity change. The first of these is a local rate of change, which is a result of
local temporal changes in velocity taking place at the point of interest. For example, if
the velocity of a river increases with time at a specific point, that would reflect a local
acceleration of the flow at that point. The second type of acceleration is the result of the
advection (transport) of a particle through a spatially-varying velocity field. As a particle
is transported through a region of velocity change, that results in a corresponding change
in the velocity of the particle. This transport-based acceleration is known as advective
acceleration, since the velocity of the particle will change due to the advection
(movement) of the particle through a spatially varying velocity field. Using our river
example again, if the banks of a river contract inward, such that the velocity of the river
increases as it flows downstream, then a particle can be viewed to be accelerating at a
point due to this advective increase in velocity. Clearly, the total acceleration of a
particle is the sum of both the local, temporal changes in velocity and the changes in
velocity due to advection through a spatially-varying velocity field.

Du u u u u
a xp  u v w 
Dt x y z  t
Dv v v v v
a yp  u v w  (3.2)
Dt x y z  t
Dw w w w w
a zp  u v w 
Dt x y z  t

The acceleration of a particle, as derived above, is a vector derivative, comprised of three

separate components of acceleration, shown in Eq. 3.2 (in Cartesian coordinates).

42
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 3

Equation 3.2 gives the components of acceleration for a particle at an arbitrary position
within the Eulerian velocity field. This means that Eq. 3.2 must also represent the
Lagrangian acceleration of that particular particle, but determined by means of an
Eulerian analysis. This is significant since this provides a clear connection between
Newton’s second law for Lagrangian particle mechanics and the acceleration of all fluid
particles within a continuously distributed fluid. This connection is important in making
use of Newton’s second law to derive the governing differential equation of momentum
for fluid motion. However, let’s first assure ourselves that the acceleration reflected by
the Eulerian equations derived above is truly identical to the comparable Lagrangian
acceleration of an arbitrary particle.

Example: Here we examine a relatively simple Eulerian flow field, from which
we will determine the Lagrangian velocity and acceleration for a specific particle moving
within that flow field. We then use this information to make a comparison of: (1) the
Lagrangian acceleration of that particular particle when it passes through a specified
location within the velocity field, with (2) the general Eulerian acceleration at the same
specified location within the velocity field.

Let’s first assume a simple two-dimensional Eulerian velocity field given by


V  (3x  t )î  (3y) ĵ , with x and y-direction velocity components u E  3x  t and
v E  3y , respectively. Here the E subscript indicates that these are Eulerian velocity
components. Now, we will use this velocity field information to establish the following
Lagrangian information for the movement of a particle within the velocity field:

x L  f1 ( x init , y init , t )  x  x init  1

a)  for L @ t  0;
y L  f 2 ( x init , y init , t ) y L  y init  2

Equation for the path Initial starting point of

of the particle the particle

b) u L and v L ;

c) a x L , a yL ; and,

d) that a x L  a x E and a yL  a yE ,

Here the L subscript indicates the Lagrangian expression for velocity or acceleration.
Note also that f1 ( x init , yinit , t ) and f 2 ( x init , yinit , t ) are the Lagrangian functions describing
the x and y location of a particular particle that started at an initial position x Linit  x init

43
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 3

and y Linit  y init at an initial time of t = 0. To derive the Lagrangian behavior of the
particle within the velocity field, we assume that whenever a particle is at a particular
position within the field, it moves at the velocity for that position, as described by the
velocity field, and thus u L  u E and v L  v E .

Solution: Lagrangian position of a particle

dx L df 1
Part a) Let u L    u E ( x, y, t )  u E ( x L , y L , t )  3x L  t  3f1  t
dt dt
df 1
Thus,  3f1  t , which is a differential equation for f1 .
dt

To solve this differential equation, we need to evaluate both the homogeneous and the
particular solutions of the equation, since the equation is linear with constant
coefficients, but is inhomogeneous. Therefore, we assume that the total solution is a
linear combination of the homogeneous and the particular solutions of the equation,
i.e.

f1Total  f1h  f1p

Thus, for the homogeneous solution:

df 1h df 1h
 3f1h  0   3dt  f1h  C1e 3 t
dt f1h
Here, C1 is an undetermined integration constant.

Considering the particular solution, we assume a general of solution of the form

f1P  A  Bt and substitute this first-order polynomial into the differential equation to
determine the coefficients A and B, which yields:

 B 1
df 1p B  3A  0  A  3   9
 3f1p  (B)  3(A  Bt )  t  
dt 1
  3B  1  B  
 3

Equating coefficients of
like powers of t
1 1
Thus, f1p    t
9 3
Now, summing f1h and f1p to get f1Total , gives:

44
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 3

1 1
f1  f1Total  C1e 3t   t
9 3

Applying the initial condition, f1  x init  1 at t  0 , we determine C1:

1 1 1 10
f 1  1  C 1 e 3( 0 )   ( 0 )  C 1   C1 
9 3 9 9

Thus, the Lagrangian x-position of a particle passing through x = 1, y = 2, @ t = 0 is:

10 3 t 1 1
xL  e   t
9 9 3

Similarly, to determine the function f2, we set:

dy L df 2
vL    v E ( x L , y L , t )  3y L  3f 2 (a homogeneous equation)
dt dt

df 2
So,  3dt  f 2  C 2 e 3 t
f2
Thus, for the initial condition f 2  y init  2 at t  0 :

f 2  2  C2e3(0)  C2  2
and,

y L  2e 3t

This is the y-location for of a particle passing through x = 1, y = 2, @ t = 0. x L and

y L now give the location of that particular particle at any time t > 0.

Now, using the solutions for x L and y L , we differentiate to obtain the Lagrangian
velocity and acceleration:

dx L 10 3t 1 1
Part b): u L   e   (10e 3t  1)
dt 3 3 3
dy
vL  L  6e 3t
dt

45
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 3

du L
Part c) a xL   10e 3 t
dt
dv
a yL  L  18e 3t
dt

Part d) The Eulerian acceleration we obtain from Eq. 3.2 in terms of the general
position of a particle within the velocity field as:

Du E u u u u
a xE   u E E  vE E  w E E  E
Dt x y z t
 (3x  t )(3)  (3y)(0)  (0)(0)  1
 9x  3t  1

Dv E v v v v
a yE   u E E  vE E  w E E  E
Dt x y z t
 (3x  t )(0)  (3y)(3)  (0)(0)  (0)
 9y

Now, since x L and y L represent the sequence of positions of a particular particle as

it moves through the Eulerian field, we can now substitute the functions for x L and
y L into the Eulerian acceleration equations, to give us the comparable acceleration
expressions for that particular particle as it passes through the velocity field.

10 1 1 
a xE  9  e 3 t   t   3t  1
9 9 3 
 10e 3 t  a xL and

a yE  9(2e 3 t )  18e 3 t  a yL

Not surprisingly, the Lagrangian and Eulerian expressions for the acceleration of the
particle reduce to identical values, which illustrates that the Eulerian expressions apply
not only for the particular particle we assessed in this example, but generically for any
particle initiating from any position at any time. You can further demonstrate this
equivalency by assuming a different initial position (e.g. xinit, yinit = 2, 2) and/or a
different initial time (and thus a different Lagrangian particle), and demonstrate that for
 
all particles a E  a L . Note that often the velocity component expressions may depend on
more than one variable, creating a series of coupled differential equations. To determine
the Lagrangian behavior of these sets of equations will generally require the use of
Laplace transforms, or (for non-linear cases) numerical solutions.

46
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 3

Thus, the Total derivative of Eqs. 3.1 and 3.2 now provides us with a tool with which we
can accurately assess the Lagrangian acceleration of any and all fluid particles within an
Eulerian flow field. We will put this tool to good use in Chapter 5, where we derive the
governing differential equations of fluid motion.

3.2 Flow Lines

In the study of fluid mechanics, we often gain important information regarding a fluid
flow by experimental observation techniques involving the introduction of a visible
marker into the fluid, such as the injection of dye into water or smoke into air. We can
also make good use of ad hoc observations of some natural contaminant in a flow, such
as wind-blown dust or suspended dirt in a water flow. Additionally, we can develop
physical insight into the voluminous results of large computational studies by
establishing the behavior of a particular type of flow line, known as a stream line, which
illustrates the directional flow patterns within a flow. In this section, we examine three
different types of descriptor flow lines, termed path lines, streak lines, and stream lines,
all of which can help us both visualize and assess flow behavior.

The first descriptor line we consider is a path line. As the name implies, this is a line
traced out by a selected particle moving within a flow over a given time interval, thus
visualizing the path that the particle follows. Note that since we follow an individual
particle when describing a particular path line, the Lagrangian position functions for the
particle, x L and y L , (like we derived in the example in Section 3.1) describe the series
of locations comprising a path line. A rough example of a path line would be a line
connecting the temporal sequence of locations that a piece of wood passes through as it
floats along a river. We could capture such a line if we were to take a photographic time
exposure of the block of wood moving with the current of a river, or of any other easily
visible object moving within a velocity field.

The second descriptor line we find useful is a streak line. Here, the name of this line is
not as clearly descriptive of its characteristics, and it takes a bit of study to appreciate
what it represents. A streak line describes the present position of all particles that
initially passed through a specific point at some earlier point in time. So, if we can
imagine coloring each fluid particle that passes through some fixed location, and then
observing the trail of colored particles that emanate away from that point in space, this
trail of colored particles will form a streak line. A reasonable approximation of such a
line can be generated by injecting dye into water or smoke into air at a fixed point (e.g.
injection of dye via a hypodermic needle into a water flow). Note, however, that to
properly illustrate a true streak line, such injection processes need to be done very
carefully to avoid disturbing the original flow by the injection process.

47
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 3

In order to obtain a mathematical description of a streak line within a given velocity field,
we have to determine a series of Lagrangian functions for particles that initiate at the
same position in space, but over a sequence of initiation times. Thus, we will need to
generate a set of Lagrangian functions, x L (xinit, yinit, tp, ti) and y L ( xinit, yinit, tp, ti). Here
xinit and yinit are the initial x and y locations at particle insertion, ti is the time of the
insertion of the particle at the initial position (xinit, yinit), and tp is the present time of
interest (where tp > ti). Thus, by varying ti, we can determine the present location of all
those particles at some arbitrary later time, tp.

Our final descriptor line is a stream line. This is a line within a fluid for which the fluid
particles lying along the line all have velocity vectors that are tangent to the line, as
shown in two-dimensions in figure 3.3. This defines an interesting line of constraint.
Since the velocity is always tangent to a stream line, then there can be no flow across
(perpendicular to) a stream line. This is a very important property of a stream line --- one
that we will exploit when dealing with inviscid (non-viscous) flows, and use to help
simplify the governing equations of fluid mechanics.

differential element along stream line of


length d s  dx î  dy ĵ  dz k̂
y

 V
ds
x
z
Stream line

Figure 3.3 Illustration of the characteristics of a stream line

Additionally, for a two-dimensional, incompressible flow, the tangential velocity

constraint implies that the flowrate between two adjacent stream lines must always
remain constant. Thus, the spacing between stream lines is a general indicator of flow
acceleration or deceleration, with the velocity of the flow between two adjacent stream
lines increasing as the stream lines converge, and decreasing as stream lines diverge.
This is a bit of a simplistic interpretation, since we will see that there can be strong
velocity variations between stream lines, particularly when they are curved. However,
the convergence or divergence of stream lines provides a good visual indication of strong
changes in the local velocity field, and as such can be used for effective qualitative
interpretation of local flow behavior.

Now, let’s examine how we can develop a mathematical equation for a stream line.
Figure 3.3 is a sketch of a typical stream line relative to a Cartesian coordinate system.
On the line, we have indicated a velocity vector that is tangent to the stream line.

48
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 3

Additionally, we also envision a differential element along the stream line of length
 
d s  dx î  dy ĵ  dz k̂ , such that d s is also tangent to the stream line.

 
Now, if V and d s are both tangent to the stream line, then they are also parallel to each
other. Recalling that the cross product of two parallel vectors is zero, we can write:

 î ĵ k̂ 
   
V  d s  0   u v w   î ( vdz  wdy )  ĵ( wdx  udz )  k̂ (udy  vdx )  0
dx dy dz 
 

Since each component of this vector equation must be identically zero, we have three
governing differential equations for a stream line given by:

vdz  wdy ; wdx  udz ; udy  vdx (3.3)

For a three-dimensional flow, we would have to necessarily satisfy all three of these
equations. However, if we consider only a two-dimensional flow field (depending only
on x and y), the equation for the z-direction, udy  vdx , is the only relevant equation,
which can be rewritten as:

dy v
 (3.4)
dx u

Equation 3.4 is the differential equation of a two-dimensional stream line in x and y.

Note that this is also the slope of the stream line, which is the ratio of the respective
velocity components.

3.2.1 Flow Line Examples: Steady Flow


For the sake of example, consider a two-dimensional velocity field given by V  xî  yĵ ,
such that u = x and v = -y. Here, we will derive the appropriate equations for a path line,
a streak line, and a stream line, all passing through the location x,y=1,1.

3.2.1.1 Path Line: Steady Flow

To establish the path line for a particle inserted at point x,y = xinit,yinit = 1,1 at t = 0, we
determine the Lagrangian functions x L  f1 x init , yinit , t  and y L  f 2 x init , yinit , t  for the

velocity field V  xî  yĵ . Using the same approach as employed in Section 3.1, and

49
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 3

assuming that the velocity of a specific fluid particle at any position within the velocity
field is the velocity specified for that position by the Eulerian velocity field function, we
equate the following:

df 1 df 2
uL   f1 vL   f 2
dt dt

or, rearranging and integrating gives:

df 1 df 2
 dt  dt
f1 f2 (3.5)
x L  f1  C1e t y L  f 2  C 2 e t

Here f1 and f2 are the general Lagrangian functions for the x and y location of any
arbitrary particle moving within this velocity field. To establish the particular functions
for the initial position, x,y = xinit,yinit = 1,1 at t = 0, we substitute into our equations for
x L and y L ,and solve for the arbitrary constants C1 and C2.

x init  1  C1e ( 0 )  C1  1 y init  1  C 2 e ( 0 )  C 2  1

This gives the particular Lagrangian functions y

describing a path line for a fluid particle starting at
position x,y = xinit,yinit = 1,1 at t = 0 as: Path line passing
3
through (x,y)=(1,1)
2
x L  e t and y L  e  t
1
A sketch of this path line is shown at the right.
x
1 2 3
3.2.1.2 Streak line: Steady Flow

Here, we want to determine the location at the present time, t = tp, of all particles that
passed through point x,y = xinit,yinit between a series of insertion times, t = ti , and the
present time, t = tp, where tp  ti.

Using the general Lagrangian functions for a path line, Eq. 3.5, that we developed in
section 3.2.1.1, we set x,y = xinit,yinit at an arbitrary insertion time, ti. Here ti is any time
prior to the present time of interest, tp. This develops a general equation for the location
of all the particles inserted before t = tp. Then, by determining the present location of all
those particles at t = tp, we can connect all those particle locations to represent a streak

50
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 3

line. Thus, for a particle inserted at location x,y = xinit,yinit at t = ti, the Lagrangian
equation for a general particle location is given by:

f1  C1e t i  x init f 2  C2 e  ti  yinit (3.6)

From Eq. 3.6, we can solve for the constants, C1 and C2 as:

x init y init
C1   x init e t i C2   y init e t i
e ti e t i

Note that C1 and C2 are constants for any arbitrary insertion time, ti. Substituting the
values of C1 and C2 into the equations 3.6 for f1 and f2, gives the general equations for the
particle location at some later time, where t > ti:

x s  x init e  t i e t y s  y init e t i e  t
(3.7)
 x init e ( t  t i )  y init e ( t  t i )

In Eqs. 3.7, xs,ys give the x,y location of fluid particles comprising a streak line. If we
now set t = tp, Eq. 3.7 gives the present position of all the streak line particles, inserted at
earlier times, ti, as:
( t p t i ) ( t p  t i )
x s  x init e y s  y init e (3.8)
y Example of a streakline composed
Eqs. 3.8 now give the present locations of all of particles passing through a
particles that passed through x = xinit, y = yinit for point (x,y) = (1,1) for ti ≤ tp.
any time t < tp. To establish a streak line at the From ti<t<tp
present time, tp, we simply plot the positions for a ti = tp
series of particles inserted at a series of earlier ti< tp

times, ti < tp. An example streak line for
 x
(xinit,yinit)=(1,1) is shown at the right.

3.2.1.3 Stream line: Steady Flow


Applying Eq. 3.4 for a two-dimensional stream line to the velocity field V  xî  yĵ , we
can write:

dy v y
   , or
dx u x
dy dx

y x

51
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 3

y
Integrating gives:
C
ln y   ln x  ln C  ln  Stream line, xy = 1
x
passing through x, y = 1, 1
or
C
y   xy  C  const.
x
x

This is the general expression for any stream line within the specified velocity field. To
determine a particular stream line, we need to specify a point that we wish that stream
line to pass through. For example, if we want to identify the stream line that passes
through the point x,y = 1,1, we substitute those position values into the general equation
and solve for the stream line constant. For this particular stream line equation, we
determine that C = 1. Thus, xy = 1 is the equation of the stream line passing through
x,y = 1,1, and defines all other points that will lie along that particular stream line. A
sketch of this stream line is shown at the above right.

3.2.2 Equivalence of Flow Lines in Steady Flow

Note that for the example shown above, the path lines, streak lines, and stream lines are
all identical lines. One can demonstrate this in several ways. For example, if the
expressions determined in Section 3.2.1.1, 3.2.1.2, and 3.2.1.3 are used to determine the
slope of the lines at any point along the line (e.g. at x,y = 1,1), the slopes of the lines will
all be identical. Alternatively, if the Lagrangian position functions for the path and streak
lines are substituted into the stream line equation, they will identically satisfy the stream
line equation. Figure 3.4 shows plots of the path, streak, and stream lines determined in
Section 3.2.1, which illustrates the coincidence of all flow lines for steady flow.

1.2 1.2 1.2

1 1 1
0.8 0.8 0.8

y 0.6 y 0.6 y 0.6

0.4 0.4 0.4
Path Line Streak Line, tp = 1 Streamline
0.2 0.2 0.2
0 0 0
0 1 2 3 0 1 2 3 0 1 2 3
x x x
Figure 3.4 Plots of a path line, streak line (for tp =1, and 0 < ti <1), and stream line

passing through x,y = 1,1 for velocity field V  xî  yĵ .

52
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 3

This coincidence of path, streak, and stream lines is characteristic of steady velocity
fields, and is quite a useful experimental tool, since a photographic image of either a set
of path lines (using a time exposure of discretely added visible particles to a flow) or
streak lines (using a regular photograph of continuously injected material into a flow,
such as dye or smoke) will reasonably represent the behavior of the corresponding stream
lines. Since the stream line pattern can provide valuable information regarding the
pressure field of a fluid, as we will discuss in Chapter 7, such images can be quite helpful
in designing and assessing geometric modifications to reduce internal energy losses in
ducting or the drag on vehicles (i.e. thus the origin of the term “streamlining” of a car).

However, if the velocity field is locally unsteady, path lines, streak lines, and stream lines
cease to be coincident, and snap-shot or time-exposure images of particle or marker
behavior cannot be used to establish stream line behavior. An example of such an
unsteady flow is the unsteady flow separation from the rear of a bluff body, such as a
cylinder or truck traveling along a highway (which we discuss in Chapter 15). While
path, streak, and stream lines can be visualized or constructed (both experimentally and
computationally) for unsteady flows, one needs to be careful not to equate one type of
descriptor line with another. I illustrate the impact of unsteady flow on path, streak, and
stream lines in the following section.

3.2.3 Flow Line Examples: Unsteady Flow

To illustrate the non-equivalence of flow lines in an unsteady flow, let's determine the

path, streak, and stream lines for the velocity field V  ( x  t )ˆi  yˆj , with x and y-
direction velocity components u E  x  t and v E   y . Note that this is similar to the
velocity field we considered in section 3.2.1, but with a linear time variation added to the
x-direction velocity component. We again consider the flow lines for particles passing
through point x,y = xinit,yinit = 1,1, at an initial time, t = 0.

3.2.3.1 Path Line: Unsteady Flow

Writing our path line equations similar to section 3.2.1.1, we have:

df 1 df 2
uL   f1  t vL   f 2
dt dt

Here, our u L equation is inhomogeneous, so we need to use a technique similar to our

example in Section 3.1. The result from the integration of these two equations is:

x L  f1  C1e t  1  t y L  f 2  C2 e t (3.9)

53
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 3

The x L expression of Eq. 3.9 contains additional terms due to the time variation; the y L
expression is the same as we derived in Eq. 3.5 in Section 3.2.1.1. For a particle released
at t = 0 at location xinit,yinit = 1,1 , C1 and C2 are:

1  C1  1  C1  2 and C2  1
Thus, the corresponding path line expressions are:

x L  2e t  1  t y L  e t (3.10)

3.2.3.2 Streak Line: Unsteady Flow

Here, we again determine the location at the present time, t = tp, of all particles that
passed through point xinit,yinit = 1,1 between a series of insertion times, t = ti , and the
present time, t = tp, where tp  ti.

Using our general Lagrangian functions for a path line, Eq. 3.9, from section 3.2.3.1, we
set x,y = 1,1 at an arbitrary insertion time, ti, where ti is again any time prior to the
present time of interest, tp. This develops a general equation for the location of all the
particles inserted before t = tp. By determining the present location of all those particles
at t = tp, we can connect all those particle locations to represent a streak line. Thus, for a
particle inserted at location x,y = 1,1 at t = ti, the Lagrangian equation for a general
particle location is given by:

f1  C1e t i  1  t i  1 f 2  C 2 e ti  1 (3.11)

From Eq. 3.11, we can solve for the constants, C1 and C2 as:

C1  2  t i e  t i C 2  e ti

Since C1 and C2 are constants for any arbitrary insertion time, ti, we now substitute the
values of C1 and C2 into the equations 3.9 for f1 and f2, which gives us the general
equations for the particle location at some later time, where t > ti:

x S  2  t i e  t i e t  1  t yS  e ti e t
and (3.12)
 2  t i e t  t i 
1 t e t  t i 

In Eqs. 3.12, x S , y S give the x,y location of fluid particles comprising a streak line. If
we now set t = tp in Eq. 3.12, this gives the present position of all the streak line particles,
inserted at earlier times, ti, as:

54
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 3

t p  t i   
x S  2  t i e
 t p t i
1 t p and yS  e (3.13)

Eqs. 3.13 now give the present locations of all particles that passed through x = 1, y = 1
for any time t < tp. To establish a streak line at the present time, tp, we simply plot the
positions for a series of particles inserted at a series of earlier times, ti < tp.

3.2.3.3 Stream Line: Unsteady Flow


Applying Eq. 3.4 for a two-dimensional stream line with V  ( x  t )ˆi  yˆj , we can write:

dy v y dy dx
  , or 
dx u xt y xt

Integrating gives:

 C 
lny    lnx  t   ln C  ln 
xt
or
C
y  yx  t   C  const. (3.14)
xt

Eq. 3.14 is the general expression for any stream line within the specified velocity field.
To determine a particular stream line, we need to specify a point that we wish that stream
line to pass through and the time of interest. For example, if we want to identify the
stream line that passes through the point x,y = 1,1 at time t = 0, we substitute those
position values into the general equation and solve for the stream line constant. For this
particular stream line equation, we determine that at t = 0, C = 1. This would give an
equation for that streamline as:

xy  1 (3.15)

However, the equation for a streamline passing through x,y = 1,1 at time t = 1, would
yield an equation for that streamline as:

yx  1  2 (3.16)

These are different streamlines, due to the time variation of the flow field, as we show in
Figure 3.5.

55
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 3

3.2.4 Non-Equivalence of Flow Lines in Unsteady Flow

Figure 3.5 shows plots of the path, streak, and stream lines determined in Section 3.2.3.

1.2

Streak Line, tp = 1
1
Streamline, t = 0
Streamline, t = 1
0.8
Path Line, 0<t<1

y 0.6

0.4

0.2

0
0 1 2 3 4

x
Figure 3.5 Plots of a path line, streak line (for tp =1, and 0 < ti <1), and stream lines
for t=0 and t=1 passing through x,y = 1,1 for velocity field

V  x  t ˆi  yˆj .

Clearly, these are all different flow lines, since the velocity field varies with time, t. Note
that the path line (black) and the streak line (red) have coincident starting and termination
points. This is because all particles start at x = 1,1, and the particle released at ti = 0 will
trace the path shown (between 0 < t < 1) in figure 3.2. So at t = tp = 1, the end points
(particles) of the path and streak lines shown will both occupy the same location.
However, for 0 < t < 1 the path and streak lines will describe different lines.

The variation in the velocity field is quite clearly evident by the stream lines displayed.
Although the lines (Eq. 3.15 and 3.16) pass through the same initial location (x,y = 1,1),
because the u-velocity increases with time, the stream lines will skew in a counter-
clockwise manner, since the flow will move more rapidly in the x-direction with
increasing time.

Clearly, the various flow lines in figure 3.5 are not coincident, as they were in figure 3.4
for a steady flow. And as such, for an unsteady flow we cannot equate a path or streak
line as representative of a stream line. As we will show in Chapters 7 and 8, the flow
along a streamline can be directly related to the pressure field associated with a flow, and

56
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 3

the surface pressure effects (e.g. lift and drag) on bodies within the flow field. Thus, we
must be careful in our interpretation of path and streak lines as steam lines (see this link,
which give a good, dynamic illustration of the variations that can occur), which has often
been done erroneously in experimental studies using particle, smoke, and dye
visualizations. We will discuss the characteristics and impact of unsteady behavior
further in Chapters 7 and 8.

3.3 The Substantial (or Material) Derivative

In Section 3.1, we derived an expression for the acceleration of a particle in an Eulerian

velocity field, which led to a derivative operator termed the total derivative. However,
this derivative operator can be employed more broadly to establish the rate of change of
any property of a material fluid particle as a result of its movement through an Eulerian
velocity field. This of course assumes that we can also develop an appropriate Eulerian
field description for the continuously distributed property (e.g. density, temperature, etc.)
within the same spatial region as the velocity field. When such a property field exists, a
particle carried through the property field by the coincident velocity field will experience
changes according to the appropriate property field description.

To assess the generic property change due to the motion of a particle, let the function
F  F(x, y, z, t ) represent the Eulerian field description of any property (either scalar or

vector) coincident within the same spatial region of an Eulerian velocity field, V( x, y, z, t ) .

Using the same approach we employed in section 3.1 for the change of a particle velocity
within a velocity field, the change in the particle property for a differential time t, as the
particle undergoes differential displacements of xp, yp, and zp, can be written using
the chain rule as:

F F F F
F  x p  y p  z p  t
x y z t

Again, since these displacements will be the product of the local Eulerian velocity
component and the differential time change, we can write:

F F F F  F F F F 
F  u t  v t  w t  t   u  v  w  t
x y z t  x y z t 

Dividing through by the time differential, t, over which the property change occurs, and
taking the limit as t0, gives:

57
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 3

F DF F F F F
lim   u v w
t 0 t Dt x y z t

We again define this limiting change in the property F by the derivative operator, D( )/Dt,
which we termed in section 3.1 as the “substantial” or “material” derivative. Here the
definition of the term should be more obvious, since this operator yields the time change
of F (a substance or material property of a particle) as the particle is carried through the

property field, F(x, y, z, t ) , by the Eulerian velocity field, V( x, y, z, t ) .

DF F F F F
Thus,  u  v  w  (3.17)
Dt x y z t

Total rate of Rate of change of F due to Local rate of change

a change of advection through the of F at a point
property F velocity field

As we commented on in Section 3.1, this form of derivative operator, D( )/Dt, accounts

not only for the local change of a property with time, but also the changes created by the
advection of a fluid particle through a medium with spatial property variations (e.g. like
experiencing a decrease in temperature as you hike up a mountain).

Expressing the above equation in vector form gives:

DF  F
 ( V  ) F  (3.18)
Dt t

Or if the property is a vector, for example in Cartesian coordinates, f  f x ˆi  f y ˆj  f z kˆ ,we
can write:
 
Df   f
 ( V   )f  (3.19)
Dt t

A simple example of the application of the substantial derivative is an examination of the

change in a particle’s position with time as the particle passes through a velocity field.

To examine this, we define an arbitrary particle position as r  xˆi  yˆj  zkˆ , where x, y,
and z are the Cartesian position coordinates of the particle. Now, we seek the change in
the position of the particle, Dr Dt , as it passes through the velocity field. Here, equating
F = r in Eq. 3.17, we write:

58
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 3

    
Dr r r r r
u v w 
Dt x y z t

u

 xˆi  yˆj  zkˆ
v
  
 x ˆi  yˆj  zkˆ
w
  
 x ˆi  yˆj  zkˆ  xˆi  yˆj  zkˆ


x y z t

Or, considering only the non-zero derivative terms,


Dr
u
 xî  
v
 yĵ
w

 zk̂  
 0  uî  vĵ  wk̂ (3.20)
Dt x y z


So, D r Dt is the velocity of the flow field itself at the point r  xˆi  yˆj  zkˆ , as it should

be. Note that in Eq. 3.20 the term r t is zero since we are dealing with an Eulerian

description, and a specified point, r  f ( x, y, z) , and the position is not a function of time.

The general form of the substantial derivative derived above can be used to express the
changes for any property of a particle moving through that property field. For example,
the expression for a change in density,  (a scalar), is given by:

D    
u v w 
Dt x y z t

 
On the other hand, a change in vorticity,  (a vector), where   x ˆi  y ˆj  z kˆ , is
given by:

    
D    
u v w 
Dt x y z  t

Note that we will use both of these expressions for density and vorticity change later,
when we derive the respective differential equations for conservation of mass, and the
transport of vorticity.

Example: If we consider coincident velocity and density fields given by:

 y
V  U o ˆi and   o 1  ax  bt 
h

Then, the change in the density within the flow field is given by:

D    
u v w 
Dt x y z t

59
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 3

or,
D y  y 
 U o  o a   0  0   o  b    o   U o  a b 
Dt h  h 

Changes due Changes due

to advective to local time
motion variations of
through the the density
spatially field
varying
density field

3.4 The Substantial Derivative: Cartesian and Cylindrical Coordinates


3.4.1 Substantial Derivative in Cartesian Coordinates: V  uˆi  vˆj  wkˆ

DF F F F F
 u v w where F  Fx , y, z, t 
Dt t x y z
    
Df f f f f  
 u v w where f  f x , y, z, t   f x ˆi  f y ˆj  f z kˆ
Dt t x y z

 Df  f f f f
x-component:    x  u x  v x  w x
 Dt  x t x y z

 Df  f f f f
y-component:    y  u y  v y  w y
 Dt  y t x y z

 Df 
z-component:    f z  u f z  v f z  w f z
 Dt  z t x y z

3.4.2 Total Acceleration in Cartesian Coordinates

    
 DV V V V V
a  u v w Total acceleration
Dt t x y z
Du u u u u
ax   u v w x-component of acceleration
Dt t x y z
Dv v v v v
ay   u v w y-component of acceleration
Dt t x y z
Dw w w w w
az   u v w z-component of acceleration
Dt t x y z

60
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 3


3.4.3 Substantial Derivative in Cylindrical Coordinates: V  vr ˆir  vˆi  vz ˆiz

DF F F v  F F
  vr   vz where F  Fr , , z, t 
Dt t r r  z
    
Df f f v  f f  
  vr   vz where f  f r, , z, t   f r ˆi  f ˆj  f z kˆ
Dt t r r  z

 Df  f r f v  f  f
r-component:     vr r    r  f   vz r
Dt
 r  t r r    z

 Df 
-component:    f   v r f   v   f   f r   v z f 
 Dt   t r r    z

 Df 
z-component:    f z  v r f z  v f z  v z f z
 Dt  z t r r  z

3.4.4 Total Acceleration in Cylindrical Coordinates

    
 DV V V v  V V
a   vr   vz Total acceleration
Dt t r r  z
Dv r v r v v v v v 2
ar    vr r   r  vz r   r-component of acceleration
Dt t r r  z r
Dv  v  v v v v vv
a    vr      vz   r  -component of acceleration
Dt t r r  z r
Dv z v z v v v v
az    vr z   z  vz z z-component of acceleration
Dt t r r  z

61
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 3

Problems

1. A particular flow has the following field characteristics (assume these are non-dimensional):

velocity: V  xˆi  ( t  2y)ˆj  zkˆ
temperature: T  2x  yz
density:   1  ex

Determine (non-dimensionally) at time t = 1:

a. The acceleration field

b. The rate of change of  at a point (2,1,1)
c. The rate of change of temperature at point (1,3,2)

2. A particular flow has the following field characteristics (assume these are non-dimensional):

velocity: V  (2 y  y 2 )î  e y ĵ  2k̂

temperature: T  (9  x 2  z2 )e yt
density:   (1  x )e  y

Determine (non-dimensionally) at time t=0:

a. The acceleration field at (0,0,0)

b. The rate of change of  at a point (1,1,1)
c. The rate of change of temperature at point (1,0,1)

3. A non-dimensional velocity field in cylindrical coordinates is given by:

1
V   2 îr  4r 2 î
r 
Determine:

a. An expression for the acceleration of a particle anywhere within the flow field.

b. The equation for a streamline passing through the point (x,y) = (0,2); plot the
streamline from (x,y) = (0,2) to (0,0).

c. How long (in non-dimensional terms) it will take a particle to go from (0,2) to (0,0).

4. A non-dimensional velocity field in cylindrical coordinates is given by:

1
V    ˆir  4r ˆi
r
Determine:

62
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 3

a. An expression for the acceleration of a particle anywhere within the flow field.

b. The equation for a streamline passing through the point (x,y) = (0,2); plot the
streamline from (x,y) = (0,2) to (0,0).

c. How long (in non-dimensional terms) it will take a particle to go from (0,2) to (0,0).

5. A non-dimensional velocity field in cylindrical coordinates is given by:

1 ˆ  r
V   2  ir  4r1  ˆi
r   3
Determine:

a. An expression for the acceleration of a particle anywhere within the flow field.

b. The equation for a streamline passing through the point (x,y) = (0,2); plot the
streamline from (x,y) = (0,2) to (0,0).

c. How long (in non-dimensional terms) it will take a particle to go from (0,2) to (0,0).

6. Consider the plane flow field,

V  xˆi  ytˆj

Develop expressions for streamlines, pathlines, and streaklines through a point (a,b)=(1,1).
Plot: streamlines for t = 0, 0.5, and 1; a pathline at t=1 for a particle introduced at t=0; and a
streakline at t = 1 for all particles which passed through point (1,1) between t = 0
and t = 1. Show all plots on one, full-page graph..

7. Consider the plane flow field,

x ˆ
V( ) i  { y(1  t )}ˆj
1 t

Develop expressions for streamlines, pathlines, and streaklines through a point (a,b)=(1,1).
Plot: streamlines for t = 0, 0.5, and 1; a pathline at t=1 for a particle introduced at t=0; and a
streakline at t = 1 for all particles which passed through point (1,1) between t = 0
and t = 1. Show all plots on one, full-page graph..

63
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 4

Chapter 4

Motion and Deformation of a Fluid

Contents

4.1 Fluid Motions ............................................................................................................... 64

4.1.1 Translation .......................................................................................................... 65
4.1.2 Rotation .............................................................................................................. 65
4.1.3 Angular Deformation ........................................................................................... 66
4.1.4 Linear Deformation .............................................................................................. 67

4.2 Cauchy-Stokes Decomposition ...................................................................................... 68

4.2.1 Use of Tensors ..................................................................................................... 71
4.2.2 Decomposition in Tensor Notation ........................................................................ 72

4.3 Vorticity ....................................................................................................................... 75

4.4 Rate of Strain and Vorticity Equations: Cartesian and Cylindrical Coordinates ................. 78
4.4.1 Rate of Strain in Cartesian Coordinates ................................................................. 78
4.4.2 Vorticity in Cartesian Coordinates ........................................................................ 78
4.4.3 Rate of Strain in Cylindrical Coordinates .............................................................. 79
4.4.4 Vorticity in Cylindrical Coordinates .............................................................................79

In order to develop differential equations of motion for a fluid, we need to understand the
general types of motion we must consider. For simplicity, in this chapter we examine
motion in two-dimensions, which is sufficient for our purposes of examining the types of
motions that can occur, and determining how we can approach the mathematical
modeling of these motions. The objective here is to provide a physical background for
the development of the governing differential equations, which is done in Chapter 5.

4.1 Fluid Motions

Initially, we consider the general motions that a fluid can undergo. To do this, we
examine an infinitesimal differential element of dimensions x by y in a Cartesian
coordinate system, where the local vector velocity at the center
 y
of the element is assumed to be V  uî  vĵ . While the
v
dimensions of the differential element are considered
infinitesimal, we accentuate the size of the element here, for y u
the point of illustration. x
x

64
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 4

Note that for the all the behaviors considered here, the velocity vector components u and
v are assumed positive relative to the respective coordinate-axes. Basically, we recognize
that a fluid element can undergo translation, rotation, and angular and linear deformation
(i.e. distortion), depending on how the velocity components change from the center of the
element to the element boundaries.

In the following sections, we examine each type of motion individually by considering

the possible behavioral changes of the differential element over an infinitesimal time
interval, t .

4.1.1 Translation

Here we assume that the velocity across the ut

element is uniform (no changes in either coordinate
direction), which will only result in a translation of
the element. Assuming that t is so small that the
y vt
velocities remain essentially constant during the v
time change, the magnitudes of the translation in y u
the x and y coordinate directions are given
x
respectively by ut and vt, as shown in figure 4.1.
x
This results in the differential element retaining its
original geometry while translating to a new position Figure 4.1 Translation of a
in space. differential element

4.1.2 Rotation

To illustrate and examine rotational motion of our differential element, we next assume
that the velocities across the element change in a manner that will result in the pure
rotation of the primary axes of the element (i.e. two orthogonal line segments fixed in the
element, and initially aligned with the reference coordinate axes). We assume that these
changes yield a positive rotation of each line segment (counter clockwise for this case, by
the right-hand rule). We again assume that t is infinitesimal, and that the degree of
rotation of the element can be characterized by the collective displacement of the two
primary axes fixed in the element. We characterize the displacement of these axes by the
displacement of their extensions at the element boundary, as shown in figure 4.2.

This displacement is the result of a change, from the center of the element to the element
boundary, of the velocity normal to a line segment comprising a primary axis. For
example, for a line segment initially oriented along the x-axis, this velocity change is
given by v/x, or the derivative v/x for the limit of an infinitesimal change. Thus, the

65
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 4

subsequent displacement of the outermost point of this line segment is v x x 2 t ,
where x/2 is the length of the line segment from the center of the element to the outer
boundary, and t is the time interval of the displacement. Clearly, this change is actually
assumed quite small, although it is exaggerated in figure 4.2.

 u  y 
   t  assume -u/y to yield positive rotation
 y  2 
rotation about centroid due to motion normal to
x-direction with positive orientation

y v
y/2  v  x 
  t
u  x  2 
time interval
change in x from the center to edge of element
x/2 change in v with respect to x-direction
x
velocity normal to x-direction

Figure 4.2 Rotation of a differential element

We recognize that rotation of the element can occur only if the two orthogonal primary
axes execute identical displacements relative to the fixed coordinate axes, and remain
orthogonal. Thus, a similar argument is made for the displacement of the line segment
initially aligned along the y-axis, with its displacement being  u y y 2 t , which
will reflect the same magnitude of displacement as the other line segment, and yield a
positive rotation of the element. Note that we need to assume  u y  (where u is
assumed positive) to reflect a displacement yielding a positive rotation of the element.

Also, although the displacements are relative to the centroid of the element, this does not
mean that the element cannot be undergoing other simultaneous motions. Clearly, a flow
can translate, rotate, and deform concurrently (as we will discuss in Section 4.2)—here
we simply illustrate how those individual motions can be envisioned geometrically, and
then described mathematically.

4.1.3 Angular Deformation

To address deformation of a fluid, we first recognize that a fluid can undergo both
angular deformation and linear deformation. Here, we consider angular deformation,
which can be characterized by the same types of displacements that we examined in
section 4.1.2 for rotation. We again assume that the change in the velocities across the
element will yield a rotation of the primary axes (i.e. orthogonal line segments) of the

66
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 4

element about the element centroid. However, for angular deformation to occur, the
primary axes of the element are assumed to move in opposition to each other, as shown in
figure 4.3 (note that this differs from the conjunctive motion of the axes, which yielded
rotation). Note that since we assumed u and v to be positive, angular deformation will
occur when the extensions of the initially orthogonal line segments both undergo positive
differential changes. These positive changes reflect angular motion of the line segments
counter to each other, which results in a subsequent loss of orthogonality of the primary
axes, as shown in figure 4.3.

 u  y 
  t
 y  2  Note that the assumed positive changes create
angular changes in opposition to each other,
resulting in deformation, rather than rotation.
y y/2

 v  x 
  t
x/2  x  2 
x

Figure 4.3 Angular deformation of a differential element

4.1.4 Linear Deformation

For linear deformation to occur, the velocity component in a specified coordinate

direction must change in that direction. In such types of deformation, the extensions of
the primary axes remain orthogonal, but are assumed to undergo positive differential
changes in the coordinate directions as shown in figure 4.4. Thus, a positive differential
change in the x-direction will be reflected by u x  , resulting in a corresponding
displacement (lengthening) of the x-directed line segment by u x x 2 t , which
reflects a linear extension of the element in the x-direction. Correspondingly, a positive
change in the y-direction, reflected by v y  will result in a linear extension in the y-
direction of v y y 2t .

Note that figure 4.4 illustrates positive extensions in both coordinate directions, and thus
an overall expansion of the element, which might be expected for a compressible fluid,
such as air. However, whereas angular deformation can only occur by rotation of the line
segments in opposition to each other, linear deformation in either coordinate direction is
not constrained by such a requirement, if the fluid is compressible. However, as we will
show later, if a fluid is incompressible, conservation of mass requires that linear

67
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 4

deformations in the coordinate directions must act in opposition to each other to assure
that mass is conserved.
 u  x 
   t
 x  2 
y

 v  y 
  t
 y  2 

Figure 4.4 Linear deformation of a differential element

Of the described motions, which will result in stresses? Obviously, only those motions
that result in a relative change in the dimensions of the line elements x , y (which
implies the presence of strain) can create a stress. Thus, only angular and linear
deformations result in the generation of a stress within a fluid. As we will show, angular
deformation is proportional to a shear stress, and linear deformation is proportional to a
normal stress.

Since fluid behavior is a combination of all of these four fluid motions (translation,
rotation, and angular and linear deformations), we need to be able to express the
cumulative motion of a fluid in a compact, mathematical manner in order to perform
appropriate modeling of fluid behavior, and thus development of the governing
differential equations. In the following discussion of the Cauchy-Stokes decomposition
of fluid motion, we illustrate how we begin to mathematically model, represent, and
examine the differential motions of a fluid.

4.2 Cauchy-Stokes Decomposition

A useful first approach to understanding the motion and deformation of a deformable

medium (e.g. a fluid) is to consider an approach for “decomposing” all potential fluid
behaviors mathematically using what is known as the Cauchy-Stokes Decomposition
Theorem, named after the two gentlemen who developed the concept (see Truesdale
(1954) ). Although the theorem is generic to three-dimensions, we will examine a two-
dimensional version here, since the two-dimensional mathematics are easier to follow and
more physically instructive. And of course, what we learn from the two-dimensional
approach is directly extendable to three dimensions, as we will discuss later.

68
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 4

y

Vp  
   Vp  dV
r p dr
x

Figure 4.5 A conceptual model for two-dimensional Cauchy-Stokes decomposition.

We begin by considering the motion of a fluid element in a two-dimensional Cartesian

  
velocity field, V uî  vĵ , where V  Vx, y , as shown in figure 4.5. We assume that a
fluid element, or particle, moves within the velocity field from point p, at a vector
 
location r  xî  yĵ , across an infinitesimal differential vector distance, d r  dx î  dy ĵ ,
as shown.

Here, we realize that as the element moves along dr , it also moves through the velocity
 
field, V , such that its local velocity undergoes a change from Vp (the velocity at point p)
    
to a new value of V  Vp  dV at its new position at r  dr . Using the chain rule, we

reason that we can express the change in velocity, dV , as:

 
 V V  u u   v v 
dV  dx  dy   dx  dy î   dx  dy  ĵ
x y  x y   x y 

Thus, we can differentially express the velocity at the new position as:
 
    V V
V  Vp  dV  Vp  dx  dy (4.1)
x y

Expanding Eq. 4.1 in vector components, and noting the relationship of the individual
terms to the four types of motion we discussed in section 4.1, we obtain:

  u u  ˆ  v v 
V  ˆi u  dx  dy   j  v  dx  dy  (4.2)
 x y   x y 

translation ? translation linear

linear deformation
?
deformation

69
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 4

How do we make these identifications in Eq. 4.2? Well, the translation terms should be
obvious, since translation is directly related to the local velocity. And the linear
deformation terms are also reasonably obvious, since these reflect changes in the
direction of each velocity component, and are thus linear changes (e.g. u x is the
change of the x-direction velocity in the x-direction). However, the other derivatives are
more problematic (and thus the question marks), since they reflect changes normal to the
direction of the respective velocity component, which could reflect components of either
rotation or angular deformation. Since these terms represent a combination of behaviors,
we attempt to separate these terms further using some mathematical manipulation.

Let's first consider the x-direction term u y dy in Eq. 4.2. If we split this term in
half, and add and subtract a term 1 2 v x  dy (this contributes nothing additional, but
allows us to manipulate the initial term), we end up with an expanded expression:

u 1 u 1 u 1 v 1 v
dy  dy  dy  dy  dy (4.3)
y 2 y 2 y 2 x 2 x
     
split in halves add -subtract same term

Now, we rearrange the terms on the right side of Eq. 4.3 and obtain:

u 1  u v  1  u v 
dy    dy    dy (4.4a)
y 2  y x  2  y x 
Similarly, we can show that:

v 1  v u  1  v u 
dx    dx    dx (4.4b)
x 2  x y  2  x y 


Substituting Eqs. 4.4 into Eq. 4.2 for V , and rearranging gives:

  u 1  u v  1  u v  
 ˆi u  dx    dy    dy 
  x 2  y x  2  y x  
  (4.5)
 V  translation linear deform. rotation angular deform.


 ˆ v 1  v u  1  v u   (Compare with sketches of
 j v  dy    dx    dx  fluid behavior in section 4.1)
  y 2  x y  2  x y  

70
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 4

In this expanded expression, Eq. 4.5, we can now correlate the mathematical terms with
the corresponding motion they represent. Thus, we have illustrated how the motions and
deformations of a fluid element moving through a velocity field can be mathematically
“decomposed” into the constituent mathematical expressions reflective of translation,
rotation, and angular and linear deformation.

4.2.1 Use of Tensors

Now, a higher-order form of vector that can help generalize Eq. 4.5 is a second-order
tensor, which is a double indexed mathematical expression reflecting (in this case) the
plane of application and the direction of action of a tensor property relative to that plane.
For example, deformation acts in directions either normal or parallel to a specified plane,
and motions of rotation can be construed to take place within a designated plane.

We should note here that a tensor is just a higher form of a vector. The word “tensor” is a
generic term applied to mathematical functions which transform according to certain laws
as a reference coordinate system is rotated about the origin. For a Cartesian coordinate
system (where unit vector differentiation has no effect), we define the rank or “order” of
a tensor according to the number of scalar components required to completely define it,
which is equivalent to the number of coordinates comprising the reference system raised
to the order power. So, for our two-dimensional example, there are two reference
coordinates (x and y). Raising the coordinate number (2) to the “zero” order (i.e. 20=1)
yields one scalar component, which is the number of components required to specify a
“zeroth-order” tensor, which is a scalar (of course), and is not dependent on the
coordinate orientation. Two raised to “first” order (21 = 2) indicates two scalar
components are required to define a “first-order” tensor in a two-dimensional system,
which is of course what we commonly term a vector. And finally, two raised to the
“second” order (22 = 4) indicates that four scalar components are required to define a
“second-order” tensor in two dimensions. For example, we can write a two-dimensional

tensor (let’s use the stress tensor)  ab in Cartesian coordinates as follows, where
subscripts a and b are indices referencing coordinate directions for the components.

ˆ ˆj 

     ˆj ˆii
ab  ˆi ˆi  xx  ˆiˆj xy  ˆj ˆi  yx  ˆjˆj yy  ˆi ˆi  xx  ˆj xy  ˆj ˆi  yx  ˆj yy  ˆi xx xy
ˆj 
 yx yy 

tensor dyad tensor dyad

Here, the first unit vector in the dyads (the tensor components) indicates the action plane
(at least for fluids), and the parenthetical vector indicates the direction of the action
relative to that surface. For example, the first dyad acts on the x-plane (a plane defined

71
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 4

by a x-direction normal). The subsequent direction of action relative to that x-plane is a

vector given by ˆi  xx  ˆj  xy , where xx acts in the x-direction, and xy acts in the y-
direction.

Tensors work well in describing multidirectional behavior, such as deformations and

stresses in a fluid. In common practice, the unit vectors are implied when writing tensors
in matrix format, such that we can express the above as:

  xx  xy 
ab  ˆi ˆi  xx  ˆiˆj  xy  ˆj ˆi  yx  ˆjˆj  yy  
 yx  yy 

For further background on tensors, and their use, you can check out a nice NASA
technical memorandum by Kolecki (2002), for a reasonable explanation of the
importance and applications of tensors.

4.2.2 Decomposition in Tensor Notation

Applying tensor notation to our decomposition expression, we now can express Eq. 4.1
as: 2nd order tensor  surface of action + direction

   u v 
    V V       x 
V  Vp  dV  Vp  dx  dy  Vp  d r  (V) at p  Vp  d r   x
u v 
x y  
 y y 
Expansion shows this is identical to Eq. 4.5, the decomposed formulation above:

  u 1  u v  1  u v  
 ˆi u p  dx    dy    dy 
   x 2  y x  2  y x  
V 
 ˆ v 1  v u  1  v u  
  j  v p  dy  
  
 dx    dx 
  y 2  x y  2  x y  

So, rewriting Eq. 4.5 in terms of tensors, gives:

 u 1  v u   1  v u 
     0   
   x 2  x y    2  x y 
V  V  d r    dr  (4.6)
 1  u v  v   1  u v  
 2  y  x    2  y  x  0 
   y     

Rate-of-Strain Tensor, Vorticity Tensor,

72
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 4


Here, E ab is known as the rate-of-strain tensor, since it mathematically reflects all the

rates of strain that can result from velocity changes.  ab is termed the vorticity tensor
(we will discuss vorticity in section 4.3 below), since it mathematically expresses all the
velocity changes that give rise to fluid rotation. Thus, we can write our final expression
for the decomposed fluid motion as:

     
V  V  d r  E ab  d r  ab (4.7)

Rate-of-strain Vorticity
tensor tensor

Note that in two dimensions these tensors are written as:

 u 1  v u 
    
 xx  xy   x 2  x y  , and
E ab    (4.8)
 yx  yy   1  u v  v 
 2  y  x  y 
   

 1  v u 
0   
  xx  xy   2  x y 
ab     (4.9)
 yx  yy   1  u v  
 2  y  x  0 
   


Here the vorticity tensor, ab , has no components in the direction of the normal to the
bounding planes (or bounding line in 2-D) of application. As we showed in section 4.1.2,
rotation (i.e. vorticity) requires a change in the velocity components acting perpendicular
to the normal of the bounding plane. Thus, a change in the velocity component normal to
a bounding plane cannot contribute to rotation (as we showed, such changes only
contribute to linear deformation). Additionally, since xy = yx in the rate of strain matrix,

this implies that the rate of strain tensor, E ab , is a symmetric tensor; conversely, since

xy = -yx in the vorticity matrix, this indicates that the vorticity tensor, ab , is an
asymmetric tensor.

If we expand our consideration of fluid motion to three dimensions (Cartesian), we of

course end up with more terms in a tensor. As was pointed out above, for a second-order
tensor the number of components is equal to the number of independent coordinates
raised to the second power. So for a full three-dimensional system (x, y, z), tensors will

73
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 4

contain nine scalar components. So, in three dimensions the rates of strain and vorticity
tensors become:

 u 1  v u  1  w u 
      
x 2  x y  2  x z 
 xx  xy  xz  
    1  u v  v 1  w v  
E ab   yx  yy  yz          (4.10)
2  y x  y 2  y z  
  zx
  zy  zz  
 1  u w  1  v w  w 
 2  z  x    
   2  z y  z 


and

 1  v u  1  w u 
 0      
 xx  2  x y  2  x z 
 xy  xz 
    1  u v  1  w v  
ab   yx  yy  yz       0    (4.11)
2  y x  2  y z  
 zx
  zy  zz  
 1  u w  1  v w  
 2  z  x     0
   2  z y  


In Eqs. 4.10 and 4.11 we include a third component of velocity in the z-direction, w, such

that V uî  vĵ wk̂ . Note that in three dimensions the rate-of-strain tensor remains
symmetric and the vorticity tensor also remains asymmetric.

Note also that a (3-D) symmetric tensor, like the rate of strain tensor, has six independent
components, since ij = ji, where i and j represent the system coordinates (e.g. x, y, or z).
For the rate of strain tensor [where xy = yx, zy = yz, and xz = zx], the six unique
components account for all the types of strain rates that can be applied to the various
surfaces of a fluid [we will revisit this when we derive the differential equation of
momentum in Chapter 5]. Conversely, a (3-D) asymmetric tensor, such as the vorticity
tensor, has only three independent components, since ij = - ji. Note that along the
diagonal the only value that can satisfy that criterion is 0, which reduces the independent
components by three. Thus, the independent vorticity components (three) reflect the
respective rotations in each respective coordinate direction.

Now that we have shown that we can mathematically decompose fluid motion into its
component behaviors, we will make use of these elements to analyze and model fluid
behavior. We show in Chapter 5 that we can use the rate-of-strain tensor to model the
stress field in a fluid, and to assess the impact of surface forces on fluid motion via

74
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 4

Newton’s second law, a critical step in the development of the governing differential
equation for fluid momentum. We will also employ the vorticity tensor, and the parallel
concept of circulation (the summation of vorticity over a fixed area), to account for the
rotational behavior of a fluid. To begin this process, in the following section we first
address the concept of vorticity, and show how we assess and describe local fluid
rotation.

4.3 Vorticity

Vorticity is the term applied to the angular rotation of a fluid. To develop a mathematical
description of the angular rotation, we examine the behavior of two infinitesimal
orthogonal line segments fixed in a fluid, as shown below. As we did in our examination
of general fluid rotation in Section 4.1.2, we again assume that these line segments rotate
in such a manner as to yield a positive rotation of each line segment (counter clockwise).
We also again assume that t is infinitesimal, and that the degree of rotation of the
element can be characterized by similar infinitesimal angular changes, as shown.

 u 
b   y t
 y 


y
 v 
y a   x t
  x 
x

Figure 4.6 Characterization of angular rotation of two orthogonal line segments

The angular velocity of each line segment is assumed to be the differential change in
angular rotation over time, t. However, to relate the angular rotation to characteristic
changes in the velocity components, we relate the angular rotation to the arc length
displacement of the line segment, identical to the process we employed in our discussion
of rotation in Section 4.1.2. Thus, for the rotation of the horizontal line segment shown
above, this displacement is given by a  v x x t , and we can write:

75
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 4

 a / x
 of line x  lim  lim
t 0 t t 0 t
 v 
 xt  / x
x v
 lim   
t 0 t x

Likewise, for the angular velocity of the vertical line segment:

u
 of line y  
y

The total angular velocity is then assumed to be the average angular velocity of the two
orthogonal line elements, with their axis of rotation about the z-axis (out of the plane of
this page), i.e.
1 1  v u 
 z  ( of line x   of line y)    
2 2  x y 

rotation about z-direction coordinate

The angular velocity for the other two coordinate directions can be derived similarly,
giving:

1  w v  1  u w 
 x     and  y    
2  y z  2  z x 

Thus, the total vector angular velocity for a fluid is given by:


  (ˆi  x  ˆj y  kˆ  z )

(4.12)
1  w v  1  u w ˆ 1  v u  ˆ 1 
    ˆi     j    k    V
2 y z  2  z x  2  x y  2

Note that this is one half the curl of the velocity field, which is the classical dynamical
vector operator relating angular velocity to the respective velocity. Since the one half is
cumbersome to carry along in equations involving fluid rotation, we generally drop this
coefficient and simply characterize fluid rotation by the curl of the velocity field, which
  
we term “vorticity,” and indicate by a lower case omega,   2    V . Note that this
 
definition of vorticity as     V applies generically for all coordinates systems;
however, care must be exercised in non-Cartesian coordinate systems. Note that

76
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 4

additional terms will appear when determining the vorticity in cylindrical coordinates,
due to the unit vector dependence upon the angular coordinate, . In spherical
coordinates, additional terms are generated by unit vector dependence upon both angular
coordinates,  and .


Example: Given a simple shear flow, V  Uy h  î, determine the components of
both the rate of strain and rotation tensors, and the vorticity. y
U

Here, noting that u  Uy h  and v = 0, we first calculate the h

x
components of the rate of strain and vorticity tensors we
developed above.

u
 xx  0  xx  0 (always)
x
1  v u  1 U U 1  v u  1 U U
 xy      0     xy      0    
2  x y  2 h  2h 2  x y  2 h 2h
v
 yy  0  yy  0
y
1  u v  U 1  u v  U
 yx       yx     
2  y x  2h 2  y x  2h

Remembering that the first index subscript prescribes the plane of action, and the second
index subscript indicates the direction of the action, we can qualitatively represent each
of the non-zero components of strain and rotation as shown below. Here we recognize
that this flow, although very simple, comprises motions of both angular deformation
(opposing rotations of primary axes) and angular rotation (like rotation of primary axes).
The result is a flow that angularly deforms at the same rate it rotates. This simultaneous,
and complimentary, angular deformation and rotation maintains a non-uniform
translation of the fluid parallel to the x-axis, as illustrated below.

yx Subsequent motion:

y y a uniform shear flow
yx Rate of strain

xy xy

Rotation
x x

With regard to the vorticity, the only vorticity component is in the z-direction (out of the
plane of this page), such that:

77
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 4

   v u  U
    V    kˆ   kˆ
 x y  h

Here the vorticity is negative relative to the reference coordinate, or of clockwise

orientation, like a ball rolling from left to right.

For some additional viewpoints on vorticity, and some further illustrations, take a look at
the Wikipedia article at this link.

4.4 Rate of Strain and Vorticity Equations: Cartesian and Cylindrical Coordinates

The following list the specific equations for all of the rate of strain and vorticity
components in Cartesian and cylindrical coordinates.


4.4.1 Rate of Strain in Cartesian Coordinates: V  uˆi  vˆj  wkˆ

u 1  v u  1  u w 
 xx   xy      xz    
x 2  x y  2  z x 
1  v u  v 1  w v 
 yx      yy   yz    
2  x y  y 2  y z 
1  u w  1  w v  w
 zx      zy      zz 
2  z x  2  y z  z


4.4.2 Vorticity in Cartesian Coordinates: V  uˆi  vˆj  wkˆ

   w v  ˆ  u w ˆ  v u  ˆ
    V  (  x ˆi   y ˆj   z kˆ )    i    j    k
 y z   z x   x y 

78
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 4


4.4.3 Rate of Strain in Cylindrical Coordinates: V  vr ˆir  vˆi  vz ˆiz

v r 1    v  1 v r  1  v v 
 rr   r   r       rz   r  z 
r 2  r  r  r   2  z r 
1    v  1 v r  1 v v r 1  1 v z v 
 r   r           z    
2  r  r  r   r  r 2  r  z 
1  v v  1  1 v z v  v z
 zr   r  z   z      zz 
2  z r  2  r  z  z


4.4.4 Vorticity in Cylindrical Coordinates: V  vr îr  vî  vz îz

  1 v z v   ˆ  v r v z ˆ 1   v ˆ
 i   rv    r

    V  (  r ˆi r    ˆi   z ˆi z )     ir     iz
 r  z   z r  r  r  

References

Kolecki, J. C. (2002) An Introduction to Tensors for Students of Physics and

Engineering, NASA/TM—2002-211716 (link: http://www.grc.nasa.gov/WWW/k-
12/Numbers/Math/documents/Tensors_TM2002211716.pdf)

Truesdell, C.A. (1954) The Kinematics of Vorticity. Indiana University Press,

Bloomington

79
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 4

Problems


1. Consider two three-dimensional flows with velocities, V  c1 xˆi  c 2 yˆj  c 3 zkˆ , and

V  c1zˆi  c 2 xˆj  c 3 ykˆ Determine the vorticity, the strain-rate tensor, and the rate of
expansion for these flows (hint: the rate of expansion is given by the divergence).

  Q ˆ
2. A two-dimensional flow in cylindrical coordinates is given by: V    ir . Compute the
 2 r 
vorticity, the strain-rate tensor, and the rate of expansion for this flow (hint: the rate of
expansion is given by the divergence).

3. A non-dimensional velocity field in cylindrical coordinates is given by:

1
V   2 îr  4r 2 î
r 
Determine:

a. The components of the rate-of-strain tensor; sketch a differential element in cylindrical

coordinates and show (as vector arrows), the magnitude and direction of the rate-of-strain
components.

b. The vorticity of a particle at (x,y) = (0,2) and (0,0).

4. A non-dimensional velocity field in cylindrical coordinates is given by:

1
V    ˆir  4r ˆi
r
Determine:

a. The components of the rate-of-strain tensor; sketch a differential element in cylindrical

coordinates and show (as vector arrows), the magnitude and direction of the rate-of-strain
components.

b. The vorticity of a particle at (x,y) = (0,2) and (0,0).

5. A non-dimensional velocity field in cylindrical coordinates is given by:

1 ˆ  r
V   2  ir  4r1  ˆi
r   3
Determine:

a. The components of the rate-of-strain tensor; sketch a differential element in cylindrical

coordinates and show (as vector arrows), the magnitude and direction of the rate-of-strain
components.

b. The vorticity of a particle at (x,y) = (0,2) and (0,0).

80
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 4

6. The velocity field for a decaying, ideal line vortex is given by:

    r 2 
v   1  exp  
 2r   4 t 

Where  is a vortex strength parameter. Let all parameters and variables be non-dimensional,
and  = 1.

Determine, and show graphs for 0  r  4,

a. The non-zero rate-of-strain components

b. The vorticity.
 4   2r 
Plot separate graphs of the normalized (1) vorticity   and (2) strain rate  .
     
Show plots on each graph for t = 0, 1, 6 (note: you will need to use L'Hopital's Rule for r
and t  0).

7. The velocity field for a decaying, ideal line vortex is given by:

 Hr   r2 

v   
2 
exp  
 8 t   4 t 

Where H is a vortex strength parameter. Let all parameters and variables be non-
dimensional, and  = 1. Determine:

a. the non-zero rate-of-strain components

b. the vorticity.

 8v     16 
Plot separate graphs of non-dimensional (1) velocity   , (2) vorticity  
 H   H 
 322  r 
,and (3) strain rate   for 0  r  4. Show 3 plots on each graph for t = 0, 1, 3

 H 
(note: you will need to use L'Hopital's Rule for r and t  0).

81
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 5

Chapter 5

Differential Equations of Motion

Contents

5.1 Background .................................................................................................................. 82

5.2 The Differential Continuity Equation: Via a Differential Analysis .................................... 84

5.3 The Differential Continuity Equation: Via the Divergence Theorem ................................ 90

5.4 Special Cases of the Continuity Equation ....................................................................... 92

5.4.1 Steady Flow ......................................................................................................... 92
5.4.2 Incompressible Flow ............................................................................................ 93

5.5 Uses of the Continuity Equation .................................................................................... 93

5.6 The Momentum Differential Equation for a Fluid: The Navier-Stokes Equation ............... 98
5.6.1 Modeling of Differential Surface and Body Forces ................................................. 99
5.6.2 The Differential Momentum Equation ................................................................ 102
5.6.3 Relating Stresses to Fluid Rates of Strain ............................................................ 103
5.6.4 Reducing the Momentum Equation to the Navier-Stokes Equation ...................... 108

5.7 Utilization of the Incompressible Continuity and Navier-Stokes Equations ................... 111

5.8 The Governing Equations for Incompressible Flow ....................................................... 112

5.8.1 Stresses on an Incompressible Fluid .................................................................... 112
5.8.2 Continuity and N-S Equations in Cartesian Coordinates ....................................... 113
5.8.3 Continuity and N-S Equations in Cylindrical Coordinates ...................................... 113

In this book, it is assumed that students have a basic background in undergraduate fluid
mechanics, particularly the use of basic control volume equations as they apply to
conservation of mass and momentum. If you feel that you don’t have a reasonable
grounding in the basics of control volume analysis, it is recommended that you review
this material in a standard undergraduate fluid mechanics text (e.g. Fluid Mechanics, Fox
& McDonald, Chapter 4), since the derivation of the differential equations of fluid
motion done in this chapter will depend on several underlying control volume concepts.

5.1 Background

To understand the capabilities of a control volume analysis versus a differential analysis,

let’s review their respective strengths and weaknesses. In both cases, we assume that we
are conducting a coupled analysis of the equations of both continuity and momentum.

82
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 5

Recall from your undergraduate studies that a control volume approach develops
equations that assume that the velocity and property distributions are known for a flow
entering and exiting a fixed region in space, such as a pipe or jet engine. In many
practical cases, the velocities (and other properties, such as density and temperature) are
often assumed to be uniform over the entrance and/or exit of the control volume. Given
the assumed flow characteristics, the solution of the appropriate integral equations is
relatively straightforward. However, the results of such calculations yield only the global
characteristics of a flow system, such as the resulting force or thrust generated by the
change in momentum across a control volume. Thus, the control volume equations of
momentum and mass provide useful, but generally approximate, results and are
reasonably easy to employ.

In contrast, a differential approach develops more sophisticated differential equations that

can yield very detailed information regarding local velocity and other property
information, requiring only the appropriate boundary and initial conditions to allow
solution of the equations. However, the differential equations are quite complex, such
that their complete solution may require sophisticated mathematical techniques or very
advanced numerical/computational approaches. The result provides detailed velocity and
property field information, which can subsequently be: (1) employed to assess local
forces on the constraining boundaries, or (2) combined with control volume approaches
to assess more accurate global characteristics, such as thrust or flow-induced drag.
Luckily, for our purposes, there are a number of simple flows that lend themselves to
complete differential analysis, for which one can develop closed-form solutions of the
governing differential equations, before branching out into more complicated flows that
will generally require substantial numerical analysis to achieve a solution.

In this and the following chapter, our objective is to develop systematically the basic
differential equations of continuity (i.e. mass conservation) and momentum (i.e. the
Navier Stokes equation). As you will see, these sets of equations are key elements for
our subsequent assessment of fluid behavior. In particular, in Chapter 6, we will
demonstrate the application of these equations for the determination of the velocity and
shear stress fields for some simple laminar flows. And in Chapter 12, we will address
solution techniques for more involved laminar flows. However, the differential equations
of continuity and momentum are the crux of the material we discuss throughout the rest
of this book. In the present chapter, we begin our assessment of fluid dynamics by the
derivation of the differential equations of both continuity and momentum.

83
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 5

5.2 The Differential Continuity Equation: via a Differential Analysis

Recalling our discussion in Chapter 1 of the characterization of forces acting on a

differential fluid element, in this section we employ that same differential element
approach for the analysis of fluid continuity, or conservation of mass. We perform our
derivation in Cartesian coordinates, but similar approaches can clearly be employed using
cylindrical or spherical coordinates; the resulting equations for cylindrical coordinate
system are shown in Section 5.8.2. The equations for spherical coordinates can be found
here. As we did in Chapter 1, we start by assuming an infinitesimal differential element
of volume, d  dxdydz , where dx, dy, y
and dz are the differential dimensions of
the element in the respective coordinate
directions, as shown in figure 5.1. dy

,u,v,w
dz
We next assume that the local density dx
and velocity (in terms of the respective
directional components) at the center of x
the element are given by , u, v, and w.
Our objective is to assess the changes in
mass both within and across this z
differential volume using an integral
Figure 5.1 The basic differential fluid
control volume equation applied to this
element in Cartesian coordinates.
differential element of the form:

 d  V  
  d0 (5.1)
t C . V . C .S .

In Eq. 5.1, C.V. implies integration across the entire control volume, and C.S. implies
integration over the entire control surface bounding the control volume. The volume of
the differential element is d  dxdydz , and the magnitudes of the bounding control
surfaces, across which the fluid can move into or out of the differential element, are given
by the respective sides of the element, dAx = dydz, dAy = dxdz, and dAz = dxdy. To make
our derivation process easier to follow, we note that we can separate the integral term
reflecting the mass crossing the control surface of the differential element into the three
integrals representing the mass crossing the x, y, and z surfaces respectively, such that:

       
 V  dA   V  dA   V  dA   V  dA
C .S . C .S . x C .S . y C .S . z

Using the x-direction surface integral as an example, we now determine the amount of
mass crossing the two x-surfaces of the differential element, in terms of the density, ,

84
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 5

 
and the x-direction velocity, u. The result is  V  dA   uˆi  dA
C .S . x C .S . x
x nˆ , where u is the x-

direction velocity component parallel to the outward normal unit vector (shown as n̂ ) for
the x-surfaces of the differential element. Note that nˆ   ˆi , depending on the direction
of the outward normal for the respective x-surface of the differential element. Also note
that this integration will only require knowledge of the x-direction velocity, u, since the
y-direction and z-direction velocities are parallel to the x-surfaces, and will contribute no
mass flow across those surfaces. To begin, we need to establish the values of the density
and the u-velocity at the point where they cross the two differential x-directed surfaces of
area dAx = dydz (the magnitude of both the left and right surfaces of the differential
element shown in figure 5.1).

To establish the changes of  and u that occur across the differential element, we start by
assuming that  and u are the base values at the center of the differential element. That
being the case, we can make use of a Taylor series expansion for both  and u to
approximate the values of density and velocity at the x-surface boundaries. Recall from
calculus that if we know (or we assume we know) the value of a function, call it F(x, y,
z), at some spatial location, x, y, z, then we can determine the value of the function at a
nearby location removed by a differential amount, call it x, as:

2 3
F 1  F 2 1  F
Fx  x , y, z   Fx , y, z   x 
2
x  
3
x 3   higher order terms  (5.2)
x 2! x 3! x

In the mathematical sense, this expression would be termed the power series expansion of
F upon x. We also note that the series must be convergent to a finite value, and not
divergent. However, convergence is always the case for real fluid systems. Now, if x is
very small, which it is since we are assuming that the differential element is
infinitesimally small, then higher powers of x, such as x2 and x3, will be much
smaller--so small, that they can be considered negligible compared to x. This suggests
that the higher order terms in the Taylor series become negligible when we let x limit to
x0, such that x  dx. We thus can approximate that:

F
lim Fx  x, y,z   Fx  dx , y,z   Fx, y,z   dx (5.3)
x 0 x

This form of the Taylor series is known as a truncated Taylor series, and is assumed
applicable in the limit of an infinitesimal differential element.

85
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 5

So, to establish property changes of  and u across the differential element in the x-
direction, we apply this truncated Taylor series to model the property values at the right
and left faces of the element. For example, for the density at the right face we write:

  dx   dx
right face      (5.4)
x  2  x 2

Where  is the assumed value at the center of the differential element, and +dx/2 is the
distance that the right face of the element is removed from the center of the element. Note
that this is a change in the positive x-direction, and thus the plus sign (for emphasis of
this point). In contrast, the density at the left face of the differential element will be:

  dx   dx
left face      (5.5)
x  2  x 2

Here, the change in sign is a result of the left face being removed - dx/2 from the center
of the element in the negative x-direction (thus, the negative value). Similar expressions
for the x-direction velocity, u, at the right and left faces of the differential element can be
written as:

u dx u dx
u right face  u  and u left face  u  (5.6)
x 2 x 2

In Eq. 5.6, u is again the assumed value at the center of the differential element.

Using these values of density and velocity at the right and left faces of the differential
element, we can now illustrate the mass flows across the x-surfaces of the element in
figure 5.2.

Note that we assume that all flows (velocities) are in the positive direction relative to the
coordinate direction x [density has no direction, being a scalar]. By doing this, we realize
that when we later solve the differential equations, that a positive or negative result for
the u-velocity will reflect the direction of the u-velocity relative to the positive direction
of the x-coordinate we employ for the solution procedure.

86
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 5

  dx  u dx  dy   dx  u dx 
   u  dy dz 
,u,v,w    u  dy dz
 x 2  x 2  Outward  x 2  x 2 
dz Outward
normal normal
dx

Figure 5.2 Representation of the differential mass flows across the x-surfaces of a
differential fluid element.

We now use the respective values of density and velocity at the x-surfaces of the element
to establish the mass flows across the element, for the x-direction only. Recall from
above that:

 
  dA 
V
C .S . x
 uˆi  dA
C .S . x
x nˆ

Remember that the dot product of the velocity relative to the direction of the surface
[defined by the direction of the outward normal ( nˆ   ˆi ) from the element surface,
relative to the coordinate system] determines whether the integrated mass flow is
considered positive or negative. And since we have assumed all velocities (for the sake
of this derivation) to be in a positive coordinate direction, the outward normal unit vector,
n̂ , will determine whether the mass flow is considered positive or negative in the
subsequent equation.

So, utilizing our expressions from Eqs. 5.4, 5.5, and 5.6, we can sum the flows of mass
across the x-surfaces (defined by their outward normal relative to the x-direction, as
shown in figure 5.2), as:

 
 V  dA     x
  dx 
 u 
2 
u dx ˆ
x 2 
  
 i  dydz   ˆi    

 dx 
x 2
 u 

u dx ˆ
x 2 
 
 i  dydz   ˆi
C .S . x
(5.7)
  dx  u dx    dx  u dx 
    u  dydz      u  dydz
 x 2  x 2   x 2  x 2 

87
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 5

Eq. 5.7 reflects the only mass flows crossing the two x-surfaces of the differential
element. Now expanding the terms on the right side of this Eq.5.7 mass balance, and
canceling and combining like terms, we have:

   u dx  dx  du dx 2 

C .S . x
 V  dA  


 u  
x 2
 u
x 2

x x 4
dydz

 u dx  dx  du dx  2
  u   u  dydz
 x 2 x 2 x x 4 
  u 
or  V  dA   dxdydz  u dxdydz
C .S . x
x x

 (u )dxdydz (5.8)
x

If we perform similar mass flowrate balances for the y and z surfaces, conservation of
mass applied to the differential control volume gives two similar expressions:

  
  dA  y (v)dxdydz

C .S . y
V (5.9)

  
 
C .S . z
V  dA  (w )dxdydz
z
(5.10)

Combining Eqs. 5.8, 5.9, and 5.10 for the mass flow rate in each of the three coordinate
directions, we obtain the total mass flow across the differential element as:

       
 V  dA   V  dA   V  dA   V  dA
C .S . C .S . x C .S . y C .S . z

  (u ) (v) (w )


C .S .
V  dA 
x
dxdydz 
y
dxdydz 
z
dxdydz (5.11)


t C
Now, the time rate of change of mass within the differential element, d , is
.V .

modeled by assuming that in the limit of an infinitesimal differential element (d0),

the density, , is relatively uniform across d. At first this assumption of uniformity
might appear inconsistent with our previous approximation of the changes across the
differential element using a truncated Taylor series. However, since we assume a linear
change across the element, the average value of the density within the differential volume

88
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 5

will be just , the assumed value at the center of the differential element. And since we
will ultimately assume that d0, we approximate that d  dxdydz  , and thus,
C.V .

 dxdydz  

t C.V .
d 
t
 dxdydz
t
since dxdydz  f(t) (5.12)

Note in Eq. 5.12 that   ( x, y, z, t ), but dxdydz is not a function of time, and thus can
be factored out of the derivative, as indicated.

Substituting Eq. 5.11 and 5.12 back into Eq. 5.1, we obtain the cumulative conservation
of mass balance for the differential control volume as:

     (u )  (v )  (w )

 d   Vd  dxdydz  dxdydz  dxdydz  dxdydz  0
t c.v . c.s . t x y z

Since all terms are multiplied by the differential volume, d=dxdydz, we can factor this
term out of the equation, leaving only the derivative functions comprising the differential
equation. It would seem that this equation is independent of the size of the assumed
differential element, since d=dxdydz does not appear in the final equation. However,
remember that we assumed that our differential is infinitesimal (i.e. d0), which
allowed us to: (1) apply the approximation of a truncated Taylor series, and (2) assume
that  varies linearly across the element.

Therefore, the differential equation for conservation of mass is:

 u  v  w   

   or    (V )  0 (5.13)
t x y z t

Equation 5.13 is also termed the continuity equation [a term that we employ throughout
this book], emphasizing the application of mass conservation to a continuous medium.
Also note this is termed a “differential” equation, since it evolves from our assumption of
an infinitesimal “differential” element with continuous properties.

We can manipulate this equation further, by expanding the mass flux terms to give:

    u v w
u v w    0 (5.14)
t x y z x y z

D 
  V
Dt

89
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 5

As we identify, the first four terms on the left of Eq. 5.14 comprise the substantial
derivative of the density, and thus the changes in density that a particle experiences as it
is transported through a velocity field. The remaining three terms represent the product
of the density and the divergence of the velocity field, such that the resulting equation can
be written in a compact form as:

D 
   V  0 (5.15)
Dt

Equation 5.15 is a generic vector representation, which is applicable in any other

coordinate system (i.e. cylindrical or spherical). We further explore the significance of
this vector form of the continuity equation in the following section.

5.3 The Differential Continuity Equation: via the Divergence Theorem

The derivation of the differential equation of continuity that was done in section 5.2
employs the application of a control volume analysis to a differential element, using
several approximations to establish the balance of the derivative changes in velocity and
density that occur across and within a differential element. That approach necessitates
some sweeping assumptions of behavior as the differential volume asymptotes to zero.
An alternate derivation of the differential equation of continuity can be developed in a
more mathematically elegant, although less physically obvious, manner using the
application of the surface integral relationships reviewed in section 2.4 to the basic
control volume equation (Eq. 5.1) of continuity. Consider again the general control
volume equation for continuity, given by:

  
t C   dA  0
d  V
.V . C .S .

For a fixed size control volume (i.e. the volume is constant), the time derivative can be
taken inside the integral, since the limits of integration over that volume will be fixed,
and thus not a function of time (Leibnitz rule), giving:

  
  t
C.V .
d    dA  0
C .S .
V

or

 
  t d   (V)  dA  0
 S
(5.16)

90
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 5

In Eq. 5.16,  implies a fixed volume in space, with S indicating the bounding surface of
that volume. However, the Gauss divergence theorem (Section 2.4.1, Eq. 2.15) allows us
to relate the surface integral of a vector property to the volume integral of the divergence
of that property, such that we can write:

 
 (V)  dA    (V)d
S 
(5.17)

Substituting Eq. 5.17 into Eq. 5.16 yields;

     
 t d    (V)d    t    (V)d  0
  
(5.18)

However, since there is no stipulation of the size of the control volume to which this
equation applies, this means that the integrand in this equation cannot depend on the
volume of integration. The only value of the integrand that satisfies this requirement is
an integrand that is everywhere equal to zero, or

 
   (V )  0 (5.19)
t

Equation 5.19 is, of course, the same as Eq. 5.13 that we obtained previously using a
differential analysis.

Another way to consider the concept of continuity is to consider a differential amount of


mass, dM, which is moving through a velocity field, V . Assuming that this arbitrary
amount of mass is distributed in a continuous manner over a differential volume, d, we
can write dM  d . Now, the change in this differential mass as it moves within the
velocity field is then given by the material derivative of dM, such that:

D
dM   D d  0 (5.20)
Dt Dt

We set the derivative in Eq. 5.20 to zero, since there can be no change in the amount of
mass as it moves through the field. Expanding the right side of Eq. 5.20 gives:

D
d   Dd  d D  0
Dt Dt Dt
or
1 Dd 1 D
 (5.21)
d Dt  Dt

91
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 5

Equation 5.21 indicates that the relative change in volume of the fluid element is
balanced by an opposite change in the density of the fluid. By substitution for D/Dt in
Eq. 5.21 from the general continuity equation, Eq. 5.15, we have:

1 Dd 1 D 
  V (5.22)
d Dt  Dt

Equation 5.22 indicates that the divergence of the velocity field represents the relative
rate of expansion of the fluid volume at a point. This provides a useful physical
interpretation of the divergence of the velocity, and will prove useful in the
simplifications of the continuity equation, as discussed in the following section.

5.4 Special Cases of the Continuity Equation

In its general form, the continuity equation is applicable to any fluid flow, regardless of
the properties of the flow. And as such, the equation is equally valid for incompressible
as well as compressible flows, and heavy viscous flows or dilute gas flows, to illustrate
some extremes. Since it contains four dependent variables ( and three velocity
components), it cannot be solved independently, and must be solved in conjunction with
the other governing equations, such as the differential momentum equation [derived in
section 5.6]. However, since the equation is a function of four independent variables
(time and three spatial coordinates), utilization of the full equation is analytically and
computationally complicated, and potentially intractable because of the degrees of
freedom inherent in three-dimensional, unsteady flows. Fortunately, many flows of
engineering utility can be configured or constrained to reduce the degrees of freedom,
which allows the application of a simplified form of the continuity equation. Two of the
constraints that are most commonly invoked are: (1) a steady flow, and (2) an
incompressible flow.

5.4.1 Steady Flow

Here, we assume that there is no time dependency of the flow, which assumes no local
time rate of change in density at a point. All incompressible and many compressible
flows, such as flow around high-speed aircraft in steady flight are well approximated by
this assumption. The subsequent continuity equation reduces to:

(u ) (v) (w ) 

     V   0 (5.23)
x y z

While Eq. 5.23 is a simpler equation, it still includes four dependent variables, and can be
quite complicated to solve.

92
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 5

5.4.2 Incompressible Flow

A further simplification is to specify the fluid as incompressible ( = constant). This is a

reasonable approximation for essentially all liquids and isothermal flows of gases at low
Mach numbers (roughly M < 0.3 for air). This incompressibility assumption is
appropriate for many industrial and domestic flow systems, such as ducting and piping
systems, flows around automobiles and ships, etc. For an incompressible flow, the
substantial derivative of the density (D/Dt) is zero, such that Eq. 5.15 reduces to:
0
D  
   V  0  V  0 (5.24)
Dt

u v w
or   0 in Cartesian coordinates
x y z


Since   V  0 , this implies that the velocity field for an incompressible flow is non-
expanding. This means that any changes in a velocity component in its respective
coordinate direction must be balanced by collective corresponding changes of each of the
other coordinate velocity components, such that no local expansion or contraction of the
fluid volume can occur. This is illustrated by Eq. 5.22 above, which indicates that:

1 Dd 1 D 
  V
d Dt  Dt


Thus, when   V  0 , a fluid cannot undergo local expansion/contraction, the density, ,
will remain constant everywhere, and the fluid will behave incompressibly. This will be
true even if the fluid could be compressed, but is not for the particular flow process under
consideration (e.g. low-speed flow of air through ducts or around bodies). In general, a
vector field that has zero divergence is termed a solenoidal field.

5.5 Uses of the Continuity Equation

The continuity equation is a particularly powerful equation, which models the balance of
changes between density and velocity. For incompressible flows, density change terms
all drop out, and the equation relates the balance of velocity changes in the respective
coordinate directions. This balance can be used to good effect, for example, to establish a
third component of velocity when the other two components of velocity are known [e.g.
if the functional behavior of two velocity components are known for a flow, then the third
velocity component can be established via the continuity equation]. This same process
also applies, of course, for two-dimensional flows, where only one known velocity

93
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 5

component is necessary to establish the second. Additionally, the continuity equation is

used to simplify the differential equation of momentum (the Navier-Stokes equation), as
will be illustrated in Section 5.6.


Example: For an incompressible flow, Vx, y, z   uî  vĵ  wk̂ , the x- and y-
direction velocities are given by u = ax and y = - by, where a and b are constants.
Determine the velocity component w, where w = 0 over the x-y plane at z = 0.


Here we apply Eq. 5.24 for incompressible flow,   V  0 , solving for the w term.

 u v w w u v
V    0     a  b
x y z z x y

Integrating, gives

w  b  a  z  f x, y 

But, w = 0 for all x,y at z = 0, thus f x, y  =0, and w  b  a  z .

Note that when a = b, then w = 0 everywhere, and this is a classic inviscid solution for a
flow impinging in the y-direction toward the x-y origin (x = 0, y = 0). We will revisit this
type of flow in Chapter 9. However, if b > a, then w will be away from the x-y plane
(w > 0 for z > 0, and w < 0 for z < 0). Likewise, if b < a, w will be towards the x-y plane.
In both of these cases, the velocity change in the z-direction balances the collective
velocity changes in the other two coordinate directions, either moving additional mass
toward the x-y plane or moving mass away from it.

Example: Consider the generic flow behavior of a two-dimensional jet, which exits a
nozzle into an otherwise quiescent flow as shown in figure 5.3.
y

u  x
u v
0  0  v is away from center
x y
Figure 5.3 Behavior of a two-dimensional jet exiting into a quiescent flow.

As experience shows, the velocity of the jet in the streamwise x-direction, u, will
decrease continuously with distance, such that along the centerline of the jet (y=0),

94
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 5

u x  0 . Based on this observation, we want to establish how the v velocity, normal to

the streamwise velocity, will behave.

Here we observe that the two-dimensional, incompressible continuity equation (in x, y)

is:
u v v u
  0 or rearranging,  (5.25)
x y y x

Thus, if u x  0 , this implies that v y  0 , or that the normal velocity, v, is away

from the axis of symmetry (in a positive y-direction for y > 0, and in a negative y-
direction for y < 0). In actuality, the flow also entrains (draws in) a certain amount of
flow from the quiescent fluid it is exhausting into, so that the streamline pattern for the
developing jet flow will appear approximately as shown in figure 5.4.

Figure 5.4 Approximate streamline pattern for a two-dimensional, steady jet.

Example: Consider a uniform, two-dimensional, incompressible flow that undergoes

a change in its cross-sectional area, such as when passing through a nozzle (decreasing
area) or a diffuser (increasing area). For simplicity, we will assume that u is uniform at
the entrance of a nozzle, with the initial velocity and cross-sectional area given by Uo and
Ao at x = 0, as shown in figure 5.5.

Uo
u(x)
y

Ao x

A(x)
Figure 5.5 An example of uniform flow through a generic nozzle

Additionally, we assume that the streamwise velocity, u(x), changes uniformly (all u-
velocity components at a cross-section will be equal in magnitude). We want to first

95
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 5

determine the behavior of the streamwise velocity, u(x), as a function of the cross-
sectional area, A(x). We then want to determine the respective behavior of the normal
velocity, v(x), and the impact of the change in cross-sectional area, dA(x)/dx, on the
velocity components.

For a uniform flow through a constrained two-dimensional duct, as shown, using control
volume considerations for an incompressible flow, we can write:

UoAo
U o A o  u ( x )A( x )  u(x) 
A( x )

  dA  
  
u  1 dA  U o A o   dx   u ( x )  dA 
Thus,  U o A o   2       (5.26)
x  A ( x ) dx  A( x )  A( x )  A( x )  dx 
 
u(x)  

Considering both a nozzle (a flow accelerator) as shown in figure 5.5, and a diffuser (a
flow decelerator), from Eqs. 5.26 and 5.25 we can make the following generalizations

dA u v
Nozzle: A( x )   0  0   0  flow toward center
dx x y
dA u v
Diffuser: A( x )   0  0   0  flow away from center
dx x y

To establish the normal velocity behavior, v(x,y), that is required to maintain a uniform
streamwise velocity, u(x), we write that:

  dA     dA  
   
u U o A o   dx   v v U o A o   dx  
  or  (5.27)
x A( x )  A( x )  y y A( x )  A( x ) 
   
   

Integrating Eq. 5.27 for v(x,y), gives:

  dA     dA  
   
  dx     dx  
vx, y   U o A o
 A( x ) 2  
dy  U o A o y  f (x)
 A( x ) 2 
   
   

If we assume symmetry about x = 0, then we surmise that v(x,0)=0 on the centerline at

96
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 5

y = 0, and thus f(x) = 0. Therefore, the general expression for v(x,y) is:
  dA  
 
  dx   u( x )  dA  UA
vx , y   U o A o y   y where u ( x )  o o (5.28)
 A( x ) 
2
A( x )  dx  A( x )
 
 

To particularize this result, let a nozzle have an area distribution given by Ax   A o e  ax ,
or an exponential nozzle. For this case we have from Eq. 5.28:

UoAo UoAo
u(x)    U oeax
A( x ) A oe  ax
and
  dA  
 
  dx     aA oe ax 
vx, y   U o A o y  U A 
o o
 y   Uoae ax y

 A( x ) 2  
2
A o e 2ax 
 
 

Note that if we examine the streamline equation for this flow (Eq. 3.4), we have:

dy v  U oae ax y dy
   ay   adx
dx u U oeax y

integrating, gives:

ln y  ax  c  y  Ce ax

Here C is a constant depending on the particular streamline. For example, along the
centerline, which passes through (x,y) = (0,0), we get C = 0, and thus the streamline is the
x-axis. However, for a streamline passing through (x,y) = (0, yo), C = yo and the
corresponding streamline will be described by:

y  yoe ax (5.29)

Equation 5.29 describes the family of streamlines entering at x = 0. Figure 5.6 shows
representative streamlines for this nozzle, with various initial starting yo locations letting
a = 1. Note that in this example, the streamlines passing through x = 0 at y = 1
represent the boundaries of the nozzle.

97
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 5

y -1
0 0.2 0.4 0.6 0.8 1

x
Figure 5.6 Streamlines for a two-dimensional exponential nozzle, Ax   A o e  ax .

5.6 The Momentum Differential Equation for a Fluid: The Navier-Stokes Equation

Continuing on with our derivation of the governing equations of fluid mechanics, we now
address the development of an equation that models the balance of momentum changes
and forces, according to Newton’s second law. Recall that Newton’s second law can be
  
generically described as  F 
t

 
mV  m a , a balance of forces and changes in

momentum. Here, we will use the differential element concept to derive a differential
form of Newton’s second law applicable to a fluid. There are several ways to approach
this derivation, one being the use of a control volume approach, similar to what we did
for the derivation of the continuity equation in section 5.3. However, a more physically
appealing approach is to employ our differential element as an infinitesimal Eulerian
region with fluid passing through it in a Lagrangian process. Since we can assess the
Eulerian acceleration of the fluid using the substantial derivative, as we showed in
Chapter 3, this provides us with a tool to establish the acceleration of the fluid
encompassed by a differential fluid element. The forces acting on the fluid can be related
to: (1) surface forces acting on the respective bounding surfaces of the differential
element, and (2) body forces associated with the mass contained within the differential
element. We again assess these differential forces by use of a truncated Taylor series.
The result of this latter process yields a stress tensor and body force vector which we
model as balancing the acceleration of the fluid, as determined from the substantial
derivative of the velocity field. The schematic in figure 5.7, schematically illustrates both
the generic process followed, and the desired end result, which is a single vector equation
which balances velocity field momentum terms with surface and body forces.

The differential equation that we will derive is termed the Navier-Stokes equation,
acknowledging the initial derivation of the equation, originally by C.L. Navier in 1822,
with a later refinement of the derivation by G.G. Stokes in 1845. This equation is the

98
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 5

keystone for the science of fluid mechanics, and forms the general basis for much that
follows in this text.
y

,u,v,w dy
Newton’s 2nd law applied to fluid element  
x
dz
z dx
Relates –

Stresses & Body Force  Acceleration

 
Fluid Deformation Tensor  Substantial Derivative
& Body Force Vector

Velocity Field Terms

+
Surface & Body Forces

Figure 5.7 The generic process for application of Newton’s second law to a
differential fluid element.

5.6.1 Modeling of Differential Surface and Body Forces

We begin by examining the surface forces, in the form of stresses, that exist at the
surfaces of an infinitesimally small differential element of volume, d  dxdydz .
Although stresses act on all surfaces of the element, to simplify our analysis, we first
consider only the x-direction stresses that act on the six surfaces of the differential
element, as shown figure 5.8. For the purposes of our differential analysis, we again use
a truncated Taylor series to represent the spatial changes in the x-direction stresses across
the differential element. Here one needs to recall the manner in which we define the
direction of a positive stress acting on a surface, as discussed in Chapter 1.

Note that we assume that all applied stresses act with a positive orientation (not
necessarily direction) relative to the coordinate system employed. By making this
assumption during the initial derivation, when we later solve the resulting differential
equation, the sign of the resultant stress will indicate whether the stress acts with a
positive or negative orientation on the respective surface of interest (we will revisit this
point when we engage in solutions of the Navier-Stokes equation in Chapter 6).

99
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 5

 yx
 yx  dy
y

y  zx


 xx  xx  xx dx
dy x

dz
dx

 yx
 zx  zx dz
z
x
Remember, xy

z
surface direction

Figure 5.8 Illustration of all x-direction stresses acting on the bounding

surfaces of a differential element.

Recall that the direction for the positive orientation of a stress depends on the product
sign of (a) the outward normal vector for a surface, and (b) the direction the stress acts on
that surface. Thus, a positively-oriented stress is indicated by either: (1) a positive
surface normal with a stress acting in a positive coordinate direction, or (2) a negative
surface normal and a stress acting in a negative coordinate direction.

For our derivation, it is a little awkward to assume a base value for each stress at the
center of the differential element, since stresses are generally associated with surfaces of
action. Thus, we assume that all base values of the assumed stresses are those acting on
the surfaces of the differential element that are closest to the origin of our coordinate
system. For example, the normal stress, xx, acting on the x-surface in the x-direction is
assumed to act on the left face of the differential element, as shown in figure 5.8.
Likewise, the shear stress, yx, acting on the y-surface in the x-direction is assumed to act
on the bottom face of the differential element. Note that since these respective surfaces
have an outward normal that points in a negative coordinate direction, each of these
assumed stresses is assumed to act in a negative x-direction to yield a positive orientation
for the assumed stress, as per our convention.

As was done previously in section 5.2, a truncated Taylor series is used to characterize
the changes in the respective stresses across the differential element, using the differential

100
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 5

dimension appropriate for the direction of expansion. For example, to characterize the
normal stress on the right facing x-surface, we expand xx across the differential distance

dx, giving a normal stress of  xx  xx dx  . This stress has a positive orientation in the
x
positive x-direction since the outward normal for the right facing surface is also in the
positive x-direction. Similarly, the x-directed shear stress on the top surface is

 yx  yx dy  , since we expand across dy in a positive y-direction. Likewise, this stress
y
has its positive orientation in the positive x-direction since the outward normal for the top
surface points in a positive y-direction. The other similarly modeled x-directed stresses
on the differential element are as shown in figure 5.8.

Having modeled our assumed x-directed stresses on the differential element, we now
 
apply Newton’s second law, dF  dm a , using the differential stresses shown in figure 5.8
to model the resultant x-direction surface forces on each face of the differential element.

To characterize a generic differential body force, we assume that this mass-based force is

characterized by some directional force field we define as  . Such a physical field could
be a gravitational field (normally the case), a magnetic field, or even possibly an electric
field (if it can cause a mass-based force for the fluid material in question). Commonly,
  
we assume that   g , where g is a gravitational field, appropriately oriented relative to
our selected coordinate system.

Having characterized all the relevant surface stresses and the body force field, we can
now model the total forces acting on the differential element. Note that to establish the
surface forces, we further assume that the modeled stresses act uniformly across the
differential area of the respective bounding surfaces. For example, the x-directed
differential force acting on the top surface of the differential element (call it dFy+dy, x) will
be a product of the x-directed shear stress times the differential area it acts on, dAy =dzdy,
yielding:

  
dFy  dy , x    yx  yx dy  dz dx 
 y 

Other x-directed surface forces can be modeled similarly from the modeled stresses.
For the body force, we assume that the both the density and the body force vector are
essentially uniform across the infinitesimal differential element. Similar to the approach
employed for derivation of the differential equation for continuity, the density is assumed
to vary linearly across the differential element (since it is assumed infinitesimal), and thus

101
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 5

we use the assumed center value of density, , when determining the body force. Thus,
the differential mass contained within the element is assumed to be the product of the
density and the differential volume, dm = (dxdydz). Since the body force is
proportional to the mass, the subsequent differential body force is modeled as the product
 
of the differential mass and the body force vector field, or dFbody   dx dydz  .

If we now sum all x-direction forces acting on the differential element, eliminating like
terms, we obtain:

body force field (e.g. gravity)

x-component

  
dFx   x dxdydz    xx  xx dx dydz   xx dydz
 x 
  yx 
   yx  dy dxdz   yx dxdz
 y 
  
  zx  zx dz dydx  zx dydx (5.30)
 z 

    
  x  xx  yx  zx  dxdydz
 x y z 

5.6.2 The Differential Momentum Equation

Substituting Eq. 5.30 into the x-direction component of Newton’s second law for the
differential element,  dFx  dm a x , where we employ the substantial derivative of the x-
direction velocity, u, to model ax, we have:

  xx  yx zx   u u u u 
 x  x  y  z  dxdydz  dxdydz  t  u x  v y  w z  (5.31a)
   

dFx dm ax = Du/Dt

Similarly, for the y and z directions, we obtain:

 xy  yy zy   v v v v 
 y  x  y  z    t  u x  v y  w z  (5.31b)
   

102
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 5

 xz  yz zz   w w w w 
 z  x  y  z    t  u x  v y  w z  (5.31c)
   

Eqs. 5.31a,b,c are the respective component equations for Newton’s second law applied
to a differential fluid element, and collectively they comprise the vector equation,
  
  
 F  t mV  m a . Note that these equations are in their most general form, and will
apply to any fluid, regardless of its properties (i.e. viscous or visco-elastic, compressible
or incompressible, etc.).

5.6.3 Relating Stresses to Fluid Rates of Strain

Since we are considering a homogeneous, continuous medium, we can reduce the stress
tensor components in the above equations from nine to six by noting that the moments
imparted about the center of our differential element by the shearing stresses must be in
balance when the material is in equilibrium (see Timoshenko and Goodier, 1970). This
allows us to equate the shearing stresses for adjacent sides of our differential element to
each other, so that:

 xy   yx ,  xz   zx , and  yz   zy

We are still far from a satisfying set of equations, since we still have six different stresses
within the governing equations. Now our objective is to represent these stresses in terms
of strain rates, which can be directly related to the fluid deformation, and thus reflected
by spatial derivatives of velocity, as was illustrated in Chapter 4. Since we are dealing
with a fluid continuum, we assume that the stress tensor for a fluid is linearly related to
the strain rate tensor of a fluid, in the same manner that the stress tensor for a solid
material is related to the strain tensor of a solid. This assumption of a fluid with a linear
stress to strain rate is termed a Newtonian fluid, after Sir Isaac Newton. Newton, in his
Principia Mathematica (1687), first hypothesized that most fluids demonstrate a linear
relationship between the applied stress and the rate of strain (although Newton expressed
it a little less concisely). This concept was later adapted by Stokes (1845), who
developed the mathematical model of a Newtonian fluid that is employed in modern fluid
mechanics, and which we discuss below.

To develop our analogy from solid mechanics, we first review the general relation of the
stress tensor to the strain tensor for a Hookean solid. Here, stresses are assumed to have
a linear relationship of the form:

103
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 5

 xy  c1 xx  c2 xy  c3 xz  c4 yy  c5 yz  c6 zz

In this most general relationship, each stress component is assumed to depend on all six
rate-of-strain components, . Clearly this is a mess, since similar equations for the other
five stress components would require 30 additional independent coefficients in order to
define completely the state of stress in the fluid. Therefore, we make an assumption of an
isotropic medium (i.e. that the material stress vs. strain properties are independent of
direction). This assumption (and simplification) reduces the number of coefficients
required to define the state of stress to two, since many of the original coefficients are
either identically zero or related to each other. Recall that these proportionality
coefficients for a Hookean solid are [for normal stresses]:

E - the modulus of elasticity--a constant of proportionality of normal stress to normal

strain; and

n - Poisson’s ratio--which relates the lateral contraction of a material to the axial

expansion (i.e. directly related to the volumetric change of a material under an
applied stress).

Using these coefficients, the stress-strain relationships for a solid can be represented as:

1
 xx 
E

 xx  n  yy  zz  
1
E

 yy   yy  n  xx  zz   Normal Stresses (5.32a)

1

 zz  zz  n  xx   yy 
E

21  n  1
 xy   xy   xy
E G
21  n  1 Shear Stresses
 yz   yz   yz (5.32b)
E G
21  n  1
 zx  zx  zx
E G

Here G is termed the modulus of elasticity in shear, and is a constant of proportionality

between shear stress and shear strain. Note that if we invert the shear strain equations in
Eqs. 5.32, and solve the three normal strain equations simultaneously to obtain explicit
expressions for each normal stress, we obtain:

104
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 5

nE E
 xx 
1  n 1  2n 
 xx   yy   zz    xx   xx   yy   zz   2G xx
1 n
nE E
 yy 
1  n 1  2n 
 xx   yy   zz    yy   xx   yy   zz   2G yy (5.33a)
1 n
nE E
zz 
1  n 1  2n 
 xx   yy   zz    zz   xx   yy   zz   2G zz
1 n

 xy  G xy
 yz  G yz (5.33b)
zx  G zx

In the three normal stress equations, Eqs. 5.33a, the first group of constants is given the
designator , and is termed Lame’s constant. Since this constant modifies the collective
strains, it is reflective of the volumetric expansion (or contraction) of the material. The
second constant is of course the modulus of elasticity in shear.

Similar arguments were made by Stokes in developing the stress-strain rate relationships
for a fluid. As pointed out earlier, a fluid cannot sustain a fixed strain, but will deform
continuously under an applied force or stress. This is what led Newton to the logical
conclusion that any force applied to a fluid will cause it to continuously deform, and that
the rate of this deformation, or strain rate, is the appropriate response of the fluid to the
applied force, or stress. What Stokes hypothesized was that a set of stress-strain rate
equations, analogous to the stress-strain equations for a Hookean solid, could be
developed for a Newtonian fluid, under the following set of postulates:

1. The fluid is continuous, and the components of the stress tensor will vary linearly
with the strain rate.
2. The fluid is isotropic, such that its properties are independent of direction.
3. When the fluid is at rest, with zero strain rates, the deformation equations must
reduce to the hydrostatic pressure condition (i.e.  xx   yy   zz  p ).

Note that our assumptions to this point satisfy the first two postulates. The third is
achieved by including a pressure term in the equations for the normal stresses.

Correspondingly, the stress equations that Stokes arrived at for a Newtonian fluid are
given in the following Eq. 5.34.

105
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 5

linear strain volume expansion shear strain

xx  p  2  xx    xx   yy   zz   xy  2 xy

 yy  p  2  yy    xx   yy   zz   yz  2 yz (5.34)
zz  p  2  zz    xx   yy   zz  zx  2 zx

In Eqs.5.34,  is termed the coefficient of viscosity,  is termed the coefficient of bulk

viscosity (sometimes the second coefficient of viscosity), and the  terms indicate the rate
of strain components. Note that the coefficient of viscosity, , is analogous to G for a
solid, accounting for viscous forces due to shear and incompressible deformation of a
fluid.  models the ratio of the applied stress to the local strain rate, and for Newtonian
fluids will remain essentially constant over a range of local strain rates (however, it can
vary, often strongly, with temperature). The coefficient of bulk viscosity, , performs the
same function as  for a solid, and relates the proportionality of the applied stress to the
volumetric expansion/contraction of the fluid. This term, as we will see, only plays a
significant role when variations in fluid density, or fluid compressibility, are relevant,
such as for flows of high-speed gases.

The inclusion of pressure in Eqs. 5.34 satisfies Stokes’ third postulate, such that if the
rates of strain vanish, we still have a normal stress acting on the fluid, caused by the
pressure acting inward on a fluid element. Note that the pressure term is taken as
negative since it will always act opposite to the outward surface normal for our
differential fluid element.

Recall that in Chapter 4, Eq. 4.9, we derived the rate-of-strain tensor (in Cartesian
coordinates) in terms of spatial variations of the fluid velocity field as:

 u 1  v u  1  w u 
      
x 2  x y  2  x z 
 xx  xy  xz  
    1  u v  v 1  w v  
E ab   yx  yy  yz         
2  y x  y 2  y z  
 zx
  zy  zz  
 1  u w  1  v w  w 
 2  z  x    
   2  z y  z 


So, substituting these velocity-based expressions for the rate-of-strain components into
the stress-rate of strain equations, Eqs. 5.34, we obtain:

106
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 5

u  u v w   u v 
 xx  p  2        xy    
x  x y z   y x 
v  u v w   v w 
 yy  p  2        yz     (5.35)
y  x y z   z y 
w  u v w   w u 
 zz  p  2        zx    
z  x y z   x z 

u v w 
To further simplify these stress equations, we substitute      V [the
x y z

divergence of V ] into Eqs. 5.35, and have:

u   u v 
 xx  p  2    V  xy    
x  y x 
v   v w 
 yy  p  2    V  yz     (5.36)
y  z y 
w   w u 
zz  p  2    V  zx    
z  x z 

To examine the impact of Stokes’ third postulate for these equations, we sum the three
component normal stresses of Eqs. 5.36 and obtain:

  
   
xx  yy  zz  3p  2   V  3   V  3p  2  3   V

However, if we assume a mechanical pressure (call it p ) that is the negative average of

the normal stresses, we can write:

xx   yy  zz 2  
p  p        V (5.37)
3 3 

Of course, if the fluid is at rest, and   V is zero, then Eq. 5.37 yields p  p , which
satisfies Stokes’ third postulate. However, when there is fluid motion, Eq. 5.37 is
problematic, since the mean pressure in a deforming viscous fluid may not be equivalent
to the thermodynamic pressure. This result presents a dilemma, although it is generally

not a problem for incompressible flows (where   V  0 always), and often has little
impact for compressible flows. Stokes, faced with this dilemma (which would greatly
complicate the momentum equation derivation), took the easy way out, and hypothesized

107
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 5

2
that     , which is known as Stokes’ hypothesis. This hypothesis, although
3
universally utilized, has been the subject of controversy (see White, 1991) and the
commentary by Gad El-Hak (1995) . However, as White notes, for most applications
Stokes’ hypothesis has little impact—with the exceptions being for shock waves, and
sound absorption and attenuation. For the purposes of this text, we will accept Stokes’
hypothesis, since it is consistent with the types of incompressible flow problems we
address here. Accepting Stokes’ hypothesis further reduces Eqs. 5.36 stresses to:

u 2   u v 
 xx  p  2    V  xy       yx
x 3  y x 
v 2   u w 
 yy  p  2    V  xz       zx (5.38)
y 3  z x 
w 2   v w 
zz  p  2    V  yz       zy
z 3  z y 
pressure linear stresses due to angular
strain compressibility deformation
rate

5.6.4 Reducing the Momentum Equation to the Navier-Stokes Equation

Now armed with the velocity-based models of Eqs. 5.38 for the fluid stresses, we return
to the differential equation for Newton’s second law, and substitute our Stokes simplified
stress models from Eqs. 5.38 for a Newtonian fluid into Eqs. 5.31. Considering the x-
direction equation first, Eq. 5.31a, we obtain:

  u 2      v u     u w 
 x    p  2    V           a x (5.39)
x  x 3  y   x y  z   z x 

If we further restrict the fluid to one of constant viscosity,  = constant, this yields:

p   u 2    v u    u w 
 x     2    V             a x (5.40)
x x  x 3  y  x y  z  z x 

Rearranging Eq. 5.40, and numbering select terms for further simplification, we have:


 x 
p
x
 2 u 2 
 2 2 
x 3 x
V 
2v
yx
  2u
 2u
 2  2 
y z
2w
zx
 a x (5.41)
1 6 2 3 4 5

108
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 5

By judicious rearrangement of the selected terms in Eq. 5.41, as indicated below, we can
make the collective identification of more compact terms which utilize the Laplacian and
divergence operators, as shown:


 x 
p
x
  2u  2u  2u    u v w  2 
  2  2  2        
x  x y z  3 x
  V  a x  
 x y z 
1 /2 3 4 1 /2 2 5 6


2u 1 

3 x
V  
Thus, the x-direction component of the momentum equation simplifies to:


 x 
p
x
1 
  2 u  
3 x
  V  a x  (5.42a)

Similarly, y- and z-direction components of the momentum equation are:


 y 
p
y
1 
  2 v  
3 y
  V  a y  (5.42b)


 z 
p
z
1 
  2 w  
3 z
  V  a z  (5.42c)

So, expressing the momentum equation for a Newtonian fluid collectively as a single
vector equation gives what is termed the Navier-Stokes equation—first derived
generically by Navier, and particularized for a Newtonian fluid by Stokes:

 
  1    Navier -
2
  p   V     V  
3
V
t
 
  V  V  DV
Dt
   (5.43)
 Stokes Eqn.
   
body  deformation compressible  advective 
force pressure stresses stresses local acceleration total
force acceleration acceleration


Now, in section 5.4.3 we showed that   V  0 for an incompressible fluid. Employing
this relationship in the equation above eliminates the compressible stresses term, giving:

 2
 DV
  p   V  
Dt

109
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 5

Since density is constant for an incompressible flow, it is often convenient to divide

through by , giving an equation of the form:


 1  DV
  p   2 V  (5.44)
 Dt

Here,   , and is termed the “kinematic” viscosity. Since  is the ratio of two fluid

properties, it is of itself also a fluid property. It is this form of the Navier-Stokes
equation, Eq. 5.44, that we will make the most general use of in this text, since it provides
a less cumbersome form for dealing with constant density flows.

Note that the vector form of the Navier-Stokes equation applies generically for any
coordinate system. Although we developed the equation in Cartesian coordinates, the
generic vector form of the equation is applicable for any coordinate system. For
example, the expanded x-direction equation in Cartesian coordinates is:

1 p   2 u  2 u  2 u   u u u u 
x    2  2  2     u  v  w  (5.45)
 x  x y z   t x y z 
  
g x if   g

However, note that the directional components of the Navier-Stokes equation expressed
in the cylindrical or spherical coordinate systems will contain additional terms that result
from the dependency of selected unit vectors employed in those systems upon changes in
angular position. For example the r-direction component equation in cylindrical
coordinates, shown below, contains “extra” terms due to the dependence of the îr and î
unit vectors upon changes in the angle .

  1 
rvr   12  v2r   v2r  22 v    vr  vr vr  v vr  v  vz vr 
1 P 2 2
 2

r    
 r  r  r r  r  z r    t r r  r z 

Extra terms due to

coordinate system
A listing of all the directional components of the incompressible Navier-Stokes equation
in Cartesian and cylindrical coordinates are given in Section 5.8, at the end of this
chapter. A listing of the Navier-Stokes equation, and its components, in spherical
coordinates can be found here.

110
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 5

5.7 Utilization of the Incompressible Continuity and Navier-Stokes Equations

For constant density flows, the continuity and Navier-Stokes equations are sufficient to
solve for the velocity and pressure fields within an isothermal fluid. As shown in figure
5.9 below (for Cartesian coordinates), these equations provide four equations (one scalar
plus three vector component equations) with four dependent unknowns (u, v, w, and p),
rendering an exact solution possible.

Continuity  Scalar equation (1 equation)

4 equations
Navier-Stokes  Vector equation (3 equations)
(momentum)

u, v, w, p

An exact solution
4 unknowns is possible

Figure 5.9 Schematic of the generic solution process for a constant density,
isothermal fluid, in Cartesian coordinates.

While we now have the equations to implement exact solutions, the solution process can
be quite complicated, particularly as the dimensional complexity of a flow increases.
And while closed form solutions are possible for one and two-dimensional flows, three-
dimensional flows generally necessitate the use of numerical solutions, employing
sophisticated computational approaches. The addition of time dependency to a flow
further compounds solution procedures, with all but the simplest of such flows requiring
both numerical solution approaches, as well as possible further modeling of the
unsteadiness properties, such as is done in the assessment of turbulent flows (see Chapter
17).

That being said, there is still much that can be learned about the behavior of fluid flows
using closed-form solution techniques. It will be the approach in this book to exploit
flow problems that lend themselves to closed-form solutions, which can then be
employed to examine the implications on the physics of fluid behavior. While modern
commercial computational programs can provide effective numerical solutions of many
useful fluid flow geometries of significant complexity, it is important that the user of
such programs understand the interplay of the physics within a fluid that influences those

111
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 5

solutions. In addition, it is through the solution and examination of simpler physical

situations that one learns the physical trade-offs inherent in the governing equations, and
develops the intuitive feel for what behavior to anticipate for a specific application, which
is the hallmark of a competent fluids engineer.

Accordingly, in the following Chapter 6 we will begin to examine the application of the
continuity and Navier-Stokes equations to selected physical situations, and illustrate how
to utilize appropriate simplifications that allow the extraction of tractable, closed-form
solutions of the velocity fields. We will do this by first considering simple solutions,
which involve one-dimensional dependence. Later, in Chapter 12, we will consider more
complicated solutions, which depend on two dimensions, or one dimension and time.

5.8 The Governing Equations for Incompressible Flow

5.8.1 Stresses on an Incompressible Fluid: Cartesian and Cylindrical Coordinates

u  u v 
 xx  p  2  xy       yx
x  y x 
v  u w 
 yy  p  2  xz       zx
y  z x 
w  v w 
 zz  p  2  yz       zy
z  z y 

v r    v  1 v r 
 rr  p  2  r   r        r
r  r  r  r  
 1 v  v r   v 1 v z 
   p  2    z        z
 r  r   z r  
v z  v v 
 zz  p  2  zr   z  r    rz
z  r z 

112
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 5

5.8.2 Continuity and N-S Equations in Cartesian Coordinates (,  constant)

u v w
  0 (5.46a)
x y z

u u u u 1 p 2u  2u  2u 
u v w   g x   2  2  2  (5.46b)
t x y z  x  x y z 

v v v v 1 p  2v 2v 2v 

u v w   g y   2  2  2  (5.46c)
t x y z  y  x y z 

w w w w 1 p  2w 2w 2w 

u v w   g z   2  2  2  (5.46d)
t x y z  z  x y z 

5.8.3 Continuity and N-S in Cylindrical Coordinates (,  constant)

1
rv r   1 v   v z  0 (5.47a)
r r r  z
v r v v v v v2
 vr r   r  vz r  
t r r  z r
(5.47b)
  1  2 2
v 

1 p
 g r    rv r   12  v2r   v2r  22  
 r  r  r r  r  z r  

v  v v v  v v v
 vr     vz   r 
t r r  z r
(5.47c)
1 p   1  2 2
 1  v  v v 
  g     rv    2 2  2  22 r 
r   r  r r  r  z r  

v z v v v v
 vr z   z  vz z
t r r  z
(5.47d)
1 p 1   v z  1  2 v z  2 v z 
  g z   r  2  
 z  r r  r  r 
2
z 2 

113
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 5

References

Fox, R.W. and McDonald, A.T. , Introduction to Fluid Mechanics (any edition), John Wiley &Sons
Wikipedia: Liebnitz integral rule; Wikiepedia: Differentiation of integral
Wikipedia: Solenoidal field
Stokes, G.G. (1845) “On the Theories of Internal Friction of Fluids in Motion,”
Transactions of the Cambridge Philosophical Society, 8, pp. 287–305.

Timoshenko, S.P. and Goodier, J.N. (1970) Theory of Elasticity, McGraw-Hill Inc.
White, F. M. (1991) Viscous Flow, McGraw-Hill Inc.

Problems

1. Do the following:

a) For a 2-D, incompressible flow, u = Ax. If v = 0 along x-axis, determine v.

b) For a 3-D, incompressible flow, u = 2x and v = -y. Determine w, if w = 0 in the x-y plane.

c) The velocity of a flow is given by: V  xyˆi  y 2 ˆj  xzkˆ
i) Is this an incompressible flow? Why?
ii) If not, determine the rate of change of density at (x, y, z) = (1, 2, 1), where  = 2.

2. Do the following:

a) For a 2-D, incompressible flow, u = Axy. If v = 0 along x-axis, determine v.

b) For a 3-D, incompressible flow, u = 2y and v = -x. Determine w, if w = 0 in the x-y plane.

c) The velocity of a flow is given by: V  xyˆi  y 2 ˆj  zxkˆ
i) Is this an incompressible flow? Why?
ii) If not, determine the rate of change of density at (x, y, z) = (2, 1, 1), where  = 2.

 1 2
 DV
3. Starting from the vector equation,   p   V  (Eq. 5.44), derive the
 Dt
The Navier-Stokes equation for cylindrical coordinates (Eqs. 5.47b-d).

114
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

Chapter 6

Simple Solutions to the Navier-Stokes Equations

Contents

6.1 Simplifications of the Governing Equations .................................................................. 116

6.1.1 Steady Flow ....................................................................................................... 116
6.1.2 Dimensionality ................................................................................................... 116
6.1.3 Fully-Developed Flow ......................................................................................... 117

6.2 Boundary Conditions ................................................................................................... 118

6.2.1 Solid Boundaries and the No Slip Condition ........................................................ 118
6.2.2 Porous Boundaries ............................................................................................. 119
6.2.3 Boundaries within a Homogeneous Fluid ............................................................ 119
6.2.4 Boundaries Between Dissimilar Fluids ................................................................. 120

6.3 Simple One-Dimensional Solutions with Parallel Boundaries ........................................ 120

6.3.1 Couette Flow ...................................................................................................... 120
6.3.2 Poiseuille Flow in a Channel ................................................................................ 125
6.3.3 Fully-Developed Flow Between Porous Plates ..................................................... 130
6.3.4 Two-Fluid Couette Flow ...................................................................................... 138
6.3.5 Two-Fluid Poiseuille Flow in a Channel ............................................................... 141
6.3.6 Falling Liquid Film on a Wall ................................................................................ 147

6.4 Simple One-Dimensional Solutions with Radial Symmetry ........................................... 150

6.4.1 Poiseuille Flow in a Pipe or a Tube ...................................................................... 150
6.4.2 Couette Flow Between Concentric Rotating Cylinders ......................................... 154
6.4.2.1 A Cylinder Rotating in an Infinite Fluid ..................................................... 157
6.4.2.2 Fluid Rotating within a Cylinder ............................................................... 157
6.4.3 Poiseuille Flow in a Duct of Annular Cross-Section ............................................. 158
6.4.4 Couette Flow Between Porous Concentric Rotating Cylinders .................................162

6.5 Summary ................................................................................................................... 172

In this chapter we begin to examine solutions of the Navier-Stokes equation. While the
combination of the Navier-Stokes and continuity equations presents a particularly
daunting set of equations, experience has shown that we can develop closed-form
solutions for a number of simple flows by the implementation of several simplifying
assumptions that reduce both the dimensionality and complexity of the governing
equations. The value of these simpler solutions is that they allow one to develop accurate
solutions for highly-simplified physical situations, which then provide insight into the
dynamics of fluid interactions, and allow examination of the behavior of associated
physical properties (e.g. shear stress and vorticity). The subsequent examination of the
predicted fluid motion, and the associated properties, allows us to understand the impact

115
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

of a moving fluid, and develop a physical appreciation for the expected behavior in
different types of flow situations.

In its full-blown glory, the Navier-Stokes equation comprises a set of three coupled,
partial differential equations of second order. As pointed out in Chapter 5, the
mathematical form of these equations depends upon the specific flow conditions and
geometry, which may render the equations elliptic, parabolic, or hyperbolic in nature.
We will comment on the specific form and relevance of the simplified equations as we
examine appropriate solution techniques, and illustrate the necessary conditions required
to establish solutions. In the present chapter, we address only steady, one-dimensional
flows, which will render the equations as ordinary boundary-value problems. For such
problems, we must be able to specify the conditions of the appropriate dependent variable
(e.g. u, v, or w, and p in a Cartesian system) at the flow system boundary, as defined by
the relevant independent coordinate (e.g. x, y, or z).

In considering appropriate solutions of the governing equations, we first discuss the types
of simplifications we can invoke, and then the types of boundary conditions that may be
employed. We then examine a number of simple solutions that are subject to specific
simplifications and boundary conditions.

6.1 Simplifications of the Governing Equations

6.1.1 Steady Flow

Many flows of practical interest can be construed as steady (i.e. not changing with time).
Examples of such flows are steady flow through a pipe, the air flow around an airplane in
steady flight, or the lubricant moving between two surfaces of a continuously rotating
sleeve bearing. In such cases, the time-dependent term in the Navier-Stokes equation
will be negligible, leaving the advective acceleration terms as the only acceleration
components in the Navier-Stokes equation. Such a condition will arise, for example,
when a fluid flows steadily through a contracting nozzle. As pointed out earlier, a fluid
particle will undergo advective acceleration within a contraction as the fluid particle
passes through the increasing velocity field; however, when we view the flow at a fixed
point, there will be no local change with time, thus the flow is steady.

6.1.2 Dimensionality

Recall that by the dimensionality of a fluid, we refer to the number of independent

coordinate dimensions (e.g. x, y, z) required to specify the functional behavior of a
dependent variable (e.g. u, v, w, p). Now, all real flows are three dimensional to a lesser
or greater extent, depending on the initial flow conditions and the geometry that the flow
passes through or around. However, certain flow geometries lend themselves to
situations where we can make a reasonable assumption of either two- or one-dimensional
flow, and thus greatly simplify the governing equations. Additionally, appropriate
dimensional constraints can render one or more of the dependent variables as null, thus
eliminating one or more of the directional equations of motion.

116
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

The flow between parallel flat plates of infinite extent is an example of a reduced
dimensional and variable type of flow. For such a flow, the sides constrain the flow to
motion only in directions parallel to the plates, which eliminates any flow normal to the
plates, as we shall see in several examples we consider in this chapter. Such parallel
plate flows are driven either by a pressure change parallel to the plates, or by the plate
motion itself. If the driving forces are constrained to a one dimensional direction, it is
relatively easy to show that changes will only occur in the velocity component in the
direction of application of the driving force(s). This will consequently reduce the Navier-
Stokes equation to a one-component equation. Thus, any changes to the dependent
velocity component will be constrained to changes only in the coordinate direction of, or
normal to, the relevant velocity.

We will examine several of these constrained types of flows, where only one velocity
component is relevant, and which undergo changes in only one coordinate direction.
While such flows are highly idealized, they present quite tractable problems that lend
themselves to straightforward analytical solutions, since the governing equation is
reduced to an ordinary differential equation (ODE). And the subsequent solutions allow
for straightforward examination of the properties of the moving fluid, such as shear stress
and vorticity, providing a basis to examine the behavioral tradeoffs that take place in a
fluid flow. Accordingly, as we relax the constraints on a system, such as the examination
of flows between converging surfaces, our more constrained solutions will provide a base
of understanding that will help us assess the generally more complicated and complex
flows that develop for less constrained and more geometrically complicated flow
systems.

6.1.3 Fully-Developed Flow

When a steady flow passes through a constant cross-section channel or pipe, we often
constrain our solution to the fully-developed flow that will exist when the flow has
reached equilibrium, such that the velocity profile no longer changes with downstream
position. Such conditions occur when a flow between parallel plates, or within a pipe or
other constant cross-section duct, has passed far enough downstream such that the flow
undergoes no further redistribution of fluid momentum. Under such a condition, the
velocity distribution of the flow will remain fixed with respect to all coordinate
directions. This characteristic greatly simplifies the governing equations, since all
advective acceleration terms are null, which essentially reduces the Navier-Stokes
equation to a balance between viscous and pressure forces. Since the basis of the Navier-
Stokes equation is Newton’s law, the presence of a fully-developed flow simply reduces
the analysis of a flow field to a balance of forces, with no fluid acceleration.

Such fully-developed flows generally yield simple governing equations and solutions.
These solutions are not only instructive on fluid flow behavior, but also form the
fundamental basis for several instruments that are employed to measure fluid viscosity.

117
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

6.2 Boundary Conditions

From your calculus courses, you should recall that the complete solution of a differential
equation requires that we be able to specify as many boundary conditions as the order of
the derivative functions for each dependent variable relative to each independent variable.
For example, consider the x-direction Navier-Stokes equation, given by:

1 p   2 u  2 u  2 u   u u u u 
x    2  2  2     u  v  w  (6.1)
 x  x y z   t x y z 

Equation 6.1 contains second-order derivatives of u relative to x, y, and z. This requires

two boundary conditions for each of these derivatives, providing two boundary
conditions on either u or the first derivative of u at specified bounding values of x, y, and
z. Thus, we require six boundary conditions for u. Note that we also require a seventh
condition in x for the pressure, and an eighth for u at a specified time, t. This latter time
condition is usually termed an initial condition, although it is still a condition for a
boundary value problem. Thus, to solve the collective set of three scalar equations
comprising the components of the Navier-Stokes equation would require an additional
eight boundary conditions for each equation, for a total of 24 bounding conditions. Of
course these 24 conditions will also satisfy the boundary conditions for the continuity
equation, our fourth equation necessary for development of a complete solution.

At this point one can be staggered by the complexity of the solution process for a fully
three-dimensional, time-dependent flow, and can appreciate the need for some degree of
simplification. Generally, as the governing equations are simplified (or eliminated
because of non-varying dependent variables), the number of bounding conditions is
reduced accordingly. In our most simplified situations, such as our first example below
for a Couette flow (Section 6.3.1), which reduces to a second order ODE, we only require
two boundary conditions. As we address more complicated flows, with either multiple
equations or additional order derivatives, we will require additional boundary conditions.
The most complex flow we address in this text is a boundary layer flow in two
dimensions, which necessitates five boundary values: two for u with y, one for u with x,
one for v with y, and one for p with x.

6.2.1 Solid Boundaries and the "No Slip" Condition

The most common boundary condition is generally a solid boundary. Boundary

conditions for a solid boundary will always be of the first kind, which indicates a
specified value of a dependent variable (e.g. u = 4) that will exist at a specified spatial
location (e.g. at y = 1). For example, the velocity of a viscous fluid adjacent to a solid
boundary will always be identical to the velocity of the boundary, as a consequence of
what is termed the “no slip” condition.

The no slip condition is a result of fluid-surface behavior at a molecular level. The fluid
molecules immediately adjacent to a solid surface fill the molecular voids in the solid
surface and mechanically bond (by adhesion) to the molecules of the solid surface. In

118
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

contrast, fluid interaction away from a solid surface is due to molecular cohesion, caused
by individual fluid molecules being mutually attracted to other fluid molecules; the
subsequent effect of molecular cohesion is reflected by the property of fluid viscosity, as
discussed in Chapter 1. Since mechanical adhesion is much stronger than molecular
cohesion, a fluid adjacent to a solid surface will move at the velocity of the surface,
creating the so called no slip condition. For example, if a boundary is stationary, the
bounding velocity is zero; if the boundary is moving in the x-direction at a velocity U,
then the boundary condition is u = U. Additionally, the velocity normal to a solid
boundary is always zero (unless the boundary moves in a normal direction).

6.2.2 Porous Boundaries

Porous boundaries are quite interesting. Such boundaries in reality are such things as
screens or (to a certain approximation) perforated plates. These boundaries differ from a
solid boundary in that fluid may penetrate perpendicular to the bounding surface. For
porous surfaces we idealize that the flow parallel to the bounding surface is still subject
to the no slip condition, and thus moves at the surface velocity in a direction parallel to
the surface. However, the permeability of the surface allows flows to move normal to
and across the surface, which requires the normal velocity to be specified at the surface.
Again, these are boundary conditions of the first kind, with the velocity value specified at
a particular spatial position or surface. An example of a porous surface (although the flow
is not a Newtonian fluid) is a moving wire belt that is used in paper making processes. In
paper making, fluidized paper pulp (generally 1% wood fiber in water by volume) flows
tangentially onto a moving belt, where the water is rapidly removed from the pulp by
suction across the moving, porous wire surface. Other quite interesting examples of
porous boundaries will be addressed in this and other chapters to follow.
6.2.3 Boundaries within a Homogeneous Fluid
Boundary conditions within a homogeneous fluid may be either of the first kind or
second kind (where the first derivative of an independent variable is specified at a fixed
position). An example of a boundary condition of the first kind would be if a fluid
extends to infinity (or at least very far from another bounding surface), at which point the
velocity asymptotes to a fixed value, such as a fluid which is at rest far from a moving
surface. Another example would be where a fluid is moving at a constant value outside a
region of large viscous change, such as the flow outside of the boundary layer that forms
when a viscous fluid flows over a solid boundary.

Examples of a boundary condition of the second kind usually occur where the shear stress
in the fluid can be considered either zero or a constant value. Since shear stress is
directly proportional (for a Newtonian fluid) to the first derivatives of velocity variables,
such a bounding condition would indicate a null or constant value of the appropriate
velocity derivative(s). This type of boundary condition can occur where a fluid is again
assumed to be either at rest or moving at a constant velocity, or where it can be surmised
that the velocity has reached either a maximum or minimum (which by definition yields a
null first derivative). Such an example is the flow through a pipe or tube, where by
symmetry it can be inferred that the velocity achieves a maximum at the centerline.

119
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

6.2.4 Boundaries between dissimilar fluids

When two fluids move adjacent to one another, this presents an interesting set of
boundary conditions, since both the velocity and the shear stress at the interface between
the two fluids must be identical, since the velocity in a Newtonian fluid cannot be
discontinuous (by definition). This results in coupled boundary conditions of the first
and second kind (via the first derivatives of velocity, related through the Stokes shear
model of Chapter 5), applied at the fluid-fluid interface. These types of conditions apply
for the common movement of adjacent immiscible fluids, such as water and oil.

An interesting two-fluid process is the coupled adjacent flow of water and air. Due to the
significant variance in densities (approximately 1000:1), it would seem that the degree of
interaction would be insignificant. And when the prime mover is water, such as a
flowing river, this is to some degree true. However, when the prime mover is a high-
speed airflow, as experienced in storms over the ocean, the result can be the transference
of significant motion and energy to the water mass, as evidenced by the generation of
ocean waves, which are essentially caused by the solar induced movement of air (i.e.
“winds”).

6.3 Simple One-Dimensional Solutions with Parallel Boundaries

The simplest types of flows are those between closely-spaced parallel plates. Here we
assume that the plates have some finite spacing, and extend to infinity in both the flow
and in the cross-flow directions. In the following, we examine several classic parallel
plate flows.

6.3.1 Couette Flow

The Couette flow is the simplest possible viscous flow. However, this type of flow has a
variety of applications, from the development of viscometers for the measurement of
fluid viscosity, to the assessment of energy dissipation in fluid bearings used by rotating
machinery. The concept, as shown in figure 6.1, is quite simple. A viscous fluid is
bounded between two parallel plates of infinite extent. One plate then moves at a
constant velocity U, while the other plate remains fixed. Here we make several
simplifying assumptions, as indicated below. Key among these is that the flow is both
steady and fully developed, which negates any fluid acceleration effects. Thus, we only
need to consider a balance of forces—and only shear forces, since (as we will show)
pressure forces are not relevant.

y U
h
x

Figure 6.1 Couette flow between a lower stationary surface and an upper moving
surface.

120
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

Since this configuration is rectangular (we will examine the axisymmetric version of this
problem later), we utilize the Navier-Stokes equation in Cartesian coordinates to develop
a solution. If we assume that the plates are infinite in extent in the z-direction, then we
can further assume that there will be no flow changes in the z-direction (i.e. into the page,
as drawn above). This assumption allows us to neglect the z-direction equation, and any
terms in the x- and y-direction equations that contain w, the z-direction velocity.

So, writing the x and y Navier-Stokes equations in two-dimensions, we have:

1 2 3 2
u u u 1 p   u  2u 
2
u v  x    2  2  x-direction N-S (6.2a)
t x y  x  x y 
1 2, 3 3 3 3
v v v 1 p  2v 2v 
u  v  y    2  2  y-direction N-S (6.2b)
t x y  y  x y 

Initial Assumptions

1. Steady flow
2. Fully-developed flow; no variations in x-direction
3.  = constant
4. Neglect body forces

We now delete non-relevant terms from the Eqs. 6.2 as a consequence of our simplifying
assumptions. Note that the red arrows indicate the terms that we neglect in the equations,
with the associated number indicating the assumption that allows us to neglect the term.

It may be initially unclear why assumption 3 results in the elimination of velocity terms
in the equations. This is a consequence of the continuity equation for an incompressible

flow. Recall for a constant density flow that the continuity equation reduces to   V  0 ,
which in two dimensions is:
2
u v
 0
x y

However, because of the fully-developed flow assumption, the continuity equation

reduces to:
2
v
 0  v = f(x) + constant  v = constant
y

Note that v will be a constant, since the fully-developed flow assumption means that the
velocity characteristics will not change with x-position. To determine the constant in the
equation, we note that the boundary condition for v at either solid surface is zero. So,

121
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

taking v = 0 @ y = 0, then v = 0 everywhere (which also means that higher derivatives of

v will also be null as well). Thus, continuity considerations lead to the five cancellations,
noted by 3, of v-associated acceleration terms in Eqs. 6.2. Thus, by utilization of the first
three assumptions, Eqs. 6.2 are reduced to:

1 p  2u
x    2  0 , and (6.3a)
 x y
1 p
y  0 (6.3b)
 y

Now, our fourth assumption above allows us to drop the body force terms represented by
x and y. However, let’s consider why this is a reasonable assumption.

If we assume that gravity acts in the y-direction (vertically, as we look at this page), then
we can write:

   x ˆi   y ˆj  gˆj

This renders Eqs. 6.3 as:

 2 u 1 p
  , and (6.4a)
y2  x
p
 g (6.4b)
y

Integrating the y-equation, Eq. 6.4b, yields:

p  gy  f ( x )  C (6.5a)

Now, if we let the flow be quiescent, then we would have a hydrostatic condition, where
p
u = 0, and Eg. 6.4a reduces to  0 . Integrating the simplified Eq. 6.4a yields:
x

p  f ( y)  C (6.5b)

Comparing Eqs. 6.5a and 6.5b, we note that since Eq. 6.5b contains no function of x, then
f(x) in Eq. 6.5a must be zero. Thus, Eq. 6.5a can be reduced to p  gy  C , or
pH y  gy  p0 , where p0 is the pressure at y = 0. PH(y) is the pressure due to
hydrostatic effects, which will exist whether or not motion is imparted to the fluid. Now,
when we allow the fluid to move, equation Eq. 6.5a becomes:

p  gy  p 0  f ( x)  p H y  Px  (6.6)

122
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

where p H y  reflects the hydrostatic pressure due to gravity, and P(x) reflects the
dynamic pressure due to the movement of the fluid. Now, substituting Eq. 6.6 back into
the original Eqs.6.4 gives:

 2 u 1 P p H P
  and   g
y 2  x y y

Or rearranging (noting =µ/) yields:

 2 u 1 P P p
 and  g  H  g  g  0
y 2  x y y

Thus, the dynamic pressure, P, can only be a function of x, the direction of fluid motion.

Now consider the assumption that the plates are infinite in the x-direction. The only way
that such an assumption is realistic is if the plates were to actually circumvent the earth,
and join again, making this a set of continuous parallel surfaces, one moving relative to
the other. Such a situation renders the fluid region between the plates connected, and as
such we could draw streamlines through the fluid, parallel to the plates. However, this
also implies that when these streamlines meet after circumventing the earth, the pressure
at that juncture point must be continuous, and thus identical. And since any point along a
continuous streamline could be the juncture point, then the pressure at all points along the
streamline must be identical. Thus, the pressure must be a constant, and the derivative of
pressure with respect to x must be zero.

P
Thus, recognizing that  0 , the Navier-Stokes equations for a Couette flow reduce to:
x
 2u d 2u
 0 or  0 , since u is only a function of y (6.7)
y 2 dy 2

Since Eq. 6.7 is a second order equation for u with respect to y, we require two boundary
conditions. Clearly the appropriate boundary conditions are:

1) u  0 @ y  0 2) u  U @ y  h.

Integrating Eq.6.7 twice yields:

du
 C1 Applying the B.C.s 1) 0  C1 (0)  C2  C2  0
dy
U
u  C1y  C 2 2) U  C1h  C1 
h

123
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

Therefore, the general equation for flow between two parallel plates, with the lower one
fixed and the upper one translating at velocity U is:
U
y y
uU (6.8)
h x
The shear stress for this flow field is given by:
0
 u v  du U  U
 yx           (6.9)
 y x  dy h h

Equation 6.9 indicates that the shear is constant across the entire flow field. Additionally,
the vorticity is given by:
0
 v u  du U
zz           (6.10)
 x y  dy h

Equation 6.10 indicates that the angular velocity is also constant, with the fluid rotating in
a clockwise direction.

An additional property of interest is the total flow rate for the fluid passing a cross
section. If we assume the depth of the flow (into the page) to be W, we can calculate the
volume flow rate, Q, crossing the region 0  y  h by noting that:

yh yh
Uy UWh
Q  u Wdy  
y 0 y 0
h
Wdy 
2
(6.11)

Equation 6.11 is the flow rate induced by the movement of the upper plate due to viscous
interaction between the fluid and the plate surface.

Although this solution is for a set of parallel plates, to a reasonable approximation the
results are representative of the behavior of a flow between two closely spaced circular
surfaces with a low degree of curvature, such as in large journal bearings or any situation
where two closely-spaced plates of limited curvature move relative to one another. How
one assesses closely adjacent surfaces with a high degree of curvature will be addressed
in section 6.4.2, where we examine simple Navier-Stokes solutions with circular
symmetry.

Before we leave this Couette problem, note that the same type of solution will apply to
the same geometry with different first order boundary conditions. For example, if we
held the top plate fixed, and translated the lower plate, the boundary conditions would be:

1) u  U @ y  0 and 2) u  0 @ y  h.
y

The subsequent solution would then be: x U

124
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

y  y
u  UU  U 1   (6.12)
h  h

Likewise, if we translated both of the plates at different velocities, we would obtain

another variation on the linear Couette flow solution.
Another consideration is the impact of the orientation of the coordinate system. Suppose
we orient the coordinate system such that the origin is at the upper surface, pointing
downward, as shown in figure 6.2. Here Eq. 6.7 is still valid. However, the boundary
conditions are now:

1) u  U @ y  0 2) u  0 @ y  h.

x U
h
y

Figure 6.2 Couette flow with an alternative coordinate system.

The subsequent solution for the velocity profile will be:

 y
u  U 1   (6.13)
 h
Equation 6.13 describes the same flow as Eq. 6.8, but is functionally different due to the
use of a different coordinate system. Also note that while Eq. 6.13 and Eq. 6.12 are
functionally identical, they represent two physically different flows. This example thus
illustrates the importance of understanding the particular coordinate system to which a
particular solution applies.

6.3.2 Poiseuille Flow in a Channel

Whereas a Couette flow is driven by the motion of the bounding surfaces, a Poiseuille
channel flow is driven by a pressure difference applied to the fluid flowing between two
parallel plates. For the purposes of this problem, we will assume that the pressure
p
change is in the x-direction, and represented by the pressure gradient term, , as shown
x
in figure 6.3.

Decreasing pressure,

y
h
x

Figure 6.3 Poiseuille flow between parallel plates; fully-developed flow with a
constant pressure gradient.

125
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

Again, because of the rectangular symmetry of this flow we utilize the Navier-Stokes
equation in Cartesian coordinates, and again assume that the plates are infinite in extent
in the z-direction (into the page), such that the flow is uniform and unchanging in the z-
direction. Thus, we can neglect the z-direction equation, and terms containing w.

For the x and y Navier-Stokes equations in two-dimensions, we have:

1 2 3 4 2
u u u 1 p   u  2u  2
u v  x    2  2  x-direction N-S (6.14a)
t x y  x  x y 
1 3 3 4 3 3
v v v 1 p  v  v 2 2
u  v  y    2  2  y-direction N-S (6.14b)
t x y  y  x y 

As indicated below, we again employ the same simplifying assumptions that were
employed for the Couette flow example. Note that here our assumption of fully
developed flow assumes that the flow is well downstream from any entrance effects (we
will discuss entrance changes later in Chapter 12) such that we again have a flow which
has no acceleration effects, and reflects a simple balance of shear and pressure forces.

Initial Assumptions

1. Steady flow
2. Fully-developed flow; no variations in x-direction
3.  = constant
4. Neglect body forces

As indicated by the arrows, we again delete the non-relevant terms from Eqs. 6.14, noting
again that the incompressible continuity equation in two dimensions yields:
2
u v
 0
x y

and
1
v
 0  v = f(x) + constant  v = constant
y

Again, at the plate boundaries, v = 0 @ y = 0 (or y = h), thus v = 0 everywhere. After

using this result, Eqs. 6.14 reduce to:

1 P  2u 1 P
  2  0 and  0
 x y  y

126
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

Note that the rationale for dropping the body force is the same as addressed for the
Couette flow, and thus the introduction of P to represent the dynamic component of
pressure. Simplifying the equations gives:

 2u 1 P 1 P 
2
  (where   ) (6.15a)
y  x  x 
P
and 0 (6.15b)
y

Integration of Eq. 6.15b indicates that P  f (x) . However, we don’t know the
P
functionality of P or , which is needed if we are to solve the x-direction differential
x
equation. We establish this functionality by taking the derivative of the x-direction
differential equation, Eq. 6.15a, with respect to x, which yields:

   2 u    1 P  1  2 P
    0 (6.16)
x  y 2  x   x   x 2

Equation 6.16 is equal to zero, since we assumed that u is fully developed, and as such its
derivatives must be a function of y only.

P  2P
Equation 6.16 indicates that must be a constant, since  0 . This is reasonable
x x 2
since the shear stress will be a function of the derivative of the velocity profile, which
will also be invariant with streamwise position. So, noting that u = f(y) only, Eq. 6.15a,
with appropriate boundary conditions simplifies to:

d 2 u 1 dP
  constant (6.17)
dy 2  dx

B.C. 1) u  0 @ y  0
2) u  0 @ y  h.

Integrating Eq. 6.17 twice yields:

du 1 dP
 y  C1
dy  dx
1 dP y 2
u  C1y  C2
 dx 2

127
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

1 dP 0
Applying the B.C.s 1) 0   C1 0  C2  C2  0
 dx 2

1 dP h 2 1 dP h
2) 0  C1h  C1  
 dx 2  dx 2
This yields a general equation for a pressure driven flow between two stationary parallel
plates as:

1 dP y 2 1 dP hy 1 dP 2 h 2 dP  y  y
u  
 dx 2  dx 2 2 dx
y  hy  
  1
2 dx  h  h
 (6.18)

The velocity profile reflected by Eq. 6.18 is parabolic (as shown in figure 6.3 above),
achieving a maximum midway between the plates, at y = h/2, or y/h = 1/2. If we term the
maximum velocity Umax, then:

h 2 dP  1  1 h 2 dP
Umax    1   (6.19)
2 dx  2  2 8 dx

Note the negative sign in Eq. 6.19. What this negative sign indicates is that in order for
Umax > 0 (i.e. flow in a positive x-direction) then dP/dx < 0. Since a Poiseuille flow is a
pressure driven flow, Eq. 6.19 indicates that the pressure must be decreasing in the
direction of flow (x, in this case), which is entirely logical and intuitive.

So, substituting Eq. 6.19 into Eq. 6.18 gives:

 y y
u  4 U max 1   (6.20)
 hh

The shear stress for this flow field is then given by:

du 4 U max  y
 yx    1  2  (6.21)
dy h  h

Equation 6.21 indicates that the shear varies linearly across the flow channel, varying
uniformly from a maximum positive value at y = 0 to a maximum negative value at y = h,
while passing through zero at y = h/2.

This shear stress result can be used to illustrate how the process used to define the
positive orientation for a surface force, as discussed in Chapter 1, applies here. Recall
that the absolute direction a stress acts, relative to the coordinate system employed,
depends on: (a) the surface of application, and (b) the sign of the calculated stress. Recall
also that the product of the signs of the surface normal direction and the stress direction
yield the directional sign of a positive stress. Thus, if a calculated stress is positive, it

128
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

will act in the direction of a positive stress; if a calculated stress is negative, it will act in
opposition to the direction of a positive stress.

So for the shear stress acting on the lower channel surface of the Poiseuille flow, from
Eq. 6.21 we calculate:
y
4 U max
 yx  at y = 0 yx h
h x
actual

The outward normal for the lower surface is oriented in the positive y-direction. Thus, to
have a positive absolute orientation, the product of the surface normal and the shear
direction must both be positive. Since the calculated shear stress value is positive, the
actual shear stress will act in a positive x-direction, as shown.

However, if we consider the upper channel surface, Eq. 6.21 gives the shear stress acting
on that surface as:

4 Umax y
 yx   at y = h yx actual
h
h x

Here, the outward normal for the upper surface is oriented in a negative y-direction.
Thus, to have a positive absolute orientation, the product of the surface normal and the
shear direction must both be negative. Since the calculated shear stress value is negative,
the actual shear stress will act in a positive x-direction, in opposition to the absolute
orientation for that surface, as shown.

Thus, the shear stresses acting on both the upper and lower surfaces are in the same
absolute direction (which we would anticipate by symmetry), even though the calculated
shear stress values are of different signs. As pointed out in Chapter 1, the process of
assigning a “positive” orientation for forces based on the sign product of the surface
normal/stress directions assures that the orientation of the force will be consistent, and
independent of the particular coordinate system employed.

The vorticity for this flow is given by:

 v u  du 4 U max  y 
z         2  1 (6.22)
 x y  dy h  h 

Equation 6.22 indicates that the angular velocity varies linearly from a positive maximum
at the upper surface to a negative maximum at the lower surface. As will be shown in
Chapter 11, vorticity is generated at both channel surfaces due to the pressure gradient —
positive at the upper surface and negative at the lower surface. These generated
vorticities then diffuse toward the symmetry plane due to viscous effects, where they

129
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

cancel each other out, thus maintaining a vorticity balance (we will demonstrate this more
thoroughly in Chapter 11).

The flow rate for this channel flow (assuming again a depth into the page of W) is given
by:
yh yh
1 dP dP W h 3 2
Q   u Wdy 
2  dx  y 2
 hyWdy  
dx
12
 Umax Wh
3
(6.23)
y 0 y 0

Equation 6.23 indicates that the flow rate for this viscous flow will be 2/3 of the flow rate
of a uniform, inviscid flow moving at Umax. Thus, the area-averaged mean velocity for a
Poiseuille flow will also be 2/3 Umax.

6.3.3 Fully-Developed Flow Between Porous Plates

Shown in figure 6.4 below is an interesting extension of the basic Poiseuille-type flow of
Section 6.3.2, consisting of a pressure gradient driven, steady flow between two parallel
porous plates with constant transverse flow across and through the plates. Since the
plates are porous, flow can move both parallel to the plates, like a conventional Poiseuille
flow, and transverse to the main flow, as illustrated in figure 6.4. This transverse flow
results in a loss of symmetry for the x-direction velocity profile, with the cross flow
skewing the u velocity profile towards the suction plate (the lower plate in figure 6.4
below).

This type of flow can be complicated if the transverse flow is allowed to vary in the x-
direction, or where the flow injected through the upper surface does not equal that
removed by suction from the lower surface (which would require the mass flow in the
main flow to change to conserve mass). However, for the present case, we will assume
that the transverse velocity is constant through both bounding surfaces, which renders
this problem quite tractable.

Transverse Flow [@ velocity -V (in y-direction)]

y h Main
Flow
x

Figure 6.4 Fully-developed laminar flow between parallel plates with injection/suction.

For this configuration, we note that the boundary conditions are given by:

u  0 and v  V @ y  0 and y  h

Here V is a constant.

130
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

The governing two-dimensional equations (x and y directions, for constant density and no
body forces) are given by:

1 2 2
u u u 1 P   u  2u  2
u v    2  2  x-direction N-S (6.24a)
t x y  x  x y 
1 2 v=const. 2 v=const.
v v v 1 P  2v 2v 
u v    2  2  y-direction N-S (6.24b)
t x y  y  x y 
2
u v
 0 continuity (6.24c)
x y

To simplify the equations, as shown by the red arrows, we assume:

(1) steady flow

  
(2) fully-developed flow   0
 x 

To simplify the equations further, we note that we can solve the simplified continuity
equation as:

v
 0  v = f(x) + constant
y

Here we note that v must be a constant since we have specified that the flow is fully
developed, which means that v  f(x). Thus, from our v boundary conditions (at either
boundary):

v  V

Since v is a constant, v  V will apply everywhere within the flow field. Thus, as
indicated, for Eq. 6.24b all v derivative terms drop out, leaving:

P
 0  P  f ( x ) only
y

Using an argument similar to that employed for the non-porous Poiseuille flow of Section
P
6.3.2, we can show that =constant. Thus, since u = u(y) only, we can rewrite Eq.
x
6.24a as:

131
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

-V
du 1 dP d 2u
v   2
dy  dx dy
du 1 dP d 2u
V   2
dy  dx dy

d 2 u V du 1 dP
or   = constant (6.25)
dy 2  dy  dx

Equation 6.25 is a second-order, linear, inhomogeneous ordinary differential equation.

du
To solve Eq. 6.25 we first substitute   , and rewrite the equation as:
dy

d V 1 dP
   const.
dy   dx

This reduces the equation to a first-order, linear, inhomogeneous equation, which we

solve for both the homogeneous and the particular solution, where:

   homo   part

We can show that,

1 dP
 part  , and
V dx
V
 y
 homo  C1e 
,

Such that:

V
du  y 1 dP
  C1e  
dy V dx

Integrating again for u, gives;

V
  y 1 dP
u  C1 e   y  C2
V V dx

Now, applying the u-velocity boundary conditions at the plate surfaces,

u  0 @ y  0 and y  h , yields the following solution for the x-direction velocity:

132
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

   y 
V

  1  e  

h dP  y  
u   (6.26)
V dx  h   h
V 
 1  e   
  
 

Vh y
To simplify Eq. 6.26, we let   (a form of Reynolds number) and y*  , and note
 h

that   . The resulting expression for the velocity profile can be expressed as:

  
Vh y 

 1  e  h  
 h 2 dP  y     h dP  y *  1  e 
 2  y*
u  (6.27)

 Vh dx h 


Vh
1  e   
  
  dx  1  e  
 

  
 

B=0

B=1
1
0.9 B=2
Velocity/Umax, B=0

0.8 B=4
0.7
0.6 B=6
0.5 B=10
0.4
B=20
0.3
0.2
0.1 Suction Direction
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Position (y*)

Figure 6.5 Channel flow velocity distributions as a function of suction parameter,

=Vh/, with dP/dx = constant for all  values.

Figure 6.5 illustrates the effect that the suction velocity, in terms of the suction parameter
, has on the velocity profile. In figure 6.5, we non-dimensionalize the velocity on the
h 2 dP
maximum velocity for a non-porous channel flow (Eq. 6.19), Umax   , which
8 dx
allows us to factor out the pressure gradient term. Note that this presumes that we
maintain identical pressure gradients for both the suction and non-suction channel flows,
which allows us to compare the effect of the transverse velocity on the fully-developed
streamwise velocity profile under the same driving condition (i.e. the pressure gradient).

133
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

As figure 6.5 clearly shows, the impact of the transverse flow is not only to skew the
velocity profile toward the suction wall (y*=0), but also to retard the streamwise flow.

To illustrate the magnitude of this retarding effect, we determine the volume flow rate as
a function of , as follows:
y* 1
  1 
 y *  e  y*  
y* 1

Q   u Whdy* 
h 3 W dP  1  e   y*
y* 1
   
h W dP  y * 
3 2
 
  dx y* 0 
 y*  dy*  
y*  0
1  e       dx  2 
1 e 
 
 
  y*  0
  1   1  1
 1  e 
h W dP  1  
3
   3

  
1  e    1 
Q       h W dP  1    (6.28)
  dx  2 1  e   1  e  
   dx  2 1  e   
   
 

To provide an appropriate reference, we non-dimensionalize the flow rate for the porous
channel, Eq. 6.28, relative to the flow rate for the non-porous channel, given in section
6.3.2 (Eq. 6.23) as:

dP W h 3
Q  0  
dx 12

1
Flowrate/Flowrate, B=0

0.9
0.8
0.7
0.6
0.5
0.4
0.3
Increasing Suction
0.2
0.1
0
0

8
10

12

14

16

18

20

Beta Parameter

Figure 6.6 The retardation of the volume flow rate through a porous
channel with increasing transverse velocity ( = Vh/).

134
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

dP
Here we again let the driving force, , be identical for both the non-porous and porous
dx
Q
channels. The result for is shown in figure 6.6, which shows a significant
Q0
retardation of the flow rate with increasing transverse flow.

Since this is a fully-developed flow, we reason that the increased flow retardation may be
due to a cumulative increase in the surface shear stresses. To examine this, we calculate
the shear stress for this velocity profile as:

 V y 
V
e
du h dP  1 
     V 
dy V dx  h  h

 1  e 


h dP   e y* 
or  1   (6.29)
 dx  1  e 

Figure 6.7 illustrates how the transverse suction velocity does affect the shear stress. In
this figure, we non-dimensionalize the shear stress for the porous channel relative to the
maximum wall shear stress for a non-porous Poiseuille flow determined in Section 6.2.2:

4 U max h dP
  0, Max  
h 2 dx

B=0
B=1
2 B=2
Tau/Tau max, B=0

1.5 B=4
Suction Direction
1 B=6
B=10
0.5
B=20
0

-0.5

-1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Position (y*)

Figure 6.7 Channel flow shear stress distributions as a function of suction parameter 

135
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6


Figure 6.7 plots as a function of the suction parameter, , and illustrates the
  0, Max
marked variations in shear stress that take place across the porous channel with
increasing .

Note that the shear stress adjacent to the suction surface (y* = 0) increases significantly
with transverse flow, whereas the shear stress is substantially reduced near the injection
surface (y* = 1). Note also that for higher  values the shear stress becomes almost
constant over a broad region of the flow.

Since there is no acceleration of the fluid for a fully-developed flow, we can perform a
two-dimensional, integral force balance on the fluid between the plates, for a distance L
in length, obtaining:

 F  P h  P h  
1 2 w yh   
L   w y o L  0 
y
wy=h
 V  V h V  x
 e   P1 P2
    
hL dP   h 
V
  h 
V
( P1  P2 )h    wy=0
V dx  1  e
    
 1  e 
    L
 
Or,
  h
V

P1  P2 h   hL dP V 1  e V   P1  P2   dP  dP   P1  P2 


(6.30)
V dx   h L dx dx L
1  e  

Equation 6.30 is the same result that we would obtain for a non-porous channel Poiseuille
flow. What this implies is that the sum of the shear stresses on the channel surfaces
exactly balances the streamwise pressure gradient. One can see in figure 6.7 that the sum
of the shear stresses on the top and bottom surfaces [note that one has to account for the
“absolute” direction of the magnitude of the stresses---both will act in the same direction]
will be the same, regardless of the transverse velocity (i.e.  y* h   y*  0 = constant for
all , including  = 0).

So if the cumulative shear stresses on the plate surfaces don’t change, and the pressure
gradient is held constant, why does the flow slow down with increasing ? The answer
lies in the x-direction momentum change that the transverse fluid must undergo in
crossing the channel.

When a fluid particle is injected at y* = y/h =1 it has zero streamwise velocity. As it

enters the channel, the collective action of the pressure gradient and viscous interaction

136
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

with the faster moving streamwise fluid causes the fluid particle to accelerate in the x-
direction as it initially proceeds across the channel. However, after the fluid particle
passes the point of maximum streamwise velocity, the fluid particle encounters a negative
velocity gradient, such that the fluid particle is now retarded in the x-direction by viscous
interaction with the slower moving streamwise fluid, as it approaches to the suction wall.
This retarding effect due to viscous interaction overwhelms the opposing effect of the
pressure gradient, causing the particle to decelerate as it approaches the suction surface at
y*=0, until it exits the suction surface with zero streamwise velocity. Thus, what
happens physically is that the fluid particle must be accelerated by the pressure gradient,
and then decelerated by viscosity as it traverses the channel. The net change in x-
direction momentum is zero, since the fluid particle enters and exits with zero streamwise
velocity. However, viscous effects subsequently dissipate all the x-direction momentum
imparted to the fluid particle by the pressure gradient.

Thus, although there is no net increase in shear at the bounding surfaces, there is an
internal dissipation of momentum that mediates the effectiveness of the driving force,
dP
. Consequently, the pressure gradient must not only balance the wall shear stresses,
dx
but must also provide the impetus to accelerate all of the injected fluid to steady state
conditions, prior to the injected fluid being removed and that momentum “lost.” The
result is a diminished flow rate, and a retarded velocity field, while experiencing the same
flow resistance as a non-suction channel.

Note that for higher  values, the velocity profile away from the injection surface
behaves almost like a Couette flow [e.g. the  = 20 profile in figure 6.5]. If we examine
the governing equation, Eq. 6.25,

d 2 u V du 1 dP
 
dy 2  dy  dx

what we realize is that the change in the shear stress in this Couette-type region,
d 2u
reflected by the term in Eq.6.25, is essentially negligible. Consequently, the
dy 2
constant pressure gradient causes the injected particle to undergo essentially a constant
streamwise acceleration as it moves across the channel, until it approaches the suction
wall and experiences the significant viscous x-direction deceleration necessary to allow it
to exit at zero streamwise velocity.

One of the problems given at the end of this Chapter concerns a Couette flow with
uniform transverse suction. In a boundary-driven Couette-type flow, a similar type of
retardation of the induced streamwise flow rate with increasing transverse flow is
observed, reinforcing that the necessity to continually accelerate newly injected fluid
particles will always reduce the effectiveness of the driving force in moving the main
flow. However, for the case of a Couette flow with transverse flow, the impact of the x-
direction momentum changes is experienced directly as an increase in the force required

137
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

to translate the surface due to the x-direction momentum lost through the suction surface,
and a consequent increase in the shear stress on the moving surface.

As a final consideration of this channel flow with transverse suction, note that if you let
V = 0, the result should reduce to the solution for a Poiseuille flow between two non-
porous parallel plates, as was derived in section 6.3.2 above. However, in order to obtain
these limiting conditions from the porous plate solution you will need to apply
L’Hospital’s rule for  0 (i.e. V0)----often multiple times to determine the limiting
values. The reader should do this as an exercise.

6.3.4 Two-Fluid Couette Flow

Consider a Couette-type flow, with two immiscible fluids (here we choose water and oil)
contained within two parallel constraining plates, and driven by the steady motion of the
top plate at a constant velocity U. For simplicity, we will assume that the two fluids are
of equal thickness, h, as shown. We also assume that the upper fluid is the oil, since oils
are always less dense than water (see this link).

U
h Oil o , o
y
h Water w , w
x

Figure 6.8 The two-fluid, Couette flow

We wish to determine the velocity behavior within the two fluids, and the resulting shear
stress on the moving plate. While a bit artificial, this flow will illustrate the use of
common boundary conditions when considering two mutually adjacent fluids.

We again utilize the two-dimensional Navier-Stokes equations, but we must apply a

separate set of equations for each of the fluids. The assumptions and simplifications that
apply for a single fluid Couette flow also apply here, and are:

1. Steady flow
2. Fully developed flow; no variations in x-direction
3.  = constant
4. Neglect body forces

Employing these assumptions, and noting that again the continuity equation yields v = 0
throughout both fluids, we reduce the governing equations to a pair of simple second
order ordinary differential equations:

d 2u o
0 (6.31a)
dy 2
and

138
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

d 2u w
0 (6.31b)
dy 2

In Eqs. 6.31, uo is the x-direction velocity within the oil, and uw is the x-direction velocity
within the water. Note that Eq. 6.31a and 6.31b each applies only within the region
occupied by the respective fluids, and the boundary conditions at the interface between
the two fluids (y = h) must match.

Here, the appropriate set of boundary conditions is:

1) u w  0 @ y  0 2) u o  U @ y  2h

du o du
3) u w  u 0 @ y  h 4) w  0 @ y  h (i.e.  o  w w )
dy dy

Note that we require four boundary conditions since we have two second-order
differential equations, each of which will yield two constants of integration. Also,
boundary condition four requires that the shear stress between the two fluids must match.
Since the shear stress (for this parallel flow) is proportional to the y-derivative of the u
velocities, this represents a boundary condition of the second kind.

Integrating Eqs. 6.31 twice yields two simple linear solutions:

u o  C1y  C2

and u w  C3 y  C4

Applying boundary conditions 1) and 2) yields:

U  C1 (2h )  C2  C 2  U  2hC1

and 0  C4

Such that:

u o  U  C1 y  2h  and u w  C3 y

Applying boundary conditions 3) and 4) at the common boundary, y = h, yields:

C3h  U  C1h and w C3  oC1

Solving for C1 and C3 simultaneously yields velocity distributions:

139
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

U y  2 h   w
uo  U  (for 2h  y  h) (6.32a)
h  w  o 
and
Uy o
uw  (for h  y  0) (6.32b)
h  w   o 

With corresponding shear stresses within each fluid of:

o U w o  w U
o   (6.33a)
h  w  o   w  o  h
and
w U o o  w U
w   (6.33b)
h  w  o   w  o  h
Note that Eqs. 6.33 indicates that o  w = constant across both fluids. This is a
reasonable result, since a force balance applied to the boundaries of the fluid indicates
that the shear at both the top and bottom boundaries must be equal, and thus the shear
stress must be in balance throughout the fluids.

Using these results, let’s consider a two-fluid flow of SAE 10 oil and water. At 20C, the
absolute viscosity of SAE 10 oil is about 100 times that of water, so the shear stress
within the oil (and at the top surface) and the water would be:

100 w  U
2
 U
o   0.99 w
101 w h h

By comparison, if the region between the plates contained all water or all oil, the
respective shear stresses would be:

w U  U 50 w U
w  or o  o  (for SAE 10 oil)
2h 2h h

So the addition of the much more viscous oil layer will only increase the shear stress of
the combined fluid layers by roughly a factor of two over what would exist with only
water—and much less than the 100-fold increase in shear stress that would occur from an
all water flow to an all oil flow. Clearly, the layer of a much less viscous fluid in
proximity to a much more viscous fluid strongly biases the wall shear stress toward the
shear stress of the less viscous fluid. In fact, our analysis indicates that the shear can
never reach more than twice the shear that would exist for the less viscous of the two
fluids. In situations where the two fluids can be easily separated, one can conceive of
using a less viscous fluid to facilitate the transport of a more viscous material at reduced
power costs. However, in practice, it would be hard to maintain a laminar environment,

140
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

and in a pipe flow one would have to be able to concentrate the more viscous fluid at the
center, with the less viscous fluid at the boundary.

6.3.5 Two-Fluid Poiseuille Flow in a Channel

Now, consider a pressure gradient driven Poiseuille-type flow of two immiscible fluids
(fluid a and fluid b) flowing adjacent to each other, constrained by stationary flat plates.
Here, we assume a channel width of h, and that fluid a is of thickness t within the
channel, and fluid b fills the remainder of the channel t < y < h, as shown.

h fluid b b , b
y
t fluid a a , a
x

Figure 6.9 A two-fluid, Poiseuille flow

We wish to determine the velocity behavior within the two fluids, and the resulting
volume flow rate. This type of flow is a bit more practical, since it can illustrate, for
example, the flow interaction between two separate, but adjacent, fluids, such as air and
water, or oil and water. This flow again illustrates the use of common boundary
conditions when considering two mutually adjacent fluids. However, it is particularly
more complicated than the Couette flow examined in section 6.3.4.

Utilizing the two-dimensional Navier-Stokes equations, we again must apply a separate

set of equations for each of the fluids. And again, the assumptions and simplifications
that apply for a single fluid Poiseuille flow also apply here, and are:

1. Steady flow
2. Fully-developed flow; no variations in x-direction
3.  = constant
4. Neglect body forces

Employing these assumptions, and noting that again the continuity equation yields v = 0
throughout both fluids, we reduce the governing equations to a pair of second order
ordinary differential equations:

d 2u a 1 dP
2
 (6.34a)
dy  a dx
and
d 2u b 1 dP
2
 (6.34b)
dy  b dx

141
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

In Eqs. 6.34, ua is the x-direction velocity within fluid a, and ub is the x-direction velocity
within fluid b. Note that Eqs. 6.34a and 6.34b each apply only within the region
occupied by the respective fluids, and the boundary conditions at the interface between
the two fluids (y = t) must match.

Here, the appropriate set of boundary conditions is:

1) u a  0 @ y  0 2) u b  0 @ y  h
du a du
3) u a  u b @ y  t 4) a  b @ y  t (i.e. a  b b )
dy dy

We again require four boundary conditions, since we again have two second-order
differential equations, each of which will yield two constants of integration.

Integrating Eqs. 6.34 twice yields two simple linear solutions:

du a 1 dP
 y  C1a
dy  a dx
1 dP y 2
ua   C1a y  C 2 a (6.35)
 a dx 2
and

du b 1 dP
 y  C1b
dy  b dx
1 dP y 2
ub   C1b y  C 2 b (6.36)
 b dx 2

Note that we will employ the boundary conditions to solve for the respective constants,
C1a , C2a , C1b , and C2b .

Applying boundary condition 1) @ y = 0 yields:

1 dP 0
2
ua  0   C1a 0  C2a  C2a  0 (6.37)
a dx 2

From boundary condition 2) @ y = h we have:

1 dP h 2
ub  0   C1b h  C 2 b (6.38)
 b dx 2

From boundary condition 3) @ y = t we get:

142
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

1 dP t 2 1 dP t 2
ua   C1a t  u b   C1b t  C2 b (6.39)
 a dx 2  b dx 2

And finally, applying boundary condition 4) @ y = t, yields:

du a dP du dP
a  t  a C1a   b b  t   bC1b
dy dx dy dx

 a C1a   b C1b  C1b  a C1a (6.40)
b

Now, combining Eq. 6.38 and 6.40, we get:

1 dP h 2 1 dP h 2  a
C2 b    C1b h    C1a h (6.41)
 b dx 2  b dx 2  b

Substituting Eqns. 6.40 and 6.41 into 6.39 gives:

1 dP t 2 1 dP t 2  a 1 dP h 2  a
 C1a t   C1a t   C1a h
 a dx 2  b dx 2  b  b dx 2  b

Solving for C1a gives:

    1 dP  t 2 h 2  b t 2 
C1a  t  a t  a h      
  b  b   b dx  2 2  a 2 
 t   a   a  h dP   b  t  2 
2
C1a h  1      1     1
h
   b   b 2 b dx  a  h  
h dP   b  t  
2

1     1
2 b dx   a  h  
C1a  (6.42)
t  a  a
1   
h   b   b

Substituting Eq. 6.42 back into Eqs. 6.40 and 6.41, we have:

h dP   a  t   a 
2

  1   
a 2 b dx   b  h   b 
C1b  C1a  (6.43)
b t  a  a
1   
h   b   b
and

143
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

h 2 dP   a  t   a 
2

  1
 h    
1 dP h 2 2 b dx 
 b   b 

C 2b   (6.44)
 b dx 2 t  a  a
1   
h   b   b

Finally, substituting Eqs. 6.37, 6.42, 6.43 and 6.44 into Eqs. 6.35 and 6.36, and
simplifying, we have:

hy dP   b  t  
2

1     1
1 dP y 2 2 b dx   a  h  
ua  
 a dx 2 t  a  a
1   
h   b   b

    b  t  2  
  1     1 
h 2 dP   b  y   y     a  h 
2

ua   (6.45)
      
2 b dx   a  h   h    a  t  a  

 1     
 
    b  h  b  
and
hy dP   a  t   a  h 2 dP   a  t   a 
2 2

  1      1   

1 dP y 2 2 b dx   b  h   b  1 dP h 2 2 b dx   b  h   b 
ub    
 b dx 2 t  a  a  b dx 2 t  a  a
1    1   
h  b  b h   b   b
   a  t   a  
2

    1 
 h     
h 2 dP  y 
2
  y  2     b   b 
ub     1     1  (6.46)
2 b dx  h  h    t  
     1  a   a  
  
    b  h  b  

To make the results non-dimensional, and a bit more appropriate for plotting, we divide
h 2 dP
Eqs. 6.45 and 6.46 through by , to give:
2 b dx

   a  t   a  
2

   
 1    
 b  y   y     b
2

ua  2
ua
  h   b   (6.47)
      
h dP  a  h   h    a  t  a  
 1    
2 b dx      h   
  b b

144
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

and
   a  t   a  
2

    1    
ub  y 
2
  y  2     b  h   b  
ub  2     1     1 (6.48)
h dP  h   
  h    1   a  t   a  
2 b dx    
    b  h  b  

To make these less cumbersome, we define:

2
 a  t  
  1   a

K ab   b  h   b
 a  t a
1   
 b  h b

Allowing us to write:

b y  2  y  
ua      K ab  (6.49)
a h   h  
 y 2   y  

u b     1     1 K ab (6.50)
 h    h  

a
Note that if  1 , then Kab  1 , and Eqs. 6.49 and 6.50 reduce to the same equation,
b
i.e.:
2
 y  y  y  y
ua  ub          1
h h h h

which is of the same form as Eq. 6.17, for a single fluid Poiseuille flow, as it should be.

Figure 6.10 shows two plots of the velocity. The first is for equal thickness fluid layers (t
 
= 0.5h), for varying ratios of a . The second figure is for a  50 (the ratio of
b b
t
viscosities of water to air), for varying ratios of .
h

Figure 6.10a shows that the effect of an increasing viscosity ratio is to not only retard the
more viscous layer (fluid a), but also to reduce the overall flow rate of the less viscous
fluid (fluid b) as well (reflected by the integral of the velocity profiles). Note that a ratio
of a b  50 is roughly that of water to air, which is probably the most common
situation for a two-fluid process. Using this typical water-air viscosity ratio, figure 6.10b
illustrates the effect of increasing the thickness of the more viscous fluid, which again

145
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

Velocity vs. y/h Velocity vs. y/h

0.25 0.25
Non-dimensional Fluid Velocity

Non-Dimensional Fluid Velocity

0.20 0.20

0.15 0.15
a/b=1 t/h=0
a/b=2 t/h=0.2
0.10 a/b=5 0.10 t/h=0.4
a/b=10 t/h=0.6
a/b=50 t/h=0.8
0.05 0.05

0.00 0.00
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
y/h y/h

(a) (b)

Figure 6.10 Non-dimensional velocity for a two-fluid Poiseuille flow between

parallel plates.
(a) t/h = 0.5, with viscosity ratios a b  a b  of 1, 2, 5, 10, and 50.
(b) a b  50 , with thickness ratios of t/h = 0, 0.2, 0.4, 0.6, and 0.8.

Shear stress vs. y/h

0.02
Non-Dimensional Shear Stress

0.01

0.00 t/h=0
0.0 0.2 0.4 0.6 0.8 1.0 t/h=0.2
-0.01
t/h=0.4
-0.02 t/h=0.6
t/h=0.8
-0.03

-0.04
y/h

(c)

Figure 6.10c Non-dimensional shear stress for a two-fluid Poiseuille flow between
parallel plates. a b  50 , with thickness ratios of t/h = 0, 0.2, 0.4,
0.6, and 0.8.

reduces the flow rate of the less viscous fluid, through reduction of the flow cross-
section. Reduction of the flow cross-section for fluid b reduces the motive force acting
on the fluid (since the pressure gradient acts over a reduced area), which means fluid b

146
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

must sustain reduced fluid deformation, resulting in lower shear with the bounding
surfaces, and consequently a reduced flow rate.

Figure 6.10c shows the shear stress distributions for the velocity profiles shown in figure
6.10b. While not intuitive, the shear within both fluids again varies linearly, but is biased
by the ratio of the fluid thicknesses.

6.3.6 Falling Liquid Film on a Wall

Consider the motion of a viscous liquid flowing down a sloped surface due to
gravitational effects, as shown in figure 6.11.

g
y Patm

x

h
, 

Figure 6.11 Schematic of a falling liquid film

For this problem, we align the coordinate system with the sloping surface, which is at a
fixed angle to the vertical of . We also assume that gravity acts vertically as shown,
and that the thickness of the liquid layer is h, which remains constant as the liquid flows
down the sloped surface. Here we assume governing two-dimensional N-S equations (x
and y directions, constant density) given by:
1 2 v=0. 2
u u u 1 dp   u  2u  2
u v  x    2  2  x-direction N-S (6.51a)
t x y  x  x y 
1 2 v=0. 2 v=0.
v v v 1 p  v  v 2 2
u  v  y    2  2  y-direction N-S (6.51b)
t x y  y  x y 
2
u v
 0 continuity (6.51c)
x y

We again simplify Eqs. 6.51, as shown by the red arrows, assuming:

(1) steady flow

(2) fully-developed flow

147
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

Using the reduced continuity equation, Eq. 6.51c, we can again show that v = 0 within the
liquid layer, which reduces the N-S equations to:

 2 u 1 p x
  (6.52a)
y2  x 
and
p
  y (6.52b)
y

Note that we have retained the total pressure in the N-S equations, since this is a gravity

driven flow. For this flow, that the body force vector,  , is given by:

   x ˆi   y ˆj  g cos  ˆi  g sin ˆj

So the y-direction equation, Eq. 6.52b, can be solved for the pressure as:

p
  g sin  p  Patm   g sin h  y f x  (6.53)
y

However, since the entire liquid layer is in contact with the atmosphere, this means that
the pressure at the surface of the liquid must be atmospheric (i.e. p = Patm = constant @ y
= h). Thus, p cannot be a function of x, which means that in Eq. 6.53 f(x) =0, and
p
consequently  0 . Correspondingly, Eq. 6.52a reduces to:
x

d2u g cos 
2
 (6.54)
dy 

The boundary conditions for this flow are interesting. Clearly, at the plate surface (y = 0)
the no slip condition requires that u = 0. However, at the interface between the liquid and
the atmosphere (y = h), we do not know the boundary velocity. However, we know that
the velocity and the shear stress for the liquid at the y = h interface must be equivalent to
that of the bounding atmosphere. However, since we assume the atmosphere to generally
be at rest, this could give us some irksome boundary conditions, and require that we solve
equations for the flow in the atmosphere as well. However, since we assume that the
liquid moves slowly, and that both the density and the viscosity of the air are much, much
less than the liquid, we ignore the boundary velocity of the atmosphere, and assume that
to a good approximation that there is essentially no shear stress between the liquid and
du
the atmosphere. Thus, we assume that at the interface,  yx  0 such that  0 at
dy
y = h, providing our second boundary condition (and of the second kind).

Now, integrating Eq. 6.54 once, gives:

148
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

du g cos  g cos  du
 y  C1  C1  h for  0 at y=h
dy   dy

Integrating once again gives:

g cos  y 2 g cos 
u  hy  C 2  C 2  0 for u=0 at y=0
 2 

Thus, the velocity profile for the liquid is:

g cos   y 2  2 Umax  y2    y   y 2 
u  hy     hy    Umax 2     
  2 h 2  2   h   h  
(6.55)

gh 2 cos 
where Umax  at y = h.
2
The corresponding shear stress within the liquid is:

du
 yx    g cos h  y  (6.56)
dy

And the volume flow rate (assuming again a depth into the page of W) is:

yh yh
g cos   y2  g cos  h 3 2
Q   u Wdy 
y '0
 y'0  hy  2  Wdy   3 W  3 Umax Wh  (6.57)

Q 2
Using the flow rate yields an average velocity of U   U max .
Wh 3

Note that the velocity, shear stress and flow rate are all a linear function of the slope
angle, as we might expect. Also notice that this is the same shape velocity profile as we
derived for the Poiseuille channel flow (i.e. only ½ of the Poiseuille flow, from surface to
centerline), only here the driving force is gravity, as opposed to the pressure gradient.
This similar profile shape is also reflected by the average velocity, which has the same
proportionality to the maximum velocity as we found for the Poiseuille channel flow of
example 6.3.2.

6.4 Simple One-Dimensional Solutions with Radial Symmetry

Many practical flows have geometries with radial symmetry. Pipes, tubes, disks, annular
regions are just some of these. Most of our water and air distribution systems use circular
cross section pipes or ducts, and most bearing surfaces of rotating and reciprocating

149
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

machinery depend on fluid lubrication via small annular regions. The following
examples illustrate some of the simple fully-developed laminar flows that occur in
geometries with radial symmetry.

6.4.1 Poiseuille Flow in a Pipe or Tube

In section 6.3.2 we examined a Poiseuille flow between parallel plates. In the present
example, we examine the same type of pressure gradient driven flow in a pipe or tube of
constant cross-section. Here we choose to use the Navier-Stokes equation in cylindrical
or radial coordinates, with the z-axis aligned with the centerline of the pipe, as shown in
figure 6.12.
r
R
z


Figure 6.12 Geometry for a Poiseuille flow in a circular cross section pipe.

We begin our analysis by assuming that there is no swirl, or flow changes in the  or
azimuthal direction, which eliminates all terms depending on  and the velocity
component, v  . Consequently, the appropriate equations are the z-direction and r-
direction components of the Navier-Stokes equations, and the continuity equation in
cylindrical coordinates, as follows:
1 5 =0 2 3 3 5

v z v v v v 1 P 1   v z  1  2 v z  2 v z 
 vz z  vr z   z     r  2  2 
z-N.S.: t z r r   z  r r  r  r 
2
z 
4 4 2 4 2 4 4 4 2
2
 2 2

r-N.S.: v r  v r v r  v v r  v z v r  v   1 P     1  rvr   12  v2r   v2r  22 v 
t r r  z r  r r r r   r 
 z r  
2 5
1
Continuity: rvr   1 v  vz  0
r r r  z

Here we again eliminate terms in the equations, based on the following set of
assumptions:

1. steady flow
2. no v motion
3. no changes in the -direction
4. constant density  reduced continuity equation
5. fully-developed flow  no z-direction changes

150
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

From the reduced continuity equation, we can write:

 C
(rvr )  0  rvr = constant = C  v r 
r r

However, since v r  0 at r = R, this implies that C = 0 and vr = 0 throughout the pipe.

Thus, the r-direction equation reduces to:

1 P
  0  P  P( z ) only
 r

Therefore, the above z-direction equation becomes:

1 dP  d  dv z  
  r 0 wher e,  
 dz r dr  dr  

d  dv z  r dP
or r  (6.58)
dr  dr   dz

Note: if we take z-derivative of Eq. 6.58, we obtain:

d  dP  d   d  dv z 
  r  0
dz  dz  dz  r dr  dr 

This term is not a function of z, thus, dP/dz must

therefore, the derivative is zero. be a constant

The appropriate boundary conditions for this flow are:

(1) vz  0 @ r  R

dv z
(2) 0 @ r0
dr

The first of these boundary conditions is obvious, since we require no slip at the pipe
boundary. The second boundary condition is of the second kind, and reflects the
expectation of symmetry for this flow about the centerline, which requires that the
velocity be either a maximum or minimum at the centerline. In either case, this is
expressed by a requirement that the first derivative of the streamwise velocity with
respect to the radius be zero at r = 0.
Integrating Eq. 6.58 gives:

151
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

dv z r 2 dP
r   C1
dr 2 dz
dv z r dP C1
 
dr 2 dz r

Applying boundary condition 2 gives:

0  0  C1  C1  0

Integrating again yields:

r 2 dP
vz   C2
4 dz

Now, applying boundary condition 1:

R 2 dP R 2 dP
0  C2  C2  
4 dz 4 dz

Thus, the velocity for Poiseuille flow in a tube or pipe is given by:

1 dP 2 2
vz  
4 dz
R r  (6.59)

Since viscosity will cause the pressure to decrease in the direction of the flow, the
negative sign in Eq. 6.59 indicates that the velocity will move in the positive z-direction
dP
when the pressure decreases in the positive z-direction (i.e.  0 ).
dx
dP
If  0 , then the direction of flow will be reversed (i.e. moving in the negative z-
dx
direction).

Now note that the maximum velocity (as we surmised) occurs at the center of the pipe.

1 dP 2 R 2 dP
Umax  vz r  0  
4 dz

R 0   
4 dz
(6.60)

Thus, we can rewrite Eq. 6.59 in terms of Umax using Eq. 6.60 as:

 r2 
v z  U max 1  2  (6.61)
 R 

152
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

Note that this flow distribution describes a parabola, with the maximum velocity at the
center of the pipe.

The volume flow rate for this pipe flow is obtained by integrating across the cross section
of the pipe. To do this requires that we integrate both radially and azimuthally over the
cross section using an appropriate differential element, dA, given by dA  rdrd . The
geometric source of this differential element is shown in figure 6.13.
r
rd  dA  (rd)(dr) = rdrd
d

dr

Figure 6.13 Differential area element in cylindrical coordinates.

Thus, the volume flow rate for the Poiseuille pipe flow is given by:

 2 r  R  2 r R
 r2   R2 
Q   v z dA    vz rdrd  Umax  d   R 2 
 1   rdr U max 2   
A 0 r 0 0 r 0   4 

Umax  R 4 dP
Q
2
2
R   8 dz
 (6.62)

Note that the average or mean velocity for this flow (the flow rate divided by the cross
1
sectional area) is U  U max , as compared to the Poiseuille channel flow, where
2
2
U  U max . This difference is due to the radial symmetry of the pipe, which biases
3
lower velocity fluid toward the region of the bounding walls, which comprises a larger
component of the cross-sectional area.

The corresponding shear stress for this flow is:

=0
 v v  r dP  2 rUmax
rz   z  r    (6.63)
 r z  2 dz R2

which is linear with the radius, r. Note however, that the shear stress at the pipe wall is:

R dP  2 U max
rz r  R   (6.64)
2 dz R

An interesting observation from Eq. 6.64 is that for a fixed pressure gradient the shear
stress at the wall will increase linearly with the pipe radius. One might think that an

153
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

increase in wall shear would reflect a corresponding decrease in the flow rate through the
pipe. However, as Eq. 6.62 shows, the flow rate will increase as the fourth power of the
radius. Clearly, as the pipe radius increases the effect of the pressure gradient, which is
applied over the cross sectional area of the pipe, will significantly outweigh the impact of
the increased wall shear which acts on only the periphery of the pipe. Eq. 6.60 further
illustrates this impact of increased pipe radius, indicating that the maximum velocity in
the pipe will increase as the square of the radius. Since the fluid nearer the center of a
larger pipe will be farther removed from the retarding viscous effects of the pipe
boundaries, it will be less restrained and reach higher values under the same forcing
condition (i.e. pressure gradient).

6.4.2 Couette Flow between Concentric Rotating Cylinders

In section 6.3.1 we examined a Couette flow between parallel plates. In this example, we
examine the same type of surface-driven flow in the annulus formed between the surface
of concentric cylinders rotating at different angular velocities. Here we use the
cylindrical Navier-Stokes equation with the z-axis aligned out of this page on the
centerline of the two cylinders, as show in figure 6.14.
o
i
r
 ro
ri

Figure 6.14 Geometry for Couette flow between concentric cylinders

We assume that there is no flow in the z-direction, which eliminates all terms depending
on z and the velocity component v z . Consequently, the appropriate set of equations is the
-direction and r-direction components of the Navier-Stokes equations, and the continuity
equation in cylindrical coordinates, as follows:

1 =0 4 2 =0 5 4 3 =0

-N.S.: v   v r v   v  v   v z v   v r v    1 p     1  rv   12  v2   v2  22 v r 

2 2

t r r  z r  r   r  r r  r  z r  

=0 =0 =0 =0 =0 =0 =0 4
2 2 2

r-N.S.: v r  v r v r  v  v r  v z v r  v   1 p     1  rvr   12  v2r   v2r  22 v  

t r r  z r  r r r r   r 
 z r  
4 3
1
Continuity: rvr   1 v  vz  0
r r r  z

154
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

We again eliminate terms in the equations, based on the following set of assumptions:

1. steady flow
2. no vz motion
3. no changes in the z-direction
4. fully-developed flow  no -direction changes
5. No pressure changes in -direction

Note that assumption 5 corresponds to the same argument we made previously in section
6.3.1, and applies in general for all surface-driven, Couette-type flows.

From the reduced continuity equation, we can again write:

 C
( rvr )  0  rvr = constant = C  v r 
r r

Again, since v r  0 at r = ro and ri, this implies that C = 0 and vr = 0 throughout the
annulus. Thus, the Navier-Stokes equations reduce to:

 1 
 rv   0 (-direction) (6.65a)
r  r r 

p  v2
 (r-direction) (6.65b)
r r
We note that v = f(r) only, which also makes the pressure only a function of r.

Integrating the simplified -direction equation, Eq. 6.65a, gives:

d
rv   C1r
dr
Cr C
v  1  2
2 r

The appropriate boundary conditions for this flow are:

(1) v  i ri @ r  ri

(2) v  o ro @ r  ro

This is the most general case, where we assume that each tube rotates at a different
angular (and thus tangential) velocity. Applying these boundary conditions gives us a
cumbersome set of equations for C1 and C2, which when solved give a velocity
relationship of:

155
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

v 
2
 2
 o ro r 2  ri   i ri ro  r 2  2
 2
 (6.66)

r ro  ri
2 2

Using the appropriate equation for  r from Section 5.8.1, the shear stress within the
fluid is given as:
=0
   v  1 v r  2 ri ro  o   i 
2 2

 r   r      (6.67)
 r  r  r  
2
r 2 ro  ri
2
 
The resulting torque per unit depth is given is given by:

2 2
ri ro
Torque=  r 2 r  r  4
r o
2
 ri
2
  o  i  (6.68)

So the torque is constant, and equal for each cylinder, as it should be.

Thus, for a set of cylinders of known geometry and rotation, one can calculate the
viscosity of the fluid by measuring the torque exerted on either cylinder. This process,
first suggested by Couette (1890), is still a popular and practical method used in
viscometry.

Note a couple of special cases for this type of flow. The first is for either a fixed inner
cylinder (i=0), or a fixed outer cylinder (o=0). Applying these constraints to Eq. 66,
gives:

v 
2
 o ro r 2  ri  2
 (inner fixed) (6.69a)

r ro  ri
2 2

or

v 
2
i ri ro  r 2 2
 (outer fixed) (6.69b)

2
r ro  ri
2

Eq. 6.69a (i = 0) is comparable to the initial parallel plate Couette solution we did in
section 6.3.1, where the outer plate moved while the inner plate was held fixed. Note that
if the cylinders are very close together, such that the annulus is small and the curvature
large, such that we can approximate that r  ri  ro, then we can simplify Eq. 6.69a as:
1 1

v 
2
o ro r 2  ri2
 o ro  o

r r  ri r  ri  o ro r  ri   ro   r  ri 
    (6.70)

2
r ro  ri
2
r ro  ri ro  ri  ro  ri   r   ro  ri 
If we let h = ro-ri, and designate a variable y = r-ri, then Eq.6.70 can be written as:

156
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

v 
 o ro r  ri   o ro y  Uy where U= o ro (6.71)
ro  ri  h h

Thus, when the curvature is quite large, the solution for the concentric cylinder Couette
flow asymptotes to the solution (Eq. 6.8) for a parallel plate Couette flow in Section
6.3.1, as it should (and Eq. 6.67 for shear stress, asymptotes to Eq. 6.9 for parallel flow).

Two other variations on this type of flow are ones where we essentially remove one or
the other of the cylinder surfaces. We do this by either expanding the diameter of the
outer cylinder to infinity (case 1) or shrinking the diameter of the inner cylinder to zero
(case 2).

6.4.2.1 A Cylinder Rotating in an Infinite Fluid

Here we let ro and o0 in Eq. 6.52, and note that ro  r 2  ro  ri2 , which  2
  2

yields:
=0 =1

v 
2
 2

 o ro r 2  ri   i ri ro  r 2
2
 2
    r  r
i i
2
o
2
 r2  r
i i
2
(6.72)

r ro  ri
2 2
  r  r
  o
2
 ri
2
 r
In Chapter 8, we will show that Eq. 6.72 corresponds to the behavior of an irrotational
vortex and discuss why this profile evolves. It is of interest to note that this is the only
example where a viscous flow with a bounding surface retains no vorticity within the
flow field.

6.4.2.2 Fluid Rotating within a Cylinder

Alternatively, if we let ri0 and i0 in Eq. 6.66, this yields a flow with no inner
cylinder:

=0 =0

v  o o
 2 2 2
 
2

 r r  ri  i ri ro  r 2  o ro  r 2 
2 2

 (6.73)
2
r ro  ri
2
   r  r 2  o r
  
  o
=0
Note that Eq.6.73 is simply a solid body rotation within the outer cylinder, which is to be
expected. Again, we will address this flow, and its temporal development, when we
discuss vorticity and vorticity transport, in Chapter 11.

Notice that we have not addressed the application of the simplified r-direction N-S
equation, which we derived above as Eq. 6.65b:

157
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

 p  v2

r r

Equation 6.65b basically prescribes the behavior of the pressure within to the fluid due to
v. This relationship indicates, as we will discuss in the following Chapter 7, that for
circular flows pressure will only change in the radial direction. Additionally, pressure
will always increase outward from the center of rotation, regardless of the velocity profile
p  v 2
(since  > 0 for all possible circular velocity fields). For example, for Case 1
r r
above, for a cylinder rotating in an infinite fluid (ro and o0), we have:
2
dp  v 2   i ri 
2 2 4
i ri
   
dr r r  r  r3

Integrating from r = ri outward gives:

p r 2 4
dr  r 1 1
 dp  p  pi  i ri
2 4
rr r 3  2i i  2  2 (6.74)
r 
p  pi i  i r 

where pi is the pressure at the surface of the inner cylinder.

As Eq. 6.74 shows, pressure continually increases outward from the cylinder, reaching a
2 2
 i ri
maximum of p  p i  as r  . A similar result of an outward pressure increase
2
is found for Case 2 as well.

6.4.3 Poiseuille Flow in a Duct of Annular Cross Section

This is a pressure driven flow in the annular cross section formed between two concentric
cylinders. The pressure gradient acts along the axis of the cylinders, driving the flow as
shown in figure 6.15 below. Such a flow is characteristic of the flow in many heat
exchangers, where one fluid flows through a contained cylindrical annulus directly
adjacent to a second fluid flowing in a central tube or cylinder. This is again a pressure
gradient driven Poiseuille-type flow, and we again apply the Navier-Stokes equation in
cylindrical or radial coordinates, with the z-axis aligned with the centerline of the
concentric cylinders, as show in figure 6.15.

158
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

ro r

ri z


Figure 6.15 Geometry for a pressure-driven Poiseuille flow in an annular cross

section

We again assume that there is no swirl, or flow changes in the  or azimuthal direction,
which eliminates all terms depending on  and the velocity component v  . The
appropriate set of equations is the z-direction N-S equation, and the continuity equation
in cylindrical coordinates:
1 4 =0 2 3 3 4
v z v v v v 1 P 1   v z  1  v z  v z  2 2
 vz z  vr z   z     r  2  2 
z-N.S.: t z r r   z  r r  r  r 
2
z 
2 4
1
Continuity: rvr   1 v  vz  0
r r r  z

We eliminate terms in the equations based on the assumptions:

1. steady flow
2. no v motion
3. no changes in the  -direction
4. fully-developed flow  no z-direction changes

The reduced continuity equation gives vr = 0 throughout the pipe. Thus, the r-direction
equation reduces to:

1 P
  0  P  P( z ) only
 r

Therefore, the above z-direction equation reduces to the same equation, Eq. 6.58, as the
Poiseuille flow in a pipe, as shown in section 6.4.1:

1 dP  d  dv z  
  r 0 wher e,  
 dz r dr  dr  

d  dv z  r dP
or r  (6.75)
dr  dr   dz

Just as with the pipe flow, we can again show that dP/dz = constant within the annulus.

159
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

The appropriate boundary conditions for this annular flow are no slip at the two annular
boundaries:

(1) vz  0 @ r  ro

(2) v z  0 @ r  ri

Integrating Eq. 6.75 once gives:

dv z r 2 dP
r   C1
dr 2 dz
dv z r dP C1
 
dr 2 dz r
Integrating again yields:

r 2 dP
vz   C1 ln r  C2 (6.76)
4 dz

Now, applying boundary conditions 1 and 2 to Eq. 6.76 gives:

2
ro dP
0  C1 ln ro  C 2 ,and
4 dz
2
ri dP
0  C1 ln ri  C 2
4 dz
Solving these simultaneously for C1 and C2 gives:

C1  
ro
2
 ri dP
2

 r  dz
4 ln o 
 ri 

C2  
2
ri dP r  ri dP
 o
 2 2

4 dz  r  dz
4 ln o 
 ri 

Substituting C1 and C2 into Eq. 6.76, the velocity for Poiseuille flow in an annulus is
given by:
  r 
 ln  
vz 
1 dP  2
4 dz 
2
 2
r  ri  ro  ri
2

 r 
 
 ri   
 ln o  
  ri  
or

160
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

  r 
 2 ln  
r dP  r
2
 r 2
 r
vz  i  2  1   o2  1  i   (6.77)

4 dz  ri   

  ri  ln ro  
r 
  i  

Non-dimensional velocity profiles for four radius ratios from 1.25 to 2.00 are shown in
figure 6.16.

0.6
ro/ri=1.25

0.5 ro/ri=1.5
ro/ri=1.75
0.4 ro/ri=2
Velocity

0.3

0.2

0.1

0
1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2
r/ri

Figure 6.16 Non-dimensionalized velocity vs. r/ri for ro/ri values from 1.25 to 2.
2
r dP
Here Velocity= v z i
4 dz

For this flow, the maximum velocity does not occur at the center of the annulus, but is a
function of the radius ratio, ro/ri, given by:

r@ U max 
r
o
2
 ri
2

r 
2 ln o 
 ri 

However, for values of ro/ri less than 3 the maximum velocity occurs very near to the
annulus midpoint [i.e. (ro+ri)/2], so radial skewing of the profiles is minimal, as one can
observe in figure 6.16.

The corresponding shear stress for this flow is:

161
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

 =0    ro 2  
 2  1 
 v v  1 dP 
rz   z  r  
 2 2 
 
r  ri  ri dP   r 
2r  o  2
 r
    i

  (6.78)
r z 4 dz     4 dz  r 
 
r ln 
ro
   i   r  ln ro  
  
  r  r   r 
  i     i   i  

Non-dimensional shear stress profiles for four radius ratios from 1.25 to 2.00 are shown
in figure 6.17.

r/ri
1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2
2.5
2
1.5
1
Shear Stress

0.5
0
-0.5
r/ri=1.25
-1
r/ri=1.5
-1.5
r/ri=1.75
-2
r/ri=2
-2.5

Figure 6.17 Non-dimensionalized shear stress vs. r/ri for ro/ri values from 1.25 to 2.
r dP
Here Shear Stress = rz i
4 dz

As figure 6.17 shows, the shear stress varies almost linearly from ri to ro, quite similar to the
behavior for Poiseuille flow in a parallel channel, that we examined in section 6.3.2. In fact,
one can show that both the velocity and shear stress equations asymptote to the parallel plate
solution when ro/ri1. However, notice that as the radius ratio increases, the shear stress at
ro will be correspondingly lower than the shear stress at ri [i.e.  r ri   r ro for ro ri  ].

Similar to Poiseuille flow in a pipe, the wall shear stress will increase with increasing
annulus gap width. However, as can be seen by the velocity profiles, the maximum
velocity and flow rate will also increase with an increasing annulus gap, since the driving
force of the pressure gradient (applied over a larger cross section area) greatly outstrips
the increase in shear stress.

6.4.4 Couette Flow between Porous Rotating Concentric Cylinders

In section 6.4.2 we examined a Couette flow between rotating cylinders with solid
surfaces. Here, we examine the impact of having a cross-stream flow between the

162
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

cylinders by having a constant injection/suction across the bounding surfaces. For the
purposes of this example, we assume a constant injection at the inner cylinder (vr = V0 @
r = ri), which is rotating at a constant angular velocity , as shown. We again use the
Navier-Stokes equation in cylindrical coordinates with the z-axis aligned out of this page
on the centerline of the two cylinders, as show in figure 6.18.


r
V0
 ro
ri

Figure 6.18 Geometry for Couette flow between concentric, porous cylinders with
transverse flow in the radial direction (vr = V0 @ r = ri).

Also, we again assume no changes in the z-direction, which eliminates all terms
depending on z and the velocity component v z , leaving the continuity equation, and the
-direction and r-direction component of the Navier-Stokes equations in cylindrical
coordinates, as the governing equations:

1 =0 4 2 5 4 3 =0
v  v v v v vv   1  2 2

 vr      vz   r   
1 p
   rv   12  v2   v2  22 v r  (6.79)
t r r  z r r    r  r r  r  z r  

=0 =0 =0 =0 =0 4
v r v v v v v 1 p   1 
2 2 2

 vr r   r  vz r      
rvr   12  v2r   v2r  22 v  (6.80)
t r r  z r  r  r  r r  r  z r  
4 3
1
rvr   1 v  v z  0 (6.81)
r r r  z

We eliminate terms in the equations, based on the following set of assumptions:

1. steady flow
2. no vz motion
3. no changes in the z-direction
4. fully-developed flow  no -direction changes
5. No pressure changes in -direction
6. vr = V0 at r = ri

Assumption 5 corresponds to the same argument we made previously in sections 6.3.1

and 6.4.2.

From the reduced continuity equation, we can write:

163
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

d C
( rv r )  0  rv r = constant = C  v r 
dr r
V0 ri
However, since v r  V0 at r = ri, this implies that C = V0ri and v r  across the
r
annulus. Thus, the Navier-Stokes equations reduce to:

v  v r v   1 
vr    rv  (-direction) (6.82a)
r r r  r r 

v r v2 1 dp  1 
vr     rvr  (r-direction) (6.82b)
r r  dr r  r r 
V0 ri 
Here, we note that v r  , and that rvr    V0 ri   0 , so we can rewrite the
r r r
equations as:

V0 ri v  V0 ri v   1 
 2   rv  (-direction) (6.83a)
r r r r  r r 

V0 ri  V0 ri  v2 1 dp
 2    (r-direction) (6.83b)
r  r  r  dr

Note that v = f(r) only, which makes the pressure only a function of r. This also lets us
write the equations as ordinary differential equations. Rearranging Eqs. 6.83 gives us:

dv     2 d 1 d
r  v    r  rv  (-direction) (6.84a)
dr  V0 ri  dr  r dr 

dp V02 ri2 v 2
  (r-direction) (6.84b)
dr r3 r
Here the pressure is only a function of r, and depends on knowing v, but is decoupled
from the differential equation for v, Eq. 6.84a. So our first order of business is to
determine v, from which we can determine p(r) from Eq. 6.84b, if we so desire.

The Eq. 6.84a for v looks a bit daunting and non-linear, but a solution can be found by
some judicious substitution. To begin, we note that we can write:

dv  drv     2 d 1 d 
r
dr
 v 
dr
 
V r
r
dr  r dr rv 
 0i  

Now, we let  = rv, and define an injection parameter

164
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

V0 ri
k so the equation becomes:


d r 2 d  1 d 
  
dr k dr  r dr 

Dividing through by r gives:

1 d r d  1 d 
  
r dr k dr  r dr 

1 d
Finally, we let   , which gives us a final equation form of:
r dr
r d

k dr

Separating variables gives:

d dr
k
 r

which integrates to give:


ln  kln r  ln C  ln r k  ln C k  ln r k  ln C1  ln C1r k 
or
1 d
  C1 r k 
r dr

Separating variables again, and integrating for  yields:

d  C1 r k1dr
r k 2
rv    C1  C2
k2
r k 1 C 2
v   C1 
k2 r

(note: k = -2 is a special case, for which the solution is: rv     C1 ln(r )  C 2 )

The appropriate boundary conditions for the present flow are:

(1) v    ri @ r  ri

(2) v  0 @ r  ro

165
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

Applying these boundary conditions gives us a set of equations for C1 and C2, which
when solved give a velocity relationship of:

r 
ln o 
v   ri
2 r o
k2
r k 2
 (note: k = -2 yields v   ri
2  ri  ) (6.85)
r r 
k 2 k2
 ri r 
o
r ln o 
r

r
Letting r  , we can write Eq. 6.85 in non-dimensional form as:
ri

v 
r*  r
k 2 k 2
 v  
ln ro*
ri
 o k 2
r  ro*  1  (k = -2 yields
ri

 r *

r * ln o* 
) (6.86)

r 

Note that the determining the shear stress within the fluid is a bit cumbersome, and is
given by:
=0
   v  1 v r 
r   r      
 ri 2ro  kr k  2
2
 k2
 (6.87)
 r  r  r  
 k2
r 2 ro  ri
k2
 
r
Or non-dimensionally (using r  again),
ri

 r 
2ro*  kr 
k 2 k2


 2 k2
r  ro*  1   (6.88)

V0 ri
A comparison of velocity and the shear stress for injection values of k  = 0, 2, 4,

and -4 (a suction through the inner surface) is shown in figure 6.19.

Figure 6.19a shows that injection through the inner cylinder increases the velocity values
at non-bounding radii within the annulus, whereas suction through the inner cylinder
decreases the respective velocity values. This change in velocity profile will impact the
volume flow rate within the annulus accordingly.

166
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

1 0
0.9 k=0
0.8 k=2 -1
0.7 k=4

r/
k=-4
v/ri

0.6 -2
0.5
k=0
0.4 -3
k=2
0.3
k=4
0.2 -4 k=-4
0.1
0 -5
1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2 1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2
r/ri r/ri
(a) Velocity (b) Shear stress

Figure 6.19 Non-dimensional velocity and shear stress vs. non-dimensional radius
for a concentric annulus with cross-stream injection (Vo) through the
inner porous boundary. The outer cylinder is fixed and inner cylinder
rotates at , with ro/ri = 2. Behavior is shown for injection parameters k
= Vori/ = 0, 2, 4, and -4 (this latter value is a suction at inner surface).

The volume flow rate (per unit depth) within the annulus can be calculated as:

ro

Q   v  dr  ri
ro
r k 2
 r k 2  dr  r  ro
k 2
r  1 
 rr
2 2
o
ln o  
ri ri o
k2
 ri
k 2
 i
 r o
k 2
 ri
k 2
  ri

 k  2

Or non-dimensionally as:

Q  ro*
k 2
1 
 
 
*
  ln r   (6.89)
ri
2 k2
 ro*  ri*
k 2 o
k  2
Figure 6.20 illustrates the impact of both injection/suction and the radii ratio ro* = ro/ri on
the induced flow rate within the annulus. To better illustrate changes, the flow rates with
suction/injection are non-dimensionalized on the flow rate for radius ratios with a non-
porous surface (k = 0), Q/Q k=0. Figure 6.20 shows that fluid injection through the inner
cylinder increases the induced flow rate, up to 27% for k = 4, ro/ri = 2. However, surface
suction on the rotating inner cylinder has a more pronounced effect, decreasing the
induced flow rate by up to 37% for k = - 4, ro/ri = 2. The effect of injection/suction will
have differing effects, depending on which surface is rotated, the radii ratio, and the
magnitude of the suction/injection, but clearly the impact of such suction, even relatively
small values, can have a substantial effect. For example, if we base the Reynolds number
for the fluid within the annulus on the inner cylinder surface velocity (vi = ri), the
cylinder radius (ri), and the kinematic viscosity (), the Reynolds number for this flow
would be ri2/, and we can show that V0/vi = k/(ri2/). Thus, for a small Reynolds

167
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

number of 100, an injection parameter of k = 4 would give V0/v = 0.04. So, a 4%

injection rate results in a 27% increase in induced flow rate (and a 4% suction rate results
in a 37% reduction in induced flow rate). Clearly, small cross-stream flows through
porous surfaces can have very significant effects on overall flow behavior.

140%

120%

100%
Q/Qk=0

80%

60%
k=0
40% k=2
k=4
20% k=-4

0%
1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2

ro/ri

Q k 0
Figure 6.20 Flow Rate Ratio  vs. radius ratios for a concentric annulus
Q k 0
with cross-stream injection (V0) through the inner porous cylinder, and
a mass balanced suction through the outer porous cylinder. The outer
cylinder is fixed and the inner cylinder rotates at . Behavior is shown
for radii ratios 1 < ro/ri < 2, and injection parameters k = V0ri/ = 0, 2,
4, and -4 (this latter value is a suction at the inner surface).

To further illustrate the global impact of a small amount of surface injection/suction, note
the effect of the injection/suction parameter k on the shear stress values shown in figure
6.19b. Note that the corresponding shear values become more uniform with injection
(k > 0), and display larger variations with suction (k < 0), which suggests that suction
through the inner cylinder should result in increased torque on the inner cylinder.

To examine this, we calculate the general torque (per unit depth) at a radius r, as:

Torque r  r 2 2r o

k2
 kr k  2 (2r)r  2r 2r
2 o
k2
 kr k  2 
r r  r 
i 2 k2 k2 i k2 k2
o  ri o  ri

Particularizing the torque to the respective cylinder surfaces, the resulting torque on the
inner and outer cylinders per unit depth is given by:

168
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

Torque  2r 2 2r o

k2
 kri
k2
 (6.90a)
r  ri i
ro
k2
 ri
k2

2 2  k  ro k  2
and Torque  2r (6.90b)
r  ro i
ro
k 2
 ri
k2

Comparing the ratio of the torques in Eqs. 6.90, we have:
k2 k2
r  1
2  k i  2  k * 
Torque r  ri 2ro
k2
 kri
k2
 ro   ro 
Torque Ratio     (6.91)
Torque r  ro
2  k ro k 2 2  k  2  k 
Equation 6.91 indicates that the torque varies significantly with the injection/suction
parameter k. If k = 0, the torques are equivalent, as we demonstrated for non-porous
concentric cylinders in section 6.4.2. However, if k > 0 [injection out of the inner
cylinder], the torque will be less on the inner cylinder, whereas if k < 0 [suction into the
inner cylinder], the torque will be higher on the inner cylinder. This result is again a
consequence of the momentum change that the injected fluid must undergo as it passes
across the gap between the cylinders, as we discussed in section 6.3.3.

7
k=0
6 k=2
k=4
5 k=-4
Torque ratio, Ti/To

0
1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2
ro/ri

Torque r  r
Figure 6.21 Torque Ratio  i
vs. radii ratios for a concentric annulus with cross-
Torque r  r
o

stream injection (V0) from the inner porous cylinder, and a mass-balanced suction
through the outer porous cylinder. The outer cylinder is fixed and the inner
cylinder rotates at angular velocity, . Behavior is shown for radii ratios 1 < ro/ri
< 2, and injection parameters k = V0ri/ = 0, 2, 4, and - 4 (this latter value is a
suction at inner surface).

169
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

Figure 6.21 shows the variation in the torque ratio of the inner to the outer cylinder (Eq.
6.91) as a function of both the ratio of the cylinder radii and the injection/suction
parameter. Note that injection reduces the torque ratio, due to a lower shear on the inner
cylinder, and higher shear on the outer (figure 6.19b indicates that the shear stress, and
thus the actual torque on the inner cylinder, is lower with injection). However, figure 6.21
shows that suction through the inner cylinder results in a much more pronounced increase
in the torque ratio (seven fold for ro/ri = 2 and k = - 4). The absolute torque on the inner
cylinder also increases by 75% over that for a solid cylinder. Note that these variations
are independent of how fast the cylinder is rotated (as long as the flow in the annulus is
laminar). This is a striking result, and quite illustrative of the impact of the transfer of
fluid momentum across the annulus.

Now, consider the pressure behavior between the cylinders. The r-direction differential
equation derived above, Eq. 6.84b, gives:

dp V02 ri2 v 2
 
dr r3 r

Substituting in for v from Eq. 6.85, we have:

4
dp V02 ri2  2 ri ro  r k 2
 
 k 2

2

dr r3 r3 r
k 2
r
k 2 2
o i 
We separate variables and integrate from ri outward, where p = pi at r = ri:

p r  4 k 2
 V02 ri2 2 ri ro  r k 2  dr
2

 dp  rr  r 3  r 3 r k2  r k2 2

  
ppi i  o i 

And after some messy integration, we get:

  2 k 4 4ro k 2 ri k 2 ri 2 k 4   2 k 4 4ro k 2 r k 2 r 2 k 4  
  ro     ro   
V02 ri2  1 1   2 ri   k k  1   k k  1  
4
p  pi       (6.92)
2  ri 2 r 2  2  2 k 2
ri ro  ri
k 2 2
   k 2
r 2 ro  ri 
k 2 2

 
 
p  pi p  pi
Defining a non-dimensional pressure coefficient as Cp  2
 1 2 2
, and letting
2 Vi 2 i ri
1

r
r   again, we can rewrite Eq. 6.91 as:
ri

170
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

 2k 4 
1   * 2 k  4 4ro* r 
k 2
4r  k2 k2
r 
  ro* 2 k  4  o  r   
2   k k  1   o
k k  1  
V   1  
C p   0  1  2       (6.93)
   

 ri   r  
 k  2 2 2 k  2 2
ro*  1 r  ro*  1 
 
 

V0
In most cases, we would expect to be very small, and thus contribute little to the
ri
pressure change. The bulk of the change is a result of the large, bracketed terms on the
right of Eq. 6.93, which reflects pressure due to variations in angular momentum of the
fluid. The behavior of the pressure coefficient from Eq. 6.93 is shown in figure 6.22.

Note that for all k values shown (and in general), the pressure increases outward, as it
should, since dp/dr > 0 for all k values. However, as the figure shows, for k > 0 [injection
from the inner cylinder] the pressure increases more rapidly outward; conversely, for k <
0 [suction into the inner cylinder] the pressure increases
less rapidly, reaching a lower value at r = ro.

One can understand this behavior from the velocity profiles of figure 6.19a. As the fluid
particles injected through the inner surface move toward the outer surface, they retain
more of their initial -direction momentum at the same radial location than would occur
through just viscous diffusion. Since the angular velocity remains higher as it nears the
outer fixed cylinder (where the fluid must exit with zero velocity), the angular
momentum of the fluid is larger, which is reflected as an increased radial pressure
difference.

The opposite occurs if there is suction at the inner surface. In that case, the acceleration
of the fluid to the inner rotation velocity as it approaches the inner cylinder occurs more
slowly, thus retarding the angular momentum, which in turn reduces the radial pressure
changes.

171
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

0.6
k=0
0.5 k=2
k=4
0.4 k=-4
Cp

0.3

0.2

0.1

0
1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2
r/ri
p  pi
Figure 6.21 Non-dimensional pressure coefficient C p  vs. non-dimensional
1
2
 i2 ri2
radius for a concentric annulus of ro/ri=2, with cross-stream injection (V0)
from the inner porous cylinder, and a mass balanced suction out of the
outer porous cylinder. The outer cylinder is fixed and the inner cylinder
rotates at , with Re = ri2/ = 100. Behavior is shown for injection
parameters k = Vori/ = 0, 2, 4, and - 4 (this latter value is a suction at the
inner surface).

6.5 Summary

Sections 6.3 and 6.4 have presented several simple analytic, closed-form solutions to the
Navier-Stokes equations. These are all one-dimensional, since they depend on only one
space variable. Most of these examples reduce the Navier-Stokes to a force balance
between shear stress and pressure. However, as illustrated by the examples in Sections
6.3.3 and 6.4.4, the introduction of a simple advection term due to a steady cross-stream
velocity by balanced injection/suction can have a significant impact on the flow behavior,
and the subsequent forces, such as wall shear and flow rates.

The inclusion of advection and time dependent acceleration term in the Navier-Stokes
equation makes the solution of the resulting equations much more complicated, requiring
more advanced mathematics and numerical solutions. In Chapter 12, we will examine
more complicated, two-dimensional solutions of the Navier-Stokes equations, for which
the flow is dependent on either one spatial dimension and time, or on two spatial
dimensions. In either case, the number of possible closed-form solutions is limited, and
the mathematics is often quite complicated.

172
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

Problems

1. A fluid flows in the x-direction between two parallel plates, a distance h apart in the y-direction.
dp
The upper surface moves with a speed U and a pressure gradient, , is imposed on the flow.
dx
dp h 2
Define P   , and:
dx 2U

(a) Determine the velocity profile, u(y), and the shear stress, (y) , for the flow.

(b) Determine the value of P which causes the shear stress on the lower plate to be zero.
(c) Calculate the volume flow rate, and Determine the value of P which causes the
volume flow rate to be zero.
dp
(d) For air at 20ºC, U = 20 cm/s, h = 1cm, and  0.3 N / m 3 , calculate the value of
dx
the shear stress on the lower plate.

(e) If the shear stress for this flow is  yh   w  0 at y=h, does this mean τw acts in the negative
x-direction on the upper plate? Explain your answer.

(f) Calculate the volume flow rate (in cm3/s per meter depth) for part (d).

2. A vertical plate of infinite extent has a film of water flowing downward on the outside due to
gravity.
y
a) Determine the velocity profile in terms of gravity (g), the kinematic
x
viscosity (), the volume flow rate per unit width (Q), and the distance g
from the surface (y).


b) If =10-5 ft2/sec. for water at room temperature, determine

the thickness of the water layer and the maximum velocity within
the water layer if the flow rate per unit width is Q=6.6x10-4 ft2/sec.

3. A thin Newtonian film of uniform thickness  is formed on the external surface of a vertical,
infinitely long, stationary rod. Assume that the flow is steady, the surface tension is zero, and
the ambient air is stationary. If gravity is g = constant, the z-coordinate is along the rod axis in
opposition to gravity, the rod diameter is R, and kinematic viscosity is :
z
(a) Calculate the velocity within the film, vz(r),
in terms of r, g, R, , and . r
g

(b) Determine the shear stress on the cylinder surface, rz,  

in terms of , g, R, and . 2R

173
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

4. A thin Newtonian film of uniform thickness  is formed on the external surface of a vertical,
infinitely-long rod. The rod moves upward (against gravity) at a velocity V. Assume that the
flow is steady, the surface tension is zero, and the ambient air is stationary. If gravity is g =
constant, the z-coordinate is along the rod axis in the opposite direction to gravity, the rod
diameter is 2R, and kinematic viscosity is : z

(a) Calculate the velocity within the film, vz(r), V r

g
in terms of V, r, g, R, , and .

(b) Determine the shear stress on the cylinder surface, rz, 2R 
in terms of , g, R, and .

(c) If R>>, determine what value of rod velocity V will

make vz = 0 at r = R+ in terms of g, R, , and 
(i.e. the outer fluid surface will be stationary relative to the z-coordinate).

5. Consider the vertically oriented moving belt shown below. The belt moves upward at velocity V
out of a vat of liquid of density  and viscosity , as shown in the figure. Above a certain belt
velocity, the shear stress acting on the fluid will cause the liquid to rise up the belt due to viscous
effects, and to be deposited in the upper reservoir of fluid, as shown. It is assumed that the
moving belt is sealed at the bottom of the two reservoirs, such that fluid will not leak out. You
are to perform an analysis of the flow to determine the velocity profile and flow rate of liquid
moving up the surface of the belt.
Patm
Assume that the flow is steady, the thickness of
the liquid is T, atmospheric pressure acts on all moving
belt upper reservoir
surfaces of the liquid, and g acts as shown.
Consider only the portion of the belt where the g fluid
flow might be considered fully developed. List layer V
the remainder of your assumptions very
carefully, and think hard about the physics of Patm T
the flow behavior since this will impact the
conditions you must place on the solution. , 
lower reservoir
Specifically: y

a. Starting with basic equation(s), develop a simplified differential equation to allow

the determination of v(x); clearly state simplifying assumptions and the boundary conditions
for the problem.

b. Using the result of part a, establish a general solution for the velocity profile, v(x), in terms
of x and constants shown in the figure.

174
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

c. Determine the flowrate per unit depth, Q, of the liquid from the lower reservoir to the upper
reservoir.

d. Determine (in terms of g, T,  and ) the minimum velocity necessary to assure that liquid
will flow upward.

e. Is there a limiting height to which the fluid can be pumped, and is this a function of the
velocity of the belt? If so, determine the maximum pumping height as a function of velocity,
V.

6. Consider the flow of a fluid of viscosity, , and density, , in the long, rectangular cavity shown
below, where L >> h. The lower surface and the ends of the cavity are flat, solid surfaces. The
upper surface is a flat belt that moves over the cavity from left-to-right at a constant speed U.
Using the coordinates shown, determine the steady state velocity profile in the central region of
the cavity, where the flow can be considered to be parallel and fully developed such that the
velocity depends on y alone. Note that a constant pressure gradient, dP/dx, exists in the central
region of the cavity.
Assume that the seal between the belt and the cavity corners is perfect, such that no fluid leaks
from or into the cavity. List the remainder of your assumptions very carefully, and think hard
about the physics of the flow behavior, since this will affect the conditions you must place on the
solution. moving belt
U (velocity of belt)

h y cavity flow
Specifically:

L
x

a. Starting with the 2-D continuity and x-direction Navier-Stokes equations, develop a
simplified differential equation which will allow the determination of u(y); clearly state
the boundary conditions for the problem.

b. Using the result of part a, establish a general solution for the velocity profile, u(y), in
terms of defined and undefined constants and y.

c. Determine the value(s) of the undefined constant(s) through the application of the
boundary conditions to yield a specific solution in terms of U, h, , dP/dx and y.

d. Since dP/dx is an unknown constant, determine a way to establish the value of the
pressure gradient in this central region of the cavity in terms of µ, U, and h only.
[Hint: consider the net flowrate, Q, at the center of the cavity]

e. Using the result of part d, eliminate dp/dx from your solution for u(y) such that u(y) is
expressed as a function of U, h, and y only. Plot the shape of the resulting velocity
profile, y/h vs. u/U.

175
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

f. Establish the y-location and the magnitude of the maximum and minimum velocity
within the cavity.

7. Consider an infinite circular pipe of radius R, which is translating vertically upward at velocity V
out of a vat of liquid of density  and viscosity , as shown in the figure. Above a certain pipe
translation velocity, the shear stress acting on the fluid at the pipe inner wall will cause the liquid
to rise up inside the pipe, much like a “viscous” pump (note that the liquid will also rise up on
the outside of the pipe, but you are to neglect those effects). You are to perform an analysis of
the flow to determine the velocity profile and flow rate of liquid inside the pipe, from 0 < r < R.

Using the coordinates shown, assume that the Patm

pipe is of infinite extent, the flow is steady, g
atmospheric pressure acts on all surfaces of the V V
liquid, and g acts as shown. List the remainder R
of your assumptions very carefully, and think
Liquid Surface
hard about the physics of the flow behavior z
since this will impact the conditions you must Patm
place on the solution. r
,  Infinite Pipe

Specifically:

a. Starting with basic equations, develop a simplified differential equation which will allow the
determination of vz(r); clearly state the boundary conditions for the problem.

b. Using the result of part a, establish a general solution for the velocity profile, vz(r), in terms of r and
constants shown in the figure.

c. Determine the volume flowrate of the liquid up the pipe, Q.

d. Determine [in terms of g, R and  (=/) ] the minimum pipe translation velocity necessary to assure
that liquid will flow upward.

e. Is there a limiting height to which the fluid can rise in the pipe, and is this a function of the velocity of
the pipe? If so, determine the height as a function of velocity, V.

176
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

8. Consider the viscous pumping of a fluid of viscosity  due to the axial motion of a circular tube
encompassing an annulus formed between an inner rod of radius ri and a surrounding tube of ro
(assume the rod and tube are of infinite length). Determine the velocity profile, vz vs r/ro, within
the annulus where the outer tube is translated in the axial direction at a velocity Vo. On one
graph, plot curves of vz/Vo vs. r/ro for ri/ro = 0.9, 0.7, 0.5, 0.3, and 0.1. Also, determine the non-
Q r
dimensional flow rate within the tube, 2
, as a function of r'  i , and on a second graph
 ro Vo ro
Q
plot vs . r' for 0  r'  1 .
 ro2 Vo
What will be the force per unit length (F/L) required to keep the rod moving at a constant speed?

Vo
r
ro
ri z


Vo

9. Consider the viscous pumping of a fluid of viscosity  due to the axial motion of a rod within an
annulus formed between the rod of radius ri and a surrounding tube of ro (assume the rod and
tube are of infinite length). Determine the velocity profile, vz vs r, within the annulus where the
inner rod is translated in the axial direction at a velocity Vi. On one graph, plot curves of vz/Vi
vs. r/ro for ri/ro = 0.9, 0.7, 0.5, 0.3, and 0.1. Also, determine the non-dimensional flow rate
Q r Q
within the tube, , as a function of ri '  i , and one a second graph plot vs . ri for
2
 ro Vi ro  ro2 Vi
0  ri  1 . Determine (mathematically, not graphically) the value of ri for which this viscous
pump will yield the highest flow rate for a fixed value of Vi. What will be the force per unit
length (F/L) required to keep the rod moving at a constant speed?

ro r

ri z
Vi 

177
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

10. An upper flat plate is translated at a velocity U parallel to a lower fixed flat plate, with a spacing
of h between the plates, creating a Couette flow. Originally the fluid between the plates is oil,
which results in a shear stress within the oil, and on the bounding plates, of  = oU/h. Now, a
layer of water of thickness t is introduced at the lower surface (w > o), as shown below. The
viscosity of the water is much less than the oil, such that o = 51w. Again assuming a steady
Couette type flow between the plates, what thickness of water, t, in terms of h, will reduce the
shear stress on the translating flat plate to ½ of the original shear stress with oil alone?

U
µo o t
y h
µw w
x

11. Consider the flow of a viscous fluid (viscosity = µ, density = ) similar to a Couette flow
between two infinite, parallel, porous plates; fluid of the same properties is uniformly injected
through the fixed, lower plate, and fluid is uniformly suctioned out through an upper plate, which
moves at a velocity U. If u = 0, v = Vo = const. at y = 0, and u = U = const., and v = Vo at y = h,
with no flow in the z-direction, derive the velocity profile, u(y), of the fully-developed flow
u y Vh
between the plates, and plot vs . for 0  Re V0  0, 1, and 3 , all on one graph. Using your
U h 
velocity expression, determine the shear stress () as a function of y, and on a second graph, plot
h y
vs . , again for Re V0  0, 1, and 3 . Using your general expressions for velocity and shear
U h
stress, determine the limiting values as Vo  0, and explain the subsequent results.
Vo

h y

x
Vo

178
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 6

12. Consider the flow of a viscous fluid (viscosity = µ, density = ) similar to a Couette-type flow
between two infinite, parallel, porous tubes of radii ri and ro, where ri < ro. The inner tube of
radius ri translates along the z-axis at a velocity U, and the outer tube of radius ro is stationary.
Fluid of the same properties is uniformly injected radially at velocity V through the inner tube of
radius ri, and fluid is uniformly suctioned out through the outer tube of radius ro, satisfying
continuity. Determine the radial velocity behavior, vr(r), and the axial velocity profile, vz(r), in
the annulus between the tubes, for this fully-developed flow.

ro r

ri
V 
z
U

v z r  v z r  1  r * k
Show that the non-dimensional velocity, v z *  , can be determined as:  ,
U U 1  ri * k
Q
and the non-dimensional flow rate, , through the annulus can be determined as:
2 ro2 U
1 1
 
1  ri * k    
1  ri * k  2 
Q* 
Q

2 k2  , where k  Vri , and r *  ri (note:
i
2
2 ro U 1  ri * k
 ro
ro
ro *   1 .)
ro

Plot two separate graphs of the results. On the first graph plot v z * vs . r * for 0.5 < r* < 1.0, and
k = 2, 4, and - 4 (i.e. suction through the inner tube). This graph will show three separate lines—
one for each k value. On the second graph plot Q* vs. ri* for 0.1 < ri* < 0.9, and k = 2, 4, and - 4
(i.e. suction through the inner tube). This graph will again show three separate lines---one for
each k value. What do these graphs reveal about the effect of fluid injection/suction on the
velocity and flow rate?

179
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 7

Chapter 7

Equations of Motion for Inviscid Flow

Contents

7.1 The Euler Equation ...................................................................................................... 181

7.2 Special Cases of the Euler Equation .............................................................................. 183

7.2.1 Flow Along a Streamline ..................................................................................... 184
7.2.2 Irrotational Flow ................................................................................................ 184

7.3 The Bernoulli Equation ................................................................................................ 185

7.4 Applications of the Bernoulli Equation ........................................................................ 186

7.4.1 Two-Dimensional Inviscid Duct Flow: Nozzles and Diffusers ................................ 186
7.4.2 Starting Flow from a Tank .................................................................................. 189
7.4.3 Inviscid Stagnation Flow .................................................................................... 191

7.5 Flow Curvature Effects: Euler s and n Equations ........................................................... 193

7.6 Applications of the Euler s and n Equations ................................................................. 197

7.6.1 Pressure Field Inside an Inviscid Vortex .............................................................. 197
7.6.2 Inviscid, Irrotational Flow Around a Bend in a Duct ............................................. 199
7.6.3 The Impact of Curvature on Particulate Motion ................................................. 201
7.6.3.1 Vortex-Induced Motion of Particles or Gas in a Liquid .............................. 202
7.6.3.2 Erosive Effects and River Meandering ...................................................... 202

7.7 Summary of Inviscid Flow Equations ............................................................................ 205

7.7.1 Euler Equation Along a Streamline or in Irrotational Flow ................................... 205
7.7.2 Bernoulli Equation (along streamline or irrotational flow) ................................... 205
7.7.3 Euler s and n Equations (streamline coordinates, unsteady) ............................... 205
7.7.4 Euler s and n Equations (integrated, unsteady) .................................................. 205

While all real flows have at least some degree of viscosity, there are many cases for which
an assumption of inviscid flow ( = 0) is a reasonable approximation. The assumption of
inviscid flow can greatly simplify the governing equations such that they are generally much
easier to solve than the fully viscous, Navier-Stokes equations, and may often yield good
approximations to the true fluid behavior. These inviscid solutions can also indicate the
limiting condition that is possible for a flow configuration (i.e. the “best” condition that
might exist for a flow geometry). It is also often the case that an inviscid solution for one
region of a flow may be selectively combined with a viscous solution within an adjacent
region of the flow, in order to provide a broader global solution. Flows around airfoils are
an example of this latter approach, where the region near the surface requires use of the
fully, viscous equations, but the region somewhat removed from the surface can be
reasonably modeled as an inviscid flow.

180
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 7

7.1 The Euler Equation

In this chapter, we begin to address inviscid flows by deriving and examining the governing
basic equations, which are generally known as the Euler equations. To begin, consider a
flow with  = 0, which we term as inviscid (i.e. lacking viscosity). However, since a real
fluid can never be truly inviscid, what are situations for which the assumption of “inviscid”
behavior is a close (or at least a reasonable) approximation? There are basically two
situations:

1. Where shear gradients are small (i.e. the flow is relatively uniform), and
2. For high Reynolds number, where the flow is well removed from solid surfaces.

Although we have not discussed the Reynolds number yet, although we alluded to it in
Chapter 6, it is defined generically as the local ratio of momentum to viscous forces. A
generic momentum for a flow is characterized by: momentum = (mass flowrate)(velocity)
 
 VL2 V  V 2 L2 . Here,  is the fluid density, V is a characteristic velocity, and L is a
characteristic length (note that L2 reflects a characteristic area). A generic viscous force for
 V
a flow can be characterized as: (shear stress)(area)=   L2   VL , where µ is fluid
 L
viscosity, and V and L are again a characteristic velocity and length. Defining the Reynolds
number as the ratio of the generic momentum and viscous forces gives:

V 2 L2 VL VL 
Re    where  
VL   

When the Reynolds number for a flow, Re, is very large, this implies that momentum effects
dominate viscous effects, particularly if the flow is well away from a bounding surface
(more on this in Chapter 13).

Given that one or both of the above situations (essentially uniform flow, or large Reynolds
number) exists, we will take the liberty to assume that the viscous terms in the Navier-
Stokes equation can be neglected. Thus, dropping the viscous terms from the Navier-Stokes
equation (in vector form, Eq. 5.43), gives us:
Neglect Neglect 
    
1
3
  2
   V   V    P  
V
t
  V  V   (7.1)

Equation 7.1 is termed the Euler equation, and applies for both compressible and
incompressible flows. Note that removal of the viscous terms reduces the equation to a
balance between pressure, body forces, and momentum changes.

While somewhat simplified, the remaining equation is still quite complicated. To assess
possible further simplifications, we expand the velocity terms as follows, using some vector
identities to illustrate the inherent motions involved in the equation:

181
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 7


curl V

 V   V2   
  P       V    V  (7.2)
t   2  
 
V   V
Now, consider the Euler equation for any arbitrary direction in space. We will develop Eq.
7.2 in a Cartesian coordinate system by taking the dot product of the Euler equation with a

differential displacement vector, given by ds  dx ˆi  dy ˆj  dz kˆ :



  V   V2     
  P        V    V    d s

 t   2  

Or expanding:

i.e.


ˆ ˆ ˆ  1  ˆ P  ˆj P  k
 i x  i  y  jz    i
ˆ P 



  x y z  


V 

2

V 

2

V 

2
 

  ˆi dx  ˆj dy  kˆ dz 
  2   
 V
  ˆi    ˆj  2   kˆ  2
 
  ˆi u  ˆj v  kˆ w  curl V 


 t x y z 
 V2 
And collecting terms: d 
dP  2 


 V     V 2   V 
2  2 

 x dx   y dy   z dz  
  x
1  P
dx 
y
dy 
P
dz  
z  t
P
 ds  
 
dx   dy    V dz 
y  2  z  2  
 x  2  
  w v   u w   v u 
 wdy  vdz     udz  wdx     vdx  udy    ( 7.3)
  y z   z x   x y 

x y z
 
where   ˆi x  ˆj y  kˆ z  vorticity    V

Note that in Eq. 7.3 we identify the gradients of pressure and V2 (which is a scalar), as well
as the components of vorticity.

If we assume that gravity is the only relevant body force, and that g (as shown below) acts
parallel to and always opposite to a directional displacement h, where h  h( x, y, z) , then
we can express in Eq. 7.4 the body force vector in components as:

182
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 7

 h h h  h h h 
   x ˆi   y ˆj   z kˆ  g ˆi  g ˆj  g kˆ  g ˆi  ˆj  kˆ   gh (7.4)
x y z  x y z 

To understand how we establish Eq. 7.4, the geometric origin of the terms in the body force

vector  is shown in figure 7.1, for the x-direction component.
g
y
dh h
dh h
 where cos    g
dx dx x x
h z
 x  g cos   g
x
Figure 7.1 Geometric illustration of the body force component in x-direction

Using the components from Eq. 7.4, we can write the body force term in Eq. 7.3, due to a
gravitational field only, as:

h h h
x dx   ydy  z dz  g dx  g dy  g dz
x y z
(7.5)
 h h h 
 g dx  dy  dz   gdh
 x y z 

Substituting Eq. 7.5 into Eq. 7.3 and rearranging gives:


1  V 2  V 
g dh  dP  d    
 d s  wdy  vdz x  udz  wdx y  vdx  udy z  0 (7.6)
  2  t

Equation 7.6 is a general equation that applies throughout the Cartesian flow field; we could
of course derive a similar expression for any other coordinate system of choice, as we will
show when we employ streamline coordinates in Section 7.5.

7.2 Special Cases for the Euler Equation

Clearly, the general Euler equation, Eq. 7.6, is simplified, but is still quite complicated for
practical applications. However, by utilizing some further flow constraints we can simplify
Eq. 7.6 even further, such that tractable solutions are possible. As we will show, these
constraints require that either: (1) we are able to follow the flow along a prescribed
streamline, or (2) the entire flow field is devoid of vorticity, and thus irrotational.

183
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 7

7.2.1 Flow along a streamline

Recall that in section 3.2 we showed that the generic equation for a streamline is given by
 
Vxd r  0 , which mathematically represents a set of differential equations 
V
for a line which is always tangent to the velocity vectors for all particles 
ds
comprising that line. Expanding this equation in Cartesian coordinates,
 
where V  uˆi  vˆj  wkˆ , and dr  dxˆi  dyˆj  dzkˆ , gives a streamline vector equation,
which can be written as:
 
Vxd r  wdy  vdz ˆi  udz  wdx ˆj  vdx  udy kˆ  0
or as component equations
vdx  udy  0, wdy  vdz  0, udz  wdx  0

Substitution of these component equations for a streamline into the Euler equation, Eq. 7.6,
results in null coefficients for the components of vorticity in the equation, negating any
influence of vorticity along a prescribed streamline, yielding:

1  V2  V 
gdh  dP  d    ds  0 (7.7)
  2  t

Equation 7.7 is Euler’s streamline equation for inviscid, compressible, unsteady flow along
a streamline, with conservative body forces (i.e. gravity only).

7.2.2 Irrotational Flow

Alternatively, we can consider Euler’s equation within an irrotational flow field. An

irrotational flow is a flow that is devoid of vorticity (more on this in Chapter 8), for which
we can write:
  
    V  curl V  x ˆi  y ˆj  z kˆ  0  x  y  z  0

Thus, all the vorticity components in the Euler equation, Eq.7.6, are zero for irrotational
flows, giving:

1  V 2  V 
gdh  dP  d    ds  0 (7.8)
  2  t

Note that Eq. 7.8 is the same as the streamline form of the equation, Eq. 7.7. However, Eq.
7.8 may apply across streamlines (not just along them) for a flow that is inviscid,
compressible, unsteady, and irrotational with conservative body forces. Note that both
Eq.7.7 and 7.8 apply throughout the flow field (assuming the respective constraints are
satisfied). We will now consider the application of both equations.

184
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 7

7.3. The Bernoulli Equation

Consider an inviscid flow either: (1) along a streamline, or (2) within an irrotational flow
field. In both cases, we can integrate Eq. 7.7 or 7.8 from a point 1 to a point 2 within the
flow field, giving:

2 2

dP  V22 V12  V 
gh 2  h 1       
  ds  0 (7.9)
1
  2 2  1 t

Equation 7.9 applies for an unsteady flow of variable density, i.e.   f P  .

However, if the density is constant (incompressible), the equation simplifies further to:

2

1 1 2 V 
gh 2  h1   P2  P1   V2  V1  
 2
2
t

 ds  0
1
or

 P2 V22   P1 V12  2 V 
   gh 2      gh 1     ds  0 (7.10a)
 2   2  1 t

or more generically,

P 1 2 V 
 gh  V    d s  constant (7.10b)
 2 t

Equation 7.10 is termed the “unsteady” Bernoulli equation, since it takes into account
locally unsteady or time-varying behavior (but we clearly have
 to know the path we follow
V 
from point 1 to 2 to evaluate the time-dependent term,   d s , in Eq. 7.10). If we further
t
require that the flow remain steady (no velocity changes with time, only spatial changes),
then we obtain:

1
gh 2  h1   P2  P1   1 V22  V12  0
 
 2
or
P2 P
 gh 2  12 V22  1  gh 1  12 V12 (7.11)
 

Equation 7.11 is the classic steady Bernoulli equation, which applies only to steady flows of
constant density. This is one of the most celebrated, most useful, and one of the most
misused equations of fluid mechanics. While providing a simple relationship relating
changes in pressure, gravitational effects, and velocity, the equation (as we noted) is subject

185
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 7

to numerous restrictions. And while many flows approximately satisfy most of these
restrictions, and their behavior can be assessed to a good approximation using the Bernoulli
equation, many flows do not satisfy one or more of the restrictions and cannot be properly
assessed using the Bernoulli equation. An example of where the Bernoulli equation cannot
be used is in the analysis of real fluid flows near a solid boundary, where viscous effects are
always important. Thus, while a useful tool, one must be careful to assure that the results
obtained using the Bernoulli equation are appropriate to the particular type of flow situation.

7.4 Applications of the Bernoulli Equations

The following section demonstrates the application of the Bernoulli equation for several
inviscid flows. Note that for all the cases discussed, viscosity will have some effect, which
will influence the actual expected behavior. However, as pointed out in Section 7.3,
Bernoulli solutions in many cases can provide a good approximation of the type of behavior
that can be expected for a real flow, and provide a limiting “best case” scenario for such
flows. As we will illustrate, these best-case flow situations can often be used as guidelines
for the real flow behavior through the use of some empirical coefficients to account for the
impact of viscosity.

7.4.1 A Two-Dimensional Inviscid Duct Flow: Nozzles and Diffusers

Steady flow Bernoulli equation, along a streamline

For a generic, two-dimensional duct with a steady inflow, we can assess the pressure
change, P2-P1, across the duct as a function of the inlet velocity, V1, the inlet and outlet
areas, A1 and A2, and the density, . Depending on the area change within the duct, the duct
can act as either a nozzle (flow accelerator) or diffuser (flow decelerator). Here, we assume
that the inlet and outlet velocities are uniform. We will also establish the non-dimensional
pressure rise across the duct based on the inlet dynamic pressure or head i.e. 12 V12  .

V1 V2 V1 V2
A1 A2 A1 A2
2 1
1 2
(a) (b)

Figure 7.2 Schematic of parameters for a two-dimensional duct flow:

(a) a nozzle (flow accelerator) and (b) a diffuser (flow decelerator)

Figure 7.2 shows the generic shape and parameters characterizing two-dimensional nozzles
and diffusers. We will employ the Bernoulli equation to understand the tradeoffs between
velocity and pressure changes in such varying area ducts for an incompressible flow.
Applying Eq. 7.10 from points 1 to 2 (we assume along any streamline), we can eliminate
the terms indicated based on the listed assumptions:

186
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 7

5 5 4
 Assume
 P2 V22   P1 V12  2 V 1) inviscid
   gh 2      gh 1    ds  0 2)   const.
 2   2  1 t
3) along S.L.
4) steady
P2  P1 1 2

 V1  V22
2
  5) neglect height changes
6) uniform flow

A1
For constant density and uniform flow, the continuity equation gives: V2  V1 .
A2
Substituting for V2 in the Bernoulli equation gives:

P2  P1 1 2   A1  
2

 2 V1 1    
   A 2  

Now, note the effect of an area change on the pressure change:

A1
 0 if  1  nozzle
  A 2  A2
P2  P1  2 V1 1  
1 2 1
  is (7.12)
  A 2    0 if
A1
 1  diffuser
A2

Here we apply the term “nozzle” to duct geometries that cause an acceleration of the flow
(V), and a consequent decrease in pressure from the inlet to the exit. Conversely, we
apply the term “diffuser” to a duct geometry that causes a deceleration of the flow (V),
with a consequent increase in pressure from inlet to exit.

Note that the term 12 V12 in Eq. 7.12 above has the dimensions of pressure (Force/Area), and
represents the change in pressure that would be experienced if all the entering momentum of
the inlet flow were converted to pressure by bringing the flow inviscidly to rest. We term
this potential pressure change the “dynamic pressure,” and utilize it to form a non-
dimensional parameter, Cp, by dividing the pressure change by the dynamic pressure
available at the inlet:

P2  P`1   A1   
2
1 
CP  1  1      1  2
(7.13)
  A 2    AR 
2
2
V1

A2
where AR  (termed the area ratio), and Cp is termed the pressure coefficient. The
A1
pressure coefficient, which is also often termed the Euler number, is a convenient
engineering relationship that removes the dimensional dependence on the actual flow
velocity, and indicates the capability of a: (1) nozzle to “convert” pressure to additional

187
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 7

momentum exiting the nozzle, or (2) diffuser to “recover” pressure from the fluid
momentum entering a diffuser.

Note that for a nozzle, AR < 1, which means that Cp < 0, or that P1 > P2. A falling pressure
in a duct flow creates a very stable type of flow, which assures that a nozzle flow will create
an acceleration of the flow pretty much in proportion to the pressure drop from nozzle inlet
to exit. However, because of viscous effects in real ducts, Eq. 7.13 will slightly under
predict Cp (i.e. a slightly higher inlet pressure will be required to achieve a desired exit
velocity). Usually, this variation, known as a nozzle discharge coefficient, CD, is taken into
account by the use of an empirical correction factor. In general, because of the stability of a
falling pressure, there is no limit on how small the area ratio (AR) of a nozzle may be.

Whereas, a nozzle behaves pretty much according to Eqs. 7.12 and 7.13, the flow process in
a diffuser is more complicated, and subject to strong variations from Eqs. 7.12 and 7.13. For
example, if AR  2 for a diffuser, Eq. 7.13 indicates that C p  34 . This means that, ideally,
3
4
 
of the dynamic pressure 12 V12 entering the diffuser could be converted or “recovered”
by the diffuser as a pressure increase. This increase in pressure could be employed
productively in a fluid machine (e.g. a gas turbine) to produce more power (in the form of
thermodynamic work,  PdV ). However, like a nozzle, viscous effects for a real fluid flow
will reduce the amount of pressure actually recovered relative to this ideal amount. What is
worse, is that if the area increase across the diffuser is too large (roughly AR > 2.75 to 3),
the actual pressure recovery capabilities of the diffuser will be much less than that predicted
by the inviscid solution, Eq. 7.13. This under prediction is caused by a reduction in the
effective exit area for the flow due a process of pressure-gradient induced flow separation (a
process we will discuss in detail in Chapter 14).

1
0.9
(a) inviscid flow
0.8
0.7
0.6
(b) viscous flow
Cp

0.5
0.4
0.3
0.2
0.1
0
1 1.5 2 2.5 3 3.5 4 4.5 5
Area Ratio=A2/A1

Figure 7.3. Non-dimensional pressure recovery (Cp) vs. diffuser area ratio (AR=A2/A1)
for: (a) an inviscid flow, and (b) an actual viscous flow.

188
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 7

When flow separation develops within a diffuser, a region of stagnant or recirculating flow
develops adjacent to one wall of the diffuser, which diverts the flow through the diffuser
away from that wall (i.e. causing it to “separate” from the solid boundary). This process,
termed diffuser separation or stall, results in a fluid blockage of a portion of the diffuser
exit. This blockage thus reduces the effective exit area, A2, which results in a sharp
reduction of the actual pressure recovery by the diffuser from the ideal recovery predicted
by Eq. 7.13. Figure 7.3 is an illustration of the degree of variation the pressure coefficient
can experience with increasing area ratios. Thus, when it comes to diffusers, a larger exit
area is not necessarily better. More information on diffuser behavior can be found here.

This example shows clearly that one must carefully assess the conditions for which the
assumption of inviscid (or at least nearly inviscid) flow is an appropriate assumption.

7.4.2 Starting Flow from a Tank

Unsteady Bernoulli equation, along a streamline
Consider a tank, as shown in figure 7.4, filled with water to a depth D. A pipe of length L
allows water to flow out of the tank. If the pipe leading from the tank is initially plugged,
how will the velocity of the water leaving the tank, V2(t), behave after the tube is unplugged
at t = 0?

To simplify this problem, we make the assumption that A1>>A2 such that the tank water
depth, D, can be considered constant. We could do the problem with h varying with time,
but it makes the final differential equation non-linear, and more complicated to solve. We
will also assume that we can follow a streamline from the free surface of the water in the
tank (1) to the exit of the pipe at (2). Of course, we also assume an inviscid flow, which is
a stretch if the pipe is of any modest length. However, the solution we obtain will indicate
the maximum velocity behavior that could be expected for such a flow, and thus provides an
upper bound on the velocity behavior for any real flow.
Patm
V10  A1
g 1

y D
L Patm
V2(t)
1' 2
s

Figure 7.4 Schematic for the starting flow from a tank through a long pipe.

189
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 7

Applying the unsteady Bernoulli equation, Eq. 7.10, along a streamline between points 1
and 2 as shown, we have:
Assume
1)   constant
P2 P1   1 V2  V2   gh 2
 h1   
V
ds  0
2) flow along streamline
 2 2 1 2
t 3) inviscid flow
1
4) uniform velocity @ 1 & 2
Patm-Patm=0 0 0 -D 5) V10 since A1>>A2
6) V is negligible from 1 to 1
t
1' 2
V V 7) V  V2 in the pipe
V2  gD   ds  
2
1
2 ds  0
1
t 1
t
0
Note that since the pressures at the beginning and end of the streamline are both
atmospheric, the pressure terms cancel (we assume height effects on the surrounding
atmosphere to be negligible). Also, since A1>>A2, the velocity V1 will be negligible relative
to V2. Finally, we break the integral of the flow acceleration into two parts: (a) from the
tank water surface to the entrance to the pipe, and (b) from the pipe entrance [s = 0] to the
pipe exit [s = L]. Since the velocity within the tank will be quite slow, we assume that the
V V2
time change in this region will be negligible, and that within the pipe  , such that :
t t
sL
V2
2 V2  gD  
2
1
ds  0
s 0
t

dV2 dV2
(L  0) since  f (s)
dt dt

Rearranging terms gives:

dV2 2gD  V22

 (7.14)
dt 2L

Separating variables and integrating Eq. 7.14:

V2 t
dV dt
0 2gD 2V22  0 2L

V2  2gD 
 tanh  t (7.15)
2gD  2L 
 

Equation 7.15 indicates that as t increases the pipe exit velocity will increase asymptotically
toward a steady state value of V2  2gD . As shown in figure 7.5, increasing the pipe
length increases the time required for the velocity to reach steady state. This delay is due to

190
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 7

the increased inertia of the water within the tube as L increases. Clearly, as tube length is
increased, the gravitational driving force must accelerate a larger mass of water within the
tube --- and since the driving force is fixed, the flow will accelerate more slowly.

V2
2gD
increasing L

0
t
Figure 7.5 Velocity behavior of the starting flow from a tank through a pipe of length L
and free surface height of b.

Obviously, most real pipe flows will also experience viscous effects, which will both
increase the time to reach steady state, and reduce the maximum flow rate that can be
achieved. Therefore, Eq.7.15 represents the upper limit of how rapidly the water flow from
the tank can accelerate, and the maximum average velocity that can be achieved at steady
state.

7.4.3 Inviscid Stagnation Flow

Consider an incompressible, inviscid, two- y

dimensional flow toward a solid boundary as
shown below. The Cartesian velocity field for this
flow is given by:

V( x, y)  Ax î  By ĵ
x
Where A and B are constants, and the velocity
components are:

u  Ax and v  By

We want to establish the pressure behavior along a typical streamline for this type of flow.

As we discussed in Chapter 5, an incompressible two-dimensional flow must satisfy the

reduced two-dimensional continuity equation of the form:

u v
  0  AB A  B
x y

So, to have a valid incompressible flow we must have V  Ax î  Ay ĵ , where
u  Ax , v  Ay .

191
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 7

To establish a streamline, we recall from section 3.2 that for a two-dimensional flow the
equation for a streamline is:
 
V  d r  udy  vdx  0 along a streamline.

So for the present case, we have:

dy v  Ay y
  
dx u Ax x

Separating variables and integrating gives:

dy dx
 y
 
x

ln y   ln x  lnC , so that

xy  C , where C is a constant (7.16)

Now, applying the steady flow Bernoulli equation, Eq. 7.11, along an inviscid streamline
gives:
neglect
P 1 2
 V  gh =constant
 2

or substituting the velocity components:

P 1 2 2

 A x  A 2 y 2 = constant
 2
 (7.17)

To specify a particular streamline, we select a particular point that we want the streamline to
pass through. For example, if we specify that the streamline pass through the origin, (x,y) =
(0,0), Eq. 7.16 reduces to:

xy = C = 0

If we specify the pressure at (x,y) = (0,0) to be Po, we can solve for the constant in Eq. 7.17
as:

Po P
 0 = constant = o
 

Thus, at any other point along this streamline, xy = 0, we have:

192
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 7

A2 2
P  Po 
2

x  y2  (7.18a)

Thus, the highest pressure along this streamline occurs at the point (x,y) = (0,0), which

happens to be the stagnation point for the flow ( V = 0), and thus Po is termed the stagnation
pressure.

  v u 
Note that there is no vorticity for this flow,   z kˆ    kˆ  0 , which means that
 x y 
this particular flow is also irrotational. Thus, the pressure equation derived above will also
apply at all points within the flow, as well as along the designated streamline. Rewriting
Eq. 7.18a, as:

2
x 2  y2  Po  P  (7.18b)
 A2

Eq. 718b indicates that the pressure field for this flow is represented by isobaric (constant
pressure) curves that are circles centered on the origin (i.e. if Po  P  constant , Eq.7.18b is
the equation for a circle).

7.5 Flow Curvature Effects: Euler s and n Equations

While the Bernoulli equation derived in section 7.3 provides a useful relationship between
the velocity and the corresponding pressure and height changes along a streamline (or
within an irrotational flow), it is often useful to understand how the pressure varies across
streamlines due to curvature within a flow. Many flows in nature and technical applications
undergo significant curvature, which can give rise to significant pressure variations in a
direction normal to the curved streamlines. For viscous fluids, the resultant cross-stream
pressure variations caused by flow curvature can play a significant role in creating
secondary flows of fluid near solid boundaries (note: secondary flows are localized flows
created by the effects of the pressure field due to the primary flow). In fact, as we discuss in
Section 7.6.3, curvature induced pressure changes are a key factor in causing the
meandering of rivers and in the development of other highly three-dimensional flows.

To address the effects of flow curvature, and assess the impact it can have on inviscid (and
viscous) flows, we will reconsider the Euler equation along a streamline, Eq. 7.7, by
employing a streamline coordinate system.

Recall that along a streamline, the general Euler equation (where   gh ) is given by:
 
1 DV V  
 gh  P    ( V   )V (7.19)
 Dt t

Now consider a two-dimensional flow along a streamline. Here we employ coordinates s

and n, where s is tangent to the streamline and n is normal to the streamline. Consequently,

193
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 7

we write the velocity magnitude as a function of s and n, such that V = f(s, n). Thus, the

vector expression for the velocity is given by V  Vs, n sˆ . Here V is the magnitude of the
velocity, ŝ is a unit vector oriented tangent to the streamline (along s), and n̂ is a unit
vector oriented normal to the streamline (along n), as shown.

Note that we would typically write the velocity vector 

 n
for such a coordinate system as V  Vs sˆ  Vn nˆ . n̂ ŝ V  Vsˆ
s
However, since the coordinates are fixed on a streamline

we have V  V sˆ (since there is no flow across a
streamline), such that Vn  0, and Vs  V . R
+
So, in streamline coordinates, the acceleration, given by applying

the substantial derivative to V  V sˆ , becomes:
 =0
DV    sˆ V sˆ V
 ( Vsˆ )  V ( Vsˆ )  Vn ( Vsˆ )  V  sˆ  V 2  sˆ V (7.20)
Dt t s n t t s s

To make proper sense of Eq. 7.20 requires that we evaluate the derivatives of the unit
sˆ sˆ
vectors, and , which we show graphically below.
s t

ŝs  sˆ s  s  magnitude
n̂
s s  R ŝs  sˆ  1 nˆ 
R
 sˆ s  s  direction

+
Thus,

sˆ sˆ  nˆ  nˆ
 lim  lim  (7.21a)
s s 0 s s 0 R R

and

sˆ sˆ  nˆ  
 lim  lim  nˆ (7.21b)
t t  0 t t  0 t t

Substituting Eqs. 7.21 into Eq. 7.20 we have:


DV  V V    V 2 
ˆ
 s V   nˆ  V   (7.22)
Dt  t s   t R 

194
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 7

Now, using Eq. 7.22, we rewrite Eq. 7.19 in sˆ and nˆ coordinates, where the del operator in
 
streamline coordinates is given by   sˆ  nˆ , which gives us the Euler streamline
s n
equation:

 h h  1  P P   V V    V 2 
 g sˆ  nˆ    sˆ  nˆ    V sˆ   V  nˆ (7.23)
 s n    s n   t s   t R 

Separating Eq. 7.23 into its vector components, and rearranging, gives:

h 1 P V V
g  V  0 s-direction component (7.24a)
s  s s t

dh 1 P V 2 
g   V n-direction component (7.24b)
n  n R t

Equations 7.24a and 7.24b are known respectively as the Euler s-equation and Euler n-
equation.

To integrate the Euler streamline equation along a streamline, we take the dot product of Eq.

7.23 with a differential element oriented along the streamline, d s  sˆ ds , and integrate the
resultant equation from a point 1 to a point 2 on the streamline:

2 2 2 2
h 1 P V V
1 g s ds  1  s ds  1 V s ds  1 t ds  0 (7.25)

dh dP dV =0
h h h
Note that we can associate, for example, dh  ds  dn  ds , since we are
s n s
constraining changes to occur only along the streamline, for which dn = 0.

Integrating Eq. 7.25 gives:

2 2
dP 1 2 V
g( h 2  h1 )  


 2 V2  V12  
t

ds  0 for unsteady,   f (P) (7.26)
1 1
or
2
1
g( h 2  h1 )  P2  P1   12 V22  V12   V ds  0
  for unsteady,   const. (7.27)
 1
t

Equations 7.26 and 7.27 are the same unsteady Bernoulli equations as Eqs. 7.9 and 7.10 that
we obtained in section 7.3.1 above, when we invoked the conditions of flow along a
streamline for the general Euler equation. This is of course as it should be, and illustrates
that the Bernoulli equation is independent of a specific coordinate system.

195
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 7

To assess the effect of flow curvature, we can use the component of the Euler equation for
the n-direction (Eq. 7.24b). Starting with the n-equation, we can demonstrate how pressure
and height vary normal to the direction of a streamline.

Let dn represent a differential element that is perpendicular to a streamline, where
 
dn  dn nˆ . We now take the dot product of dn with Eq. 7.23, giving:

h 1 P V2 
g dn  dn  dn  V dn (7.28)
n  n R t

dh dP
 
Since we will only be integrating in the normal direction, n, we can associate d    dn ,
n
since we are constraining the changes to only occur normal to the streamline, such that ds = 0.
r 
For an arbitrarily curved streamline, the integration of the velocity V  Vsˆ
terms could be pretty complicated. So, to simplify our n̂ ŝ s
considerations of curvature, we will constrain our assessment to a
set of streamlines, each of which has a circular radius of curvature
R. We also let dn  dr , where r is a variable originating at the R
origin and oriented along the radii of curvature of the streamlines,
+
so Eg.7.28 becomes:

1 V2 
gdh  dP  dr  V dr (7.29a)
 R t

Integrating Eq. 7.29a between two points along the radius, and letting  = constant, gives:

2 2 2 2
1 V2 
 gdh 
1
1 dP  
1
R
dr   V dr
1
t
(7.29b)
2 2
1 V2 
g( h 2  h1 )  ( P2  P1 )   dr   V dr
 1
R 1
t

or further constraining the flow to be steady, we get:

2
1 V2
g(h 2  h1 )  ( P2  P1 )   dr (7.30)
 1
R

Note that: (1) the velocity, V, in Eqs. 7.29 and 7.30 is the magnitude of the velocity along
the streamlines, V=Vs=V(r,s), and (2) these equations only apply normal to streamlines
when r is parallel to n,   constant, and Eq. 7.30 further requires that the flow is steady.
Additionally, the centers of curvature must remain fixed or the equation does not apply
(basically, all flow streamlines must describe a circle, or a circular arc). As we will see,

196
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 7

generally we set R = r when integrating the right hand side of Eqs. 7.29 and 7.30, since the
origin of the radius of curvature, and the coordinate r will be the same for a set of circular
arc streamlines.

7.6 Applications of the Euler s and n Equations

The following section demonstrates the application of the Euler s and n equations for several
inviscid flows. Most real flows have some degree of curvature, which can affect the cross-
stream pressure distribution. Recall that we showed this pressure variation with curvature
for several of the examples we did in section 6.4. These cross-stream pressure differences
can have a significant effect on what are termed secondary flows, which are local flows that
are not coincident with the main or primary flow. Secondary flows generally occur where
viscosity retards the primary flow to such an extent that a cross-stream pressure field can
cause movement of fluid normal to the primary flow direction. These secondary flows
commonly develop near solid surfaces, where viscous effects are the strongest. The
consequences of such secondary flows are the development of cross-stream variations in the
overall velocity field, leading to significant three-dimensionality of the flow field, and often
increased energy losses and flow separation. We'll discuss such losses in more detail in
Chapter 15.

Euler’s n-equation, which is used to establish pressure changes across streamlines for
inviscid flows, can also be used to approximate how cross-stream pressure changes result
from curvature in real fluid flows.

7.6.1 Pressure Field Inside an Inviscid Vortex

Steady flow Euler n-equation, fixed center of curvature

One of the simplest, and most instructive, examples of curvature-induced pressure changes
is a circular vortex rotating in an inviscid environment. Consider a radially symmetric
vortex of radius R rotating with solid body rotation, with a velocity distribution given by
V(r) = cr, where c is a constant. Assuming that the flow is steady, and gravitational effects
are not relevant, Euler’s n-equation, Eq. 7.29a, for a fixed center of curvature, simplifies to:

1 dP V 2 Assume
 V(r) = cr
 dr r
1. Euler n-eqn. applies
R 2. P  P(r )
Substituting the solid body velocity r
distribution gives: 3. P=Po at r=0
Po
4. circular center of
curvature such that
dP (cr)2 R=r

dr r 5. neglect h changes
6. steady
Separating variables, and integrating outward from the
center of the vortex, r = 0, where we assume a pressure of
P = Po, we obtain:

197
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 7

P r

 dP    c rdr
2

P0 r 0
r
r2 2 r2
P  Po   c   c2
2 r 0
2

So, the pressure distribution within the vortex is given by

r2
P  Po   c 2 (7.31)
2
R2
and the pressure at the outside of the vortex, r = R, is given by PR  Po   c 2 .
2
Thus, for a vortex with solid body rotation, the pressure increases with increasing radius,
with the lowest pressure occurring at the center of the vortex.

Now, consider a vortex with a velocity profile that decreases inversely with radius, such that
c
V r   . Here, the same assumptions and the form of Euler’s n-equation apply.
r
Substituting the velocity profile into the Euler n-equation gives:
V(r) = c/r
2
 c2 
 2 dr
dP V 2  r  c2
   3 1 R r
dr r r r

Integrating from a point 1 to a point 2 along

the radius, we have:

P2 r2 r2 2 2
dr 2 1   c2  c2 V V
P r  r r 3
2
dP  P2  P2   c   c   
 2 r2    2
 2   2   1
1 1
 r1 2 r2 2 r1 2 2

Where V1 and V2 are the velocities at the respective points, and

V12 V2
P2  P1    2 (7.32)
2 2
Note that Eq. 7.32 indicates that
P2  P1   1 V 2  V 2   0 , which is the same result we
2 1
 2
would obtain integrating Euler’s s-equation for an irrotational flow. This is because this
particular type of vortical velocity profile contains no vorticity. This type of vortex is thus
termed an irrotational vortex.

However, note that there is a problem with this description of the pressure field, since if
c c
r1  0 this would require that V1     . Actually, a better way to describe this
r1 0

198
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 7

pressure field is to let one limit of integration be the pressure at an infinite radius, such that
P = P as r  , and integrate from infinity inward, which gives:

P r
dr  c2  c2  c2 V2
 dp  P  P   c  r3
2
      
P r 
2 r 2 2  2r2 2
or
 c2 V2
P  P   P   (7.33)
2 r2 2

Equation 7.33 indicates that the pressure will decrease from P toward the center.

Again, as r  0, the velocity will increase infinitely, and Eq. 7.33 indicates that the pressure
would decrease infinitely, which is not physically possible. In a real flow of this type,
viscosity would cause a “readjustment” of the velocity profile to a solid body rotation near
the center of the vortex, which allows V  0 as r  0, such that the pressure at r = 0 will be
a finite value. Such vortices, which combine an irrotational outer flow with a solid body
inner rotation, are termed combined Rankine vortices. These types of combined vortices
provide reasonable models of the pressure behavior within actual viscous vortical flows,
such as hurricanes and bathtub vortices. The analysis of such a combined vortex flow is
given as a problem at the end of this chapter.

Note that the two different vortical flows just examined illustrate a characteristic that is
generic for all vortical flows, whether they are inviscid or viscous. For all vortical flows,
dP
 0 , which implies that the pressure within a vortex will always decrease toward the
dr
center of rotation. As we discuss in the following example, cross-stream, radial pressure
dP
gradients, , resulting from streamline curvature, can play a particularly important role in
dr
creating secondary flows near constraining solid boundaries, such as in the vicinity of bends
in pipes or ducts, and even in rivers.

7.6.2 Inviscid, Irrotational Flow Around a Bend in a Duct

Consider inviscid, irrotational flow around a two-dimensional bend whose bounding

surfaces are concentric, circular arcs, as shown in figure 7.6.
Assume
r2
1) inviscid
r1 2)   constant
3) steady flow
4) irrotational flow
5) neglect height changes

Figure 7.6 Two-dimensional flow around a circular arc bend or elbow.

199
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 7

From the Euler s-equation for irrotational, steady flow, we have:

1  V2 
gdh  dP  d   0 (7.34)
  2 

Integrating Equation 7.34 within the bend flow field, since the flow is assumed irrotational,
gives:

P 1 2
 V  constant over the flow field (7.35)
 2

If we now differentiate Eq. 7.35 with respect to the radius, r, we have:

1 P V
V 0 (7.36)
 r r

Equation 7.36 will be valid anywhere for irrotational flow within the bend.

However, for any streamline passing around the bend, Euler’s n-equation will also apply in
the form:

1 P V 2

 n R

where n is coincident with the radius, r, of the streamline, and R = r is the radius of
curvature of any streamline within the bend, such that:

1 P V 2
 (7.37)
 r r

Equating the radial pressure gradients in Eq. 7.36 and 7.37 we have:

1 P V 2 V dV
  V  V
 r r r dr

Separating variables and integrating from the inside of the bend (V = V1 at r = r1) to an
arbitrary point within the bend, yields,
2
V
dV dr
r
V r P
V V  r r  ln V1   ln r1 1
1 1

or

200
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 7

r1 C
V  V1  where C  V1 r1 = constant. (7.38)
r r

Substituting Eq. 7.38 into Eq. 7.37 for the radial pressure gradient, indicates that

dP C2
 3 0
dr r

Integrating from r1 to r2 gives:

C2 1 2   r1  
2
C2
P2  P1   2   2  2 V1 1      0 since r1<r2 (7.39)
r1 r2   r2  

Equation 7.39 indicates that the pressure is always higher toward the outside of a bend or
elbow.

Recalling our results for inviscid vortices in section 7.4.3, we can generalize that for curved
streamlines,

dP V2
 0 (7.40)
dn r

Equation 7.40 implies that pressure will

always increases in the direction of Pout Pout> Pin
curvature, as illustrated to the right.
Pin
This characteristic of the radial pressure r
increasing in the direction of curvature
holds true for all flows, inviscid or
viscous, and for streamlines moving along non-circular paths. As discussed before, cross-
stream pressure gradients that result from these curved streamline patterns are responsible
for the development of secondary flows near bounding surfaces in real fluid flows.

7.6.3 The Impact of Curvature on Particulate Motion

In sections 7.6.1 and 7.6.2, we illustrated how cross-stream pressure will always increase in
the direction of radial curvature. The practical implications of this effect are manifold, from
the creation of cross-stream secondary flows near boundaries, to the creation of swirling
three-dimensional flows. Additionally, cross-stream pressure gradients due to flow
curvature effects can provide a strong driving force for the motion of solid particulates or
gas bubbles suspended in a fluid. The following two qualitative examples illustrate the
impact of flow curvature on particulate motion.

201
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 7

7.6.3.1 Vortex-Induced Motion of Particles or Gas in a Liquid.

Since pressure always increases outward from the center of fluid rotation, this pressure
difference (the local magnitude reflected by the radial pressure gradient) provides a driving
force toward the center of rotation. In essence, this is a driving force toward the center of
curvature proportional to the local angular acceleration. The
effect of the pressure gradient is to generate a surface force
acting inward, toward the center of curvature on any particle
within the flow. That surface force will be proportional to the
fluid mass displaced by the particle. However, a particle
rotating in the flow will also be subject to a centrifugal force
acting outward from the center of curvature, but proportional
to the mass of the particle. Consequently, for a particle that is
denser than the fluid, such as most dirt particles, the
centrifugal force is dominant, and the particle will move
outward. This effect is the basis for the operation of cyclone
separators, which are used to remove heavy particulate matter
from fluid and air flows. A schematic of a cyclone separator
is shown at the right. Conversely, a particle that is less dense
than the fluid will move inward, since the pressure gradient
force will dominate. This effect can be observed, for example,
by the bubbles in a carbonated beverage moving toward the
center of a glass when the beverage is stirred. Of course, if the
particle density is equal to that of the fluid (i.e. neutrally
buoyant), the particle will maintain it position and move with
the fluid.

Centrifugal General Particulate Behavior in Rotating Flow

effects particle  fluid  moves outward
Pressure particle  fluid  moves inward
gradient

7.6.3.2 Erosive Effects and River Meandering

When there are substantial variations in topographical elevation, rivers and streams tend to
follow the contour of the lowest point of land, or valleys. However, when a river or stream
flows across a relative flat area, such as prairie or open field, flow curvature can have a
significant impact on the path of the river/stream due to the development of a secondary,
cross-stream flow that carries eroded soil particles from the outside to the inside bank of a
river or stream.

Consider the plan-view schematic shown in figure 7.7a of a river rounding a curve. The
curved flow in the river will result in a cross-stream pressure gradient , with pressure
increasing outward from the center of curvature such that the pressure within the main body
202
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 7

of the flow will be higher at the outside bank of the river than the inside bank. If the flow is
fast enough, this increase in pressure will also be visually evidenced as an outward increase
in the cross-stream height of the river surface.

As the streamwise cross-section A-A of the river in figure 7.7b shows, the flowing water
will erode the outside bank of the river, with the heavy soil particulates settling toward the
riverbed. When the soil particles approach the riverbed, they will enter a viscous fluid
boundary layer just above and adjacent to the riverbed (We will cover boundary layers later
in Chapter 13). Within this boundary layer, the flow of the water in the streamwise
direction will be strongly reduced (by the no slip effect at the riverbed), and so will the
associated hydrodynamic forces which drag the eroded soil particles in the streamwise
direction. However, the cross-stream pressure gradient created by curvature of the main
flow will be “impressed” on both the boundary layer, and the soil particles within this layer.
This radial pressure gradient creates a lateral force on the particles, which pushes the
particles and the local flow adjacent to the riverbed toward the inside riverbank, causing the
particles to “migrate” across the river and be deposited on the inside bank of the river. The
dotted arrows in figure 7.7a roughly model the particle migration from the outer to the inner
bank of the river.

River collection of particles as a

A erosion sand bank or beach

riverbed
sand bank Plow
A Phigh migration of particles in the
boundary layer

migration of particles in the Streamwise view, A-A

boundary layer

(a) Plan view of a river (b) Cross-section view, looking downstream

Figure 7.7 Illustration of the migration of eroded particulate material across a river or
stream bed due to curvature-induced, cross-stream pressure gradients.

As this process of cross-stream transport continues, eroded soil particles from the outer
riverbank will continue to migrate and deposit on the inner riverbank, forming a beach or
sand bank. This continuing erosion of the outer bank and deposition of material on the inner
bank will create additional curvature of the river, which will increase the process of
erosion/transport. The long-term result is the development of an increasingly curved,
sinuous path for the river. This continued development of river curvature is generally termed
“meandering.”

The Mississippi River in the United States, and the Nile River in Egypt are excellent
examples of meandering caused by this process of erosion and secondary flow transport, all
of which can be explained by the presence of radial pressure gradients. Clearly, if a river is

203
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 7

initially straight, this process requires some initial perturbation to start the development of
curvature, but existing irregularities in the land contours are generally sufficient to initiate
this natural process.

Figure 7.8 is an example of this meandering river effect, from a picture I took from an
airplane at 36,000 feet. Note how the sand bars (light yellow color) occur on the inside of
the bends in the river, and in some cases form small islands at the inside of a bend in the
river. Given enough time, and without interference by human engineering, these bends in
the river will continue to accentuate, often causing the river to cut around the land to meet
on itself and form a separate channel. And this is all due to the processes inferred from the
Euler n-equation.

Figure 7.8 An image of river meandering, clearly showing sand bars (appearing as
light yellow) formed on the inside of bends (curves) in the river.

To illustrate the movement of heavier particulates by secondary flows due to a radial

pressure gradient, try this simple experiment. Fill a circular container (any glass, coffee cup,
or teacup will work) with water, and then drop in some fine dirt particles (heavy enough that
they settle to the bottom of the vessel—sand works well). Now stir the water until it is
moving in a relatively steady circular motion. Remove your stirring implement, and watch
what happens to the dirt on the bottom of the vessel. What you will observe is that as the
flow slows, due to viscous effects, all the dirt particles will concentrate in the center of the
vessel—proof of the presence of a secondary flow along the bottom of the vessel toward the
center of curvature due to the imposed radial pressure gradient by the main flow. This is
often called the "tea leaf phenomena or paradox", since it has long been observed that tea
leaves within a cup of stirred tea display this central concentration behavior. Click on the
above link to see a demonstration of this tea leaf concentration, which was explained
originally by no other than Albert Einstein.

204
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 7

7.7 Summary of Inviscid flow Equations

7.7.1 Euler Equation Along a Streamline or in Irrotational Flow


1  V2  V 
gdh  dP  d    ds  0
  2  t

7.7.2 Bernoulli Equation (along streamline, or irrotational flow)


 P2 V22   P1 V12  2 V 
   gh 2      gh 1     ds  0 Unsteady
 2   2  1 t

 P2 V22   P V2 
   gh 2    1  1  gh 1  Steady
 2   2 

7.7.3 Euler s and n Equations (streamline coordinates, unsteady)

h 1 P V V
g  V  0 s-direction component
s  s s t

dh 1 P V 2 
g   V n-direction component
n  n R t

7.7.4 Euler s and n Equations (integrated, steady)

 P2 V22   P V2 
   gh 2    1  1  gh 1  along streamline
 2   2 

2
1 V2
g(h 2  h1 )  ( P2  P1 )   dr normal to streamline
 1
R

205
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 7

Problems

1. An inviscid, two-dimensional flow field is given by V  x 2 ˆi  2xyˆj . Assuming that all values are
dimensionless, determine:

a) If this an incompressible flow; if so, let  = 1(dimensionless);

b) The vorticity for the flow field;

c) The equation for a streamline passing through a point 1 at x,y = 1,1;

d) The value of y at a point 2 lying on the streamline of part c at x = 2;

From our streamline, the value of y for x = 2 is given by:

e) If the pressure at point 1, P1, is zero, determine the value of pressure at point 2, P2;

f) Using point 1 in part e), could you determine the value of pressure at a point x,y = 2,2? If
you can, determine the pressure; if you can't, explain why you can't.

2. Consider a liquid jet that flows vertically upward in a gravitational field, as shown below. Assume the flow is
uniform and inviscid. Determine an expression for the change in the jet cross-section with distance y from the
jet orifice. What is the limiting height for this jet? Why?

y
Liquid Jet

g
Vo

Ao

3. Consider a liquid jet that falls vertically downward in a gravitational field, as shown below. Assume the flow
is uniform and inviscid. Determine an expression for the change in the jet cross-section with distance y from
the jet orifice.
Ao

Vo
g
Liquid Jet

206
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 7

4. An inviscid flow of constant density, , passes through a straight pipe of length L. The pressure at the
outlet of the pipe remains constant at P2=0. Determine [in terms of V0, , L,  and t ] how the inlet
pressure to the pipe (P1) must vary if the velocity in the pipe is varies sinusoidally as
V=Vo Sin(t).
1 2

P1(t) = ? V=Vo Sin(t)

P2=0
x
L

5. An inviscid flow of constant density, , passes through a straight pipe of length L. The pressure at the
outlet of the pipe (P2) remains constant at P2=0, while the inlet pressure to the pipe (P1) varies as
P1=P0(t/to) for 0  t  to, after which P1=0=constant for t > to.
a) Determine an expression for the velocity through the pipe as a function of P0, , L, to and t.
b) What is the maximum velocity achieved, and when in time is it reached?
c) Plot V/Vmax vs. t/to from 0 < t < 2to.

1 2
P1=P0(t/to), for 0  t to
V(t) ?  = constant P2=0, for t  0
P1=0, for t > to
L

6. An inviscid flow passes through a 2-D channel, followed by a 2-D channel, that undergoes an area
change over the last half of the channel, as shown. Each half of the channel is L long, and the exit
area ratio of the end of the channel is given by exit area (W2) divided by the inlet area of the
beginning of the channel (W1), such that AR=W2/W1. The channel inlet velocity increases as
V1=Vo(t/to) for 0 < t  to, after which V1=Vo=constant for t > to.
Channel area ratio=AR=W2/W1 2
1 1’
V1

L L
x

Determine the pressure ratio, C P 

P2  P1  , as a function of time, t. On a single graph, plot Cp for
1 2
Vo
2
AR=2, 1, and 0.5, over the time period 0 < t < 2to. For purposes of dimensional correctness in
plotting the behavior, assume that to = 2L/Vo

Briefly, explain what phenomena causes this unusual behavior, and what its implications are for
start-up flows.
207
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 7

7. Apply the unsteady Bernoulli equation to the U-tube manometer of constant diameter shown.
Assume that the manometer surface levels differ from the equilibrium level ho by an amount .
The initial surface deflection at t = 0 is o, and the initial surface velocity is zero. Using the
unsteady Bernoulli equation, obtain a differential equation for (t) and, solve for  in terms of the
properties shown. Also, determine the interface velocity, V(t) in terms of parameters shown.

g

2  o = Initial deflection

1 h V = 0 is initial velocity
ho

Let the heights of the respective surfaces be h1 = ho - , and h2 = ho + , and the total length of the
fluid column in the manometer be S = 2ho + L.

8. A circular combined Rankine vortex is present in an inviscid fluid, as shown.

P

The vortex consists of a core of radius R, which is in solid body rotation with velocity
r R
V1  V0   , and a portion outside the core (r  R) that has a velocity V 2  V0   . The
R r
density of the flow is , and the pressure far away from the vortex center is P.

Determine the following:

a) The vorticity for the entire flow field, as a function of r. Plot or make a nice labeled sketch
of  vs. r/R.

b) The location of the minimum pressure, Pmin, and the value of that pressure in terms of , V0, and
P. Neglect any height changes.

208
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 7

9. A circular vortex is present in an inviscid fluid, as shown.

r  r
V1  V0   V 2  V0  2  
 R
R
P

R
The vortex consists of a core of radius R, which is in solid body rotation with velocity
r  r
V1  V0   . The portion outside the core (R  r  2R) has a velocity V 2  V0  2   . The
R  R
density of the flow is , and the pressure far away from the vortex center is P, including at r =
2R. Determine the following:

a) The vorticity for the entire flow field, as a function of r. Plot or make a nice labeled sketch
of  vs. r/R.

b) The location of the minimum pressure, Pmin, and the value of that pressure in terms of , V0, and
P. Neglect any height changes.

10. A circular vortex is present in an inviscid fluid, as shown.

r V0  4R r 
V1  V0   V 2    
3  r R
R
P

R
The vortex consists of a core of radius R, which is in solid body rotation with velocity
r V  4R r 
V 1  V0   . The portion outside the core (R  r  2R) has a velocity V 2  0   .
R 3  r R
The density of the flow is , and the pressure far away from the vortex center is P, including at r
= 2R. Determine the following:

a) The vorticity for the entire flow field, as a function of r. Plot or make a nice labeled sketch
of  vs. r/R.

b) The location of the minimum pressure, Pmin, and the value of that pressure in terms of , V0, and
P. Neglect any height changes.

209
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 7

11. A circular vortex in an inviscid fluid has the following vorticity distribution,  vs. r, where,  =
o for r  Ro, and  = 0 for r  R.


vorticity distribution

o

r
R

If the velocity distribution is V  v  ˆi only, determine the following:

a) The velocity distribution, v(r), in terms of o and r, for r  R (assume that v(0) = 0).

b) The velocity distribution, v(r), in terms of o, R, and r, for r  R.

c) The location of the minimum pressure, Pmin, and the value of that pressure in terms of , o, R
and P, where P = P when r >> R. Neglect any height changes.

210
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 8

Chapter 8

The Stream Function and Vorticity

Contents

8.1 The Stream Function ................................................................................................... 211

8.1.1 The Relation of Streamlines to Flow Rate ........................................................... 213
8.1.2 The Concept of a Stream Tube ............................................................................ 214

8.2 Vorticity and Circulation ............................................................................................. 214

8.2.1 Vorticity and Circulation for a Fluid in Solid-Body Rotation .................................. 216
8.2.2 Vorticity and Circulation for an Irrotational Vortex ............................................. 217
8.2.3 Vorticity and Circulation for a Fully-Developed Channel Flow .............................. 218

8.3 Vortex Lines and Vortex Tubes .................................................................................... 220

8.3.1 Vortex Lines ....................................................................................................... 220
8.3.2 Vortex Tubes ..................................................................................................... 221
8.3.2.1 Solid-Body Rotation Along a Vortex Tube ................................................ 224
8.3.2.2 The Tornado as a Vortex Tube ................................................................ 225

In previous chapters we have developed methods to describe and visualize local fluid
velocity and rotational behavior. In chapter 2 we introduced the streamline for describing
the behavior within a velocity field. In chapter 4 we introduced vorticity, as a property
reflecting fluid rotation. In this chapter we extend those processes to assess more global
behavior, in terms of both the flowrate behavior within a fluid, and a parallel method for
assessment of the collective rotation within a fluid flow.

8.1 The Stream Function

In many situations the incompressible differential continuity equation (Eq. 5.24) can be
reduced to two dimensions in Cartesian coordinates (e.g. x and y) such that:

u v
 0 (two-dimensional,  = constant) (8.1)
x y

Equation 8.1 still leaves us with two independent velocity variables, u and v. As with
most differential equations, we look for solution techniques that take advantage of
constraining conditions. In the case of two-dimensional flow, we can reduce the number
of dependent variables from two to one by defining a function ( x, y) , such that this
function will be a solution to the two-dimensional continuity equation if u and v can be
expressed as:

211
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 8


u (8.2a)
y
and

v (8.2b)
x

If such a function ( x, y) exists, then when we substitute Eq. 8.2 for u and v into the
continuity equation, Eq. 8.1, we obtain:

u v  2  2
   0 (8.3)
x y xy yx

Equation 8.3 implies that such a function, ( x, y) , is an exact solution of the continuity
equation.
 
Now recall that the equation for a streamline is given by V  d r  0 , which for two-
dimensional flow simplifies to:

udy  vdx  0 (8.4)

Substituting Eq. 8.2 into Eq. 8.4 gives:

 
dy  dx  0 (8.5)
y x

Now if we also differentiate the function  , the chain rule gives us:

 
d  dy  dx (8.6)
y x

Comparing Eq. 8.6 with Eq. 8.5, which we obtained from the streamline equation, this
indicates that d = 0 along a streamline, or that  must be a constant along a streamline.

We term  the stream function of the flow field, from which we can derive the velocity
components, plus other useful relationships for the flow field. We will see later in
Chapter 9 that we can use the stream function to assess the behavior of inviscid,
irrotational flows. Additionally, in Chapters 12 and 13, we will demonstrate how we can
use the functionality of the streamfunction to reduce the order of the Navier-Stokes
equations to help us solve for the behavior of two-dimensional, viscous flows
(unfortunately, at the expense of generating higher-order terms).

212
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 8

8.1.1 The Relationship of Stream Lines to Flow Rate

Note that since d is an exact differential, the integral of d between any two points in a
flow field will be independent of the path of integration (remember your calculus). Now
consider the flow between two streamlines, as shown in figure 8.1.
v
 2
V 
2
u dA
dAx dy 1

1
-dx
dAy

Figure 8.1 Geometric schematic of flow between two streamlines.

From control volume continuity, we know that the volume flow rate is the integral of the
velocity field normal to the cross-sectional area spanning the streamlines, or:

  2
Q  flowrate   V  dA
1

The component of the velocity crossing a surface is often termed the velocity “flux.” So
the flow rate is the integral of the velocity flux over a prescribed area. If we examine this
flow rate integral for two-dimensional flow, using the general configuration shown in
figure 8.1, we can generalize the velocity and the differential area (which in two-
dimensions is really just a cross-sectional distance) as:
 
V  uî  vĵ and dA  dA x ˆi  dA y ˆj  dy ˆi  ( dx )ˆj


(note that dA y  dx , since the projection of dA is in the negative x-direction)

  
2 2
Thus, Q   uî  vĵ  dyî  dxĵ   udy  vdx 
1 1

Substituting for u and v from Eq. 8.2, we get:

2 2
   
Q    dy  dx    d   2  1 (8.7)
1
y x  1

Since the integral of Eq. 8.7 is path independent, this means that the flowrate between
any two streamfunctions (i.e. streamlines) is a constant.

213
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 8

8.1.2 The Concept of a Stream Tube

Extrapolating the two-dimensional flow rate/streamline relationship to three-dimensions,

we note that a collection of streamlines will form what we will term a stream tube, with
the surface surrounding these streamlines being termed the stream surface. This is like
closely packed glass fibers in an optical transmission cable.

out
in

Figure 8.2 Example of a stream tube

For such a stream tube, as shown schematically in figure 8.2, we note that the velocity

vector, V , will be tangential to the bounding surface of the stream tube, which means
that no flow can cross the stream tube surface, except across a cross-section of the stream
tube. Thus, the volume flowrate throughout a stream tube must be a constant,
   
i.e. Q   V  dA    dA
V (8.8)
Ain Aout

Depending on the behavior of the streamlines within the stream tube, the geometry of the
stream tube may change, resulting in variations of the streamwise cross-sectional area of
the stream tube. Additionally, the geometry of the stream tube may vary with time, as
the initial streamlines vary with time. However, regardless of how the stream tube varies
geometrically with time, the instantaneous flow rate through the stream tube will always
remain constant across any streamwise cross section.

8.2 Vorticity and Circulation

As discussed in Chapter 4, vorticity is the term given to the local rotational behavior of a
fluid, which for simplicity is defined as twice the local angular velocity of the fluid.
Given the Greek designator , vorticity is defined mathematically as:
  
  2  angular ve locity    V  curl V

Note that vorticity is a vector property and is expressed as derivative components of the
velocity field relative to the coordinate system employed. For example, in the Cartesian
coordinate system (u, v, w), we write:

  x ˆi  y ˆj  z kˆ

214
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 8

w v u w v u
Where, x   , y   , and  z  
y z z x x y

In radial coordinates, the vorticity equations are a bit more complicated (due to non-zero
derivatives of unit vectors). In radial coordinates, the equations of vorticity are:

 
    V  ˆir r  ˆi   ˆiz z

 1 v z v    v v  1  v 
Where, r     ,    r  z , z   rv    r ˆiz
 r  z   z r  r  r  

Whereas vorticity reflects the local rotation of a fluid at a point, we can also assess the
cumulative rotation within a prescribed region, which we define as the circulation
strength, . We define the circulation of the fluid as the summation of the vorticity that
crosses, and is enclosed by, a prescribed surface area. We first establish the flux of
vorticity (i.e. the component of vorticity normal to, and thus crossing, the area in
question) by taking the dot product of the vorticity with respect to the differential area
vector. We then integrate the resulting vorticity flux over the prescribed area, such that:
    
     dA   (  V)  dA   V  ds (8.9)
A A C  C
V

Stokes Theorem dA 


Note that by making use of Stokes theorem (Section 2.4.4), we can establish a very useful
relationship that relates the circulation to the velocity tangent to the boundary of the area.
This relationship, which employs a line integral of the tangential velocity around a circuit
bounding the prescribed area, C, bypasses the need to determine the local vorticity field,
and in many cases makes the determination of the circulation mathematically much less
involved.

Because circulation is the summation of the vorticity flux across a designated area, it is
analogous to volume flow rate, since it represents the summation of the flux of a vector
property crossing a designated area or surface (defined by a bounding loop C). As such,
circulation is a measure of the “strength” of the rotational behavior across a specified
surface area, just as flow rate is the measure of the “volume” of flow crossing a specified
surface area. Thus, the circulation strength over a prescribed region may reflect either
vorticity broadly distributed over a large region (such as a slowly rotating tropical
depression over the ocean) or a concentration of strong, concentrated vorticity within
some portion of the same large region (e.g. a tornado). As we will discuss in Chapters 11
and 12, the effects of regions of less concentrated or more concentrated vorticity can have
comparable effects on the local velocity field, if the total circulation of the regions is also
comparable.

215
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 8

8.2.1 Vorticity and Circulation for a Fluid in Solid Body Rotation

As we discussed in Section 7.6.1, flows with solid body V(r) = Cr

rotation have a velocity field of the form:
R

V  v  ˆi  Cr ˆi (where C is a constant).  r

This type of flow is characteristic of the steady flow within a

steadily rotating cylindrical vessel, and the flow near the core
of many natural vortices (e.g. a hurricane). Since this is a
cylindrically symmetric flow, we apply the equation for
vorticity in cylindrical coordinates:
0
  1  1 v r 
    V  ˆiz  ( rv  ) 
 r r r  

1   1 
 ˆiz  (Cr 2 )  ˆiz  2Cr   ˆiz ( 2C)
 r r  r 

Thus, the vorticity for this flow is a constant across the vortex, as it should be for a solid
body rotation. Now, consider the circulation strength for a vortex of radius R. Here, we
apply Eq. 8.9, using an area integral,
r
R
   r
     dA where dA ˆiz (rd)dr . 
A

The integration limits for a circle of radius R are 0    2 and 0  r  R , so we have:

2 R R
r2  R2 
 
2
 
0 0
 ˆiz (2C )  ˆiz rdrd  2C 0 2
 2C ( 2)
 2
  ( 2C)  R 2

(8.10)
0

Alternatively, we could have noted that at r  R , v  CR ˆi , and that a differential


element at the boundary of the circle can be defined as ds  Rd ˆi . Thus, using the 
line integral approach from Eq. 8.9, we would calculate the circulation as:

  2 2
ˆ ˆ
   V  d s   (CR ) i  ( Rd) i   CR 2d  CR 2 2  ( 2C)( R 2 ) (8.11)
C 0 0

Equation 8.11 is, of course, the same result we obtained in Eq. 8.10 using an area integral
of the vorticity.

216
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 8

8.2.2 Vorticity and Circulation for an Irrotational Vortex

V(r) = A/r
2
Again, as we discussed in Section 7.6.1, the velocity field for
 A
an irrotational vortex is given by V  v ˆi  ˆi (where A is
r 1 R r
a constant). For this type of velocity profile, the vorticity, in
cylindrical coordinates, is:

  1   1  A 
    V  ˆiz  ( rv )  ˆiz  ( r )  0 (8.12)
 r r   r r r 

Equation 8.12 indicates that this is a vorticity-free flow, which suggests that the flow
should (logically) be irrotational, and thus have no circulation. Clearly, if we consider an
 
area integral of the vorticity, we would have      dA  0, since our calculations
A

indicate that   0 everywhere.

However, if we consider the line integral around a circle of radius R, then at r = R, the
 A
velocity is V  ˆi , and the differential vector element along the circumference is
R

ds  Rdˆi . Thus, the circulation can be calculated as:


  2  A 
   V  d s     Rd  A( 2)  0 ! (8.13)
o
R

Equation 8.13 clearly contradicts Eq. 8.12. Why the discrepancy? The answer is that for
this irrotational-type of vortex there exists one point of vorticity: a singularity at the
center of rotation, r  0, where an infinite spike in vorticity exists. This illustrates that
the curve surrounding the vortex distribution contains all the vorticity, regardless of how
small an area the vorticity is contained within, as illustrated by figure 8.3.

No Vorticity
Vorticity

1 2
1  2

Figure 8.3 Example of two bounded regions of different area, containing the same
region of vorticity, and thus having identical circulation.

This suggests that we don’t have to be too exacting when choosing our boundaries from
which to determine the circulation for a concentrated area of vorticity located within a
region of otherwise irrotational fluid. Accordingly, we can use a boundary that

217
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 8

encompasses not only the area containing the vorticity, but also any additional adjacent
area that is devoid of vorticity. For example, if we assume that the velocity distribution
within a hurricane is roughly symmetric, we could hypothetically determine the
circulation strength of the hurricane by using the velocity at a boundary located well
outside of the rotational core, if the additional area encompassed is reasonably
irrotational.

But if there is a concentration of vorticity within the boundaries of r = R, why didn’t our
integration of the vorticity over the area give us the same answer as Eq. 8.13? Let’s
reexamine that integration, by using the expression for the vorticity as:

  1   1  A  ˆ 1  
    V  ˆiz  ( rv  )  ˆiz  ( r )  i z  ( A ) (8.14)
 r r   r r r   r r 

Now, we integrate for the circulation over the area of a circle of radius R.

  2  R ˆ 1   ˆ
2 R
  R
2
     dA    iz  ( A )  iz rdrd     ( A ) drd A 0   A( 2) (8.15)
A
0 0  r r  0 0
r  0

Equation 8.15 gives the same result as Eq. 8.13 using the line integral. Since the vorticity
was isolated at the origin as a discontinuity, we have to integrate the vorticity equation
before specifying the velocity function, to reveal the true value of the circulation.

This reinforces the point made above that the circulation is indicative of the cumulative
vorticity within the specified boundary, regardless of how it is distributed. The lesson
from this? Use the line integral calculation to determine the circulation whenever
possible, since it will always reveal the true circulation.

8.2.3 The Vorticity and Circulation for a Fully-Developed Channel Flow

In Section 6.3.2, we determined the velocity profile for a Poiseuille, pressure-driven flow
between the parallel surfaces of a channel. The relationship, based on the maximum
centerline velocity is given in figure 8.4 below.
y

4 U max
h u  hy  y  uy 
2

h2
x

Figure 8.4 Velocity profile for a fully-developed channel flow

Since this is a viscous flow, we expect that this flow will contain significant vorticity
distributed throughout the flow, with an accompanying amount of circulation.

218
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 8

Here, the vorticity (in Cartesian coordinates) is given by:

ˆi ˆj kˆ
      v u  ˆ u
  V     k  ( zero terms)   kˆ
x y z  x y  y
u v w

 4U max
  z kˆ   h  2ykˆ (8.16)
h2

Equation 8.16 indicates a linear distribution of vorticity across the width of the channel.
Examining three values of vorticity, at the two boundaries and the centerline of the flow,
we have:

4U
@ yh

   max
2
h  2h kˆ   4Umax kˆ 
h h
h 
@ y 0
2
4U
@ y0

   max
2
h  0kˆ   4Umax kˆ 
h h

Thus, the vorticity is positive in the upper half of the channel, and negative in the lower
half of the channel, and linearly distributed between these extremes, being null at the
center of the channel.

To determine the circulation, we need to define a region or surface over which we want
to establish the circulation strength. Since the streamwise extent of a fully-developed
channel flow is indeterminate, let’s examine two flow regions of an arbitrary length L:
one region will lie below the centerline (region I), and the other region above the
centerline (region II), as shown below. Here, we again determine the circulation strength
using both (a) an integral of vorticity over a defined area, and (b) a line integral of the
velocity that follows the area boundary.

For region I, using an area integral we have: L

  
I     dA dA  dydx kˆ II h/2
C B
A
h
x L y 2 y I h/2
4U max x
    h  2 ydydx D A
x 0 y 0 h2
L
h
4 U max

h2
hy  y  2

0
2
x   U max L
0

219
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 8

To determine the line integral, we employ an integration path around the region I,
employing the junction points labeled above. Note that for the x-y coordinate system
shown in figure 8.4, the integration path will follow a counter-clockwise direction,
according to the right hand rule (which has the z-coordinate pointing out of the page).
The direction is defined, of course, by the integration limits we impose on the sequenced
integrals.
  B C  D  A  
I   V  d s   VAB  dyˆj   VBC  dx ˆi   VCD  dyˆj   VDA  dx ˆi , but V  u( y)ˆi
A B C D
yh x 0 y 0 x L
2
  u( y)ˆi  dyˆj   Umax ˆi  dx ˆi   u( y) ˆi  dyˆj   (0) ˆi  dx ˆi  U max L
y 0 x L yh x 0
2

0 Umax (0-L) 0 0

-Umax L

The area and line integral results are identical, as they should be.

Likewise, for region II, using either an area or line integral, we obtain (prove this
yourself).

II  Umax L

Now if we consider the entire region between the channel walls, the total circulation of
the combined regions I and II will be simply the sum of the individual circulation for both
regions. Thus,

TOT  I  II  Umax L  Umax L  0 or no net circulation

You should prove this result yourself by integrating over the collective regions I and II.
This result illustrates that despite containing a significant amount of vorticity, a region
can have a null amount of circulation if the total amount of positive and negative vorticity
offset each other. Note that the behavior of circulation is again similar to volume flow
rate, where equal amounts of positive and negative velocity flux across a specified
surface would yield a null net flow rate across the surface. We will return to discuss the
sources of the vorticity for this, and other viscous flows in Chapter 11.

8.3 Vortex Lines and Tubes

8.3.1 Vortex Lines


Vortex lines are lines in a flow field that are tangent to the vorticity vector,  , just like a

streamline is tangent to the velocity vector, V . We define such lines in the same manner
asa streamline, by setting the cross product of the vorticity with a differential element
d s equal to zero, i.e.:

220
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 8

 
V  ds  0  streamline equation
 
  ds  0  vortex line equation

For example, in Cartesian coordinates the vorticity is given by   x ˆi  y ˆj  z kˆ , and
we can write a expression for a generic differential element along a vortex line as

d s  dx î  dy ĵ  dz k̂ . Thus, the expanded equation for a vortex line is:

  d s  y dz  z dy ˆi  z dx  x dz  ˆj  x dy  y dx kˆ

 
(8.17)

However, if we constrain a flow to two dimensions (e.g. x and y), Eq. 8.17 reduces to:

  d s  x dy  y dx )kˆ
 
(8.18)

However, we note that x and y  0 , since there can be no rotation in the x or y

directions for a two-dimensional flow in the x-y plane. Since such a flow can have non-
zero values of z , and thus contain vorticity, our above result implies that vortex lines
(actually points) could only run perpendicular to the x-y plane. So for a two-dimensional
flow in a plane (with no depth of flow), there can be no vortex lines, only point-wise
vorticity values. Thus, unlike streamlines, a flow must be three-dimensional for vortex
lines to have any physical relevance.

Now recall that in Chapter 2 we indicated in Eq. 2.14 that for a vector field, F , the vector
 
function   (  F)  0 . If we now apply that relationship to a velocity field, V , then we
   
have   (  V)  0 . Since     V , this means that     0 , which implies that the
vorticity field is a solenoidal vector field --- a vector field for which the rate of expansion
(i.e. divergence) of vorticity = 0. Note the similarity to the solenoidal velocity field for

an incompressible flow, where the continuity equation reduces to   V  0 . This
common solenoidal aspect of vorticity and an incompressible velocity field implies that
vortex lines and streamlines (in incompressible flow) obey similar rules, as we will
discuss in the following section.

8.3.2 Vortex Tubes

A vortex tube is like a stream tube --- i.e. a collection of vortex lines bounded by a
common surface. As shown in figure 8.5, a vortex tube is similar to a stream tube, in that
no vorticity will cross the surface of a vortex tube because of the requirement that the
vorticity is always tangent to a vortex line.
Vortex Tube Surface
Vortex
Lines

Figure 8.5 Schematic of a vortex tube

221
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 8


We noted in section 8.3.1 that     0 everywhere within a vorticity field. Thus, if we

integrate    over any volume of a vortex tube,  , the result must also be zero. To
examine how this can give us an appreciation of the circulation within a vortex tube,
consider the behavior of vorticity for the model vortex tube shown in figure 8.6.

i.e     d  0

Asides Volume, 
Aout

Ain


Figure 8.6 A portion of a vortex tube, with the corresponding areas bounding the
tube indicated.

For a typical vortex tube shown in figure 8.6, we can employ the Gauss Divergence
Theorem, Eq. 2.17, to assess the relationship of the flux of vorticity in and out of the

tube. To do this, we write the volume integral of the divergence of the vorticity,    ,
and use the Gauss Theorem to express the equivalence of vorticity flux in terms of
surface integrals over the bounding surfaces of the vortex tube as:

0    dA sides
        
     d     dA       dA     dA    dA  0

 A A in A sides A out

(8.19)

Gauss Divergence Theorem

Recall that in the previous section, 8.3.1, we showed that vorticity is solenoidal, such that
 
   =0 everywhere. Thus, it follows that the volume integral of    must also be zero
 
as well. With regard to the surface integrals of vorticity flux (   dA ) in Eq. 8.19, only
the vorticity which enters and exits a stream tube contributes non-zero values. By our
definition of a vortex tube, the side boundaries of a vortex tube are formed by vortex
lines. Since the vorticity is tangent to a vortex line, it follows that the vorticity is always
tangent to the sides of the vortex tube, and thus there can be no vorticity flux across the
sides of a vortex tube.

out
 dA out
in
dA in

Figure 8.7 Illustration of the normal direction of the surfaces for a vortex tube
relative to the vorticity entering and exiting the tube.

222
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 8


Noting the directional normal vectors for dA are oriented outward from the entrance and
exit surfaces of a vortex tube, as shown in figure 8.7, we perform the dot products
indicated in Eq. 8.19 and obtain:

    in dA in    out dA out  0 (8.20)

A in A out

or
 in  out  0

Equation 8.20 implies that in  out , or that the circulation within the vortex tube (i.e. the
“strength” of the vortex tube) remains constant throughout the vortex tube. This was first
proven by Helmholtz, and as we will show in Chapter 10, this even extends to the
temporal change for a vortex tube in an inviscid flow.
Consider the analogy of circulation to flow rate. As we know, the flow rate, Q, along a

stream tube in an incompressible velocity field, V , will remain constant:
 
Q   V  dA  constant along a stream tube
A


And for the corresponding vorticity field,  , the circulation strength along a vortex tube
is also constant:
 
     dA  constant along a vortex tube
A

Note that as the cross-sectional area of a vortex tube decreases, the average vorticity
within the tube will increase; conversely, if the cross-section of a vortex tube expands,
the average vorticity must correspondingly decrease. However, a vortex tube cannot
shrink to zero area, since the average vorticity must remain finite. This constancy of
circulation associated with a vortex tube implies that the vortex tube, or vortex filament
[as we would call a vortex tube when the tube gets quite small], and by association a
vortex line, cannot terminate in a fluid. For this to be true requires that a vortex tube or
vortex line must either: (1) form a closed circuit [e.g. a vortex “ring’ or loop, like a
smoke ring blown by a smoker], (2) extend to infinity [a hypothetical circumstance], (3)
end at a fluid-fluid boundary [e.g. at an air-water boundary, like the surface of a lake], or
(4) end at a rotating solid boundary [such that the rotation of the boundary matches the
fluid vorticity]. Accordingly, a vortex tube cannot end at a stationary boundary, although
one could imagine some strange form of vortical flow, with a zero vortex line at its center
[and thus no velocity gradient]. This is discussed further in Section 8.3.2.2 below, and
Section 10.4.

223
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 8

8.3.2.1 Solid-Body Rotation Along a Vortex Tube

While the assumption of solid body rotation is a bit artificial, it can be used to illustrate
the process of constancy of circulation. We will examine this type of flow again later in
Chapter 12 when we address the concept of vorticity transport. Here we assume that a
flow enters a vortex tube with a solid body rotation, such that the vorticity is constant at
an initial value o across a cross-section Ao, as shown.

Ao A
o 

We designate the vorticity at any other cross-section of the vortex tube as  , where the
corresponding cross sectional area is A. The circulation strength within the vortex tube is
thus established by the initial vorticity and area as   o A o . If we assume that the flow
remains in solid body rotation within the vortex tube, such that the vorticity is constant at
any cross section, we have:

o A o 
  o A o   dA  A or    (8.21)
A
A A

Thus, the vorticity will change inversely with the area of the vortex tube. Although
Equation 8.21 was determined for a solid body rotation, the same equation applies for
most other vorticity distributions, and applies exactly if we consider the average
vorticities at a cross-section, as indicated in Eq. 8.22.

A  
(8.22)
A  

Equation 8.22 indicates that as a vortex tube narrows, the vorticity will increase. This is
roughly what happens when a slowly rotating fluid passes into a narrowing conduit, such
as occurs when a bathtub is drained. The slowly rotating flow in the tub is constricted
into a narrow exit, with a resulting marked increase in the local vorticity, reflected by a
strongly rotating “bath tub vortex." Click the following link to see such a vortex in a
kitchen sink. Note that what you see is the air core inside the vortex. Alternatively, if
the vortex tube expands, the vorticity magnitude in the vortex tube will decrease. An
example here is the trailing vortices (i.e. vortex tubes) from aircraft wings, which will
undergo a decrease in vorticity as they slow and begin to expand as they trail away from
the aircraft. However, the process of vortex tube expansion is a very unstable process,
that degenerates very rapidly into a chaotic process termed “vortex breakdown,” in which
the vortex tube expands quite suddenly, rapidly dispersing and decreasing the local
vorticity of the vortex tube. It should be noted, however, that this vorticity dispersion
process is much more three-dimensional than the process reflected in the simple model
above (as seen in the video clip link above), and the mechanics of this break down
process are still poorly understood.

224
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 8

8.3.2.2 The Tornado as a Vortex Tube

As we pointed out in section 8.3.2, a vortex tube cannot end at a stationary boundary,
since the rotation of the boundary, and thus the vorticity would be zero. However, this
raises the question of how such things as tornadoes “end”, as they impinge upon the
earth. The answer is that the vortex “tube” comprising a tornado does not end at the
earth’s surface, but is comprised of a collection of vortex lines that [where they appear to
impinge on the earth] extend laterally away (toward infinity) from the concentrated core
we view as the tornado. The general configuration of a tornado vortex tube is illustrated
in Figure 8.8.

sense of
rotation of
tornado
vortex lines

Figure 8.8 Model of a tornado as a vortex tube, comprised of vortex lines extending
laterally along the earth surface. Here the vortex tube is shaded grey.

Clearly, the creation of a tornado, which is a concentrated vortex tube, requires the
accumulation of vorticity distributed across a broad area of land into the tornado. The
source of this vorticity is due to high shearing winds, as well as strong changes in density,
which can cause both the modification and creation of vorticity. The creation and
sources of vorticity is a topic we address further in Chapter 12.

Problems

1. The stream function for a particular flow is given by

 y 
   2 2

x y 

Assume all properties are dimensionless and do the following:

a) plot the stream function = 1 for all y0

b) determine the velocity and vorticity vectors for the flow field

c) determine the pressure differences P1  P2  between P1 at (x,y) = (0,1) and P2 at both
(x,y)=(0.5,0.5) and (0,0.5); let  = 1‚

225
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 8


2. If V  y 2 ˆi  xzˆj  x 2 kˆ , determine the circulation around a rectangle with vertices at (x, y, z)=
(0, 0, 0), (3, 0, 0), (3, 0, 3) and (0, 0, 3) using both a line integral and an area integral.

3. A circular combined Rankine vortex is present in an inviscid fluid, as shown.

P

The vortex consists of a core of radius R, which is in solid body rotation with velocity
r R
V1  V0   , and a portion outside the core (r  R) that has a velocity V 2  V0   .
R r
The density of the flow is , and the pressure far away from the vortex center is P.

Determine the following:

a) The vorticity for the entire flow field, as a function of r. Plot or make a nice labeled
sketch of  vs. r/R.
b) The circulation for,
i) a circle of radius R.
ii) a square of side dimensions 4R, with the vortex in the center of the square.

4. A circular vortex is present in an inviscid fluid, as shown.

r  r
V1  V0   V 2  V0  2  
 R
R
P

The vortex consists of a core of radius R, which is in solid body rotation with velocity
r  r
V1  V0   . The portion outside the core (R  r  2R) has a velocity V 2  V0  2   .
R  R
The density of the flow is , and the pressure far away from the vortex center is P, including
at r = 2R.

226
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 8

Determine the following:

a) The vorticity for the entire flow field, as a function of r. Plot or make a nice labeled
sketch of  vs. r/R.
b) The circulation for,
i) a circle of radius R
ii) a square of side 4R, with the vortex in the center of the square

5. A circular vortex is present in an inviscid fluid, as shown.

r V0  4R r 
V1  V0   V 2    
3  r R
R
P

The vortex consists of a core of radius R, which is in solid body rotation with velocity
r V  4R r 
V 1  V0   . The portion outside the core (R  r  2R) has a velocity V 2  0   .
R 3  r R
The density of the flow is , and the pressure far away from the vortex center is P, including
at r = 2R.

Determine the following:

a) The vorticity for the entire flow field, as a function of r. Plot or make a nice labeled
sketch of  vs. r/R.

b) The circulation for,

i) a circle of radius r = 3R/2

ii) a square of side 4R, with the vortex is in the center of the square

227
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 8

6. A circular vortex in an inviscid fluid has the following vorticity distribution,  vs. r, where,
 = o for r  Ro, and  = 0 for r  R.


vorticity distribution

o

r
R

If the velocity distribution is V  v  ˆi only, determine the following:

a) The velocity distribution, v(r), in terms of o and r, for r  R (assume that v(0) = 0).

b) The velocity distribution, v(r), in terms of o, R, and r, for r  R.

c) The circulation for,

i) a circle of radius R.

ii) a square of side 3R, with the vortex in the center of the square.

7. The velocity fields in a circulation preserving flow for two different vortices of radius R1 are
given by:
vortex #1: V  2.117V0 (r * r *4 ) ,

vortex #2: V  6.75V0 (r *2 r *3 ) ,

where r*=r/R1. Determine the pressure, vorticity and circulation as a function of r* for each
V
vortex for 0 < r* < 1 (let p = po at r*=1). Comparatively plot the velocity (as  ), pressure (as
V0
p  p0  R 
2
), vorticity (as z 1 ), and circulation (as ) for both these vortices vs r*, for 0
V0 V0 2R 1 V0
< r* < 1, on four separate graphs. Which vortex has: The lowest pressure? The highest
circulation? The highest vorticity? How do you explain these variations?

228
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 8

8. A solid rod of radius R rotates steadily in an infinite fluid. The surface velocity (v) of the
rotating rod is Vo at r = R and v = 0 for r  .
Determine:
a) The velocity of the fluid surrounding the rod, if the fluid is fully developed at steady state:

b) The circulation C at r = 2R and r = 100R, using a line integral of the velocity.

c) The vorticity  within the fluid at steady state.

d) The circulation A within the fluid by integrating the vorticity of part c) over the area
encompassed from R  r   .

Explain how parts b) and d) can both be correct.

9. A circular combined Rankine vortex is present in an inviscid fluid, as shown.

P

The vortex consists of a core of radius R, which is in solid body rotation with velocity
r R
V1  V0   , and a portion outside the core (r  R) that has a velocity V 2  V0   .
R r
The density of the flow is , and the pressure far away from the vortex center is P.

Determine the following:

a) The vorticity for the entire flow field, as a function of r. Plot or make a nice labeled
sketch of  vs. r/R.

b) The circulation for,

i) a circle of radius R.
ii) a square of side dimensions 4R, with the vortex in the center of the square.

c) The value of the pressure at the center of the vortex, Po, in terms of , V0, and P. Neglect
any height changes.

229
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 8

This combined velocity profile exists at the beginning and end of a vortex tube in this inviscid
flow, where R is the radius at any cross section of the tube (i.e. the tube contains only the solid
body portion of the vortex, V1 , and the flow outside the solid body rotation is the same as V2
above). If the center of the vortex tube is a streamline along the vortex axis, the beginning and
end radii of the vortex tube are Ra and Rb, and Rb=Ra/2 (with corresponding velocities at the edge
of the stream tube of V0a and V0 b ), will fluid flow along the axis of the vortex tube, and if so,
which way and why?

Extra credit: If the velocity on the streamline at a is VSa  0 , determine the velocity at b, VSb , in
terms of V0a .
Ra Vortex tube
Rb Streamline

b
a

230
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

Chapter 9

Two-Dimensional Potential Flow Theory

Contents

9.1 Basic Approach ........................................................................................................... 233

9.2 The Stream Function, (x,y) ......................................................................................... 233

9.3 The Potential Function, (x,y) ...................................................................................... 234

9.4 The Pressure Field ....................................................................................................... 236

9.5 The Relationship Between  and  .............................................................................. 237

9.6 The Complex Potential Function, (z) ......................................................................... 239

9.7 Elementary Functions ................................................................................................. 241

9.7.1 Uniform, Parallel Flow ........................................................................................ 241
9.7.2 Source and Sink Flows ........................................................................................ 241
9.7.3 Point Vortices .................................................................................................... 243
9.7.4 The Doublet (dipole) .......................................................................................... 244
9.7.5 Corner and Wedge Flows ................................................................................... 246
9.7.5.1 Concave Corner or Wedge Flow ............................................................... 248
9.7.5.2 Convex Corner Flow ............................................................................... 251
9.7.6 Biasing the Origin Location for Elementary Functions .......................................... 252

9.8 Superposition Flow Simulations .................................................................................. 255

9.8.1 Flow Around a Circular Cylinder ......................................................................... 255
9.8.2 Flow Around a Rotating Cylinder ........................................................................ 262
9.8.3 Flow Around a Half-Body ................................................................................... 267
9.8.4 Flow Around a Two-Dimensional Rankine Body .................................................. 274

9.9 Bodies Translating in a Potential Flow ......................................................................... 280

9.9.1 Translation of a Cylinder in a Quiescent Fluid ...................................................... 280
9.9.2 Pressure Field for a Translating Cylinder .............................................................. 284

9.10 Accelerating Bodies in a Potential Flow: Virtual/Added Mass ..................................... 286

9.10.1 Acceleration of a Cylinder in a Quiescent Fluid ................................................ 286
9.10.2 The Added Mass for an Accelerating Cylinder .................................................. 288

In this chapter we deal with what is variously called “ideal flow”, “irrotational flow”, and
“potential flow.” Basically we consider a flow that we assume to be incompressible,
inviscid, and irrotational (i.e. without vorticity). Obviously, no real flow truly meets
these restrictions, particularly flows close to solid boundaries. However, certain types of

231
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

flows come close to satisfying these assumptions, such that assessment of such an
idealized flow will yield reasonably practical results. And at worst, these assumptions
can often provide solutions that are good starting points, or limiting types of flows, for
more complicated viscous flows.

Recall in Chapter 7 that we examined incompressible inviscid flows, which led us to the
Euler and Bernoulli equations. While we were restricted in how we could apply these
equations, we determined that there were a number of situations for which we could
effectively assess “bounding” behavior (i.e. the best case scenarios). The same situation
applies for potential flows—the solutions are restrictive, but provide the best case,
limiting behavior that could be expected in the absence of viscosity.

So what are some examples of where we might utilize potential flow? In regions where a
uniform flow is impinging at high speed upon a solid object. In such situations, as we
will discuss in Chapter 13, the Reynolds number of the flow (a non-dimensional ratio of
inertia to viscous forces, Re = UL/) will generally be large, and the region of viscous
behavior and deformation of the flow will be constrained to a thin region adjacent to the
object, which is termed the “boundary layer.” However, the flow outside this boundary
layer region will behave approximately in an inviscid manner, and will be generally
devoid of vorticity. So, for cases where Re  , the potential flow solutions can prove
quite useful for flow in regions exterior to the boundary layers. As an example, consider
the flow around an airfoil, as shown in figure 9.1.

Boundary Streamlines
Layer

Flow

Figure 9.1 Streamline flow around an airfoil, showing a boundary layer due to
viscous interaction with the airfoil surface (dotted line).

In cases where the boundary layer is minimal, the potential flow solution around the basic
airfoil geometry can provide remarkably accurate information on the lift behavior for a
range of airfoil shapes and angles of attack. However, the results for most practical airfoil
flows is limited in application by the effects of viscosity, which will modify the resulting
potential flow field due to the boundary layer development. This development of a
boundary layer (as shown in figure 9.1) will cause a displacement of the streamlines due
to growth of the boundary layer. As we will discuss in Chapter 13, the developing
boundary layer basically “displaces” fluid, and thus the streamlines, away from the
surface.

Practical solutions for real flows over airfoils can often be determined by modeling the
flow as a viscous boundary layer region adjacent to the airfoil, and an inviscid “outer”
flow for the region outside the boundary layer. This two-region flow is then solved

232
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

iteratively, by assuming that the velocity and pressure fields of the inviscid flow provide
the boundary conditions for the velocity and the pressure gradient at the outer edge of the
boundary layer flow. The final potential flow solution can then be used to determine the
resultant lift forces on the airfoil (by integrating the pressure field around the airfoil), and
the boundary layer solution will yield the shear stress and help determine the drag forces
on the airfoil.

Additionally, potential flow solutions indicate how an ideal flow would behave. These
idealized results can then be used to identify regions of the flow where we can anticipate
that the inclusion of viscous effects will strongly modify the ideal solution (e.g. regions
where strong flow deceleration is indicated). We also touch upon this in Chapters 15.

9.1 Basic Approach

As we discuss above, the constraints we place upon our modeled potential flow are that it
be incompressible, inviscid, and irrotational. Additionally, we require that the fluid have
conservative body forces (i.e. gravity), such that the Bernoulli equation applies, which
allows us to determine pressure variations based on the potential flow velocity field
solutions. Since the Bernoulli equation is derived from the momentum equation (via the
simplified Navier-Stokes, and subsequently the Euler equations), we need only satisfy the
constraints of continuity and irrotationality ( = 0 everywhere) in order to develop a
solution for the velocity field.

Recall that the incompressible continuity equation is represented by:


V  0 (9.1)

and the irrotationality constraint is expressed by:

 
  V  0 (9.2)

Thus, our approach will be to develop a solution technique that assures simultaneous
satisfaction of both of these constraints. We will demonstrate that through the effective
use of complex mathematics we can satisfy both constraint equations 9.1 and 9.2 by use
of either: (a) the stream function concept we developed in Chapter 8, or (b) an alternative
function termed the potential function.

9.2 The Steam Function,  (x, y)

Recall that in Chapter 8 (Section 8.1) we introduced ( x, y) as the stream function---a

function that identically satisfies the two-dimensional continuity equation, and provides a
functional expression representing all streamlines for a particular flow field. In section
8.1, we showed that the velocity field is related to  by:

233
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

   
u and v   , or V  ˆi   ˆj (9.3)
y x y x

Substituting Eq. 9.3 into the continuity equation (Eq. 9.1), gives:

 u v  2   2 
V     0 thus ,  satisfies continuity
x y xy xy

Now, for an irrotational flow, we know that the vorticity of the velocity field must be
zero, so substituting Eq. 9.3 into Eq. 9.2 gives:

 v u  2  2
V  0    2  2
x y x y
or
 2  2
 0 (9.4)
x 2 y 2

We note that Eq. 9.4 is the well-known, two-dimensional Laplace equation. Given the
appropriate boundary conditions, the solution of Eq. 9.4 will yield the function (x,y),

which can then be differentiated to yield the velocity field, V . However, the stream
function can only apply in a two-dimensional plane. In order to establish a more
universal function which is also applicable for three-dimensional flows, we address the
solution of the two coupled equations, Eq. 9.1 and 9.2 (continuity and irrotationality), in
the reverse order to derive an equation for what we term the potential function, .

9.3 The Potential Function,  (x, y)

Here, we now stipulate a potential function, x, y  , for which the velocity field is given

by the gradient of the potential function, V   . For two dimensions (x and y), the
potential function is related to the velocity field components as:

  ˆ  ˆ  
V  u ˆi  v ˆj    i j  u and v  (9.5)
x y x y

Substituting the velocity functions of Eq. 9.5 into the vorticity equation, Eq. 9.2, gives:

   v u    2  2  ˆ
    V    kˆ    k  0
 x y   yx yx 

Thus,  satisfies the irrotationality condition exactly.

234
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

The potential function must also satisfy two-dimensional continuity, Eq. 9.1, which
means:

 u v  2  2
V     0 (9.6)
x y x 2 y 2

Equation 9.6 is also a Laplace equation, for the potential function, . Again, the solution
of Eq. 9.6 will yield a function (x,y), from which we can also determine the velocity
field via Eq. 9.5.

Note that the potential function can satisfy the irrotationality constraint in three
dimensions, not just two dimensions. To demonstrate this, consider a Cartesian system in
three dimensions, such that:

  ˆ  ˆ  ˆ   
V  u ˆi  v ˆj  wkˆ    i j k  u  , v  , w  (9.7)
x y z x y z

Noting the vorticity in three dimensions is:

   w v   u w   v u 
    V    ˆi    ˆj    kˆ (9.8)
 y z   z x   x y 

Substituting from Eqs. 9.7 into Eq. 9.8, we have (for a continuously differentiable
function):

   2  2 ˆ   2  2 ˆ   2  2  ˆ
  V     i     j    k  0
 yz zy   zx xz   yx xy 

Thus, irrotationality is exactly satisfied in three-dimensions, and continuity gives us:

 u v w  2  2  2
V       0
x y z x 2 y 2 z 2
Or

2  0 (9.9)

Equation 9.9 is again the Laplace equation, applicable in three-dimensions. The

satisfaction of irrotationality and continuity in three dimensions allows the potential
function to be utilized to simulate flows in three dimensions, such as axisymmetric flows.

235
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

9.4 The Pressure Field

Since  and  are both solutions to the equations for continuity and irrotationality (Eqs.
9.1 and 9.2), we could solve for either function or both. Once we have determined one or
more of these functions, they can be used to establish the velocity field by the derivative
functions of Eqs. 9.3 or 9.5 above. Since the flow is assumed irrotational, we can apply
the incompressible form of the Bernoulli equation (7.10) everywhere to establish the
pressure field.

P 1 2 V 
i.e.  gh  V    d s  constant (9.10)
 2 t

Revisiting the development of Eq. 9.10 that was done in Chapter 7 (section 7.2), we note
that we can write the Euler equation for an irrotational, incompressible fluid as:

P  1 2  V
 gh   V   0
  2  t

or

P 1 2  V
  gh  V   0 (9.11)
 2  t

Recall that we can write the general expression for the velocity in terms of  as:

  ˆ  ˆ  ˆ
V  uˆi  vˆj  w kˆ  i j  k  
x y z

Using this relationship, we can write:


V    
    (9.12)
t t  t 

since x, y, z are independent variables, and not a function of time (a fourth independent
variable).

Substituting Eq. 9.12 into Eq. 9.11 allows us to collect all the terms of Eq. 9.11 inside the
gradient operator, such that:

P 1  
  gh  V 2    0 (9.13)
 2 t 

236
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9


By taking a dot product of the Euler equation with d s  dx î  dy ĵ  dz k̂ , as was done in
section 7.1 for an irrotational, unsteady flow, Eq. 9.13 can be integrated to give:

P 1 
 gh  V 2   C(t ) , (9.14)
 2 t

In this modified version of the Bernoulli equation, takes care of the time dependency,
t
and C(t) is a function of time and must be determined from a known initial condition.

The key here is to note that we can employ the potential function both to determine the
velocity field and to establish the pressure field by use of Eq. 9.14, the potential function
form of the unsteady Bernoulli equation. We will show later in sections 9.9 and 9.10 that
Eq. 9.14 can be used effectively (amongst other applications) to establish the “virtual” or
"added" mass for a body accelerating through a quiescent fluid medium. The
virtual/added mass is the amount of fluid mass that that moves along with a moving body,
and must also be accelerated when there is a temporal change in the velocity of the
moving body. This added mass can significantly affect the forces required to
accelerate/decelerate the body). This effect is nominal for heavy solid bodies, such as
cars, passing through air (where air is much less dense), but is important for lighter
bodies passing through a liquid medium, such as water, where the fluid is equivalent to or
heavier than the body (e.g. ships, submarines, and even fish).

9.5 The Relationship Between  and 

Note that the total differentials of  and  can be written in two-dimensions as:

 
d  dx  dy (9.15a)
x y
and
 
d  dx  dy (9.15b)
x y

Here we note that along a line of  = constant we have d  0 , and correspondingly

d  0 along a line of  = constant. Thus, if we set d = 0 and d = 0 in Eqs. 9.15a and
9.15b, and solve for the slope of lines of constant  and , with the help of the velocity
identities of Eqs. 9.3 and 9.5, we obtain:

 
dy v dy u
  x  and   x  
dx   u dx   v
y y

Slopes of lines of constant and

237
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

u dy 1
Comparing, we have:   (9.16a)
v dx  dy
dx 

Recall from calculus that the relationship of the slopes shown in Eq. 9.16a is indicative of
two orthogonal (perpendicular) lines. So, we infer that  lines are orthogonal to  lines.

Alternatively, we could take the dot product of the gradients (in two-dimensions) of 
and . Recall that the gradient of  and  will yield vectors which will always be normal
to lines of constant  and . Substituting the relationships for the velocity components
from Eqs. 9.3 and 9.5, we have;

  ˆ  ˆ    ˆ  ˆ     
     i j   i  j     vu  uv  0 (9.16b)
 x y   x y  x x y y

Equation 9.16b again indicates that lines of equipotential ( = constant) and streamlines
( = constant) are orthogonal at all points within a flow field.

Note the similarity of potential flow theory to electric field theory, magnetic field theory,
heat conduction, solid mechanics, etc., where lines of constant potential (e.g. voltage,
temperature,  ) are normal to lines of constant flow (e.g. current, heat flow,  ) . Thus,
the general solutions to Laplace’s equation will satisfy all these types of similar physical
phenomena.

Although we can solve for  or  by direct solution of the Laplace equation, determining
solutions for other than simple geometries can often be quite challenging. For flow in
and around some simple geometries, standard closed-form solutions of the Laplace
equation can be developed. However, more complicated flow geometries will generally
require use of numerical computation approaches. These types of analytical or numerical
solutions are generally known as the direct method of solution (i.e. given a geometry,
determine the flow field). However, by the use of complex mathematics we can develop
solutions for a number of elementary flows, which can be subsequently employed in a
process of linear superposition to “simulate” more complicated flows. Such a process of
linear superposition is possible because both  and  will be solutions to the Laplace
equation, which is a linear differential equation. Recall that for linear differential
equations, all solutions of the equation are directly additive [e.g. 1  2  total , where
1 and 2 are separate solutions of the same differential equation, but subject to different
boundary conditions]. So, we can combine any number of solutions for the Laplace
equation, and their cumulative sum will also be a solution to the Laplace equation.

This process of linear superposition is generally termed the indirect or inverse method,
wherein we employ judicious superposition of selected elementary solutions to create and
simulate more complicated flows. Simple superposition solutions can prove quite
effective in simulating the flow behavior for a number of useful flow geometries

238
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

(e.g. flow around a cylinder or an ellipse). However, the simulation of the flow around a
more complicated body shape using this inverse technique may require extensive
numerical optimization of an array of elementary solutions. Additionally, by the use of
conformal mapping techniques, simple flow geometries, such as the flow over a cylinder,
can be "mapped" to simulate more geometrically complicated geometries, such as flows
around airfoils. We don't cover conformal mapping techniques here, but these are fairly
standard extensions of complex mathematics---see this link.

9.6 The Complex Potential Function,   z 

Since we have shown that both  and  are scalar point functions, satisfy the Laplace
equation, and are orthogonal to each other, we can make use of complex mathematics to
manipulate and solve for these functions. Using the rules of complex mathematics, we
can construct in two-dimensional Cartesian coordinates a complex potential function
defined as  :

(z)  ( x, y)  i( x, y) , where z  x  iy (the complex variable) (9.17)

d
Note that  will be an analytic function of z since  and will exist and be single-
dz
valued except at selected points. Therefore, from complex variable theory, and our
previous definitions of u and v in terms of derivatives of  and  [Eqs.9.3 and 9.5], we
can write:

d 

dz x
 i

y
where i  1 
(9.18a)
   
 i  i   iv  u
x x y y
or
d
 u  iv  W  the conjugate of the complex velocity (9.18b)
dz
and thus W  u  iv  the complex velocity. (9.19)

real part of W imaginary part of W

Since many flows are better suited to cylindrical symmetry (such as vortical flows), it is
often more practical to express the velocity field for such flows, as well as the stream and
potential functions, in cylindrical coordinates. Note that in cylindrical coordinates:

 y
( z )  ( r , )  i( r , ) , where z  rei , and
r  x 2  y 2 ,   arctan  (9.20)
x
Note that: z  x  iy  r cos  ir sin  rcos  i sin  rei

239
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

From the velocity schematic below, we can relate u and v to the velocity components in
the r,  directions as:
y
y W
u  v r cos   v  sin v
r v vr (9.21)
v  v r sin  v  cos 
 x  x
u

Substituting Eq. 9.21 into Eq. 9.18b for the complex conjugate velocity gives:

W  v r cos   v sin  iv r sin  v cos   vr  i v cos   i sin

Or
W  v r  i v ei  (9.22)

One could express W in terms of cylindrical coordinates, and extract the terms for vr and
v. However, a simpler alternative is to note that from complex variable theory (Brown
and Churchill, 2009), the Cauchy-Riemann conditions (also see this link) in cylindrical
coordinates can be shown to be:

 1  1  
 and  (9.23)
r r  r  r
and that

d        
   i  cos   i sin    i  e  i   W (9.24)
dz  r r   r r 

Comparing, W for Eq. 9.22 and 9.24, we have:

   
W    i  e  i   v r  i v e  i 
 r r 

So by comparison, and using the Cauchy-Riemann equations (Eq. 9.23), we obtain the
cylindrical components of velocity in terms of  and  as:

 1 
vr  
r r 
9.(25)
 1 
v   
r r 

To illustrate how we incorporate the complex potential function and its component
potential and stream functions, we will first examine some elementary flows that satisfy
the complex potential conditions. We will then illustrate the use of these elementary
flows to synthesize and simulate more complicated flows using the inverse method of
superposition.

240
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

9.7 Elementary Functions

9.7.1 Uniform, Parallel Flow

The uniform flow function is a real building block of potential flow simulations, and
simulates a uniform, parallel flow. This function generally provides a simulation of the
approach flow to various simulated bodies, such as flow around a cylinder or airfoil.
Here, the complex potential function is given by:

( z )  a  ibz  ax  by   iay  bx  (9.26)

  y
 = const.

Where a and b are constants. To determine the

velocity of this simulated flow, we differentiate Eq.
x
9.26 to obtain the complex velocity.

d
 a  ib  W  u  iv (conjugate of the complex velocity)
dz

ua 
So by identity,  W  u  iv  a - ib (the complex velocity)
v  b

Thus, the values of a and b determine the angle of the vector velocity relative to the
coordinate system.

Special cases are:

i) Flow parallel to the coordinate axis:

x  axis  (z)  az (9.27a)

y - axis  (z)  -ibz (9.27b)

y
ii) Flow at an angle  to the x  axis :

x
( z )  Uze i ( where U is a constant )
(9.27c)
 Uz(cos  i sin)

9.7.2 Source and Sink Flows

These flows simulate flows that either emanate uniformly from a point (a source), or are
drawn uniformly into a point (a sink), and are represented by a complex potential
function given by:

241
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

Q
( z )  lnz  (9.28)
2

Here, Q is a constant representing the “strength” of

the function. If Q is positive (+), Eq. 9.28 represents
a source; if Q is negative (-), Eq. 9.28 represents a
Q>0
sink.
=const.
Since source and sink flows are radially symmetric, we
make use of the complex variable in cylindrical
coordinates, z  rei , such that: =const.

Q
( z )  ln r  i    i
2
Where,
Q Q
  and   ln r (9.29)
2 2

From Eq. 9.25 for cylindrical coordinates, the velocity components are given by:

1    1 
vr   and v   
r  r r r 

So, for a source/sink flow, the velocities in cylindrical coordinates are:

Q
v  0 , v r  (9.30)
2r

Since v  0 , there is no azimuthal velocity (in  direction). Thus, this flow simulates
flow radially outward or inward. If we solve for the volume flow rate (two-dimensional)
for this flow, we have:

   Q ˆ 
Q   V  dA where, V  ir and dA  r d ˆir
A
2r r

  2
Q
Q   V  dA   d  Q
A 0
2

Thus, the strength of this flow, Q, is the volume flow rate (a constant).

Note that the complex potential for this flow displays a singularity and is not analytic (i.e.
continuous, with a finite derivative) at the origin, z = 0, where the radial velocity is
undefined.

242
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

9.7.3 Point Vortices

This function represents an irrotational vortex located at a >0

point within the flow (nominally located at the origin, =const.
unless the position is shifted by biasing the complex
variable, z — see Section 9.8.1). For a point vortex
located at the origin, the complex potential function is =const.
given by:

  
( z )  i ln z  i ln r  i    i ln r     i (9.31)
2 2 2
So,
 
 ln r and    (9.32)
2 2

Here  is a constant reflecting the rotational “strength” of the vortex. Note that due to
radial symmetry we have again represented the complex variable, and its constituent
stream and potential functions, in cylindrical coordinates. Here, the cylindrical velocity
components are given by:


v r  0 , v  (9.33)
2r

The rotational direction of this simulated vortex corresponds to the sign of . If  > 0,
the rotation is counter-clockwise; alternatively, if  < 0, the rotation is clockwise. The
right-hand rule gives the sense of rotation.

Basically, Eq. 9.31 simulates an irrotational vortex with a singularity at the origin, since
(z) is undefined, and thus not analytic, at z = 0. To demonstrate this, we differentiate
Eq. 9.31, giving us the conjugate complex velocity as:

d    i 
 W  i  i e  i cos   i sin   sin  i cos 
dz 2 z 2 r 2 r 2 r
So,

W  u  iv  sin  i cos   u   sin and v

cos 
2 r 2 r 2 r

Thus, for a finite , we note that W and W are infinite at z = 0 (r = 0). Since the flow
field outside of the origin (z = 0) is irrotational, this implies that all the vorticity for this
function is concentrated at the singular point, z = 0, such that a point vortex violates the

requirement of   V  0 at the origin. We will discuss later how the insertion of one or
more point vortices can add rotation to a simulated flow, which allows for the simulation
of lifting body behavior.

243
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

9.7.4 The Doublet (dipole)

The doublet is a mathematical construct that represents the linear superposition of a
source of strength + Q and a sink of identical, but opposite, strength, -Q, which are
brought together at the same point (in this case, the coordinate origin). To understand
how we develop this hypothetical function, consider the schematic below, which shows
the general streamline pattern for a source +Q located at x = -a, and a sink -Q located at
x = +a.
y
Source of +Q Sink of -Q

a a
The complex potential function for this superposed source/sink combination is given by
(see Section 9.7.6 to see how we use superposition and biasing to create this function):

Q Q Q za
s (z)  lnz  a   lnz  a   ln  (9.34)
2 2 2  z  a 

We designate the strength of a doublet as s = Qa/, where s is held constant while the
value of a is shrunk to zero in the limit [i.e. the source and sink are brought together in
such a way that they produce a new complex potential function, (z)]. So, substituting
s
Q into Eq. 9.34, rearranging, and taking the limit as a0, we obtain:
a

 lnz  a   lnz  a  0
( z )  lim s ( z )  s lim 0
a 0 a 0
 2a 
which is indeterminate. To establish the limit, we apply L’Hospital’s Rule,
differentiating with respect to a:

 1 1 
 z  a   z  a    z  a   z  a  2z 2z s
( z )  s lim   s lim  s lim 2 s 2 
a 0
 2  a 0
 2z  a z  a  
  
a 0 2 z  a 2
2z z
 

Thus, the complex potential function for a doublet (centered on the origin) in Cartesian
coordinates is:

244
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

s s s  x  i y  sx  i sy sx  sy
( z )       2 2
 2 2
and   2 (9.35)
z x  iy x  iy  x  i y  x  y x y x  y2

Or in cylindrical coordinates is:

s s s ei s s s
( z )   i   cos   i sin    cos  and    sin (9.36)
z re r r r r

The shape of the streamlines ( = constant) for a doublet is established by noting that if
we let  = constant, from Eq. 9.35 we can write:
 sy
 = constant (9.37)
x  y2
2

Rearranging Eq. 9.37 gives:

2 2
2 sy2 2  s   s 
x y   0 or x   y     (9.38)
  2   2 
s
Equation 9.38 represents a set of nested circles of diameter D  with their centers at

s
x  0, y  . All the streamlines will be tangent to the x-axis at y = 0, as shown in
2
figure 9.2 below. Note that the  > 0 and  < 0 streamlines give reflected circle patterns
about the x-axis.

Flow pattern for s > 0 x

Streamlines ( = constant) are circles

y

Flow pattern for s < 0 x

Figure 9.2 Doublet streamline patterns for s > 0 and s < 0.

245
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

Again, like the point source and point vortex, the doublet flow displays a singularity and
is not analytic at the origin, with the velocity at the origin being similarly undefined.

The velocity components for a doublet flow are given by:

s s
vr   2
cos  and v  2 sin (cylindrical coordinates) (9.39)
r r

u

s x 2  y2  and v  
2sxy
(Cartesian coordinates) (9.40)
x 2
y 
2 2
x 2
 y2 
2

Note that since we asymptote the source/sink pair to z = 0 along the x-axis, the function
s
 (z)  has it axis along the x, or real axis. However, if we asymptote the source/sink
z
s
pair to z = 0 along the y (imaginary) axis, the corresponding doublet becomes  (z)  i ,
z
with streamlines appearing as nested circles tangent to the y axis, as shown below.
y

Flow pattern for s > 0 x

 = constant

Although a bit cumbersome, and generally not worth the effort, one can also configure
the complex potential function for doublets with axes lying off the primary axes of the
complex plane.

While not a particularly useful function by itself (in fluid mechanics), when the doublet is
superposed with a uniform flow, the resulting pattern is an exact simulation of inviscid,
two-dimensional flow over a cylinder, as we will show in Section 9.8.2.

9.7.5 Corner and Wedge Flows

These are flows which simulate the flow either around a corner, or impinging upon a
wedge shape. The complex potential function for both of these types of flow is given by:

A n A
z  x  i y 
n
( z )  (9.41)
n n

Here, A and n are constants. As we will illustrate, A controls the strength and direction
of the simulated flow; the type of flow pattern that this function simulates depends on the

246
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

magnitude of the exponent, n. In essence, this potential function reflects the conformal
mapping of a uniform parallel flow (Section 9.7.1) onto either a larger or smaller plane.

Because explicit velocity components for u and v can only be determined in the Cartesian
coordinate system for integer powers of n, it is often more functional to represent this
complex potential function using cylindrical variables as:

A i n Ar n in A r n
( z ) 
n
 
re 
n
e 
n
cos n  i sinn (9.42)

which gives:

A rn A rn
 cos n and   sinn (9.43)
n n
For these flows, streamlines are lines of constant , with  = 0 representing the bounding
geometric streamline. The bounding streamlines for  = 0 are simulated by straight lines
passing through the origin at angles of  = 0 and  = /n. Of course symmetry also
dictates that lines along the angles  = 2/n,  = 3/n,  = 4/n, …… can also represent
a “bounding” streamline. Which of these bounding streamlines we employ depends on
the particular portion of the velocity plane that is relevant to our simulation, as we will
demonstrate.

Note that the complex velocity for this function is given by:

d
 W  Az n 1  Ax  i y   Ar n 1e i ( n 1)  Ar n 1 cos ( n  1)  i sin ( n  1) (9.44a)
n 1

dz

W  u  iv  Ar n 1cos(n  1)  i sin(n  1)

So that

W  u  iv  Ar n 1cos(n  1)  i sin(n  1) (9.44b)

Thus, the Cartesian velocity components (in terms of cylindrical variables) are:

u  Ar n 1 cos(n  1) and v  Ar n 1 sin(n  1) (9.45)

However, the cylindrical velocity components are given by:

1  
vr   A r n 1 cos n and v    A r n 1 sinn (9.46)
r  r

Examining Eqs. 9.45 and 9.46 indicates that the value of the exponent, n, markedly

affects the value of the velocity, V , at and near the origin, r = 0. Note that at r = 0,

247
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9


when n > 1, r n 1  0 and V = 0 at the origin, which is a stagnation type of flow.

However, at r = 0 with n < 1, r n 1   and V will become infinite at the origin. We will
discuss and illustrate this curious variance in behavior, and the implications for the type
of flow this function simulates in the following two sections.

A n
As we mentioned above, ( z )  z maps (termed conformal mapping) a uniform flow
n
into a smaller or larger region of the complex flow plane. For n > 1, the flow is mapped
into a smaller portion of the complex plane, simulating a stagnation type flow behavior.
However, for n < 1, the flow is mapped into a larger portion of the complex plane,
simulating flows which have infinite velocity at the origin.

The reader should also note two further points. First, when n = 1, the complex potential
function of course reduces to (z)  Az , which is just a uniform, parallel flow, as
described in section 9.7.1. Secondly, the sign of the constant A dictates the direction of
the simulated flow impingement. For A > 0 the flow will initiate from the left to right
(depending on the value of n); for A < 0 the flow will be reversed, and initiate from the
right to left.

9.7.5.1 Concave Corner or Wedge Flow (stagnation flow)

These flows are represented by exponent values n  1, and simulate inviscid flow patterns
characteristic of a concave corner or a wedge, as shown. Note that this can be viewed as
a wedge flow because of the symmetry of the solution about the x axis, which simulates
the junction of two symmetric concave corners, as shown in figure 9.3b.

V=0

V=0
(a) Concave corner (b) Wedge

Figure 9.3 Illustration of a simulated (a) concave corner flow, and (b) wedge flow

These types of flows are characterized by the generation of a stagnation point at the
origin, and characterize a broad variety of impingement flows.

Example: Let n = 2: Simulates an impinging flow in a corner or normal to a flat plate.

A 2 r2
Using Eq. 9.41, (z)  z , giving a stream function   A sin2 .
2 2
To establish the boundaries of the geometry, we set  = 0, which means that:

sin2  0  2  0, , 2, 3,

248
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

 3
such that   0, , , ,…. All represent possible body boundaries, as shown in figure 9.4.
2 2

Note: A< 0, as shown y

y

=0 =0
=/2  =/2
x

=0 =0 =0

=0 =3/2 =0

(a) Concave, 90 corner (b) Flat plate impingement

Figure 9.4 Simulated flow impinging in a 90 corner, and on a flat surface.

The selection of the boundaries in figure 9.4, since the flow is symmetric within each
quadrant, depends on the type of flow to be simulated. Figure 9.4a represents an imping
flow in one corner, or quadrant of the flow field. Since viscosity is not an issue for these
simulated flows, the combination of two adjacent 90 corners (concave corners), like that
shown in figure 9.4a, will simulate the impingement upon a semi-infinite flat plate
oriented perpendicular to the impingement direction, as shown in figure 9.4b.

To establish the velocity components for the flow depicted in figure 9.4, we differentiate
the complex potential function:

d
W  Az  Ax  iAy  u  iv
dz
u  Ax
v   Ay

Note that the sign of A reflects the direction of the simulated flow. If A < 0, the
streamlines will be as shown in figure 9.4 (impinging from the right); if A > 0, the
streamlines would be reversed, exiting to the right, as if the flows are converging from
top (and bottom for figure 9.4b), and deflecting to the right (i.e. the arrows would be
reversed in figure 9.4).

Example: Let n=3/2, simulating impinging flow on a shallower corner or pointed wedge.

3
2A 2
For this flow, Eq.9.41 gives  (z)  z , with a corresponding stream function,
3
3
2A 2 3
 r sin  .
3 2

249
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

To establish the boundaries of the geometry, we set  =0, which means that:

3 2 4
sin   0    0, , , 2 (  0)
2 3 3

The possible boundaries represented by this flow are shown in figure 9.5.

Note: A < 0, as shown y

y

=0 =0
 = 2/3  = 2/3
x

=0 =0 =0

=0  = 4/3 =0

(a) Concave corner (b) Pointed wedge

Figure 9.5 Simulated impinging flow in (a) a 120 corner, and (b) a pointed wedge.

Note that these flows are again symmetric between the upper and lower half of the
complex plane. Here the flow simulated in figure 9.5a utilizes only the upper half of the
plane to simulate a concave ( < 120) corner. Again, the combination of the symmetric
flow patterns for both the top and bottom planes will simulate the behavior of a pointed
wedge, as shown in figure 9.5b. The complex conjugate velocity for this flow in
Cartesian coordinates is:

1
d
W  Az 2  A x  iy  u  iv (9.47)
dz

Equation 9.47 is not functionally useful for establishing the velocity field, since the
function is implicit. However, in cylindrical coordinates we can establish, from Eq. 9.44a:

1  1
d i   
 W  u  iv  Ar 2 e 2  Ar 2 cos  i sin 
dz  2 2

such that we can develop explicit velocity component equations:

1 1
 
u  Ar cos and v  Ar 2 sin
2
2 2
Note that for r0, u = 0 and v = 0, so this is again a stagnation type of flow.

250
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

9.7.5.2 Convex Corner Flow (infinite corner velocity)

A n
For exponent values n  1, the complex potential function (z)  z simulates flow
n
patterns which are characteristic of what we will term a convex corner as shown in figure
9.6 below.

V=
Convex Corner
Figure 9.6 Schematic of inviscid flow over a convex corner.

These types of flows are characterized by bounding streamlines which simulate wall
angles of  > 180, resulting in an infinite velocity at the origin for the simulated flow.
Since real flows abhor infinite velocities, a real flow would adapt to these convex corner
flows through the generation of significant viscous forces, which would significantly
modify both the velocity and the pressure fields, both near and downstream of the corner
region. However, examination of these inviscid flows allows one to anticipate problems
that can be expected in real flow situations where rapid velocity changes occur, such as in
the vicinity of sharp, convex corners. Such simulations can also assess (via modified
inviscid solutions) ways that such regions might be avoided, or minimized in real flow
situations. Interestingly, the flow problems encountered for sharp, convex corner
geometries are similar to the stress concentration problems which accompany sharp
corner geometries in the cutting or fabrication of metals and other solid materials.

Example: Let n=2/3 and n=1/2, respectively, simulating flows over:

(a) a 90 step (a 270 corner), and

(b) a 360 flow around the end of a finite plate.

y y

Note: A<0, as shown =0

=0
x x
=0 =0
=0 =
=0
=3/2
(a) Convex, 270 Corner (n=2/3) (b) Convex, 360 Corner (n=1/2)

Figure 9.7 Simulated flow over (a) 270 and (b) 360 convex corners.

251
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

Again, using Eq. 9.41, these flows are simulated respectively by:
2
3A 3 1
 (z)  z and (z)  2Az 2
2
with corresponding stream functions given by Eq. 9.43 as:
2
1
3A r 2 3 
 sin and   2A r 2 sin
2 3 2
The corresponding boundaries of the geometries, for  = 0, are:

2 3
sin  0    0 and for a 270 corner
3 2


sin  0    0 and  for 360 flow around a finite plate
2

Since all flows for n < 1 will yield implicit velocity fields in Cartesian coordinates, we
again express the Cartesian velocities in cylindrical coordinates using Eq. 9.45 as:

1 1
   
u  Ar 3
cos and v  Ar 3 sin for a 270 corner, and
3 3

1 1
   
u  Ar 2
cos and v  Ar 2 sin for 360 flow around a plate.
2 2

Here, the velocity field is dominated by the inverse powers of r, which results in an
infinite velocity at r = 0, and very high accelerations near the origin.

9.7.6 Biasing the Origin Location for Elementary Functions

Except for the uniform, parallel flow of section 9.7.1, all of the other elementary
functions we examined had their origin located at x,y = 0,0 (z = 0) in the complex plane.
However, when performing a linear superposition process it is often useful, and
necessary, to locate the functional origin of a superposed elementary function at a
location other than the origin of the complex plane. To locate an elementary function at a
position removed from the complex origin, we simply substitute the modified complex
variable (z'), biased to the desired origin for the function (zo), such that z'=z-zo. For
example, a source having its origin at zo would be written:

Q Q
( z )  ln z  ln z  z 0 
2 2

252
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

This seems simple enough, but this can tricky when trying to establish explicit stream and
potential functions, from which to derive the velocity components. If we substitute
z=x+iy and zo=xo+iyo into the complex potential function for the source, we obtain:

Q Q
( z )  ln z  lnx  x 0   iy  y0 
2 2
for which  and  are implicit. To allow the determination of explicit functions for 
and , we need to employ cylindrical coordinates biased to the origin of the function, xo,
yo, where z  r ei , and

 y  y0 
r  x  x o 2  y  yo 2 and   tan 1  
 x  x0 

Thus, the complex potential function can be written as:

Q Q Q
( z )  ln z  ln r ei  ln r  i
2 2 2
Q 1  y  y 0 

ln x  x o   y  yo   i tan 
2 2
      i
2   x  x0 

And thus,

Q Q  y  y0 
 ln x  x o 2  y  yo 2  and  tan 1  
2   2  x  x0 

Clearly, this process makes the determination of the appropriate functions a bit more
cumbersome. The determination of the velocity components for these off-origin
locations can be obtained either by differentiation of the complex potential function or
through the appropriate differentiation of the component functions  and , whichever is
simpler. For example, to obtain u for the above example of a two-dimensional source
located at zo=xo+iyo we can write:

d Q 1 Q 1 Q x  x 0   iy  y0 
  

dz 2 z  z 0  2 x  x 0   iy  y0  2 x  x 0 2  y  y0 2 
Q x  x 0   iy  y0 
  u  iv  W
2 r2
thus,

Q x  x 0  Q y  y 0 
u and v
2 r2 2 r 2

Alternatively, we could differentiate  or  via the Cauchy conditions:

253
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

 Q  
ln x  x o   y  y o  
2 2
u 
x 2 x 
 

 
1

Q x
x  x o   y  y o  2
2 2
Q x  x o  Q x  x o 

2

x  x o   y  yo 
2 2 2
1
 2

2 x  x o   y  y o 

2

2 r  2 
and similarly,

 Q y  y 0  Q y  yo 
v 

y 2 x  x 0   y  y0 
2 2

2 r  2 
Which, of course, yield the same functional expressions for u and v.

Note that expressions for  and  at zo for point vortices are quite similar to the
expressions for the source/sink flows, and thus the expressions for position-biased stream
and potential functions, as well as the velocity components, will have a form similar to
the point source/sink flows.

However, the doublet is a bit different. The location of a doublet at zo yields a complex
potential function of:

s s sx  x 0   i sy  y0  sx  x 0  sy  y0 

( z)     i
z  z 0 x  x 0   iy  y0  x  x 0   y  y0 
2 2
r 2
r2

For the velocity we note:

sx  x 0   i y  y0 
2
d s s
    W  u iv
dz z  z0  x  x 0   iy  y0 
2 2

x  x 0 2  y  y0 2
2

From which one can show,

u

s x  x 0   y  y 0 
2 2
 and v
2 s x  x 0 y  y 0 
r4 r4

Considering corner/wedge flows, the complex potential function for these geometries
centered on zo is given by:

A
( z )  z  z0 n  A x  x 0   i y  y0 n
n n
or
A rn
( z )  cos n  i sinn
n
where

254
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

 y  y0 
r  x  x o 2  y  yo 2 and   tan 1  
 x  x0 

Either of these formulations is cumbersome, particularly for establishing the boundary of

simulated bodies via a constant stream function, . For the Cartesian form, the stream
and potential functions can only be expressed as explicit algebraic functions of x and y
for integer values of n. As we will see, establishing explicit expressions are needed to
establish the bounding streamlines for a simulation, and thus the body shape. However,
when dealing with fractional values of n, the cylindrical form of the complex potential
function must be used, which renders the stream and potential functions trigonometric,
such that positional expressions for x,y lying on bounding streamlines are generally
implicit, and must be dealt with iteratively.

9.8 Superposition Flow Simulations

We have shown in sections 9.2 and 9.3 that both  and  are solutions to the Laplace
equation. Since the Laplace equation is a linear differential equation, that means that all
solutions for  and  are directly additive by a process of linear superposition. That is,
the sum of a set of separate solutions is still a solution to the equation. As we mentioned
in section 9.5, this process of linear superposition is termed the indirect or inverse
method, since it does not allow us to directly determine the velocity field for a specified
geometry. Instead, we employ this indirect approach to superpose a set of judiciously
selected flow functions ( or ), examine the velocity patterns they collectively simulate,
and then associate those patterns with flows around/through geometries which
approximate the bounding streamlines. Taken to its extreme, this process can be used
iteratively to develop flows for somewhat complicated geometries through the optimized
superposition of a large number of elementary functions. In this section we illustrate the
basic superposition approach to simulate several simple geometries. However, having
called them “simple,” one may often find that the reduction of the superposed solutions to
equations describing the respective body shape, velocity, and pressure distributions can
be somewhat challenging to do in closed form, as we will demonstrate.

9.8.1 Flow Around a Circular Cylinder

A very common flow that is readily simulated using superposition techniques is a

uniform flow passing around a circular cylinder (i.e. like flow around a wire or a rod).

In order to simulate the inviscid flow over a cylinder, as shown in figure 9.7, we combine
a uniform flow (Section 9.7.1), (z)  Uz , with a doublet function (Section 9.7.4),
s
 (z)  , giving a combined complex potential function of:
z

s
 (z)  Uz  (9.48)
z

255
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

R
 + x

Figure 9.7 Inviscid flow over a cylinder, simulated by superposition of a uniform

flow and a doublet.
As shown in figure 9.7, here we assume that: (1) U > 0, which gives the left-to-right
uniform flow shown in figure 9.7, and (2) s > 0, such that the doublet flow will emanate
outward, to the left of the origin, as shown. This combination creates a simulated flow
pattern where the uniform flow and the doublet impinge upon each other along the x-axis,
creating a stagnation point reflecting the leading edge of a simulated cylinder.
To determine the stream and potential functions for this flow, we can use either Cartesian
or cylindrical coordinates. In the following, we derive the stream and potential functions,
and the velocity components in both systems. Thus, using z  rei , Eq. 9.48 becomes:
s
( z)  Ure i  e i
r
s
 Ur cos   i sin  cos   i sin
r
(9.49)
 s  s
  Ur   cos   i  Ur   sin
 r  r

 

Alternatively, using z  x  i y , we obtain:

( z )  Ux  i y  
s
 Ux  i y  
s x  i y 
x  i y  x  i y  x  i y 
s
 Ux  i y   2 x  i y 
x  y2  
 s   s 
 x U  2   i y U  2  (9.50)
 
x  y2    
x  y2  

 

For the cylindrical velocity components, we have (using the Cauchy relations):

1  1  s  s
vr    Ur   cos    U  2  cos  (9.51a)
r  r  r  r 

  s
v      U  2  sin (9.51b)
r  r 

256
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

And in Cartesian coordinates, the complex velocity components are:

 s 2sy2 s  2y2 
u U 2  U 1   (9.52a)
y x  y2 x 2  y2   2
x 2  y2  x 2  y2  

 2sxy
v  (9.52b)
x 
x 2  y2 2

A comparison of Eqs. 9.51 and 9.52 suggests that it is generally easier to perform our
analysis for this configuration in cylindrical coordinates due to the radial symmetry of
this flow.

Using Eq. 9.51, we recognize that if the flow is on and around a cylinder, then at the
surface of the cylinder the radius will be a constant of r = R, and the radial velocity, vr, on
this surface will be zero (i.e. the velocity will be tangent to the cylinder surface, with v 
the only relevant velocity component).

Thus, we set Eq. 9.51a equal to zero, for r = R:

R
 s 
v r r  R  0   U  2  cos   
 R 
Stagnation
Which indicates that: Points

s s  3
U 0  R or cos   0    , , (9.53)
R2 U 2 2

The result for  in Eq. 9.53 is not particularly useful alone. However, Eq. 9.53 indicates
s
that the radius of the simulated cylinder, R  , is a function of the respective ratio of
U
the strengths of the doublet and the uniform flow, as might be expected. Changing the
strength of the doublet, or uniform flow, will change the simulated diameter of the
cylinder. However, as we will see, the dimensionless characteristics of this flow
(e.g. v U, vr U, Cp  p  p  12  U2 ) are invariant with changes in the flow or doublet
strength.

To determine the streamline that defines the boundary of the cylinder, we assess the value
s
of the streamline for r  R  :
U

 s
   Ur   sin
 r

257
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

 
 s 
 r R


s
 UR   sin  U
R 
s
U

s 
sin   sU  
sU sin  0
 
 U

Thus, the bounding streamline will be given by  = 0. To establish the extent and shape
of the bounding streamline, we now set  = 0 in Eq. 9.49, and solve for the streamline
boundaries.

 s
  0   Ur   sin
 r

s
So, either Ur   0 or sin  0 along the   0 streamline.
r

s
or r  R and   0 or 
U

Therefore,  = 0 occurs either: (1) on the cylinder surface (r = R), or (2) along the x-axis
for r > R, as shown in figure 9.8.
=0
y
=0 =0

x
 

Stagnation
Points

Figure 9.8 Location of the  = 0 streamline, and the respective stagnation points
lying on that streamline.

The stagnation points on the cylinder will, of course, occur where V  0 or where
s
v r and v  0 . So, if we let R  , which assures that v r  0 , we then set v  0 in
U
Eq. 9.51b and solve for the resulting angle:

 
 s 
v   U   sin  2U sin  0  sin  0  stag  0 and 
  s 
  
 U 

258
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

Thus, two stagnation points occur on the x-axis at both the leading and trailing edge of
s
the cylinder (i.e. x stag   R   and y stag  0 ).
U

The portion of the   0 streamline that does not lie on the cylinder surface, lies along the
x-axis, at either  =  or 0 for r > R, or along |x| > R for y = 0. Along the x-axis, the
velocity is simply the u component of the complex velocity, Eq. 9.53a, with y = 0. \

s
u  0  U  along the x-axis
x2

Thus, for x  R , we have:

u  0  U

Therefore, the velocity is essentially constant well upstream and downstream of the
cylinder, representing an impinging (and departing) uniform flow well away from the
cylinder body.

To determine the maximum velocity on the cylinder surface, we note that v  is the
s
velocity on the cylinder surface (since vr = 0 on the cylinder). For r  R  , Eq.
U
9.51b becomes:

v   2U sin  (9.54)

The maximum velocity will be where the derivative of Eq. 9.54 along the surface (with
respect to ) is zero, so

dv 
 2U cos   0    90o or 270o
d

Thus, the maximum velocity on the cylinder is v   2U (we could also have inferred
this by inspection). Note that the negative sign in Eq. 9.54 simply indicates that the
velocity moves in the negative  direction (clockwise) for 0 <  < , and positive 
direction (counter clockwise) for  <  < 2.

To establish pressure changes, both within the flow field and along the cylinder surface,
we apply the Bernoulli equation, Eq. 7.11. Note that since the flow is inviscid and
irrotational, Bernoulli applies either along a streamline or between any two points within
the flow field.

259
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

So, applying the Bernoulli equation between a point far upstream of the cylinder (r >> R),
to any other point within the flow field, gives:

P  12 V 2  P  12 U 2 (9.55)

In Eq. 9.55, P is the local pressure and V is the local velocity at any point within the
irrotational flow. P is the static press far upstream of the cylinder where the flow is
uniform at velocity U.

Now, on the cylinder surface we have V 2  v2  ( 2U sin)2 . Substituting this V 2 into
Eq. 9.55, and rearranging, gives:

 
P  P  12 U 2  12 V 2  12 U 2  12  4U 2 sin2   12 U 2 1  4 sin2   

P  P  12 U 2 1  4 sin2  
Defining a non-dimensional pressure coefficient, Cp, gives:
P  P
Cp  2
 1  4 sin2  (9.56)
1
2
U

The general shape of Cp() on the cylinder is shown in figure 9.9 below.
Cp vs 
1
0
-1
CP

-2
-3
0 30 60 90 120 150 180

 (degrees)

Figure 9.9 General behavior of the surface pressure coefficient for inviscid flow
P  P 2
over a cylinder, where Cp 
1 U 2
 1  4 sin  .
2

Now, consider the forces that act on the cylinder due to the imposed pressure forces,
where drag is a force parallel to the impinging flow (here, the x-direction), and lift will be
a force normal (perpendicular) to the impinging y
flow direction (here the y-direction). 
dA  Rdˆir
Thus,

x
Drag   dFx    PA x

260
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

Where,


dA  dA x ˆi  dA y ˆj  Rdˆir  Rd ˆi cos   ˆj sin  
Thus, dA x  R dcos   differential area projection in x-direction
and dA y  R dsin  differential area projection in y-direction

Thus,

2
  
Drag    P  12 U 2 1  4 sin2  R cos  d
0
2
sin3 

  P  U R sin 0
1
2
2
 2 2
 2U R
3 0
0

This curious result is known as d’Alembert’s paradox, and applies only for a truly
inviscid flow (which, in reality, does not exist). One can observe, from the plot of Cp in
figure 9.9 that the reason for the absence of pressure drag is that the pressure forces are
balanced between the front and rear of the simulated cylinder---in a real flow there is no
such balance of the pressure forces, as we will learn and discuss later in Chapters 14 and
15.

To develop a non-zero drag force on a body, which our experiences and experiment
indicate must occur, requires the presence of fluid viscosity. Viscosity gives rise not only
to shear forces at the body boundary, but is also responsible for the development of what
is termed flow “separation,” wherein a fluid flowing along a body will be brought to rest,
such that the surface streamline subsequently separates away from (i.e. moves away
from) the body surface. Such flow separation creates a region of stagnant or recirculating
fluid adjacent to the body, which strongly modifies both the velocity field around the
body, and the pressure distribution imposed on the body. Thus, drag for real flows can
result from both surface shearing forces and unbalanced pressure forces acting on the
body. However, our assumption of inviscid behavior yields a fluid drag of zero on this
and other (as we will see) simulated body shapes---a result that runs counter to our
experiences. This discrepancy between the zero drag predicted by an inviscid analysis
and the finite drag that exists for all real flows is, of course, the basis for d’Alembert’s
paradox.

Note that the lift forces acting on the simulated cylinder (i.e. the summation of the forces
acting normal to the impinging flow direction) can also be calculated for the simulated
cylinder. The result, not surprisingly, is:

Lift =  dFy    PdA y  0

261
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

This should not be unexpected, since the pressure forces on the simulated cylinder are
balanced top to bottom, as well as back to front, because of the symmetric pressure
distribution indicated in figure 9.9. However, as we will show in the following section,
the addition of rotation (circulation) to the simulated flow will modify this vertical
pressure symmetry, and result in a non-zero lift force.

9.8.2 Flow Around a Rotating Cylinder

This example is similar to the analysis of a stationary cylinder in section 9.8.1. However,
we now simulate rotation of the cylinder by the additional superposition (addition) of a

point vortex function,  z   i ln z , to the superposed uniform flow and doublet
2
functions employed in section 9.8.1 (i.e. uniform flow + doublet + vortex).
y >0
R
 + x +

Here we assume U > 0, s > 0, and  > 0, which will simulate a cylinder rotating counter
clockwise, with an impinging flow from left to right. The corresponding complex
potential function for this flow is:

s 
z   Uz   i ln z (9.57)
z 2

yielding a stream function (in cylindrical coordinates) of:

 s 
 r ,    Ur   sin  ln r (9.58)
 r 2

For this configuration, the radial velocity is the same as for a non-rotating cylinder

1   s
vr    U  2  cos  (9.59)
r   r 

However, the azimuthal velocity includes a new component that accounts for the
magnitude of rotation,

  s 
v      U  2  sin 
r  r  2 r
Rotation
(9.60)
Component

262
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

As we did for the non-rotating cylinder, we set v r  0 and solve for the radius of the
s
simulated body. The result is R  , which is the same value we obtained for the
U
non-rotating cylinder.

s
The bounding streamline for the body is obtained by setting r  R  in Equation
U
9.58, which gives:

 
 
 r R 
 U

s
U

s 
s 
sin 

2
ln
s
U
  
sU  sU sin 

2
ln
s
U
 
 U

 s
 r R   ln (9.61)
2 U

Note that this bounding streamline varies with the strengths of the uniform flow, the
doublet, and the rotation. That the bounding streamline is not zero is interesting, since
this indicates that the streamlines above and below the real axis will not be symmetric,
but will skew to one side of the x-axis or the other, depending on the direction of cylinder
rotation.

To determine the position of the stagnation points on the rotating cylinder, we let r = R
(which satisfies v r  0 ), set v  0 (Eq. 9.60), and solve for the stagnation angles:

 s  
v r  R    U  2  sin  0
 R  2 R
 U
 U  Usin  0
2 s
or

sinStag  (9.62)
4 Us

From Eq. 9.62, we can establish the angle(s), Stag, on the cylinder where the stagnation
points occur. As indicated in figure 9.10, depending on the collective strengths of the
uniform flow, the doublet, and the rotation, we may theoretically have two, one, or no
stagnation points lying on the cylinder. Note that the direction of the cylinder rotation
dictates whether the stagnation points occur on the upper half (when  > 0) or lower half
(when  < 0) of the cylinder.

case “c”
+1 case “b”
sinStag case “a”
263 0
case “a”
-1 case “b”
case “c”
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

(a) if sinStag  1  2 stag. pts.

(b) if sinStag  1  1 stag. pt.
(c) if sin Stag  1  No stag. pts. (on the r  R surface)

Examples ( > 0) :
(a)

(b)

(c)

Figure 9.10 Illustration of the number of stagnation points occurring on a rotating

cylinder, and the general streamline behavior, depending on the
strengths of U, s, and  in Eq. 9.62. (a) two stagnation points, (b) one
stagnation point, and (c) no stagnation points (on the cylinder surface).
Here,  > 0, so stagnation points are in upper half of the plane. If  < 0,
the stagnation points would occur in the lower half of the plane (and the
streamlines would be mirror reflections—top to bottom—of the above
patterns.

Notice that for U > 0 and s > 0, the stagnation points will skew toward the upper surface
( >  > 0) for clockwise rotation ( > 0, as shown), and toward the lower surface (2 > 
> ) for counter-clockwise rotation ( < 0). Also note that when the stagnation angle
equation exceeds unity, sin Stag >1, this indicates that the stagnation point cannot occur
on the body surface (r = R), but must occur outside of the simulated body, as illustrated in
figure 9.10(c) above. Such a stagnation point occurring within the fluid is a particular
type of critical point termed a saddle point (with flow moving toward and then away from
the point). A critical point is essentially a point within a flow field where the slope of the
streamline is indeterminate. For example, a stagnation point qualifies as a critical point
since the equation of the streamline intersecting a stagnation point is dy/dx=0/0, or
indeterminate.

The velocity on the surface of the rotating cylinder, where v r  0 , is given by:

264
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

 
 s     
v s    U   sin   2U sin   U  2 sin   (9.63)
R
U  s 2R 2R  2 Us 
 U 

The pressure distribution on the surface, determined by substituting Eq. 9.63 into the
Bernoulli equation is:

 v2 
P  P  12 U 2 1  2 
 U 

    
2

P  P  U 1    2 sin 
1
2
2
 
  2RU  

Here, P is again the static press far upstream of cylinder where the flow is uniform at
velocity U. Defining a non-dimensional pressure coefficient, Cp, gives:
2 2
P  P      
Cp  1 2
 1    2 sin    1    2 sin   (9.64)
2
U  2RU   2 Us 

Cp for Rotating Cylinder Cp for Rotating Cylinder

=0.25 =0.5
2 2
Cp (upper) Cp (upper)
1 Cp (lower) 1 Cp (lower)
Cp No Rotation Cp No Rotation
0 0
0 0.5 1 1.5 2 2.5 3 3.5 0 0.5 1 1.5 2 2.5 3 3.5
-1  -1 
Cp
Cp

-2 -2

-3 -3

-4 -4

-5 -5

-6 -6

(a) (b)

Figure 9.11 Comparison of Cp vs.  (from Eq. 9.64) for (a) = 0.25
2 Us

and (b) = 0.5. Cp with no rotation ( = 0) is shown for
2 Us
comparison.

The general shape of Cp() on the upper and lower surface of the cylinder is shown in
figure 9.11 for two cases of increasing rotation.

265
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

Although the presence of rotation skews the streamlines, and creates a pressure
differential between the upper and lower surfaces, as figure 9.11 shows, the pressure
distribution between the front and back of the cylinder remains symmetric, so like the
non-rotating cylinder there will be no drag force on the cylinder.

Drag force =  dFx    PdA x  0

However, it is clear that the variance in surface pressure between the upper and lower
surfaces must create an unbalanced lift force on the cylinder normal to the direction of
the impinging uniform flow. This lift force can be calculated as:

Lift force   dFy    PdA y    P  12 U 2C p  dA y

where, dA y  R dsin is the differential area projecting in the y-direction.

Thus,
2
  2 sin 2 
Lift Force    P  12 U 2  12 U 2  4 sin2    2 2 2  R sind
0  RU 4 R U  

which, yields the rather simple result that:

U
Lift Force =    U (9.65)

Equation 9.65 indicates that a counter-clockwise rotation (  > 0 ) will create a lift force
in the negative y-direction (downward). From the pressure distribution shown in figure
9.11, this is clearly a result of the generation of a lower pressure on the bottom surface
and a higher pressure on the upper surface of the cylinder.

Now note that the circulation around the cylinder (from Eq. 8.9) is given by :

 2 2
  
   V  ds   v Rd     2U sin   Rd   (9.66)
0 0
2R 

Since our lift result in Eq. 9.65 showed that lift =  U , Eq. 9.66 indicates that the lift
on the cylinder is directly proportional to the circulation about the cylinder. It turns out
that this result that lift is proportional to the circulation about a body, is generic, and
applies for all inviscid flows about bodies, including airfoils, regardless of the source of
the circulation.

9.8.3 Flow Around a Half-Body

One of the simpler superposition solutions would seem to be a uniform flow combined
with a source, which simulates flow over what is called a half body, shown in figure 9.12.

266
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

Flow ymax at x

-xs Q>
0 x

Figure 9.12 Schematic of flow over a half-body, simulated by the combination of

a uniform flow and a source.

While this is a simple superposition, the calculations to establish the body shape are
much less straightforward than we encountered in sections 9.8.1 and 9.8.2 using the
superposition of a uniform flow and doublet to simulate flow over a cylinder. For the
present simulation we superpose a uniform flow of strength U > 0 moving left to right,
given by z   Uz , and a source of strength Q > 0 located at z = 0, given by
Q
z   ln z . Correspondingly, the superposed complex potential function is:
2

Q
z   Uz  ln z (9.67)
2
Here, we have two options: either to employ Cartesian variables or cylindrical
variables—or both. We will expand  in both variables, and consider which better
serves our purposes.

Expanding  using z  r ei we have:

Q  Q   Q 
 z   Uz  ln z   Ur cos   ln r   i  Ur sin      i (9.68)
2  2   2 

Whereas, expanding using z  x  i y we have:

Q  Q   Q  y 
z   Uz  ln z   Ux  ln x 2  y 2   i  Uy  tan 1      i  (9.69)
2  2   2  x 

Here, the complex velocity components are given by:

d Q 1
 U  W  u  iv
dz 2 z

267
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

and
Q 1
W  u  iv  U  where z  x  iy (9.70)
2 z

The respective Cartesian velocity components can be shown, in both Cartesian and
cylindrical variables, to be:

Q x  Q  cos  
uU  2   U
2    (9.71a)
2  x  y  2  r 

Q y  Q  sin 
v      (9.71b)
2  x 2  y 2  2  r 

To establish the shape of the body, we note that a stagnation point must occur on the
body, where the complex velocity, W (and so also W ), is zero. Thus, setting Eq. 9.70 to
zero, we have:
Q 1 Q
WU  0  z  x iy  
2 z 2U
Q Q
Thus, x stag   , ystag  0 or rstag  , stag   is the location of the stagnation
2U 2U
point for the body. Correspondingly, the value of the stagnation streamline, s , which
passes through the stagnation point, is given by:
0
Q Q
 s  Ur sin    (9.72)
2 2

Now, the location of the points along a streamline that describes the boundary of the body
(b = s, since s must lie on the body boundary) is given by setting  from Eq. 9.68 and
9.69 equal to Eq. 9.72:

Q Q
s    b  Urb sinb  b (cylindrical coordinates) (9.73)
2 2
or
Q y  Q
Uy b  tan 1  b   (Cartesian coordinates) (9.74)
2  xb  2

Solving Eq. 9.73 for rb and Eq. 9.74 for xb, we have:

Q   b 
rb  (9.75)
2 U sinb
and

268
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

yb
xb  (9.76)
  2Uy b 
tan 1  
  Q 

In Eqs. 9.75 and 9.76, the subscript “b” indicates that these equations define coordinate
pairs lying on the bounding streamline, and thus the surface, as shown in figure 9.13.
Either of these expressions allows us to determine the boundaries of the body, although
Eq. 9.75 is probably a bit more convenient.
y

rb
yb
b x

xb
Figure 9.13 The geometric relationship of xb and yb to rb and b.

From the general body shape shown in figure 9.13, we infer that the maximum thickness
of the half body must occur far downstream of the stagnation point, at either xb   or
b ≈ 0.

Employing rb from Eq. 9.75 in cylindrical coordinates to obtain the half-body thickness,
yb, we have:

Q
y b  rb sinb    b  (9.77)
2U

The maximum body thickness is obtained by substituting b = 0 into Eq. 9.77, giving:

Q
ymax  (9.78)
2U

Substituting Eq. 9.78 into Eq. 9.77, and dividing through by ymax, allows us to express
the non-dimensional half-body thickness as:

yb

  b   1  b (9.79)
ymax  

In Cartesian coordinates, the corresponding non-dimensional streamwise body coordinate

from Eq. 9.76 is:

269
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

 yb 
 
xb  ymax 
 (9.80)
ymax   y 
tan 1  b 
  ymax 

Substituting Eq. 9.79 into Eq. 9.80, yields the an expression for the corresponding non-
dimensional streamwise body coordinate as a function of b :

xb

  b  (9.81)
ymax  tan b 

The figure 9.14 shows the shape of the bounding stream line, determined by choosing
selected values of b, calculating yb/ymax and xb/ymax from Eqs. 9.79 and 9.81, and
plotting the resulting shape. Alternatively, one could select a series of yb/ymax values,
calculate xb/ymax from Eq. 9.80, and plot the shape directly. In either case, one has to be
careful in incrementing b or yb/ymax, since xb/ymax varies quite rapidly once yb/ymax > 0.6.

Yb/Ymax
0.5
Cp(x)
yb/ymax
Cp CD(x-xo)
CD
0
-0.5 0 0.5 1 1.5 2

-0.5

-1
x/ymax

Figure 9.14 Half-body surface stream line (b = Q/2), pressure coefficient [Cp(x)],
and cumulative drag coefficient [CD(x-xo)]

Another characteristic we can establish from this simulation is the pressure distribution or
pressure coefficient (Cp) over the body.

From the Bernoulli equation, we can write;

P  12 V 2  P  12 U 2 where V2  u 2  v2  WW (9.82)

Here, P is the static pressure of the upstream impinging flow ( = 0, or y = 0 and x <<0),
where u=U, v=0. Rearranging Eq. 9.82, we can write the non-dimensional pressure
coefficient as:

270
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

P  P V2
Cp   1  (9.83)
1
2
U 2 U2

Equation 9.83 allows us to express the pressure changes along the body non-
dimensionally, the way we did for the cylinder flow. We are again faced with which
variable system to employ to expand this expression. Using cylindrical variables, where
z  rei and z  rei we can determine V 2 as:

2
 Q 1  Q 1 QU 1 QU 1  Q  1
V 2  WW   U   U 
2
U    
 2 z  2 z  2 z 2 z  2  z z
2
QU ei QU ei  Q  1
U 2
   2 (9.84)
2 r 2 r  2  r

And substituting Equation 9.75 for rb, into Eq. 9.84 we have:
cosb

U 2 sin b ib U 2 sin 2 b  2 sin b cos b sin 2 b 

V2  U2 
  b 

e  e ib    U 2

 1 
 
 
2 
(9.85)
  b 2     b    b  

To simplify Eq. 9.85, we let b    b , and note that sin b  sin  b   sin b , and
cosb   cos  b    cosb , which lets us write:

 2 sin  b cos  b sin 2  b 

V 2  U 2 1    (9.86)
 b  2b 

Substituting Eq. 9.86 into Eq. 9.83 for the pressure coefficient, Cp, we have:

P  P 2 sin  b cos  b sin2  b

C p  b   2
  (9.87)
2 U
1
b  2b

Alternatively, we can write V 2 in Cartesian variables as:

2
QU x Q 1
2
V U  2
  (9.88)
2
 x y 2
 2
 2  x  y
2
  
Substituting into Eq. 9.88 from our expression for ymax, Eq. 9.78, (rearranging as
Q  2Uymax ) we have:

271
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

2 y max U 2 x 2
y max U2 1  y 2max  x 1 
V2  U2    U 2
1  2   (9.89)
  x 2  y2  2 x 2  y2

 2
 x y
2
    y max   

Substituting Eq. 9.89 into Eq. 9.83, yields a second equation for Cp on the bounding
surface as:
 xb 1
2  
y max  x b 1  y max  
2
pp
Cp x b , y b   1 2   2    (9.90)
2
U  x 2
b  y 2 
b   y max 

   2
 x   y 
2

 b    b  
 y max   y max  

To use Eq. 9.87, one must determine the locus of the surface using Eqs. 9.79 and 9.81 to
correlate Cp with the location on the body; for Eq. 9.90, Eq. 9.80 is used to correlate the
location of Cp on the body. Either approach works, but again care must be used in
incrementing b or yb/ymax. The previous Figure 9.14 shows Cp as a function of xb/ymax,
superposed with the body shape.

Note that the pressure force per unit depth in the streamwise direction, or pressure drag,
is given by:
yb  ymax

FD   P  P dA x   P  P dy  b
A yb o

Or in terms of a non-dimensional drag coefficient, CD:

y ymax

FD
 P  P dy  b y b  ymax
 dy 
 C p x b , y b  b
y 0
CD  1 2
 2
  (9.91)
2 U y max U y max  y max
1
2 y b 0 

To integrate Eq. 9.91 with respect to y requires that we substitute our expression above
for xb/ymax=f(yb/ymax) in Eq. 9.80 into Eq. 9.91 for Cp, which makes the integral in Eq.
9.91 much too complicated. Accordingly, we employ our expression for Cp(b) in Eq.
9.87, and note from Eq. 9.79 that we can write yb=f(b), where b    b , along the
surface as:

Q b y max  b dy b d b
yb     (9.92)
2U  y max 

Substituting Eq. 9.92 into Eq. 9.91, and noting that the appropriate limits of integration
are from b = 0 ( = ) to b =  ( = 0), we have:

272
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

  
1 b 1  2 sin  b cos  b sin 2  b 
 
  b 0  0 
CD  C  d     d b (9.93)
 2b 
p b b
b

Identifying that (not quite self-evident):

 2 sin  b cos  b sin 2  b  d  sin 2  b 

     
 b  2b  d b   b 

We can rewrite Eq. 9.93 and integrate, giving:



1 d  sin 2  b  1  sin 2  b  1 0 0
CD     d b         ? (9.94)
 0 d b   b    b  0    0 

Equation 9.94 is indeterminate at b = 0. To assess the integral, we apply L’Hospital’s

rule to the general result of Eq. 9.94 as b  0:

lim 
 sin 2  b 
  lim

d sin 2  b lim
2 sin  b
0
 b 0
 b  b  0 d b   b 0 1

Thus, the value of Eq. 9.94 is:

1
CD  0  0  0 (9.95)


Now the result of no drag indicated by Eq. 9.95 might seem a bit unusual, since one
would assume that the streamwise pressure force would be unbalanced for this geometry.
However, recall that inviscid flow around a circular cylinder also displayed no drag. The
process for the cylinder was a bit more obvious, since the pressure forces on the leading
edge of the cylinder were clearly balanced by the pressure forces on the trailing edge.
Here, the process is a bit more subtle, as shown in figure 9.14.

Note that when Cp > 0, a pressure drag force component is applied in the positive x-
direction. However, such a condition only exists for a short distance after the leading
edge (-.159  xb/ymax  -0.318). Thereafter, for xb/ymax  -.159, Cp < 0 on the bounding
surface, which results in a negative drag force, or suction force, acting in the negative x-
direction. Since CD  0 as x  , this implies that the initial pressure drag forces on the
leading portion of the body are identically balanced by the suction forces over the trailing
portion of the body. This is illustrated in figure 9.14 by the behavior of CD(x-xo), which
shows the cumulative drag as a function of x from the leading edge (where
xstag/ ymax = -0.318). The velocity on the body surface reaches a maximum, and thus Cp
reaches a minimum, at approximately xb/ymax = 0.33 (Vmax = 1.26U), and then decreases

273
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

back to zero as V  U as x . However, the deceleration of the flow is quite gradual
over the remainder of the body, which ideally extends to infinity.

As we will discuss later, flow deceleration and the accompanying pressure rise (known as
an “adverse” pressure gradient) are not well tolerated by real viscous flows. Real bodies,
in real flows, will of course be of finite extent, which requires that at some point the body
must either: (1) terminate at a finite thickness, or (2) taper back down to yb = 0, such as
occurs for airfoils. Such geometry changes are normally accompanied by local adverse
pressure gradients in the direction of the flow. The application of an adverse pressure
gradient to a real viscous flow will generally cause the flow to undergo a process of flow
“separation” (a process that was referred to earlier with regard to the cylinder flow) at
some point along the body. The consequence of such a flow separation is a sharp
divergence from the flow pattern behavior predicted by an ideal, inviscid flow, with the
consequent generation of non-zero pressure drag. We will discuss the development and
ramifications of flow separation on flow drag for real viscous flows later in Chapters 14
and 15.

Without offering further examples or proof here, we can comment that the general result
of zero pressure drag obtained for inviscid flow over a cylinder in sections 9.8.1 and
9.8.2, and in the present section for flow over a half body, applies in general for
superposition solutions incorporating any number of sources or sinks. Regardless of how
many source/sink combinations we superpose to change the shape of a simulated body,
the net pressure drag on any resultant "body" will always be zero. Thus, in the absence of
viscosity, the flow of an ideal fluid around any simulated body will always satisfy the
d’Alembert’s paradox of zero drag.

9.8.4 Flow Around a Two-Dimensional Rankine body

In order to simulate a closed body in a uniform flow using sources and sinks, one must
assure that the total strength of the sources is equal to the total strength of the sinks (i.e.
 Q   Q  0 ). When we superpose a single source and single sink of strength Q
sources sinks
with a uniform flow, the shape of the body will be elliptical in shape, with the shape of
the body depending on both the strengths of the source/sink and the separation distance
between the source and sink. In the present example, we superpose: (a) a uniform flow of
strength U moving left to right, given by z   Uz , (b) a source located at z = -zo of
Q
strength Q, given by z   ln z  z 0  , and (c) a sink located at z = +zo of strength
2
Q
-Q, given by z    ln z  z 0  . The superposed complex potential function is:
2

Q Q
z   Uz  ln z  z 0   ln z  z 0  (9.96)
2 2

274
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

To simulate a closed body, it is clear that the source must be upstream of the sink,
otherwise the flow will not exhibit a stagnation point at the leading edge. And to
simulate a symmetric body, the source and sink must be located along an axis parallel to
the impinging uniform flow. For the present simulation we assume a uniform flow
parallel to the real axis (x), and let zo = a, which gives:
Q Q
z   Uz  ln z  a   ln z  a  (9.97)
2 2
The general configuration for this simulation is termed a Rankine body, and is shown in
figure 9.15:
y
Simulated Rankine Body
ym

U
+Q -Q
x
-xs xs
a a

-ym

Figure 9.15 Elliptical Rankine body shape simulated by superposition of an

equivalent source (+Q) and sink (-Q) with a uniform flow (U).

Since the logarithm terms in Eq. 9.97 lend themselves best to cylindrical variables, we
define two new complex variables as;

z1  z  a  r1ei1 (9.98a)
and
z 2  z  a  r2ei 2 (9.98b)

where,
 y 
r1  x  a 2  y2 and 1  tan 1   (9.99a)
xa
and
 y 
r2  x  a 2  y2 and  2  tan 1   (9.99b)
 x a 

Substituting Eqs. 9.98 into Eq. 9.97, we have:

Q Q Q  z1  Q  r1ei1 
z   Uz  ln z1   ln z 2   Uz  ln   U( x  iy)  ln 
2 2 2  z 2  2  r2ei 2 

275
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

Q  r1  Q
z   U( x  iy)  ln   i 1  2  (9.100)
2  r2  2

And substituting from Eq. 9.99 to express Eq.9.100 in terms of x and y yields:

 Q  x  a 2  y 2   iUy  Q tan 1 y   tan 1 y  


z   Ux  ln        (9.101)
 2  x  a 2  y 2   2  xa  x  a  


From Eq. 9.101, we can identify the stream and potential functions for this flow as:

Q  x  a   y2 
2

( x , y)  Ux  ln (9.102)
2  x  a 2  y2 
 
Q  1  y  1  y 
( x , y)  Uy   tan    tan   (9.103)
2  xa  x  a 

We determine the velocity components by differentiating (z) from Eq. 9.97 for W  u  iv
(this is easier than differentiating  or  with respect to x or y). Thus,

d Q 1  Q 1  Q 1  Q 1 
WU    U 
dz 2  z  a  2  z  a  2  x  a   iy  2  x  a   iy 
  

Q  x  a   iy  Q  x  a   iy 
WU    
2  x  a 2  y 2  2  x  a 2  y 2 
Or separating the real and imaginary parts

 Q  x  a   Q  x  a   
W  U     
 2  x  a 2  y 2  2  x  a 2  y 2  
 Q  y  Q y  
 i     
 2  x  a   y 2  2  x  a   y 2  
2 2

Thus, by identification we have the Cartesian velocity components:

Q  x  a    x  a     (9.104a)
u  U  2
 2 
 x  a   y   x  a   y  
2 2
2  
Q  y   y 

v  2
 2 
(9.104b)
 x  a   y   x  a   y  
2 2
2  

276
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

The stagnation points (ys and xs) for the simulated body occur where u = 0 and v = 0. We
note that v = 0 along the x axis (y = 0). Thus, to determine the stagnation points, we set u
= 0 and let y = ys = 0 and x = xs in Eqs.9.104 (which makes v = 0 in Eq. 9.104b), and
solve for the resultant xs by setting u = 0 in Eq. 9.104a:

Q  x s  a    x s  a   
u  U     0
2  x s  a 2   x s  a 2  

After some algebra, we determine xs as:

Q
x s  a 1  (9.105)
Ua
At this point, it is advantageous for plotting purposes to define non-dimensional variables
x y Q
as x  and y  , and a relative source strength term as   . Thus, the
a a Ua
stagnation points for the Rankine body are xs   1   and ys  0 .

Now, in terms of x and y the stream function equation (Eq. 9.103) becomes:

Q  1  y  1  y  
( x, y)  Ua y   tan    tan   (9.106)
2   x  1   x  1 

Since ys  0 at the stagnation points, Eq. 9.106 indicates that the stream function
equation for the body surface is ( xs , ys ) =0 (the same as for the non-rotating cylinder
flow). Thus, the equation for the body shape is given by:

Q  1  yb   y 
( xb , yb )  ( xs , ys )  0  Ua yb  tan    tan 1  b   0 (9.107)
2   xb  1   xb  1 

In Eq. 9.107, y'b and x'b are coordinates on the body surface.

Rewriting Eq. 107 in terms of  gives:

  y   y 
yb  tan 1  b   tan 1  b   0 (9.108)
2  xb  1   xb  1 

The maximum thickness of the body, ym , occurs when xb  0 , and is (unfortunately)
given implicitly by:

ym 

2
  

tan 1 ym   tan 1  ym   0  ym  2 tan 1 ym   
2

[Here we substitute the identity: tan1  ym     tan1 ym ]

277
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

Or,
 2 ym 
  (9.109)
   2 tan ym 
1

This implicit equation for ym must be solved either iteratively for a given value of , or
by choosing a desired value of ym , and solving for the required  value.

To determine the body shape from Eq. 9.108 appears to be a messy iterative process.
However, we can rewrite Eq. 9.108 as:

 y  2 y  y 
tan 1  b   b  tan 1  b 
 xb  1    xb  1 
Or,
 2 y  yb
tan b  
yb  2 y  y     xb  1
 tan  b  tan 1  b   (9.110)
xb  1    xb  1  1   yb  tan 2 yb 
 x  1    
 b   

tan a  tan b
[using the trigonometric identity: tan a  b   ]
1  tan a tan b
Solving Eq. 9.110 for xb gives:

2 yb
xb2  1  yb2  (9.111)
 2 y 
tan b 
  
Equation 9.111 now allows the explicit determination of the body shape.
The velocity along the body surface can be determined from Eqs. 9.104, using Eq. 9.111
to specify the surface coordinates. In terms of x and y , we can define non-dimensional
velocity components as:

u  x  1   x  1  
  (9.112a)
u   1   2
 2 
 x  1  y   x  1  y  
2 2
U 2     

v 
 y   y 

v    2
  2 
(9.112b)
 x  1  y   x  1  y  
2 2
U 2 

So, knowing xb and yb from Eq. 9.111, from Eqs. 9.112 we can calculate the
dimensionless velocity components on the body, ub , vb , and thus the pressure coefficient
 
Cp, using Vb2  U 2 ub2  vb2 , so that:

278
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

p  p Vb2
Cp  1
U 2
 1 
U 2

 1  ub2  vb2  (9.113)
2

Rankine Bodies
Beta = 0.25, 0.5, 1

Y' 1
Beta=0.25
Beta=0.5
0.8
Beta=1

0.6

0.4

0.2

0
-1.6 -1.2 -0.8 -0.4 0 0.4 0.8 1.2 1.6
X'

(a) Non-dimensional Rankine body shape (upper half)

Pressure Coefficient for Rankine Bodies

Beta = 0.25, 0.5, 1
1

0.5
Cp, B=0.25

Cp Cp, B=0.5
0
Cp, B=1

-0.5

-1

-1.5

-2
-1.6 -1.2 -0.8 -0.4 0 0.4 0.8 1.2 1.6
X'

(b) Non-dimensional Pressure Coefficient

Figure 9.16 Body shape and pressure coefficient for inviscid flow around Rankine
Q
bodies with relative source strengths of   = 0.25, 0.5, and 1.0.
Ua

Figure 9.16 shows the upper half body shape and the pressure coefficient for Rankine
bodies with =0.25, 0.5, and 1 as determined from Eq. 9.111 and 9.113.

279
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

Note that the shape of the body, although symmetric, varies markedly with the value of ,
as does the pressure coefficient. In particular, the minimum pressure coefficient for the
largest  value (1.0) occurs at the center of the body, whereas for the minimum  value
(0.25), there are two minimums occurring near the leading and trailing edge. As we
pointed out earlier in this chapter, when a real flow passes a point of minimum pressure,
and experiences an increasing pressure, viscous effects can cause the development of a
flow separation region, a marked departure from inviscid flow simulations, and a marked
increase in drag. For the present simulation of a Rankine body, we would expect our
small  = 0.25 simulation to experience a flow separation sooner than the large 
simulation, since the point of minimum pressure is predicted to occur much nearer the
leading edge of the small  = 0.25 simulation body. We will address the sensitivity that
real fluids have to pressure gradients when we discuss boundary layers in Chapter 13 and
14, and drag in Chapter 15.

9.9 Bodies Translating in a Potential Flow

In the previous section, we superposed elementary functions to create simulations of uniform

flows around stationary bodies. However, what if we want to simulate the behavior of a
body translating through a quiescent potential fluid? This will change the potential function
of the flow field surrounding the body, although (as we will see) the pressure on the body and
the relative velocity field remain the same. Basically, we take a flow that was simulated with
a uniform approach velocity, and delete the uniform flow function from the initial complex
potential function (this is like imposing a constant velocity of opposite sign to the superposed
function). However, one must first determine how the constants for the other elementary
function(s) relate to the uniform flow velocity. In the following section, we examine the
translation of a cylinder through an otherwise stationary fluid.

9.9.1 Translation of a Cylinder in a Quiescent Fluid

To illustrate the movement of a body through an inviscid, quiescent fluid, we will make
use of our superposition of section 9.8.1, flow around a cylinder. Now, since we can
make better sense of this type of translating flow with a body moving in a positive
direction, we will modify the superposed function to have the initial function simulate a
flow from right-to-left (section 9.8.1 simulates the flow from left-to-right).
y

R
x + 

Figure 9.17 Simulated inviscid flow over a cylinder, passing right-to-left (i.e.
moving in the negative x direction).

280
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

The appropriate function for this right-to-left flow, as shown in figure 9.17, is:

s
( z )   Uz  (9.114)
z

Note that this is just the negative of Eq. 9.48, which reverses the direction of the flow
field, with the cylinder stationary within the flow field. To simulate the translation of the
cylinder at a positive velocity U to the right, we add +Uz to the superposed function,
which gives a new function:

s s
( z )   Uz   Uz   (9.115)
z z

You may note that Eq. 9.115 is simply a doublet of strength s. This yields the flow field
in a reference frame translating at a velocity U in the positive x direction, as shown at the
right. However, what is the value of s for the doublet function in
Eq. 9.115? Recall that we showed in section 9.8.1 that the radius of R U
the simulated cylinder is given by:

s
R  s  UR 2
U

Thus, substituting s  UR 2 for s into Eq. 9.115, we get:

s UR 2
( z )     (9.116)
z z

Expanding Eq. 9.116 in both r,  and x, y variables, we have:

UR 2
cos   i sin   U R 2 x 2iy
2
( z )  
r x y  
 UR 2   UR 2   UR 2 x   UR 2 y 
   cos    i sin      2   i 2 
 r   r   x y
2
  
  x y
2
 
   i

We will only need to use the potential function in this and the following section, so we
note that for the translating cylinder:

 UR 2   UR 2x 
    cos      2  (9.117)
 r   x y
2
  

281
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

Using the Cauchy relations, we can derive the velocities in both cylindrical and Cartesian
as:

 U R 2
vr   2 cos  (9.118a)
r r

1  U R 2
v   2 sin  (9.118b)
r  r

And in Cartesian coordinates, the complex velocity components are:

u 

 U R 2 x 2  y 2  (9.119a)
x 
x 2  y2
2

 2U R 2 xy
v  (9.119b)
y 
x 2  y2
2

Examining the resultant flow field is quite interesting. Let's examine the simulated
velocity at three different points: (1) the leading edge of the cylinder at (x,y) = (R,0), (2)
the trailing edge of the cylinder at (x,y) = (-R,0), and (3) the top of the cylinder (x,y) =
(0,R).

Leading edge: u 

UR 2 R 2  0   U; v
2 U R 2 R 0 
0
R 2
0  2
R 2
0 
2

Trailing edge: u 

U R 2  R   0
2
 U; v
2U R 2  R 0
0
 R  2
0 
2
R 2
0 
2

Top of cylinder: u 

UR 2 0  R 2   U ; v
2 U R 2 0 R 
0
0  R  2 2
R 2
0  2

These results are shown graphically in figure 9.18.

y
- U

R x
U U

Figure 9.18 Velocity vectors at (1) Leading edge, (2) Trailing edge, and (3) Top of
cylinder moving at velocity U left-to-right.

282
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

At the three points shown in figure 9.18, the y-direction velocity is zero, so only the u-
direction velocity is relevant. Note that the velocity of the fluid at the leading and trailing
edges of the cylinder are u = U, which is as expected, since the fluid velocity at the
leading an trailing edge in a potential flow should be identical, and reflect the simulated
velocity of the cylinder of U. However, at the top of the cylinder, the velocity of the fluid
is u = - U --- exactly the opposite of the cylinder velocity. Now consider that since the
coordinates are fixed, these are the velocities that would be observed by a fixed observer
as the cylinder traveled by. Thus, a fixed observer would see a flow at the top of the
cylinder in the negative x direction as the cylinder passed by. While this may seem
strange, consider that for a fixed cylinder in a uniform flow of velocity U, the u velocity
will increase from u = 0 at the leading edge to 2U at the top of the cylinder (in section
9.8.1 the velocity change was from 0 to -2U, because the flow was in the opposite
direction). Thus, one might expect a total change in the velocity from the leading edge to
the top of the cylinder of u  2U , as our result shows.

This backflow behavior is not just a manifestation of potential flows, but occurs to some
extent in many real fluid flows. For example, if you stand close to a passing large truck,
train, or other vehicle (just not too close!) you will notice a flow of air in the opposite
direction to that of the traveling vehicle. Now, a real flow is not a potential flow, but it is
similar from the leading edge to the thickest point, so the same basic behavior will occur
(although somewhat diminished by viscous effects).

Another example is an ocean wave---which can be roughly modeled as a translating half-

cylinder. As a wave moves across the ocean on an otherwise still day, if you are floating
on the ocean, you can feel a reversed flow of air as the wave passes by. This is
accentuated as the wave height increases. Waves travel at velocities roughly proportional
to the square root of their height, so a large wave moves faster than small waves.
Consequently a large wave has a stronger back flow of air over its face. On a calm day,
if you watch a large wave just as it breaks upon approaching shore, you will see a spray
of water blow backwards from the wave. This is again the result of a backflow of air
created by the translating wave. The following image shows this backwards spray over a
breaking wave.

This backflow of air over a wave is also put to good use by sea birds to aid in flight over
water. Pelicans are the masters of harnessing these backflows of air over waves (they
must have had good instruction in fluid mechanics). Next time you are at the seashore,

283
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

and pelicans and modest size waves are present, notice that a pelican will often conserve
energy by flying very close to the moving wave, and gliding along its face to take
advantage of the backflow of air over the wave (see the previous picture above). To see a
video example of this pelican "wave" riding behavior click this video link. It is really
pretty nifty how the pelicans swoop and soar, catching the updraft off of one incoming
wave after another. They are the original "surfers."

9.9.2 Pressure Field for a Translating Cylinder

It would seem that the determination of the pressure field for a translating body in a
potential flow would be fairly straight forward. However, because the body is moving
relative to the fixed reference frame, the calculation is a bit tricky. Recall from section
9.4 that the Bernoulli equation for an unsteady flow is given by Eq. 9.14 (neglecting the
term for height changes, since we assume g is perpendicular to the flow) as:

P 1 2 
 V   C(t ) , (9.14)
 2 t

Now, Eq. 9.14 assumes that the flow field for this equation is fixed relative to the
reference coordinates. However, the function  in Eq. 9.117 for the translating cylinder
is not stationary, but is translating at the cylinder velocity U. Thus, while the flow field is
fixed relative to the cylinder, it is translating at a velocity U with respect to the x-
coordinate. Thus, while the flow field is fixed relative to the cylinder, it is moving
relative to the fixed x-y reference frame. What this means is that we must take into
account the translation relative to the reference frame when determining the pressure

behavior. To do this requires that we determine in a translating reference frame. This
t
means that we must take into account both the temporal and the advective changes in .
We can do this be recalling that the substantial derivative gives the time changes for a
property in a translating field. We note that we can consider  as a flow field "property",
since it is directly associated with the flow field (i.e. its derivatives define the flow field
velocities). So, we realize we need to rewrite Eq. 9.14 as:

P 1 D
  V2   C(t ) , (9.120)
 2 Dt

From Chapter 3, Eq. 3.17, we can write the substantial derivative of  as:

D  
 u rel  (9.121)
Dt x t

Note that in Eq. 9.121 we only include the advective acceleration term for the translation
of  in the x-direction, u rel . since  does not translate relative to the y-coordinate.
However, what is u rel ? This is the velocity observed by the cylinder relative to the flow,

284
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

which is u rel = -U, since that is the velocity an observer in the cylinder would perceive
relative to the coordinate system--the fluid would appear to move toward the cylinder in a
negative x direction.

Still with me? If the translation velocity U and  are steady (we'll consider the case
where they aren't in section 9.11), Eq. 9.121 reduces to:

D  
 U   Uu since u  (9.122)
Dt x x

Now, substituting Eq. 9.122 back into Eq. 9.120, noting V 2  u 2  v 2 , gives:

P 1 2
 
 u  v 2  Uu  C(t )
 2
(9.123)

To determine C(t), we note the far upstream from the translating cylinder (x >> R), the
flow will be quiescent and will have a pressure P0 = constant where u = 0, v = 0. Thus,
P
C(t) = 0 . Substituting into 9.123, we do some rearranging, as shown:


P 1 2 P

 u  v 2  Uu  0
 2 

P 1 2 1 P 1
 2
 2

 u  2 Uu  v 2  U 2  0  U 2
 2

P 1 2 P 1
 
 u  2 Uu  U 2  v 2  0  U 2
 2  2

P 1 P 1
 u  U   v 2  0  U 2
2
(9.124)
 2  2

Checking some selected points for the pressure reflected by Eq. 9.124, we showed above
that u = U, v = 0 at the leading and trailing edges of the cylinder, (x.y) = (R,0) = (-R,0),
and u = -U, v = 0 at the top of the cylinder, (x,y) = (0,R). Checking the pressures
predicted by Eq. 9.124, we have:

Leading and Trailing edges: @ (x,y) = (R,0) and @ (x,y) = (-R,0)

PLE 1 P 1
 U  U   0  0  U 2
2

 2  2
PLE P0 1 2
  U
  2

285
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

Top (and also the bottom) of cylinder: @ (x,y) = (0,R)

PTOP 1 P 1
  U  U   0  0  U 2
2

 2  2
PTOP P0 3 2
  U
  2

Note that these values are identical to what we found for a flow around a stationary
cylinder in section 9.8.1. So, the pressure field around a translating cylinder is
unchanged from that for a comparable flow around a stationary cylinder. This will also
be the case for any other simulated potential flow geometry translating through a
quiescent potential fluid.

9.10 Accelerating Bodies in a Potential Flow: Virtual/Added Mass

Now, consider what happens if we specify that the velocity of a simulated body varies
with time, such that U(t). As the body accelerates or decelerates, the surrounding fluid
will also undergo a commensurate acceleration or deceleration as well. The
acceleration/deceleration of the surrounding fluid requires an additional force above that
required to accelerate/decelerate the body alone. This additional force is reflected by a
virtual or "added" mass of fluid proportional to the body displacement. In this section,
we will examine how a time-changing velocity impacts the pressure field on a body, and
how that is reflected in the amount of virtual or added mass. Note that fluid mechanics
literature uses the terms virtual and added mass interchangeably. To simplify our
discussion, we will simply use the term added mass

9.10.1 Acceleration of a Cylinder in a Quiescent Fluid

We will again consider a translating cylinder as our example of a body translating in an

inviscid fluid at a velocity U in the positive x-direction, but such that U is a function of
time, U(t). We can revisit Eq. 9.121, the substantial derivative of .

D  
 u rel  (9.121)
Dt x t


In section 9.9.2, we showed that advective part of Eq. 9.121 was u rel
 uU . This
x
advective component remains the same for a time-varying velocity. And we can

determine from our previous expression for , Eq. 9.117, as:
t

 R 2 x dU R2 dU
 2  cos 
t 
x  y dt
2
r  dt

286
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

Thus, Eq. 9.121 becomes:

D R2 dU
 uU  cos  (9.125)
Dt r dt

Substituting Eq. 9.125 into the unsteady Bernoulli equation, Eq. 9.120, and noting that
P
C(t) = 0 , where P0 is the pressure in the quiescent fluid well-removed from the

translating cylinder, we can rearrange the equation to give:

P 1 P 1 R2 dU
 u  U   v 2  0  U 2 
2
cos  (9.126)
 2  2 r dt

The last term in Eq. 9.126 is the additional momentum change due to the acceleration of
the fluid surrounding the cylinder. Note that if we examine the pressure at the leading
and trailing edge for a cylinder accelerating linearly at
U( t )  U 0 1  at  , where U 0 and a are constants, Eq. 9.126 gives: R U(t)

Leading edge: @ (x,y) = (R,0), u = U, v = 0, and  = 0

PLE 1 P 1
 U  U   0  0  U 2  R cos0U 0a 
2

 2  2

PLE P0 1 2
  U  RU 0 a
  2

Trailing Edge: @ (x,y) = (-R,0), u = U, v = 0, and  = 

PTE 1 P 1
 U  U   0  0  U 2  R cosU 0 a 
2

 2  2
PTE P0 1 2
  U  RU 0 a
  2

These results indicate that as the cylinder accelerates, the leading and trailing sides of the
cylinder are subject to markedly different pressures. This is a different result from that for the
constant velocity translation we examined in section 9.9.2. For the accelerating cylinder,
pressures will be higher on the leading surface, and lower on the trailing surface. This
pressure difference results in an additional force on the cylinder due to acceleration or
deceleration of the fluid surrounding, and moving with, the cylinder. As we will show in the
following section, this additional force can be modeled as an "added mass"--an equivalent
amount of fluid mass that would have to be accelerated/decelerated to account for the force
due to the unbalanced pressure difference over the cylinder surface.

287
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

9.10.2 The Added Mass for an Accelerating Cylinder

To determine the force required to accelerate a cylinder in a quiescent flow, we note that
this will just be the drag force, acting in the x-direction on the cylinder. We can
determine this force by integrating the pressure field, similar to what was done in section
9.8.1. Thus, the drag force will be given by:

Drag   dFx    PA x y

Where, dA  Rdˆir

dA  dA x ˆi  dA y ˆj  Rdˆir  Rd ˆi cos   ˆj sin     x
Thus, dA x  R dcos  is the differential area projection
in the x-direction. From Eq. 9.126, we have the pressure
of the surface as:
1 1 1 dU
2 2

P r  R  P0  U 2   u r  R  U  v r R R cos 
2 2 2 dt
 (9.128)

Now, we can most easily perform the integration along the surface of the cylinder
(r = R, 0 <  < 2) in cylindrical coordinates. To do perform this integration, we must
express the x and y-direction velocities in cylindrical coordinates. Using our expressions
for u and v from Eq. 9.119, and noting that on the cylinder surface that
x = Rcos, and y = Rsin, we can write u and v as:

u r R 

UR 2 x 2  y 2   UR cos   sin   Ucos 
4 2 2
2

 sin  2  U cos 2 (9.129a)
x 2
y 2 2
 R cos   sin  
4 2 2 2

2U R 2 xy 2U R 4 cos  sin  (9.129b)

v r R    2Ucos  sin   U sin 2
x 2
 y2 2
R4

Now, substituting Eqs.9.129 into Eq. 9.128, we get (using a few trigonometric identities
along the way):

1 1 1 dU
2 2

P r  R  P0  U 2  U 2 cos 2 2  2 cos 2  1  U 2 sin 2 2 R cos 
2

dt
1 dU
 P0  U 2  U 2 1  cos 2 R cos 
2 dt
(9.130)
1 dU
 P0  U 2  2U 2 sin  2 R cos 
2

dt

1 dU

 P0  U 2 1  4 sin  2 R cos 
2 dt


288
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

Finally substituting Eq. 9.130 into the drag equation , Eq, 9.127, and integrating from  =
0 to :

2
1 dU 


Drag    P  U 2 1  4 sin2   R cos 
2

dt 
R cos  d
0

2 2
 1 2 2 sin3  1 1  dU
  P  U R sin  0  2U R
2
 R 2    sin 2 
 2  3 0 2 4  dt 0

dU
 0  0  R 2
dt
dU dU
Drag  R 2
dt

  R 2
dt
 (9.131)

Notice that the first two terms in Eq. 9.131 are identical to those in the drag calculation in
section 9.8.1 for flow around a stationary cylinder, and satisfy d'Alembert's paradox.
However, the last term in Eq. 9.131, that associated with the acceleration of the cylinder,
does not equate to zero, and is proportional to the mass of fluid displaced by the cylinder
(per unit depth). Notice also that the Drag force is negative, indicating that it is acting in
dU dU
opposition to the direction of acceleration (negative if  0 , and positive if  0) .
dt dt
Thus, m added  R 2 is the mass of fluid per unit depth whose acceleration must be
accounted for when the cylinder is accelerated. Note that if the cylinder is much denser
than the fluid (e.g. a metal rod in air), the added mass has little effect. However, if the
mass of the fluid is heavier, or comparable to the mass of the cylinder, the added mass
can have a significant impact (e.g. a wooden rod being moved through water).

The added mass for the cylinder is equivalent to the mass of fluid displaced by the
cylinder. One might be tempted to think that all other bodies would have added masses
equivalent to the mass of the fluid displaced by the body. However, that is not the case.
For example, the added mass for a sphere is equivalent to ½ the displaced fluid (i.e.
2
m added  R 3 ). And the added mass for a square rod of sides w is roughly
3
m added  1.17w 2 , or 1.17 times the displace mass. A good table of added mass values
for a number of geometries can be found in a Naval report by C.E. Brennen (1982),
available electronically through the Caltech library. Note that added mass is quite
relevant to the U.S. Navy, and others who make use of ship and boat transport, since the
mass of a ship and the water is sails in are closely equivalent, and thus significantly
influences the forces on a ship during acceleration/deceleration, and changes of direction.

Note that we have only considered added mass in one dimension. In reality, translating
geometries can experience added mass effects in multiple directions, which can make the
issue quite complicated. Brennen's report touches on a number of the issues of added
mass associated with multiple motions of a body.

289
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

Problems

1. The stream function for an inviscid flow is given by (x, y, t )  V0 x  t y  t  , where Vo is
a constant =1.

a) Show that  represents a potential flow solution.

b) Determine the corresponding form of (x, y, t) [note: let =0 at x,y = 0,0].

c) Employ  to help determine a general expression for the pressure difference, P2-P1,
between any two points in the flow field, (x1, y1) and (x2, y2).

d) Determine the non-dimensional pressure difference (P2-P1)/Vo2 between:

Case #1: (x1, y1)=(0, 0) and (x2, y2)=(1, 0); and

Case #2: (x1, y1)=(0,0) and (x2, y2)=(0,1).

e) Plot (P2-P1)/ Vo2 vs. t for both cases on the same graph for 0 < t < 2.

f) Also plot the streamline and the potential lines passing through (x, y)=(1,1) at times
t=0 and t=1 (plot each time on separate graphs); plot over the region 0 < x < 2 and 0 <
y < 2. What is characteristic about the behavior of the streamline and potential lines
in these two graphs?

2. The stream function for an inviscid flow is given by (x, y, t )  V0 x  t y  t  , where Vo is
a constant = 1.

a) Show that  represents a potential flow solution.

b) Determine the corresponding form of (x, y, t) [note: let =0 at x,y = 0,0].

c) Employ  to help determine a general expression for the pressure difference, P2-P1,
between any two points in the flow field, (x1, y1) and (x2, y2).

d) Determine the non-dimensional pressure difference (P2-P1)/Vo2 between:

Case #1: (x1, y1)=(0, 0) and (x2, y2)=(1, 0); and

Case #2: (x1, y1)=(0,0) and (x2, y2)=(0,1).

e) Plot (P2-P1)/ Vo2 vs. t for both cases on the same graph for 0 < t < 2.

f) Also plot the streamline and the potential lines passing through (x, y)=(1,1) at times
t=0 and t=2 (plot each time on separate graphs); plot over the region 0<x<2 and
0<y<2. What is characteristic about the behavior of the streamline and potential
lines in these two graphs?

290
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

3. In the section 9.8.1 example, it was shown that one could represent the flow around a
cylinder by superposition of a uniform flow and a doublet, to give the complex potential
function as:
s
 (z)  Uz 
z
Here the uniform velocity approaches from the left, where x < 0. The stream function in
Cartesian variables is:

 s 
  y U  2 
 
x  y2  

Assume a doublet strength s = U, and:

a. Determine the velocity field components u and v for this stream function in terms of x
and y:
P  P
b. Determine the pressure coefficient C p  1 2
as a function of x, y variables (don't
2
U
try to simplify—it's messy).

c. Simplify your expression for Cp to give C p  f x  along the  = 0 streamline (y=0),

for x < -R (the stagnation point of the cylinder).

d. Using your result from part c, determine the non-dimensional pressure gradient
dP
dC p
coefficient,  1 dx 2  f x  along the  = 0 streamline (y=0), for x < -R (the
dx 2
U
stagnation point of the cylinder).

e. On one graph, plot u/U, Cp, and dCp/dx vs. x along the  = 0 streamline (y=0), for
-4R < x < -R (use Eq. 9.53 to establish the value of R).

4. The flow field near a tornado can be approximated as the superposition of a sink and a
irrotational vortex, with the steam function given by:

Q 
  ln r
2 2

For Q = -2 and  = -2, determine the velocities in both radial and Cartesian coordinates,
and an expression for the pressure, where P = Patm for r . Create a plot of the streamlines
that pass through (x,y) = (2,0), (0,2), (-2,0), and (0,-2). Show the streamlines in the region -2
< x < 2, -2 < y < 2.

Is this a realistic approximation as r  0? Justify your answer.

291
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

5. Uniform flow U approaching from the left encounters a sink of strength Q = - A, located at
x,y = 0,0. This roughly approximates the behavior of a flow approaching a drain. The
A
complex potential function for this superposition is given by z   Uz  ln z .
2
y

Uniform Q<0
Flow
x

Determine the following:

a) The stream function, , in terms of both r,  and x, y;

b) The velocities u and v in terms of x, y, and r, ;

c) The location of a stagnation point in the flow, if there is one;

Now, let U = 1and A = 1 and make a nice engineering graph of:

d) The streamlines  = 0, -0.05, and -0.30 all on one graph. Show all streamlines for
the region -0.6 < x < 0.3 and 0 < y < 0.6 only. Note: to plot each streamline, you will
have to determine r = f(), calculate r for a series of  values, then determine the
corresponding series of x and y locations.

6. A point vortex with circulation strength  =+A is located at x = 0, y = 1; a similar point vortex
with  = -A is located at x = 0, y = -1. Determine the velocity field V  uˆi + vˆj and the
p
pressure coefficient Cp along the x-axis, where C p  2
].
 A 
1
 
2 

Plot Cp for y = 0, -3 < x < 3. Let p = 0 at x = 0, y = 0.

7 . Consider a vertical wall with an impinging jet, as modeled by the complex potential function
(z)=(A/2)z2, with A= -2 (flow right-to-left). Determine the following, considering all
quantities dimensionless (simplify all answers):

a) Expressions for the stream function  in terms of both x, y and r,;

b) The velocity components vx, vy (in x, y) and vr, v (in r,);

292
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

c) The location of the stagnation point for x0, and the value of  which passes through
this stagnation point;
d) An expression for p-ps as a function of r,, where ps is the stagnation pressure (let the
fluid density be  =2);
Plot the following:
e) Graphs on x, y axes of:
i) Streamlines = 0 and -2 for the region 3 > x > 0 and 3 > y > 0;
ii) The pressure P-Ps and the pressure gradient dP/dx along y = 0, 3 > x > 0;
iii) A line of constant pressure p-ps= -16 for x > 0.

8 . Consider a flow over a 90° step (a convex, 270° corner), as modeled by the complex
2
3A 3
potential function  z   z , with A= -2 (flow right-to-left).
2
y

Note: A<0, as shown

Convex, 270 Corner (n=2/3)

Determine the following, considering all quantities dimensionless (simplify all answers):

a) An expressions for the stream function  in terms of r,;

b) Expressions for the Cartesian velocities u and v in terms of r,;
c) An expression for P-P in terms of r,, where P is where r;
Now, Create nice engineering graphs of:
d) The streamlines  = 0, -1, and -2 all on one graph. Show all streamlines for x < 10
and y > -10 only. Note: to plot each streamline, you will have to determine r = f(),
calculate r for a series of  values, then determine the corresponding series of x and y
locations.
e) A plot of P-P vs. x along the streamline  = 0 from x = 10 to x = 0.2. Why wouldn't
you plot the pressure at x = 0?

9. Consider a vertical wall with an impinging jet, as modeled by the complex potential function
(z) = (A/2)z2, with A= -2 (flow right-to-left) [problem 7, above]. Now, a source of
strength Q = 4 is located at the origin (x, y = 0, 0). Use superposition to simulate the effect
of the source on the vertical wall with an impinging jet modeled in problem 7 above. For this

293
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 9

modified flow, specify the appropriate complex potential function and determine the
following:

a) Expressions for the stream function  in terms of both x, y and r,;

b) The velocity components vx, vy (in x, y) and vr, v (in r,);
c) The location of the stagnation point for x  0, and the value of  which passes
through this stagnation point;
d) An expression for p-ps as a function of r,, where ps is the stagnation pressure (let the
fluid density be  = 2);
Plot the following:
e) Graphs on x, y axes of:
i) Streamlines  = 0 and -2 for the region 3 > x > 0 and 3 > y > 0;
ii) The pressure P-Ps and the pressure gradient dP/dx along y = 0, 3 > x > 0;
iii) A line of constant pressure p-ps = -16 for x > 0;

10. Consider a circular cylinder of radius 1 meter and density of 500 kg/m3, which is moved back
and forth within a vat of inviscid water ( =1000 kg/m3) at a velocity U = U0sin(t), where
U0=2 m/s and =/2 s-1. Determine or show the following:

a) An expression for the displacement (in m) of the cylinder (call it x) as a function of

time [let x = 0 at t = 1].

b) An expression for the pressure difference ( in N/m2)between the leading and trailing
edge of the cylinder (Ple - Pte)as a function of time.

c) The force (per unit length of cylinder, in N/m) as a function of time that must be
applied to the cylinder to maintain the velocity [hint: you must include both the
cylinder mass and the added mass in the calculation].

d) Three separate graphs on one sheet of: (1) the cylinder displacement, x, (2) (Ple - Pte),
and (3) force, all for 0 < t < 4.

294
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 10

Chapter 10

Changes in Circulation and Kelvin’s Theorem

Contents

10.1 Circulation Changes for a Translating Fluid Region ..................................................... 296

10.2 Examples of Regions Showing Temporal Changes in Circulation ................................. 299

10.2.1 Diffusion of a Viscous Vortex .......................................................................... 299
10.2.2 Regions of Baroclinic Density Variations........................................................... 301
10.2.3 Baroclinic Variations in the Atmosphere .......................................................... 301
10.2.4 Baroclinic Variations Due to Local Heating ....................................................... 304
10.2.5 Baroclinic Variations Due to Density Stratification ........................................... 306

10.3 Kelvin's Theorem for a Translating Fluid Region ......................................................... 308

10.3.1 Examples of Circulation Preserving Flows ........................................................ 309

10.4 Helmholtz's Theorems for Inviscid Flow ..................................................................... 312

As we showed in Chapter 8, the circulation of a region of fluid reflects the summation of

the vorticity over the area of the region, which can be shown by Stoke’s Theorem to be
equivalent to the integration of the velocity around the bounding circuit of the region.
Since both vorticity and circulation are functions of the velocity field of the fluid (i.e. a
property field), they are also properties of the flow field, and are subject to changes as
they move within the flow field. The respective changes of circulation and vorticity due
to transport through the flow field can accordingly be assessed by applying the
substantial derivative to each property, and then using our previously-derived equations
of momentum (i.e. Navier-Stokes) and continuity to assist in simplifying the resulting
expressions. In this chapter we address the temporal changes in circulation that take
place for an initial vorticity-bearing quantity of fluid, such as a decaying, viscous vortex,
as the fluid moves within a flow field. In the following Chapter 11, we will develop the
vorticity transport equation, which models how changes in vorticity take place within a
flow field, and explore the different modes of behavior that can cause these changes in
vorticity.

As we will show in section 10.3, the additional assumption of inviscid flow for a flow
with circulation gives rise to what is known as Kelvin’s Theorem. This theorem, derived
initially by Lord Kelvin in the 1800’s, states that for an inviscid flow the circulation of an
initial quantity of fluid will not change with time. Although limited in application, since
real fluids are viscous, Kelvin's theorem will allow us to consider the limiting behavior of
a fluid due to changes in circulation.

295
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 10

10.1 Circulation Changes for a Translating Fluid Region

Consider the circulation contained within a region of fluid bounded by a contiguous

circuit (any closed path that starts and ends at the same point within a fluid, like a circle).
At any instant, we know from the development of the concept of circulation in Chapter 8
that the circulation within that circuit is given using Stokes theorem as:
 C
V
  
   V  ds  dA  (10.1)
ds 

But what happens to the circulation within this initial region of fluid as time progresses?
Is there a temporal change in  as the fluid particles comprising the original fluid region
move within the fluid? To examine this, we take the substantial derivative of Eq. 10.1 (in
Cartesian coordinates), which yields:

D D   D 
Dt Dt 
 V  d s   (udx  vdy  wdz ) where ds  dx î  dyĵ  dzk̂ (10.2)
Dt

In Eq. 10.2 we assume we can follow the original bounded region of fluid, such that the
initial mass of fluid within the region will remain constant (although the shape of the
bounding circuit and the properties within the circuit can change). Recall that Leibniz’s
rule states that if the limits of integration are not a function of time [   f (t ) ], then the
derivative can be performed inside the integral. In the present case, we assume that our
initial region of fluid is constant (i.e. of fixed volume), although the boundary of the
volume can vary. However, since the volume of material does not change, this implies
the limits of integration also do not change. Thus, by Leibniz's rule we can move the
substantial derivative inside the integral, such that we perform the substantial derivative
on the integrand of Eq. 10.2, which allows us to write Eq. 10.2 as:

D D
  ( udx  vdy  wdz ) (10.3)
Dt Dt

To assess the circuit integral of Eq. 10.3, we will examine one element of the integrand,
and extrapolate the resulting behavior to the other two elements. So, expanding the
substantial derivative of the first term of the integrand in Eq. 10.3 (the x-direction
component), we obtain:
dx
Ddx  A
dx'
B
Note that is A’ B’
Dt
the rate of change of
dx  dx '
Ddx 
the projection of dx,  u  u '  du
D( udx ) Du
 dx u which is simply du, dt
Dt Dt Dt the change in the x-
direction velocity, du.
du

296
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 10

D( udx ) Du Du 1
Dt
 dx
Dt
 udu  dx
Dt
 2 d u2   (10.4)

In Eq. 10.4, similar to what we showed in section 3.3, the substantial derivative of the
differential change in the x-direction (dx) is simply the differential of the x-direction
velocity (du).

We know from the x-direction component of the Navier-Stokes equation (Eqs. 5.44 and
5.46a) that:

Du 1 P
  x   2 u (10.5)
Dt  x

Du
Substituting from Eq. 10.5 into Eq. 10.4, we get:
Dt

D( udx )  1 P  1
     x   2 u dx  d( u 2 ) (10.6a)
Dt   x  2

Similarly, the other components of the integrand in Eq. 10.3 become:

D( vdy )  1 P  1
     y   2 v dy  d( v 2 ) (10.6b)
Dt   y  2

D( wdz )  1 P  1
    z   2 w dz  d( w 2 ) (10.6c)
Dt   z  2

Substituting Eqs. 10.6 into Eq. 10.3 and collecting terms gives:
dP dB

D 1  P P P 
    dx  dy  dz   x dx   ydy  z dz 
Dt   x y z 
(10.7)
1
  
   2 udx   2 vdy   2 wdz  d u 2  v 2  w 2
2

 
 2 V  d s V2

After identification of the collected terms, as indicated in Eq. 10.7, as the total
differentials of pressure, body force, and V2, the time change in circulation can be
separated into three parts as:

297
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 10

 
D
Dt
 
dP 

1 
   dB  dV 2      2 V  d s
2
  (10.8)
 

= 0, if dB is a conservative body force

If we assume conservative body forces (e.g. a gravitational field), the second circuit
integral in Eq. 10.8 will not depend upon the path of integration. Thus, the result of the
second circuit integral above must be zero, which reduces Eq. 10.8 to:

 
D
Dt
 
dP


   2V  ds  (10.9)

= 0 if  = f(P)

In Eq. 10.9, the rate of change of the circulation now depends on the variation of fluid
density and the viscosity. If the density of the fluid is only a function of pressure [ i.e.
that  = f(P) ], the fluid is termed a barotropic fluid, within which lines of constant
pressure (i.e. isobars) are parallel, or aligned, with lines of constant density (i.e.
isopycnals). If   f(P), the fluid is termed a baroclinic fluid, which means that the
isopycnals within the fluid are misaligned with the isobars [e.g. this often occurs when
there are significant thermal variations within the fluid]. So, if a fluid is barotropic, which
dP
is true for most single, isothermal fluids in a gravity field, then the integral  is also

not dependent upon the path of integration, and also becomes zero, such that Eq. 10.9
reduces to:

 
D
Dt

   2V  ds  (10.10)

Equation 10.10 is generally applicable for any Newtonian fluid that is barotropic and has
a body force due only to gravitation. This is the most general equation for temporal
changes in circulation for a defined region of fluid moving within a broader flow field,
and indicates that changes in circulation for such a fluid are due to the viscous behavior
of the fluid. There are a number of practical flows that satisfy this situation, particularly
single, isothermal, incompressible liquids, and polytropic gases [i.e. for which
Pvn  P n  constant ]. However, for fluids where density variations exist due to
dP
dissimilar fluids or when f(P), we then have   0 [more on this in section 10.2.2].

In situations of significant thermal or salinity variations, such as density stratifications
within the atmosphere or the ocean, changes in the circulation of a defined material
region can and will take place, as we will discuss in section 10.2.2.

298
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 10

Basically, circulation of an initial amount of advecting fluid can only change due to:

1) Viscous effects (e.g. real flows near surfaces), or

2) Changes in density that are not direct functions of pressure changes,   (P) ,
3) Non-conservative body forces.

Generally, since the body force for most practical flows is due to gravity, circulation
changes in real fluids will generally only occur due to: (1) viscous effects, and (2) density
variations.

10.2 Examples of Regions Displaying Temporal Changes in Circulation

10.2.1 Diffusion of a Viscous Vortex

Consider a symmetric, viscous vortex, with its vorticity initially concentrated within a
region of fluid of radius, R, as shown schematically in figure 10.1. We will assume the
vortex to be stationary relative to the center of rotation, and to have only an azimuthal
velocity (i.e. v). Here, the effects of viscosity will cause the vorticity within the initial
region of fluid to diffuse radially outward, and eventually migrate across the boundaries
(r = R) of the fluid region. When the vorticity diffuses across the fluid boundary, there
will be a corresponding loss of vorticity from the originally defined region of fluid (Area
= R2), and thus a loss of circulation from the region.
V(r) d
Fluid circuit, across which 0
dt

r Vorticity diffuses
R outward by viscosity.
No radial velocity

Figure 10.1 Example of the temporal diffusion of vorticity outward from an initial
point vortex due to viscosity.

If the vorticity within the vortex is all concentrated into a small region such that, rvortex << R,
then the viscous vortex will initially be enclosed within the bounded region and  will
initially remain constant within the region r  R. Changes will only occur when the vorticity
eventually diffuses outward to and across the boundaries of the original region, R.
However, it should be noted that eventually, given enough time, all of the vorticity will
diffuse outward and across the larger bounding region, r = R. At that extended point in time
D
(t  ),  0 , and = 0, since all of the vorticity will have diffused out of the bounded
Dt
region, assuming that R is finite.

As a further example, consider an initially rotating viscous fluid inside a circular

container with solid boundaries at r = R, as illustrated in Figure 10.2. One might think of

299
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 10

this type of rotating flow as similar to that created within a cup of coffee or tea by the
vigorous stirring of the fluid. Since the no slip condition applies at the outer solid
boundary, that means that the circulation within the confined fluid is zero, since v = 0 at
 
the boundary, which gives the circulation as total   V  d s  0 .

Velocity
profile

R
Stationary
container positive
(e.g. coffee cup) total = 0
negative

Figure 10.2 The circulation behavior of a viscous fluid contained within a circular
container.

This lack of circulation and the zero velocity at the container boundary means that the
total circulation within the container can never change, and that total = 0 for all time.
Since this situation clearly involves a viscous flow, with plenty of initial vorticity, how is
that possible? The key here is to recognize that such a flow contains equal amounts of
positive and negative vorticity within the region r  R, and that these vorticity
components diffuse and decay in such a way as to effectively cancel each other at all
times. In essence, if we envision the circulation within the bounded region as the sum of:
(1) the circulation associated with all elements of fluid with positive vorticity, positive,
and (2) the circulation associated with all elements of fluid with negative vorticity,
negative, then for any instant of time:

total  postive  negative   dA  0

all fluid in region
(10.11)

Where postive   positive

all positive vorticity fluid
dA (10.12a)

and, negative   negaitive

all negaitive vorticity fluid
dA (10.12b)

From Eq. 10.11 we note that:

positive = - negative (10.13)

This implies positive > 0 and negative < 0, such that positive and negative always effectively
offset each other, yielding a net circulation of zero. We note that initially, for our stirred
coffee or tea, these values of positive and negative must be substantial. From consideration
of Eq. 10.13 and Eqs. 10.12, we can infer that positive  0 and negaitive  0 , and that these
initial vorticities must also be substantial. However, after an extended period of time, we

300
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 10

know that the fluid within the cup will come to rest due to viscous effects, such that the
velocity and vorticity within the cup will be zero everywhere. When the fluid is at rest,
all the vorticity must have been dissipated, such that positive = negative= 0. What can be
inferred from this assessment is that the initial distributions of positive and negative
vorticity within the cup must cross-diffuse in such a manner as to continually cancel each
other, eventually resulting in the dissipation of all the vorticity of both signs, as the fluid
is brought to rest. We will revisit this question of cross-diffusion of vorticity in Chapter
11, as well as to examine how the vorticity is initially created within such flows.

10.2.2 Regions of Baroclinic Density Variations

In Section 10.1 we discussed how circulation can change due to density variations.
Assume for convenience that we have an inviscid fluid with conservative body forces.
This would reduce Eq. 10.9 to:

D dP
  (10.14)
Dt 

As we discussed previously, if  = f(P), or conversely P = f(), the result is that the

circuit integral in Eq. 10.14 is independent of the path of integration, and will equal zero.
These types of flows, as we discussed, are termed barotropic. What this means is that
lines of constant pressure (i.e. isobars) and lines of constant density (i.e. isopycnals) are
parallel, or aligned. We can describe this process mathematically as   P  0 , which
means that the gradients of the density and pressure are aligned. An example of such a
fluid would be an isothermal body of water, or a uniformly compressible gas. Note that
isobaric behavior is not necessarily constrained to isothermal flows, since it could also
exist if the isotherms (i.e. lines of constant temperature) in the fluid are also parallel the
isobars and isopycnals [often a difficult situation to achieve].

However, when the isobars and isopycnals do not align with each other, such as when
there are lateral variations in temperature, then   dP  0 , and a change in circulation will

result. A flow where the isobars and isopycnals do not align is termed baroclinic.
Examples of baroclinic flows include circulation induced by density stratifications within
the world’s oceans or atmosphere, where thermal gradients created by solar heating of the
ocean/land can result in the generation of winds within the atmosphere or water currents
in the ocean (salinity gradients within the ocean can also cause this). This is also the case
in forest fires.

10.2.3 Baroclinic Variations in the Atmosphere

One of the more instructive examples of the generation of circulation by density

variations is what is known as the sea breeze problem (Green 1995). Consider a
simplified model of the density variations in a coastal air layer spanning a region
comprised of: (1) the sea, and (2) the land bordering the sea, as shown in Figure 10.3.

301
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 10

 increasing  increasing
P increasing P increasing

Land Sea Land Sea

(a) Model (b) Simplified model

Figure 10.3 Modeling of the sea breeze problem

Here we assume a time during mid-day, when solar radiation has had appreciable time to
act upon both the land and the sea. Since land will warm more quickly than water, due to
better solar absorption, a significant temperature difference will develop between the land
and the sea. Consequently, the air over the sea will be cooler, and thus of higher density,
whereas the air over the land will be warmer, and thus less dense. As this density
variation within the air layer develops, the isopycnals (lines of constant density), will
become sloped relative to the earth, and thus misaligned with the lines of constant
pressure (we assume the isobars are parallel the earth), as modeled in Figure 10.3 (a). To
make the calculation of circulation simpler, we will simplify the model even more, by
assuming the isopycnals to be perpendicular to the isobars, as shown in Figure 10.3 (b).
While this is an extreme situation, it is easy to perform the integration, and is still a
reasonable, although extreme, model of the process.

Employing our simplified model, we now consider the circulation for a region one
kilometer high by 20 kilometers wide, stretching equally from the land to sea, as shown
in Figure 10.4.
 increasing
L S

L=20 km
C  B Ph P increasing
y
h=1km

x D A Po

Land Sea
Figure 10.4 Atmospheric circulation for a region 20 km wide by 1 km high

Using Eq. 10.14, the change in circulation for the defined circulation loop can be
calculated by integrating segmentally around the circuit shown in figure 10.4, starting at
point A, as follows:

302
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 10

0 0
B C D A P P P P
D dP dP dP dP dP h
dP h dP o dP o dP
           
Dt  A
S B
 C
 L D
 Po
S Ph
 Ph
 L Po

Ph P
dP o dP
 
 Ph L
 (10.15)
Po S

 P  Po  Po  Ph  1 1 
  h    Po  Ph   
 S L   S L 

Note that by assuming that the isopycnals are perpendicular to the isobars, only the
integration of pressure in the vertical direction contributes to the change in circulation,
since the pressure is assumed to not change laterally. If we now assume a 5 C difference
in the air temperature, from 20 C (sea) to 25 C (land), this reflects air densities of
approximately S = 1.205 kg/m3, and L = 1.185 kg/m3. We also assume Po-Ph  11 kPa,
roughly the vertical atmospheric pressure change over one kilometer, assuming a constant
temperature.

Thus, Eq. 10.15 becomes:

D  1 1 
 Po  Ph   
Dt  S  L  (10.16)
3 2
N
 11kPa 1000 2 1 kg2  m  1  1  m  154 m2
m  kPa s  N  1.205 1.185  kg s

If we assume the average tangential velocity at the periphery of the defined circulation
region is V, then the circulation for that region is approximated as:
D
   V  ds  Vx 2h  L  . Now, assuming that is relatively constant over a period
Dt
of time, we can approximate:
0
t
D D
0 Dt dt  Dt t  t  0  Vx 2h  L (10.17)

Here we assume initially quiescent conditions, with Vinitial = 0, and thus 0  0 . Solving
Eq. 10.17 for the velocity, and substituting from Eq. 10.16, we can determine the wind
velocity after one hour under these baroclinic conditions:

D t m 2  1  hr
V   154 2   3600 s  1  km  13.2 m
Dt 2h  L  s  42  km hr  1000  m s

This is roughly 30 miles per hour! Since this value is negative, this implies that the
velocity must move in opposition to the presumed counter-clockwise rotation for
circulation. Accordingly, this means that the airflow would be toward the land at lower

303
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 10

elevations, and toward the ocean at higher elevations. This would typically be what is
known as an “on shore” wind, due to the warming of the air over land during the day. In
the evening, since the land cools faster than the sea, the air over the land will cool more
quickly, resulting in a reversal of the flow, with flows from the land to the sea at lower
elevations, in what is termed an “off shore” wind.

Obviously, the assumed skewing of the density isopycnals is extreme, and would
probably be much less severe. Also, for a real atmospheric flow, viscosity will mediate
the wind effects. However, this example should make clear the importance of
baroclinicity in generating circulation, and thus vorticity. These thermally-induced
variations in atmospheric density [along with Coriolis effects] basically control the
world’s wind and weather patterns, including the generation of ocean waves, which are
the result of prolonged wind interactions with the ocean surface.

10.2.4 Baroclinic Variations Due to Local Heating

Other examples of baroclinic generation of circulation are natural convection due to a

heated vertical surface or air currents generated by a flame. In the case of a heated
vertical plate, the isobars will be horizontal, but the isopycnals are essentially vertical—
almost identical to the model we employed for the sea breeze problem. The situation
with a flame (let’s assume a candle) is also similar. Figure 10.5 illustrates (roughly), the
velocity field associated with a flame. Here, we assume that the flow field is two-
dimensional, but similar conclusions can be drawn for an axisymmetric flame. Note that
we assume that the air adjacent to the flame is quiescent before the candle is lit, and thus
has no circulation and no vorticity. However, after the candle is lit, both the air that
undergoes the combustion process, and the air in proximity to the flame, become less
dense, and circulation is generated.

L>0 R<0

Before flame  =0

After flame  0

Flame

Figure 10.5 Generation of circulation by a flame.

This generation of circulation is also indicative of the accompanying generation of

vorticity. However, note that the flow field in Figure 10.5 illustrates that the circulation
for the region to the right of the flame, R, is less than zero because the vorticity
generated would have a clockwise rotation (for this figure, the right-hand rule assigns
counter-clockwise rotation as positive). However, the circulation for the region to the left
of the flame, L, is greater than zero because the vorticity generated would have a

304
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 10

counter-clockwise rotation. Further note that the total circulation will be the sum of both
regions, i.e. total=L+R. If the flow is symmetric, it is obvious that L = -R, and total =
0! This is an interesting revelation, but clearly explainable. Although the flame will
generate significant vorticity, it will generate equal amounts of both positive and negative
vorticity, such that the cumulative positive vorticity (if one considers a large enough
region surrounding the flame) will directly offset the cumulative negative vorticity. This
is not dissimilar to the processes discussed in section 10.2.1 regarding the effects of
viscosity in dissipating the vorticity within a contained, viscous fluid.
As a further extension of this example, consider a candle burning in air as a viscous fluid
 
within a closed room. Here the equation that applies is:
D
Dt
 
dP

    2 V  ds ,  
since we must consider both baroclinic effects and viscous effects impacting the
 
generation of circulation. However, if we note that total   V  d s  0 , since the
boundaries of the room will be stationary, then the change in circulation within the room,
D
room  0 , since the circulation within the room can’t change due to the no slip
Dt
condition. Thus, the effects of vorticity generation by the flame due to baroclinicity,
must either be: (1) balanced, due to symmetry, or (2) offset by viscous effects such that
the total circulation within the room never changes. This balancing act between
baroclinic effects and viscosity also assures that the velocities within the room, associated
with the generation of vorticity, will never reach high values [note that without the
presence of viscosity, the local velocities within the room could continually increase due
to forcing by the density variations].
As a final aspect of temperature-induced circulation changes, consider the spread of a
forest fire. By its very nature, a forest fire creates large local density variations, with
isopycnals at sharp angles to the earth, whereas the isobars are generally aligned with the
earth. As illustrated in figure 10.6, this creates strong circulation, which results in strong
updrafts of high temperature gasses, flanked by strong down and side drafts that feed
down draft

updraft down draft updraft

fire-induced
circulation
down draft
fire
side draft
fire fire-induced
circulation
Induced uphill flow

(a) (b)

Figure 10.6 Creation of up and down drafts in proximity to fires, due to circulation
created by density variations: (a) on a flat surface (b) on a sloped surface.

305
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 10

oxygen to the fire, thus intensifying the combustion process. Figure 10.6(a) is a balanced
situation on a flat surface, with counterbalancing inflows holding the fire in place.
However, when fires occur on a sloped surface, the inflows are unbalanced, such that the
downhill circulation will be higher than the uphill circulation, causing such fires to
migrate onto an upslope due to a skewing of the baroclinic process such that the fire
creates its own uphill “wind”, as depicted in figure 10.6(b). This induced uphill airflow
can cause a fire to accelerate up the slope of a hill or mountain. This generation of
combustion-induced winds is extremely dangerous to fire fighters when the terrain
becomes constrained, such as in a canyon, often resulting in a situation known as a “fire
storm”, which usually results from a confluence of the fire with plentiful fuel on hilly
terrain.

10.2.5 Baroclinic Variations Due to Density Stratification

The effects of strong density stratification (e.g. due to strong salinity variations in sea
water, or dissimilar fluids) on the generation of circulation, and thus vorticity, can be
illustrated by considering the behavior of two adjacent horizontal layers of inviscid,
immiscible (non-mixing) fluids where a  b . If the two fluid layers are suddenly
exposed to a uniform, lateral pressure difference or gradient, then Euler’s equation
indicates that the acceleration of the two layers will be inversely proportional to the
individual densities, such that a velocity gradient will be created across the interface, and
thus vorticity and circulation must be created.

P1 P2

C  B

fluid a, a Ua
fluid b, b Ub

D A
x
L

Figure 10.7 Circulation for a two layer, immiscible vertical density variation exposed
to a lateral pressure gradient.

Consider the model above, and apply Euler’s streamline equation (Eq. 7.7) in the x-
direction :

1  V 2  V 
dP  d    ds  0 (10.18)
  2  t

306
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 10

Here, we will let V = U, where U is the x-direction velocity within either of the two fluids.

V  dU
Since the fluids are constrained to move in the x-direction, we also have  ds  dx .
t dt

Substituting into Eq. 10.18 and rearranging, we have:

dU 1  U2  1
dx   dP  d    dP  UdU
dt   2  

dU 1 dP dU
Or  U
dt  dx dx

However, since the velocity within either fluid will increase uniformly with time, but will
dU
not vary in the x-direction,  0 within each fluid, and we have for either fluid:
dx

dU 1 dP
 (10.19)
dt  dx

In Eq. 10.19, the pressure difference is assumed to decrease linearly with x, and to be
uniform across the two fluid layers, such that the pressure gradient is constant, and given
dP P2  P1
by   0 . Assuming that both fluids start from rest, we integrate Eq. 10.19
dx L
over time for fluid a to obtain the velocity behavior:

t U t
dU a
1 dP
t 0 dt U dU
dt  
 a dx  dt
a 0 t 0

1 dP
U a  0   t  0
a dx

t dP t
Ua    P1  P2  where P1 > P2 (10.20a)
a dx a L

Likewise, for fluid b, we get:

t
Ub  P1  P2  (10.20b)
b L

307
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 10

From Eqs. 10.20, we can see that if a  b , then Ua > Ub. Likewise, if a  b , then Ua
< Ub. In either case, each fluid will accelerate uniformly with time, but at differential
rates according to their density, such that a shear layer will be generated at the fluid
interface, resulting in the generation of circulation and vorticity.

If we apply Eq. 10.14 about the circuit A-B-C-D shown in figure 10.7, we have:
0 0
B C D A
D dP dP dP dP dP
       since dP = 0 from A-B and C-D
Dt  A
 B a C  D  b

P P
D 1
dP 2 dP 1 1
     P1  P2   P2  P1 
Dt 
P2 a

P1 b
a b

D  1 1 
   P1  P2  where P1>P2 (10.21)
Dt  b a 
D
Equation 10.21 illustrates that if a  b , then  0 , which means that negative
Dt
circulation and negative (clockwise) vorticity are generated. Conversely, if a  b , then
D
 0 , and positive circulation and positive (counter-clockwise) vorticity are
Dt
generated. Basically, under the same applied pressure (force) difference, the lower
density fluid will accelerate faster than the higher density fluid, thus creating a velocity
mismatch at the fluid interface. This velocity mismatch reflects generated vorticity,
which cumulatively reflects the change in circulation.

Although not a perfect model, this example demonstrates, how vorticity can be generated
within the ocean or the atmosphere when lateral pressure gradients act on a vertically
stratified fluid. Examples of this situation are density stratifications near the sea surface
due to strong thermal or salinity changes (although in reality these flows are a bit more
complicated). As we will discuss in Chapter 11, once vorticity has been generated it can
undergo a number of modifications, which can greatly change its character and
complexity.

10.3 Kelvin’s Theorem for a Translating Fluid Region

If we additionally restrict a material region to the flow of an ideal, inviscid fluid ( = 0),
the right side of the circulation change equation, Eq. 10.10 becomes zero, due to the
absence of viscosity. This further assumption of inviscid flow indicates that a region of
fluid moving within an inviscid flow, and satisfying the constraints of:

1) barotropic fluid, with

2) conservative body forces, and
3)   0

308
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 10

D
will undergo no change in circulation,  0 , such that   constant over time. This
Dt
D
statement is known as Kelvin’s Theorem. Fluids that satisfy Kelvin’s Theorem (  0)
Dt
are accordingly termed circulation preserving fluids, since there is no change in
circulation as the initial material region of fluid moves within a flow field.

Kelvin’s theorem indicates that if a region of inviscid flow starts from rest, or is
uniformly absent of vorticity at some initial time, then the circulation of this fluid region
will remain zero as it moves through the flow field. However, if the region has some
initial circulation, due to previous effects not satisfying the initial assumptions for the
flow, then the circulation of the region will remain constant at its initial value.

10.3.1 Examples of Circulation Preserving Flows

Potential flows, as were discussed in Chapter 9, are examples of circulation preserving

flows. For example, a material region of fluid approaching a cylinder [i.e. the potential
flow process modeled in section 9.8.1], would come from a region of uniform flow (far
upstream), and thus have no circulation. Since the flow is irrotational, and inviscid, the
material region, as it passes around the cylinder, and then departs down stream, according
to Kevin’s theorem will not change its circulation, and thus remain without circulation.
This is also the case for the flow over a rotating cylinder, that we modeled by the
additional superposition of a potential vortex [section 9.8.2]. Here again, the circulation
within a fluid region does not change as the region passes from the upstream region of
uniform flow, around the cylinder, and continues downstream. Recall that the source of
vorticity in this flow, and thus the source of circulation, is confined to the origin of the
superposed point vortex, which is contained within the modeled cylinder boundaries, and
does not contribute circulation to the modeled material flow region. Thus, the circulation
of a material fluid region, however defined, will thus remain zero as the region moves
within the flow field.

Now consider the variation of circulation along a vortex tube. As we showed in section
8.3.2.1, the circulation at the entering and exiting cross-sections of a vortex tube will be
constant, such that at any instant:
Ain
in   indA in   out dA out  out Aout
A in A out
in in out out

Thus, the circulation at any cross-section along a vortex tube must be constant. This
relationship holds regardless of the type of fluid, as long as the boundaries of the vortex
tube are comprised of vortex lines [so this also holds if the flow is viscous or baroclinic].
However, if viscous or baroclinic effects are present, Eq. 10.9 indicates that the
instantaneous circulation within the vortex tube can change within the original material
region [the material volume will remain constant, but the shape of the tube may change].

309
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 10

D
However, for a circulation preserving flow,  0 , which means that the circulation at
Dt
all cross-sections along the vortex tube will remain constant with time

A further example of the behavior of a circulation preserving fluid is illustrated by the

impact of “stretching” a vortex tube (i.e. an increase in the tube length). Such stretching
processes occur when a fluid with streamwise rotation passes through a nozzle, where the
streamwise velocity is accelerated, and thus the fluid is “stretched” in the streamwise
direction. Consider a section of a vortex tube, shown initially in Figure 10.8a below.

initial stretched


initial
stretching of
vortex tube  stretched

stretched = initial
stretched > initial
(a) initial vortex tube (b) stretched vortex tube

Figure 10.8 The effect of stretching on a vortex tube of constant cross-section

For simplicity, assume that the vortex tube is of constant cross-section and in solid body
rotation such that, initial = constant. This vortex tube has an initial circulation at any
cross-section of initial = initialAinitial. If this constant volume section of the vortex tube is
then stretched [due to a spatial change in velocity along its axis], it will undergo an
increase in length, and a commensurate decrease in cross-sectional area, as shown in
Figure 10.8b.

If the fluid is circulation preserving, the circulation will not change during this stretching
process, such that stretched = initial. Since the circulation of the stretched vortex tube will
remain constant, the total vorticity encompassed at a cross-section must increase, since
[assuming solid body rotation again]:

stretched Astretched  initialAinitial

 A 
or stretched  initial  initial  (10.22)
 A stretched 

From Eq. 10.22, it is clear that stretched > initial, since Astretched < Ainitial.
This is obviously a simplistic model, but it illustrates that the generic stretching of a
vortex tube will result in intensification of the vorticity within the tube, due to
preservation of angular momentum. In a real vortex, the vorticity will most likely not be

310
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 10

uniform [i.e. in solid body rotation], but the vortex lines will undergo a similar
intensification

Note that the opposite effect of stretching takes place if a vortex tube undergoes
compression [i.e. a reduction in the length of the vortex tube segment]. Such a
compression is accompanied by an expansion of the tube cross-section, and a reduction in
the vorticity magnitude. Such processes occur when a flow undergoes a deceleration,
such as in a diffuser, where the streamwise velocity is decreasing with streamwise
distance into the diffuser. A similar analysis to Eq. 10.22 yields:

 Ainitial 
compressed  initial   (10.23)
A 
 compressed 

Equation 10.23 indicates that compressed < initial, since Acompressed > Ainitial. So,
compression of the vortex tube results in a reduction in the local vorticity.

One caveat is that vortex compression is an unstable process, such that when a vortex
tube undergoes more than a mild compression, the vortex tube will be come unstable and
“break down” or degenerate into a complicated tangle of vorticity. This process, as we
discussed previously in section 8.3.2.1, is termed vortex breakdown, and generally results
in a loss of angular momentum, and a rapid increase in the complexity of the original
vortex tube. Vortex breakdown is not well understood, and causes the rapid, and
complicated, dispersion of the vorticity of the initial vortex tube over a broader cross-
section. While vortex stretching is a generally stabilizing process, the viscosity of a real
fluid can cause vortex breakdowns to develop in even accelerating, or stretched flows. It
has been determined experimentally that vortex tubes may break down into a variety of
different shapes, including closed and open bubbles, spirals and cone shapes. In general,
at least seven different forms or types of breakdown have been identified. See some
examples in videos shown here.

It is relevant at this point to remind the reader of the analogy of circulation to flow rate,
and vortex lines to streamlines. Recall that for an incompressible flow the flow rate is the
area integration of the velocity across a material cross-section, and circulation is the area
integration of the vorticity across a material cross-section. The flow rate within a stream
tube will also be identical at any cross-section, as will the circulation at any cross-section
of a vortex tube. The analogy falters for the temporal changes in flow rate and
circulation, since we do not have a similar corollary to Kelvin’s theorem for flow rate.
For a circulation preserving flow, the temporal flow rate within a stream tube can change
due to a temporal change in pressure, whereas circulation cannot change with time along
a vortex tube.

Streamlines and vortex lines are also defined similarly [lines with their tangent parallel to
the local vector property]. However, note that the velocity along a streamline can change
due to pressure/height changes [Euler’s equation shows us that], whereas the vorticity
along a vortex line may or may not remain constant. For a circulation preserving fluid,

311
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 10

the vortex lines are associated with a fixed set of fluid particles such that the vortex line
will move with the set of fluid particles defining the vortex line. The vorticity of each
fluid particle will thus be “frozen” at a constant value within the fluid. However, the
influence of viscosity allows the vorticity of individual fluid particles to change, such that
a vortex line will not be necessarily associated with the same fluid particles, nor will the
fluid particles retain a fixed value of vorticity. Thus, if the fluid is not a circulation
preserving fluid [due to viscous or baroclinic effects] the circulation and the vortex lines
may both change with time.

10.4 Helmholtz’s Theorems for Inviscid Flow

Helmholtz (1858) developed three theorems that govern inviscid flows, which preceded
Kelvin’s theorem. Helmholtz dealt with similar concepts for circulation preserving
fluids, but also addressed some issues that Kelvin didn’t. These three theorems are as
follows [paraphrased from Helmholtz].

(1) An element of fluid, which is originally without rotation, will remain without
rotation.
(2) Any element of fluid on a vortex line will remain on that vortex line, regardless of
the translation of the line.
(3) The product of the section and angular velocity of an infinitely thin vortex
filament is constant throughout its whole length, and retains the same value
during all displacements of the filament. Hence, vortex filaments must be closed
curves, or must have their ends in the bounding surface of the fluid.

Since an inviscid fluid [in a gravitational, barotropic environment] will be a circulation

preserving fluid, application of Kelvin’s theorem to an infinitesimal element within the
fluid implies that the circulation of that element will not change, from which the first
theorem of Helmholtz follows directly. A corollary [from Kelvin’s theorem] would be
that if an element of fluid originally has initial rotation [i.e. circulation], it will retain that
rotation throughout its movement within the fluid.

Helmholtz’s second theorem is demonstrated by use of the result of section 8.3.2.1,

where it is shown that the circulation at any cross-section along a vortex tube must
remain constant. Now, imagine shrinking a vortex tube to the point where it is of
infinitesimal cross-section, creating a vortex filament, which will contain: (a) a fixed line
of fluid particles, and (b) a single vortex line. From Kelvin’s theorem, we know that the
circulation along the inviscid filament must remain constant, and will not change as the
line moves with the fluid. One can think of a linked set of symmetric fluid particles,
which have both translational and angular velocities, where the angular velocities of the
particles comprising the line are initially identical. If gravity or a pressure gradient is acts
on the flow field, the translation of the individual particles can change, but the angular
velocity will remain unaffected, and the vortex line will move with the original fluid
particles. One might think of a set of pearls frictionlessly rotating on a deformable string.
In the absence of friction, external body or pressure forces may move the pearls about,

312
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 10

but the rotation of each pearl, and their orientation relative to each other, will be
maintained along the deformable string.

Helmholtz’s third theorem is again a variation on what was shown in section 8.3.2.1, but
also adds the need for a vortex tube or line to be contiguous. As discussed in section
8.3.2.2.2, vortex lines cannot summarily “end” within a fluid; there must be some aspect
of closure of the line, either upon itself [e.g. a vortex ring or closed loop], or it must
extend to infinity [which again actually assumes that the line must eventually close upon
itself]. This is true for both inviscid and viscous fluids. A frequently cited exception to
this vortex closure law is the hypothesis that a vortex line may end at the free surface of a
fluid [e.g. at an air-water interface]. Such “ending” of a vortex line in water is often
reflected by the dimpled swirls one sees on the surface of a stream. However, this
supposed termination of the vortex line at the air-water interface only means that the
water vortex line is matching the local vorticity [i.e. angular velocity] of a vortex line
within the air on the opposing side of the interface. Thus, a vortex line in water “ending”
at the water-air interface is contiguous to a vortex line within the air, which must
somehow “close” upon itself (albeit in a complicated manner).

As a corollary, if the interface is a fluid-solid boundary, a vortex line within the fluid
could only end at the boundary if the angular velocity of the solid boundary matches the
angular velocity (vorticity) of the vortex line. This latter situation is relatively rare.
However, this situation can occur for a fluid contained within a tank rotating about its
axis. If the fluid and tank rotate in equilibrium, both the fluid and tank will be in steady,
solid body rotation. The fluid will have a constant vorticity field, with vortex lines that
extend from the surface of the fluid and end at the rotating surface of the tank.

Generally, when a real, viscous fluid interacts with a solid boundary, vortex lines will be
generated, in the form of closed vortex “loops” either around a body [as for flow around a
sphere] or within a duct [e.g. flow within a pipe]. Figure 10.8 illustrates the general
orientation of several such vortex lines for flow around a sphere and through a
rectangular duct.

Generation of contiguous “loop”

vortex lines at fluid-duct boundary
Generation of contiguous “loop”
vortex lines at fluid-sphere boundary

Flow Vortex lines

translate with Vortex lines
streamwise flow translate with
streamwise flow
Flow

(a) (b)

Figure 10.8 Contiguous nature of vortex lines for flow (a) around a sphere, and (b)
through a rectangular duct.

313
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 10

The process of vorticity generation at a solid boundary is a topic that certainly merits
further discussion. We will revisit the behavior, dynamics, and generation of vorticity,
circulation, and vortex lines further in the following Chapter 11, when we derive the
governing differential equation for vorticity.

References

Green, S. I. (1995). Fluid Vortices: Fluid Mechanics and its Applications, Kluwer
Academic Publishers, Dordrecht.

Problems

1. A combined Rankine vortex is present in an inviscid fluid, as shown.

P

The vortex consists of a core of radius R, which is in solid body rotation with velocity
r R
V1  V0   , and a portion outside the core (rR) that has a velocity V 2  V0   . The
R r
density of the flow is , and the pressure far away from the vortex center is P.

Determine the following:

a) The vorticity,  for the entire flow field, as a function of r. Plot a graph of R/Vo vs.
r/R from 0 < r/R < 2.
b) The circulation for,
i) a circle of radius R
ii) a square of side dimensions 4R, with the vortex in the center of the square.
c) The value of D/Dt at r=R.

314
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 10

2. A vortex in an inviscid fluid has the following vorticity distribution,  vs. r, where,  = o
for r  Ro, and  = 0 for r  R.


vorticity distribution

o

r
R

If the velocity distribution is V  v  ˆi only, determine the following:
a) The velocity distribution, v(r), in terms of o and r, for r  R (assume that v(0) = 0).
b) The velocity distribution, v(r), in terms of o, R, and r, for rR.
c) The circulation for,
i) a circle of radius R.
ii) a square of side 3R, with the vortex in the center of the square.
d) The value of D/Dt at r=R.

3. A vortex is present in an inviscid fluid, as shown.

r  r
V 1  V0   V 2  V0  2  
 R
R
P

R
The vortex consists of a core of radius R, which is in solid body rotation with velocity
r  r
V 1  V0   . The portion outside the core (R  r  2R) has a velocity V 2  V0  2   . The
R  R
density of the flow is , and the pressure far away from the vortex center is P, including at r =
2R.
Determine the following:

a) The vorticity for the entire flow field, as a function of r. Plot a graph of R/Vo vs. r/R
from 0 < r/R < 2.
b) The circulation for,
i) a circle of radius R
ii) a square of side 4R, with the vortex in the center of the square
c) The value of D/Dt at r=R.

315
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 10

d) A section of an inviscid vortex tube of length L with the velocity profile shown above, is
stretched by acceleration of the fluid along the axis of the vortex tube to a length 4L, while
retaining the same rotational velocity profile. If R = R1 and Vo = Vo1 for the tube of length
L, and R = R2 and Vo = Vo2 for the tube of length 4L, determine the value of R2 in terms
of R1, and Vo2 in terms of Vo1 for the stretched tube of length 4L.

R1 Vo1 stretching of R2 Vo2

vortex tube 
4L

4. A vertical heated plate, 1 m in length, heats the adjacent H L

air such that the temperature varies uniformly across a 4 w
cm thickness, from ambient (20 C) 4 cm from the plate
PL
surface to 30 C at the plate surface. As shown in the g
model at the right, assume that the isopycnals are parallel
to the plate surface, and that the pressure varies with
height according to P=gh, where the density is the 
ambient density, and the pressure is constant perpendicular
to the plate surface. Treating air as an inviscid fluid, what h
will be the maximum velocity within the air after 1
minute? In your opinion, is this answer reasonable?
Explain your rationale for your opinion. y

x PH
5. The velocity field for an Oseen line vortex is given by:

 increasing
A   r 2 
V  1  exp  
2r   4t 

Let all parameters and variables be non-dimensional.

a. Determine the circulation, , for curves of radius r = R and r = . For each case,
determine  at
t = 0, t = R2/4, and t = .
b. Again for curves of r = R and r = , and at t = 0, t = R2/4, and t = , determine D/Dt,
both directly (by differentiation) and by using Eq. 10.10. Show that these are identical.
c. Based on your results of part a and b, explain (briefly) what is physically happening to
the vorticity and circulation of the flow field.

316
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 10

6. The velocity field for a viscous vortex is given by:

1   r 
V  1  exp  
2r   t 

Let all parameters and variables be non-dimensional.

a. Determine the circulation, , for curves of radius r = R and r = . For each case,
determine  at
t = 0, t = R and t = .
b. Again for curves of r = R and r = , and at t = 0, t = R, and t = , determine D/Dt
directly (by differentiation using the substantial derivative of )
c. Plot three separate graphs:
i. V vs. r from0 < r < 10 for t = 0.1, 1, and 5;
ii.  vs. r from0 < r < 10 for t = 0.1, 1, and 5; and
iii. D/Dt vs. r from 0 < r < 10 for t = 0.1, 1, and 5.
d. Based on your results of part a, b, and c explain (briefly) what is physically happening
to the vorticity and circulation of the flow field.

7. An inviscid fluid of thickness h and length L, as shown, varies in density according to

y
  o   . It is initially at rest. At time t = 0, a constant pressure difference is
h
applied over the length L, with a pressure P1 at x = 0, and P2 at x = L , with P1 > P2.

P1 P2
D C
y
h
A B
x
L

Neglect gravity effects, and determine:

a) Using the unsteady Bernoulli equation along a streamline, an equation for the x-
direction velocity within the fluid, u = f(y, t), for t > 0 and 0 < y < h. Plot of u(y,t)
for t=1, 2, 3 over 0<y/h<1 for (P1-P2)/L=1, o  1 , and  =0.2
D
b) The time rate of change in circulation, , for the region 0 < y < h and 0 < x < L
Dt
in terms of P1, P2,  o , and  .

317
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

Chapter 11
The Vorticity Transport Equation

Contents

11.1 Derivation of the Vorticity Transport Equation ........................................................... 319

11.2 The Physical Relevance of the Terms of the Vorticity Transport Equation .................... 321
11.2.1 Temporal Changes in Vorticity ........................................................................ 321
11.2.2 Transport (Advection) of Vorticity by the Velocity Field ................................... 321
11.2.3 Diffusion of Vorticity ...................................................................................... 323
11.2.4 Generation of Vorticity by Stretching and Tilting ............................................. 324

11.3 The Two-Dimensional Vorticity Transport Equation ................................................... 328

11.4 One-Dimensional Vorticity Solutions .......................................................................... 330

11.4.1 Flow Over a Flat Plate with Suction ................................................................. 331
11.4.2 Poiseuille Flow in a Parallel Channel ................................................................ 336
11.4.3 Steady Flow Outside a Rotating Rod ................................................................ 338

11.5 Sources of Vorticity ................................................................................................... 340

11.5.1 Vorticity Generation by Viscous Effects ........................................................... 340
11.5.2 Vorticity Generation by Acceleration .............................................................. 344
11.5.2.1 Acceleration of a Rotating Rod ........................................................... 344
11.5.2.2 Acceleration of a Circular Vessel ........................................................ 347
11.5.3 Vorticity Generation by a Streamwise Pressure Gradient ................................. 349
11.5.4 Vorticity Generation by Surface Transpiration ................................................. 350
11.5.5 Vorticity Generation by Baroclinic Effects ........................................................ 354

11.6 Circulation Changes in a Fixed Reference Plane .......................................................... 357

11.7 Vorticity Transport Equations: Cartesian and Cylindrical Coordinates .......................... 363
11.7.1 General Vector Equation ................................................................................. 363
11.7.2 Cartesian Equation Components (3-D) ............................................................. 363
11.7.3 Cartesian Equation (2-D in x, y) ....................................................................... 363
11.7.4 Cylindrical Equation Components (3-D) ........................................................... 364
11.7.5 Cylindrical Equation (2-D in r, ) ...................................................................... 364

In Chapters 8 and 10 we dealt with vorticity and circulation, and the behavior of
circulation for both viscous and inviscid fluids, as well as barotropic and baroclinic
fluids. However, those considerations where primarily concerned with the behavior of
either material regions or vortex tubes. In this chapter, we derive a differential equation
to model the behavior of vorticity as it is transported within an incompressible flow field.
Using this vorticity transport equation, we can assess the behavior of the vorticity field
within a fluid, as well as address the processes of vorticity generation and destruction. In
addition, we will employ the vorticity transport equation to reexamine the behavior of

318
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

circulation for a defined spatial region within a fluid, rather than a material region. This
will act as a spring board to begin to address the behavior of viscous flows adjacent to
solid boundaries.
11.1 Derivation of the Vorticity Transport Equation
To obtain an equation for the transport of vorticity, we begin with the Navier-Stokes
equation (Eq. 5.44), since we have shown that the Navier-Stokes equation models the
“changes” within a velocity field. Since vorticity is defined as the curl of the velocity
field, we take the curl of the Navier-Stokes equation and examine the result.


 DV 1  
   P  B   2 V 
 Dt  
 0 

DV
Dt
1
    P  g  h      2 V   (11.1)

We first use the Eq. 2.16b vector identity to expand the first term on the right hand side
of Eq. 11.1:
0
1 1 1 1 1
   P     P     P    2   P     P 
     

Recalling that     0 (see Eq. 2.16a), where  is a scalar, and noting that P and h are
scalar properties, we have   P and   h  0 , which reduces Eq. 11.1 to:

DV  1  
   2   P     2 V. (11.2)
Dt   

Expanding the left side of Eq. 11.2 using the definition of the substantial derivative and
 
vector identity Eqs.2.16g, a, and h, plus solenoidal behavior of V and  yields:
 
DV  V  
     V  V 
Dt  t 

 V 1   
   
  V2  V    V   
 t 2 

 V 1   
 
    V2    V    V

 
 t 2
  0, since V2


t
  V  
is a scalar 


DV   

Dt

t
 
 V        V (11.3)

319
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

 
Here we are able to eliminate    V 2 , because V 2 is a scalar.

Using the vector identity Eq. 2.16n, the last term on the right side of Eq. 11.2 becomes
 

   2 V   2   V   2  (11.4)

Substituting Eqs. 11.3 and 11.4 into Eq. 11.2 gives:


   1
t
 
 
 V        V   2   P   2 (11.5)
 

D
Dt

If we assume a barotropic fluid, where the gradients of density and pressure are parallel,
 
then   P  0 , and our final equation becomes [moving the    V term to the right
hand side of the equation]:
 
D   
Dt

t

  

 V      2     V (11.6)

Equation 11.6 is commonly known as the Vorticity Transport Equation, since it models
all the processes by which vorticity can be transported and modified. The constraints on
this equation are a constant density, barotropic fluid with constant viscosity. Note that a
unique aspect of this equation is that by taking the curl of the Navier Stokes equation, we
have eliminated any dependence of the fluid property of interest (vorticity) on pressure or
height changes, as long as the only body force is conservative (i.e. gravity) and the fluid
is barotropic.

The physical relevance of each of the terms Eq. 11.6 is indicated below.
    
D
Dt


t
 V   
  2     V (11.7)

total time local rate advection viscous production of

rate of of change of vorticity diffusion vorticity—by
change of of vorticity by velocity of vorticity stretching or
vorticity field tilting of existing
vorticity

320
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

11.2 The Physical Relevance of the Terms of the Vorticity Transport Equation

In the following section, we examine the processes modeled by the individual terms in
the vorticity transport equation. As indicated in Eq. 11.7, the terms represent the: (1)
local rate of change, (2) advection (transport), (3) viscous diffusion, and (4) production of
vorticity. Note that production of vorticity is a process unique to the property of
vorticity, since neither the continuity nor the Navier-Stokes equations include processes
for “production” of either mass or momentum.


11.2.1 Temporal Changes in Vorticity,
t


The term in Eq. 11.7 reflects the local temporal changes of vorticity at a point within
t
the flow field. One might think of the decay of vorticity at a point over time in a swirling
cup of coffee or tea. After the initial generation of the vorticity within the cup by stirring,
the vorticity at any point within the cup will slowly decrease due to viscous effects. If we
focus on a single point within the fluid, we will note temporal changes in the local


vorticity as reflected by the term. Figure 11.1 illustrates such a change in the
t
vorticity at a point for the decay of a viscous vortex.

Initial vorticity distribution


 A later vorticity distribution
t

Figure 11.1 An example of the temporal change of vorticity in a viscous

vortex
 
11.2.2 Transport (Advection) of Vorticity by the Velocity Field, V  
 
The V   term in Eq. 11.7 reflects the transport of vorticity, or advection, by the action
of the velocity field. Note that this term is similar to the advection of momentum term
 
V  V in the Navier-Stokes equation, and has a similar function of transporting a
material (i.e. fluid) related property by physical translation of the fluid. As the material
derivative [Section 3.3] illustrates, when fluid particles move within the velocity field,
they also carry with them any associated property, such as mass (i.e. density), momentum
(i.e. velocity itself), energy (i.e. temperature), or angular rotation (i.e. vorticity).
However, since the differential equations are Eulerian-based, changes in the respective
property are reflected by the spatial gradient of the property change [i.e. the spatial
changes of the property with position within a non-uniform property field]. Thus, as fluid
undergoes translation within a spatial region, it will experience a change in the value of a
local property due to its motion within the property field. One might think of this as

321
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

similar to the change in temperature that we experience when we walk from a shaded to a
sunny area. On a molecular level (i.e. using a Lagrangian view), it is of course a bit more
complicated, since the cumulative distribution of a property [i.e. the property “field”] is
really a reflection of the concentration of particles carrying that particular property.
However, from an Eulerian viewpoint (viewing the fluid as a continuum), we perceive
the local value of a fluid property as changing due to the translation of the fluid within
that respective property field.
 
Thus, when considering the Eulerian view of a fluid as a continuum, the V   term
reflects changes in the vorticity at any point due to the transport of vorticity by the fluid
towards or away from that point. This transport process is one means of physically
redistributing existing vorticity within the region of interest [i.e. the “field”]. Note that

since vorticity is a vector, and not a scalar,  is not exactly a gradient. It actually
reflects three separate gradients—one for each directional component of rotation [e.g. for

a Cartesian coordinate system,   x ˆi  y ˆj  z kˆ ]. Thus,
x ˆ x ˆ x ˆ
x  i j k reflects the changes in the x-direction component of
x y z
vorticity,  x , in each respective coordinate direction. However, the actual changes in
 x depend upon whether fluid is transported in a particular direction by the respective
 x
directional velocity component [e.g. a change of  x in the x-direction, , will only
x
occur if material is carried in the x-direction by the x-direction velocity component, u].
Thus, the product (i.e. dot product) of the directional velocity with the corresponding
directional vorticity gradient gives the rate of transport of the components of vorticity.

   x     y y y   z   

V    u  v x  w x ˆi   u v w ˆj   u  v z  w z kˆ
 x y z   x y z   x y z 

Transport of x-direction Transport of y-direction Transport of z-direction

vorticity vorticity vorticity

Note that a change in a particular directional component of vorticity due to advection will
only occur when a vorticity gradient exists in the direction of transport. Also, note that
even though a vorticity gradient may be present, no change in a particular component of
vorticity will occur due to advection if there is no velocity component in that direction.
Consider the example of the stationary, decaying vortex, as was illustrated above in
Figure 11.1. Although a significant gradient of vorticity exists in the radial direction,
since there is no radial velocity, there is no advection of vorticity for that flow.

Figure 11.2 below illustrates a simple example of vorticity transport for a vortex
translating through an otherwise uniform flow in the x-direction at a velocity u. Here

322
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

z z
u > 0 and  0 inside the vortex, and u > 0 and  0 outside of the vortex.
x x

Clearly, vorticity transport is taking place, but only within the region where z  0 .
x
Note that as the vortex moves along the x-axis, the vorticity transport at a point will
change from 0 to < 0 and back to 0 at any fixed point along the x-axis.

   
   
V    u v w y z=0
z z
x y z u 0
x
z<0 
x x
 
 u z kˆ  v z kˆ u(x) u(x)
x y
0
(For 2-D Flow)

Figure 11.2 Example of the transport of vorticity along the x-axis by a compact
vortex translating in an otherwise uniform x-direction flow.


11.2.3 Diffusion of Vorticity, 2

The 2 term in Eq. 11.7 models the viscous diffusion of vorticity due to fluid
viscosity. This term is similar to the diffusion transport term in the Navier-Stokes
equation, and is again associated with the redistribution of existing vorticity within the
vorticity field.

The diffusion of vorticity, like the diffusion of momentum, is the result of the interaction
of the fluid particles. In the case of vorticity, this interaction is reflected as fluid friction,
or viscosity, which causes a reduction in the local vorticity (i.e. angular momentum).
Such angular momentum losses are a result of local shear stresses created by spatial
velocity gradients. The shear stresses act as a damping agent, causing a decrease in the
local angular momentum, and thus a decay and redistribution of the vorticity within the

fluid. Assuming a constant viscosity, the 2 term reflects the transport of vorticity
due to viscous effects, where  represents the diffusivity of vorticity for a given fluid (i.e.
the fluids capacity to diffuse vorticity). Note that this term is directly proportional to the

Laplacian of the vorticity, 2 , which deserves some assessment.

In Cartesian coordinates, we can write the Laplacian term as (see Eq.2.16ℓ):


 
 2     x ˆi  y ˆj  z kˆ    x ˆi    z ˆj    z kˆ

323
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

Thus, the Laplacian represents the divergence of the gradient of each vorticity
component. To assess the dynamics of this term, consider a two-dimensional flow (e.g.

in the y-z plane) such that 2     ˆi , similar to that shown in Figure 11.3 for a
x
viscous vortex tube aligned along the x-axis.

outward
diffusion
 2
 x  x  ˆ
2 2
 2x  2
 2 
i x
 x y z 

Figure 11.3 Example of the transport of vorticity by radial diffusion from a vortex
aligned with the x-axis.

Here, x reflects the gradient, or spatial variations, of the x-direction vorticity, which is
  
given by x  x ˆi  x ˆj  x kˆ . Taking the divergence of this term gives,
x y z
 2x  2x  2x
  x    . Recalling that the divergence reflects the cumulative
x 2 y 2 z 2
expansion of a vector field, this term is indicative of the cumulative “balance” of the
spatial changes in the gradient of the vorticity field. As we discuss below in Section
11.5.1, the gradient of the vorticity can be viewed as proportional to the “flux” of
vorticity within the fluid. Thus, the Laplacian of the vorticity for a “balanced” flow (i.e.
one where   x  0 ) demonstrates that the flux of vorticity in one direction will be
offset by changes in the other directions, similar to how the velocity field for an

incompressible flow behaves (where   V  0 , by continuity). Thus, in the absence of
other transport effects, the Laplacian indicates that viscosity will have the effect of
redistributing existing vorticity in such a way that angular momentum is decreased, but
the total (i.e. the total circulation) of each component of vorticity is conserved. Again,
consider the viscous vortex shown in Figure 11.1. In the absence of advective transport,
the vorticity within this vortex will redistribute, and the angular momentum will
diminish, but the total vorticity, or circulation (taken over an infinite region) will be
conserved.
 
11.2.4 Generation of Vorticity by Stretching and Tilting,    V

 
This last term in Eq. 11.7,    V , is unique in that it models: (1) the generation of new
vorticity, and (2) the reduction of existing vorticity. This can be the result of either: (1)
the generation (reduction) of vorticity by the stretching (compression) of existing
vorticity in one direction, or (2) the reorientation (i.e. tilting) of existing vorticity from
one direction into another direction. To illustrate how this occurs, let’s expand the

   V term in Cartesian coordinates.

324
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

   V    x     ˆ
 

 y  u i  vˆj  wkˆ
 z
 x y z 
 u u u ˆ  v v v ˆ  w w w  ˆ
  x  y  z i   x  y  z  j   x  y  z k
 x y z   x y z   x y z 

Now, if we ignore the effects of viscous diffusion, and only consider the changes for the
x-direction, we can write a simplified version of Eq. 11.7 as:

Dx u u u
 x  y  z (11.8)
Dt x y z

Equation 11.8 indicates that as a fluid moves within a velocity field, the vorticity in the x-
Dx
direction can materially change (i.e.  0) due to the influence of gradients of the x-
Dt
direction velocity (u) on existing components of vorticity [here we illustrate the x-
direction changes, but the same process equally applies in y and z as well]. The first term
on the right hand side of Eq. 11.8 indicates that changes in  x will occur when the local
flow undergoes acceleration/deceleration along the x-axis. Note that an acceleration (i.e.
u
> 0) will result in a “stretching” of the fluid in the x-direction, with a commensurate
x
increase in the x-direction vorticity,  x . However, a deceleration of the flow along the x
u
-axis (i.e. < 0) will result in a contraction or compression in the x-direction, which
x
results in a reduction in the existing x-direction vorticity.

   u(x) increasing with x

  V V V
  V  x   y  z x
x y x x

(a) Stretching
ˆi  u   ˆi  u  
x y y x and y
x y u(y) components

 u
  0
Stretching

Tilting +
 x 
 u 
Compression   0
 x  (b) Tilting

Figure 11.4 Schematic of the generation of new x-direction vorticity due to (a)
“stretching”, and (b) “tilting” of existing vorticity by local velocity
gradients.

325
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

The second and third terms on the right hand side of Eq. 11.8 indicate that changes in  x
can also occur when the vorticity components orthogonal to the x-direction ( y or z )
are “tilted” into the x-direction by a gradient of the x-velocity (u) normal to the x-
u u
direction (i.e. or ). For this process, there is no generation (or reduction) of
y z
vorticity, but only a reorientation of existing vorticity components from one direction to
another. Figure 11.4 shows schematically how such processes of stretching and tilting of
vorticity can take place.

To illustrate that a stretching process does result in the generation of “new” vorticity,
let’s consider a further simplification of Eq. 11.8, such that we only consider a fluid
element with vorticity x under pure stretching, such as shown in Figure 11.4a. For this
situation, we can write Eq. 11.8 as:

Dx u
 x (11.9)
Dt x

u d (dx )
We now consider that is the rate of x-direction stretching or straining (i.e. ) of
x dx
a differential material element dx lying along a vortex line of strength  x . Now, we can
u
express via the material derivative as:
x

u du 1 D(dx )
  (11.10)
x dx dx Dt

u du
Here we let  for pure stretching in the x-direction only. Substituting Eq. 11.10
x dx
into Eq. 11.9 gives:

Dx u du x Ddx 
 x  x  (11.11)
Dt x dx dx Dt

Mathematically, we note that we can write:

D  x  1 Dx x Ddx
   2
Dt  dx  dx Dt dx Dt

Or rearranging,

 x Ddx D x D  x 
  dx   (11.12)
dx Dt Dt Dt  dx 

326
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

Substituting Eq. 11.12 into Eq. 11.11 gives:

Dx Dx D  
  dx  x 
Dt Dt Dt  dx 

or

D  x 
 0 (11.13)
Dt  dx 

Equation 11.13 indicates that, in a pure stretching process, the strength of the vorticity of
a fluid particle increases (or decreases) proportionally to the length of dx. Thus, the
generation of new vorticity, or the reduction of existing vorticity, is directly proportional
to the local stretching or contraction of a fluid element with initial vorticity x.

Alternatively, consider an inviscid vortex tube with its axis in the x-direction, and
subjected to an accelerating flow along its axis (such as in a nozzle), such that its cross-
sectional area decreases, as shown in Figure 11.5.

D u(x) increasing with x

x d
in out

L
Vortex tube
Figure 11.5 Schematic of a vortex tube undergoing stretching in a converging
nozzle.

Let us assume that: (1) the stream-wise vorticity is uniform across the tube cross section
at both the entrance and exit of the vortex tube, and (2) the section of the fluid within the
tube of length L, as shown in Figure 11.5, will elongate during transit through the tube to
a section of length  , where   L . During this transit, the diameter of the fluid section
will likewise contract from diameter D to d. As we showed in Section 8.3.2.1, the
circulation at any cross-section of a vortex tube will remain constant such that:

D2 d2
in  out or in   out  assuming constant vorticity at each cross-section
4 4
Thus,
2
D
out  in   so out  in if D > d (11.14)
d

327
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

Since the volume of the fluid section of length L must equal the volume of the fluid
section of length l, then:

D2 d2 D2 
 L    
4 4 d2 L

and substituting into Eq.11.14 gives:


out  in (11.15)
L

Therefore, the vorticity increases as the fluid within the vortex tube is stretched.
However, is there vorticity generated during this process? To answer this, we note that
the total amount of vorticity contained within each fluid section is equal to the product of
the vorticity times the fluid volume, such that:

D2
 L  Volume weighted vorticity for section L = in  L
4
d2
  Volume weighted vorticity for section  = out  
4

However, since the volumes of the fluid sections must be equal, then a ratio of the total
vorticity within the respective volumes is:

  out 
  (11.16)
 L in L

Thus, as Eq. 11.16 indicates, the stretching of a vortex tube (  1 ) will generate new
L

vorticity, whereas the contraction of a vortex tube (  1 ) will result in an overall
L
decrease in the existing vorticity.

11.3 The Two-Dimensional Vorticity Transport Equation

The vorticity transport equation, Eq. 11.7, like the Navier-Stokes equation, is a
particularly complex vector equation, which usually requires sophisticated numerical
techniques and significant computer resources to resolve for three-dimensional flows.
However, there are several tractable and relevant two-dimensional flow situations where
the vorticity transport equation can be applied and solved in closed form. In this section,
we examine the reduced two-dimensional vorticity transport equation. In Section 11.4,
we consider some simple applications of this reduced equation.

328
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

If we restrict ourselves to flows in a two-dimensional plane, we will have only one

component of vorticity—that in the direction normal to the plane. For purposes of
example, consider a two-dimensional flow in the x-y plane of a Cartesian coordinate

system, as shown in Figure 11.6. For this case, the velocity will be given by V  uî  vĵ ,

and the vorticity will be    kˆ . z

z
x-y plane
y
z

Figure 11.6 Two-dimensional vortical flow (x-y plane)

 
For this flow, note that the stretching/tilting term in Eq. 11.7,   V , becomes:

  u v u v
  V  z ˆi  z ˆj  0 since ,  0 for a flow in the x-y plane only.
z z z z

Therefore, the vorticity transport equation for this two-dimensional flow reduces to single
scalar equation:

Dz z z z   2z  2z 

 u v   2  (11.17)
Dt t x y  x y 2 

In a more generic vector form, applicable to all coordinate systems, the two-dimensional
vorticity equation can be written:
D  
Dt

t
 
 V      2  (11.18)

The direction of , and the specific velocity components, comprising Eq. 11.18 will
depend upon the particular coordinate system employed, but Eq. 11.18 will still reduce to
a single, scalar equation.

The key observation from Eqs. 11.17 and 11.18 is that for a two-dimensional flow,
material transport of vorticity will balance the diffusion of vorticity.

Two further simplifications can be made for: (1) flows where advection is irrelevant, and
diffusion dominates, such as the decaying viscous vortex shown in Figure 11.1, and (2)
steady flows, where there are no local temporal changes. The simplified equations for
these types of flows are:

 
Diffusion dominated flows:   2  (11.19)
t

329
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

Steady flows: V     

2
(11.20)

Expanding Eqs. 11.19 and 11.20 for a Cartesian coordinate system (x-y plane) gives:

z   2  2z 
Diffusion dominated flows:   2z  (11.21)
t  x y 2 

z    2  2z 
Steady flows: u  v z   2z  (11.22)
x y  x y 2 

Expansion for cylindrical coordinates (r- plane) gives:

z 1   z  1  2z 

Diffusion dominated flows:   r  2 2 
(11.23)
t  r r  r  r  

z v z 1   z  1  2z 

Steady flows: vr    r  2 2 
(11.24)
r r   r r  r  r  
Since the vorticity transport equation contains both velocity and vorticity terms, to solve
a problem for which the advection terms are relevant may require that we initially
substitute the velocity derivatives comprising the vorticity, which can make the resulting
equation quite messy. However, there are a number of useful solutions of the vorticity
equation that are independent of the advection terms, and capitalize upon the vorticity
equation being decoupled from the local pressure. In the following sections, we examine
several simple examples that demonstrate the application of the two-dimensional vorticity
transport equation. These examples entail situations for which changes in vorticity
depend upon only one independent variable (e.g. y). In Chapters 12 and 13 we will
consider more complicated examples where vorticity is dependent on more than one
independent variable (e.g. y and t), and thus require more sophisticated mathematical
approaches.

11.4 One Dimensional Vorticity Solutions

Here we examine three simple flows: two with planar symmetry and one with radial
symmetry. Two of these are flows we considered previously in Chapter 6.
Similar to our solutions of the Navier-Stokes equation in Chapter 6, the key elements
required to solve the vorticity transport equation for a particular flow situation are: (1) an
appropriate integration technique for the differential equation, and (2) specification of the
appropriate boundary/initial conditions for the flow in question. When dealing with
vorticity, the appropriate boundary conditions may not be evident. Quite frequently, the
boundary conditions may not be able to be specified in terms of vorticity, but may
frequently be specified in terms of velocity boundary conditions (since vorticity is a
derivative function of the velocity field, this generally works).

330
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

In reality, the most commonly applicable vorticity boundary condition is at certain

bounding locations where there is no vorticity. For example, for portions of a flow well
away from a solid surface, where we expect (or suspect) that a flow asymptotes to a
uniform flow, we can generally specify that the vorticity (and most likely the spatial
derivatives of the vorticity) within these regions far from a solid boundary will be zero.
However, near solid boundaries we have no comparable condition to the velocity no-slip
condition to fall back on, and thus must depend upon the use of velocity boundary
conditions at these locations. We will demonstrate this process of establishing boundary
conditions as we consider the following examples.
11.4.1 Flow Over a Flat Plate with Suction
As we will discuss in Chapter 13, an unconstrained flow along a solid surface results in
the development of what in fluid mechanics is termed a boundary layer. This is a layer of
fluid adjacent to the solid surface within which viscous effects dominate, and the flow
undergoes an adjustment from no slip at the surface to essentially the velocity of the outer
flow. This region is generally quite thin (relative to the surface length). Once a flow
encounters a solid surface, this boundary layer thickness will generally expand
continually as the flow passes along the surface. This continual expansion, due to the
viscous diffusion of momentum (and vorticity) away from the surface, makes a
developing boundary layer a complicated problem to solve, since changes continually
take place both away from and along the solid surface. However, if we consider a flow
over a porous surface with suction through the surface, we can in essence balance the
outward diffusion of vorticity with the advection of vorticity back toward the surface due
to the suction. Once this balance develops (at some point along the surface), the
thickness of the “boundary layer”, or viscous readjustment region adjacent to the surface,
will remain of constant thickness, and the flow will become fully developed (such that
there are no changes in the streamwise direction). In the present example, we derive the
corresponding velocity and vorticity field for this fully-developed viscous region with an
external uniform flow, u = U = constant, passing over a flat plate with uniform suction
of v = -V through the surface, as shown in Figure 11.7.

y
Fully U Layer within which
developed viscous effects dominate
flow

V
Figure 11.7 Uniform flow over a flat plate with suction.

For this flow, we employ a Cartesian coordinate system, as shown in figure 11.7. Note
that since we only consider the region of fully-developed flow, the location of the x-
direction origin is irrelevant (this will not be the case when we examine boundary layer
growth over a non-porous surface in Chapter 13). As shown in figure 11.7, the origin of
the y-axis is located at the plate surface.

331
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

From the continuity equation (in 2-D) we have:

u v
 0
x y
u
Since the flow is fully developed, this implies that  0 , which gives:
x
v
 0  v  constant
y

Since v = -V at y = 0, then v = -V everywhere.

Now, for a steady flow, Eq. 11.22 applies:

0 0
      2z 
2
u z  v z   2z  (11.25)
x y  x y 2 

We simplify Eq.11.25 by noting that for a fully-developed flow, all derivatives of

z  2z
vorticity in the streamwise (x) direction drop out, i.e. ,  0.
x x 2

Thus, Eq. 11.25 simplifies to:

z  2  d
v   2z  z  z ( y) only   and v = -V
y y y dy
dz d 2z
V   (11.26)
dy dy 2
Advection of Diffusion of
vorticity vorticity away
towards plate from plate

Rearranging Eq. 11.26, we have a second-order, linear differential equation for vorticity
as a function of y:

d 2z V dz
2
 (11.27)
dy  dy

Integrating Eq. 11.27 once gives:

d 2z V d
 dy 2     dyz

332
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

dz V
  z  C1 (11.28)
dy 

To determine the integration constant C1 in Eq. 11.28, we note that the outer vorticity
dz
boundary conditions for this flow are z  0 and  0 as y   [since the velocity
dy
asymptotes to a uniform flow]. Applying these boundary conditions (simultaneously) to
Eq. 11.28, gives:

0  0  C1  C1  0

Now, separating variables and integrating again:

dz V
 z
   dy


Vy

z  C2e 
(11.29)

We now have a problem. If we apply our boundary condition of z  0 as

y    0  C2 e    C2  ? So C2 is indeterminate in vorticity form.

Carrying C2 forward, we now substitute for the vorticity in Eq. 11.29 in terms of the
 v u  du
appropriate velocity derivatives, by letting z       [since v  f (x) ].
 x y  dy
This gives:

Vy
du 
z    C 2e 
(11.30)
dy

and integrating Eg. 11.30 gives:

Vy
 
u  C2 e 
 C3
V

We now can apply a velocity boundary condition of u  U  as y   , which gives C3:

U   C 2 (0)  C3  C3  U 

Our final boundary condition is the no slip condition, u  0 for y  0 , which allows us
to determine C2.

333
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

1
 V
0  C 2 e 0  U   C 2   U 
V 
So that the final expression for the velocity is:

 
Vy

u  U  1  e   (11.31)
 
V
and substituting C 2   U  into Eq. 11.30 for the vorticity, we have:


Vy
du V 
z    U e 
(11.32)
dy 

Figure 11.8 shows the general behavior of both the velocity and the vorticity within this
boundary region
y

u, z
UV U



Figure 11.8 Behavior of the streamwise (u) velocity and the vorticity ( z ) for a fully-
developed flow over a flat plate with constant suction, V.

Note that if we substitute Eq. 11.32 into the terms of the original differential equation,
Eq. 11.26, we have:

Vy
dz V2  y
V
dz V3  Advection
  U 2 e  , so V  U 2 e 
 toward plate
dy e dy 

d 2z V3 
Vy
d 2 V3 
Vy
Diffusion
  U  e 
, and  2z   U  2 e 
 away from plate
dy 2 3 dy 

These are of course equivalent (as they should be), and illustrate that diffusion of
vorticity away from the plate is balanced by the advection of vorticity back toward the
plate.

It is worthwhile to note that the governing parameter in this flow is the ratio of the suction
velocity (V) to the kinematic viscosity (). As we discussed in Section 11.2.3, the
kinematic viscosity represents the diffusivity of vorticity. For this flow, V is proportional

334
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11


to the advection of vorticity. Note also that the ratio has the units of length. Thus, if
V
the suction velocity (V) increases or the viscosity () decreases, the thickness (i.e. a
“length”) of the vorticity-bearing region becomes thinner, and thus moves closer to the
porous plate. Conversely, the vorticity-bearing region will become thicker, and extend

farther from the plate if V decreases or  increases. Additionally, the smaller the ratio ,
V
the sooner the flow will become fully developed, as shown in Figure 11.9. In the limit, as
V 0, the diffusion of vorticity will no longer be constrained by the suction, and the
vorticity bearing layer will continuously expand with distance along the plate, yielding a
flat plate boundary layer—a flow behavior we will address in Chapter 13.

Fully Developed


Figure 11.9 The effect of the ratio on the development of the vorticity-bearing
V
layer for a flat plate with suction.

As a final observation, note that the shear stress in this fully-developed region is given
by:
Vy
du V  V 0
w    U  e 
 U  e  U  V (11.33)
dy  
y 0 y 0  


Equation 11.33 is quite interesting, since the shear stress is only a function of the suction
velocity, and not the viscosity! However, this is a bit misleading. What this indicates is
that the shear stress at the surface is really equivalent to the momentum “lost” from the
main flow due to suction. This momentum is lost, because elements of fluid with a
momentum of U  enter the outer edge of the fully-developed region, are drawn across
the region by the suction velocity V, and are removed from the region through the porous
surface, having dissipated all of the original momentum due to the viscous interaction.
For a more viscous fluid, the fully-developed region will be thicker, and thus the shear
gradients will be smaller and this dissipation of momentum occurs more slowly (although
it will take the flow longer to become fully developed). For a less viscous fluid, the
fully-developed region will be thinner, with higher shear gradients, and the dissipation
will occur more rapidly (with the flow becoming fully developed more rapidly).
Interestingly, if the products of the density, flow velocity, and suction velocity for two
different fluids are identical, then the momentum lost during the transit across the fully-

335
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

developed region will also be identical, and as will the wall shear stresses, regardless of
the viscosities of the two different fluids.

11.4.2 Poiseuille Flow in a Parallel Channel

In Section 6.3.2 we solved the problem of a pressure-gradient driven flow between two
parallel plates, termed a Poiseuille flow. In this example, we demonstrate how we can
solve the same problem by employing the vorticity transport equation. What is interesting
here is that the vorticity transport equation has no pressure term. Therefore, the question
arises as to how we incorporate the pressure into the solution. The answer, as we will
show, is by use of a boundary condition that evolves from the Navier-Stokes equation.

Reconsider the basic flow behavior of a Poiseuille flow—a fully-devleoped, laminar flow
between parallel surfaces, with a constant, decreasing pressure gradient in the x-direction,
as shown below.
Decreasing pressure,

y
h
x

Here we again utilize Eq. 11.22, the vorticity transport equation for steady flow in
Cartesian coordinates.
0 0 0
z z   2z  2z 
u v   2  (11.34)
x y  x y 2 

From our solution in Section 6.3.2, using the continuity equation we showed that v = 0
for this fully-developed flow. Additionally, since the flow is fully developed, all x
derivatives are also zero. Applying these conditions, and noting that the vorticity is only
a function of y, reduces equation 11.34 to:

d 2z
0 (11.35)
dy 2

Thus, the only relevant term is the vorticity diffusion term. Integration of Eq. 11.35 is a
simplistic double integration, giving:

z  C1y  C2 where C1 and C2 are integration constants. (11.36)

The only vorticity boundary condition available to us is that on the symmetry plane, at y
u
= h/2, z  0 (where by symmetry,  0 ). Applying this condition gives:
y

336
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

C1h Ch
0  C2  C2   1
2 2

And Eq. 11.36 becomes:

C1h
z  C1y 
2
 v u  du
We also note for this flow that z       , so:
 x y  dy
du  h
  C1  y  
dy  2

and integrating again gives:

 y 2 hy 
u  C1     C3 (11.37)
 2 2 

Equation 11.37 requires two other boundary conditions. One of these is the no slip
condition we employed for the Navier-Stokes solution in Section 6.3.2:

u = 0 @ y=0 (no slip)

Which yields:

C3 = 0

So Eq. 11.37 becomes:

C1 2
u
2

y  yh  (11.38)

Note that we cannot use the boundary condition on the opposing wall of:

u = 0 @ y=h (no slip)

C1 2 C
This would give 0  
2
 
h  h 2   1 0 , which makes C1 indeterminate.
2

Thus, to establish C1 requires a third boundary condition, which we obtain from the
Navier-Stokes equation. Recall from Section 6.3.2 that the x-direction Navier-Stokes
equation reduces to:

d 2 u 1 P
 = constant (11.39)
dy 2  x

337
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

Employing equation 11.39 as a third “boundary” condition (although it applies at any

point within the flow field), gives:

1 P 1 P 2 h 2 P  y  y
C1  
 x
and u
 x
 
y  yh    1
 x  h  h
(11.40)

Equation 11.40 is, of course, the same result, Eq. 6.17, that we obtained in Section 6.3.2.

Note that for this flow:

du h P  y 
z     2  1 (11.41)
dy  x  h 

y
z h
x

Figure 11.10 The vorticity distribution for a Poiseuille flow

Thus, although this flow clearly transports vorticity, the only “changes” in vorticity
transport are due to viscous diffusion normal to the plates, in the y-direction. As Eq.
11.41 and figure 11.10 show, there must be “sources” of vorticity at either plate surface
(positive at the top, negative at the bottom), with vorticity generated at the plate surfaces,
and then diffusing toward the center, where the opposing sign vorticities mutually cancel
each other. The processes that result in this generation of vorticity at the plate surfaces
are discussed in detail in Section 11.5.

11.4.3 Steady Flow Outside of a Rotating Rod

The solutions of problems with rotational symmetry follow a similar procedure to that
employed for problems with planar symmetry. These flows generally vary with the
radial direction, r, and possess azimuthal (angular) symmetry. These types of flows may
be bounded at only one radius, or enclosed between two bounding surfaces of different
radii. The simplest of these types of flows is the viscous, steady flow external to a
rotating solid rod, which we examine in the following example.

Consider the 2-D flow outside a rotating circular rod. Assume that the cylinder is in a fluid of
infinite extent with kinematic viscosity  and density ; the cylinder is of radius R and rotates at a
constant angular velocity . Note that this flow is a variation on the generic type of flow we
examined in Section 6.4.2.

Consider a schematic of the flow, defined in cylindrical coordinates as shown below. We

assume the flow is two-dimensional, steady, with no changes in the azimuthal ()

338
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

 
direction. Consequently, V  Vr   v r ˆir  v  ˆi , and the 2-D continuity equation in r and
 yields:
0
1 1 v 
rvr   0
r r r   

d r
rv r   0  rv r  const R
dr
but v r  0 at r  R ,  v r  0 everywhere

Now considering equation 11.24, the steady vorticity transport equation in cylindrical
coordinates:
0 0 0
z v z 1   z  1  2z 
vr    r  2 2 
(11.42)
r r   r r  r  r  

Noting that vr = 0, derivatives in the -direction are zero, and z = z(r), Eq. 11.42
reduces to:

d  dz 
r 0 (11.43)
dr  dr 

This flow is again dependent only on viscous diffusion. Integrating Eq. 11.43 twice
gives:

dz
r  C1
dr (11.44)
z  C1 ln r  C 2

Since this fluid is of infinite extent, we reason that z  0 as r  , which requires that
both C1 and C2 must be zero [ since ln ()   ]. Thus, we have the surprising result
that z  0 everywhere. We now note that in cylindrical coordinates,
0
1 1 v r 1 d
z  rv    rv 
r r r  r dr
So, for z  0 we have:

1d
z  rv   0
r dr
d
rv   0
dr
C
v  3
r

339
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

Now, since v  R @ r = R, this yields C3  R 2 , and:

R 2
v  (11.45)
r

Note that this flow contains no vorticity! We note that this velocity behavior is the same
functional form as the irrotational point vortex we examined in Section 9.7.3 as a
potential flow solution. However, that was for an inviscid fluid, whereas the present flow
is a viscous fluid.

Consider the shear stress for this flow, which is given by (see Section 5.8.1):

   v  1 v r  d  v  d  R 2   R 2 
r   r        r     r    2 2 
 r  r  r   dr  r  dr  r 2   r 

This is an interesting finding, since one might assume that if the fluid displays a shear
stress, then there must also be vorticity, which is not the case here.

This absence of vorticity in a viscous flow raises the question of whether this flow ever
contained vorticity (e.g. when the cylinder first started to rotate). And if vorticity was
present, where did it come from and where did it go? The answers, as we will discuss in
the following section, are: (1) the flow did contain vorticity when the rod initially started
to rotate, (2) the vorticity was generated by the initial acceleration of the rod, and (3) the
vorticity that was initially generated all diffused to infinity as the flow came to a steady
state. In the following section, we discuss the sources or mechanisms of generation of
vorticity in a viscous flow.

11.5 Sources of Vorticity

Throughout Chapter 8, 10 and the present chapter, we have developed the governing
equations and processes that assess the effects of vorticity, its transport, and modification.
However, the presumption in these evaluations is that vorticity is already present (or not
present, in the case of irrotational flows), and that it somehow arose through the action of
viscosity or baroclinic effects. Just how vorticity is generated within a fluid is the topic
of this section. As we will see, there are three processes of vorticity generation that
require viscous interaction with a bounding surface, and one where vorticity is created
because of baroclinic behavior.

11.5.1 Vorticity Generation by Viscous Effects

Vorticity in a viscous, barotropic [ P  P() ] fluid can only be generated by fluid

interaction with a bounding surface, where viscous deformation of the fluid creates
gradients of vorticity. Just like the diffusion of heat in a solid or fluid depends on a

gradient in temperature to transfer energy as heat [i.e. q  kT ], a vorticity gradient

340
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

[actually the gradient of each directional component of vorticity, e.g. z ] is necessary for
the viscous transfer (i.e. diffusion) of vorticity. At first, this may sound a bit confusing,
since there really isn’t a physical property, like heat, that is proportional to the vorticity
gradients. However, just as we can view the gradient of a temperature field, T , as
proportional to the flux of heat energy (which is reflected by the temperature within the
material itself), we can view the gradient of a vorticity component (e.g. z ) as
proportional to the flux of that vorticity component. By flux, we mean the rate of transfer
of a property per unit area. When this property flux, whether it is heat or vorticity, occurs
adjacent to a bounding surface, this is construed as an infusion of “new” property into the
fluid, which we consider an addition to the fluid. In the case of vorticity, we choose to call
this flux across the boundary a “generation” of vorticity, since no vorticity actually existed
within the boundary, unlike heat.

Note that a property will diffuse in the direction of a decreasing property (i.e. "down" the
gradient of the property). For heat transfer, heat will diffuse in the direction of
decreasing temperature. Likewise, vorticity will also diffuse in the direction of
decreasing vorticity.

Also, the total vorticity flux really reflects the vector sum of the gradient of each of the
directional components of vorticity. However, to demonstrate the processes that lead to
viscous generation of vorticity, we will restrict our assessment to a two-dimensional flow,
with only one component of vorticity.

Consider the flow of a viscous fluid parallel to a flat surface. Such flows will develop a
region of viscous deformation adjacent to the surface (i.e. a boundary layer), where the
velocity profile adjusts from no slip at the surface to the free stream velocity of the outer
flow, as shown in Figure 11.11.
y

Figure 11.11 A typical velocity profile for a viscous flow adjacent to a flat surface.

To assess the behavior of the fluid at the surface boundary, we evaluate the x-direction,
two-dimensional Navier-Stokes equation in Cartesian coordinates (Eq. 5.45) at the
bounding surface (i.e. at y = 0). So, at the surface, we have:

0 0
 u u u   2 u  2 u  1 P 
  u  v    2
 2  gx  (11.46)
 t x y  x y   x y  0

341
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

Since the streamwise velocity, and its derivatives don’t vary along the plate, we have
u  2 u
u, ,  0 at y=0, allowing us to delete the terms indicated in Eq. 11.46, giving the
x x 2
simplified Eq. 11.47.
(neglect)
0
 2u u u 1 P
 2
 v   gx (11.47)
y y  0 t y  0 y y  0  x

Note that in Eq. 11.47 we leave open the possibility for transpiration across a porous
surface (v ≠ 0). By neglecting the gravity component (assuming gravity acts in the y
direction) and noting that the vorticity at the surface (y = 0) is:
0
v u
z   ,
x y

we can further simplify Eq. 11.47 to yield Eq. 11.48, which describes the flux of vorticity
at the surface:

Acceleration Pressure Gradient

z 1 u 1 P v
   z (11.48)
y y 0
 t y  0  x  y  0

Boundary
Flux of vorticity @ surface
Transpiration

As discussed above, at the bounding surface the flux of vorticity is indicative of vorticity
generation. As Eq. 11.48 illustrates, for a solid boundary (v = 0) there are two possible
sources of vorticity:

1) acceleration (of the surface or the adjacent fluid), and

2) a streamwise pressure gradient.

However, if the surface is porous, this also opens up the possibility for vorticity
generation by transpiration at the surface v y 0  0 . 
z
Note that vorticity is only generated when  0 . However, recall from the above
y
discussion that vorticity will diffuse in the direction of a decreasing vorticity gradient.
z
Consequently, the implication of the sign of on the type of vorticity that is generated
y
is counter intuitive. That is:

342
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

z
1) when  0 , Positive rotation vorticity will be diffused into the fluid (i.e.
y
generated) from the boundary, whereas
z
2) when  0 , Negative rotation vorticity will emanate from the boundary.
y

However, this generation process is relative to the direction of the normal for the
generating surface. If the outward normal for the generating surface is positive (relative
to the defining coordinate system), the convention is as just stated. However, if the
outward normal is in a negative coordinate direction, the sign of the vorticity generation
z
will be reversed [e.g. if  0 , and the surface normal is in a negative direction, the
y
vorticity generation will be negative, relative to the coordinate system]. If we consider
each of the terms on the right side of Eq. 11.48 as “sources” of vorticity, one can see that
a flow could have several sources of either positive or negative vorticity, depending on
the sign of each term.

We can perform a similar assessment of vorticity generation for a circular cylinder of

radius R in cylindrical coordinates, starting with the two-dimensional Navier-Stokes
equation in the r- plane. Here we consider the limiting conditions of

v   2 v 
v , ,  0 @ r=R, the cylinder surface, which yields:
  2
 v v 1 P  1 
  vr      rv 
 t r r  r  r r  r R

or z

Acceleration Pressure Gradient

z 1 v 1 P v r v

   (11.49)
r rR  t r R R   r rR

Boundary
Transpiration
Vorticity flux at cylinder surface

Eq. 11.49 is an expression for the flux of vorticity at a cylindrical surface of radius R.
Again, just as for a flat plate, the sources of vorticity are:

1) acceleration (of the surface or the adjacent fluid),

2) an azimuthal pressure gradient (in the  direction), and
3) transpiration (if the boundary is porous)

343
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

Note that the sign of the first two terms on the right-hand side of equation 11.49 is
positive, whereas those same terms were negative in equation 11.48. This is simply a
consequence of the way that the coordinate axes in the cylindrical coordinate systems are
defined.

Vorticity generation by viscous processes is a quite common, everyday occurrence,

particularly those processes associated with accelerating surfaces and pressure gradients.
Often, the vorticity generation process is the result of the action of both localized
acceleration and pressure gradients adjacent to a surface. Examples are the acceleration
and movement of vehicles (e.g. cars, planes, ships), the paddling of a canoe, and the
flight of a bird. We most frequently notice this generation of vorticity as a by-product,
which is the “shedding” or release of this generated vorticity into the fluid as discrete
vortices (more on this in Section 15.4). Although these vortices are often not visible,
their effects are often perceived by secondary effects such as pressure pulsations, noise,
or deformation of the free surface of a liquid. Anyone who has driven behind a large
truck has experienced the buffeting of your vehicle, which is the result of the truck
shedding large, discrete vortices that are felt as oscillatory pressure changes when they
impact your vehicle. The dimpling of the surface of a lake by your canoe paddle is due to
vortices shed from your paddle as it both accelerates and generates a pressure gradient as
it passes through the water. And if you have ever heard the thump, thump of a helicopter,
or the similar (but softer) sounds of a goose, duck, or bird as they come in for a landing,
or the buzzing of a bee, then you were hearing the effects of generated vorticity as it is
shed as vortices from the rotor or wings.

While the above processes are examples of where vorticity generation is commonly
observed, they are actually quite complicated processes. To assess the processes of
vorticity generation by viscous effects, we require some much simpler examples.
Therefore, in the following three sections we will examine and assess some very simple
examples that will illustrate more clearly the process of viscous vorticity generation by
surface acceleration, streamwise pressure gradients, and transpiration.
11.5.2 Vorticity Generation by Acceleration

11.5.2.1 Acceleration of a Rotating Rod

Consider an initially stationary, solid circular rod in a viscous fluid, which at time t = 0
suddenly starts rotating at a constant angular velocity  , such that:

v=R

t  0  v  0 R
r
t  0  v   R

As we showed above, for a radially symmetric system such as this, Eq. 11.48 models the
vorticity flux as:
0 0
z 1 v 1 P v r v
  
r r  R  t r  R R   r r  R

344
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

For this type of flow, the flow is azimuthally symmetric such that there will be no change
in pressure in the azimuthal () direction, and the solid surface renders v r r R  0 . Thus,
the vorticity flux for this situation is only a function of the acceleration of the rod.

z 1 v
 (11.50)
r r R  t r R

Note that Eq. 11.50 indicates that when:

v 
 0  vorticity is generated
t
v 
 0  no vorticity is generated
t
Therefore, the generation of vorticity can only occur during the initial acceleration of the
rod. After the rod reaches steady state, no more vorticity is generated, and the vorticity
that was generated can only diffuse away from the rod.
As we showed in the example of Section 11.4.3, the steady state velocity field outside of
a rod rotating at a constant speed in an infinite fluid contains no vorticity. However,
vorticity was originally present, since Eq. 11.50 indicates that a finite amount of vorticity
must have been generated within the fluid when the cylinder underwent its initial
acceleration. However, after the cylinder reached steady state, no more vorticity was
generated, and over time the vorticity generated during the initial acceleration all diffused
outward to infinity, leaving the flow without vorticity.
How much vorticity is generated? We can examine this question by assessing the
vorticity initially present within a region of fluid extending well away from the cylinder
surface. Immediately after the rod rotation is initiated, the vorticity that is generated will
still remain within a region of fluid very close to the cylinder. Fluid well away from the
rod (for purposes of example, let’s assume a circle of radius 3R) will not have noticed
that the cylinder is rotating, since the generated vorticity will not have had sufficient time

to diffuse outward, and thus the fluid at r = 3R will still be quiescent ( V  0 ).

3R


R
rod

fluid

Figure 11.12 Illustration of fluid region 3R outside rod of radius R.

If we calculate the circulation within the region r = 3R, as shown in Fig. 11.12, we have:

     2
initial     dA   V  d s   0 3Rd  0 (11.51)
A c 0

345
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

 
since V  vˆi , ds  rd ˆi  3Rd ˆi , and v r 3R  0 . However, when this flow reaches a
steady state, we know from Section 11.4.3 (Eq. 11.31) that within the fluid:

R 2
v 
r

Thus, the circulation (at any radius) for this steady state flow is:

    2  R 2 2
steady state   V  d s   rd   R 2d  2R 2 (11.52)
c 0
r 0

Note that steady state : (1) depends only on the cylinder radius, not the radius of the fluid
region, and (2) is a non-zero value. However, how can steady state be a non-zero value
when the vorticity within this flow field was shown to be zero in Section 11.4.3? The
answer is that while there is no vorticity within the steady state flow, the boundary we are
integrating around also contains the rod, which does have “vorticity”, in the form of a
solid body rotation. As we showed in Section 8.2.2, the circulation within a specified
region is the sum of all the vorticity within that region. So, for this example, we can
write that:

total  rod  fluid (11.53)

Now, if we only consider the rod, the region of r  R, the circulation at any time after the
rod reaches steady state is given by:

   2
rod   V  d s   R  Rd  2R 2 (11.54)
c 0

Substituting Eq. 11.53 back into Eq. 11.52, we have:

total  2R 2  fluid (11.55)

Thus, when the flow has reached steady state, we can set Eq. 11.52 equal to Eq. 11.55,
and solve for the circulation within the fluid:

total  2R 2  fluid  steady state  2R 2 (11.56)

Equation 11.56 indicates that in the steady state flow, fluid  0 . This supports the fact
that at steady state all the vorticity within the fluid will have diffused away to infinity, as
we showed in Section 11.4.3. However, if we set the initial circulation (Eq. 11.51) within
our region r = 3R equal to the total circulation of Eq. 11.56, we have:

346
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

total  2R 2  fluid  initial  0 (11.57)

Solving Eq. 11.57 indicates that fluid  2R 2 ! This represents the total amount of
vorticity (i.e. circulation within the fluid) that was generated by the initial acceleration of
the cylinder, and that initially resided within the fluid close to the rod. This is also the
cumulative amount of vorticity that eventually diffuses out to infinity. Since the
circulation for the fluid is negative, this indicates that the vorticity generated within the
fluid must have a negative orientation (i.e. vorticity with a clockwise sense of rotation).
As the rod accelerates in the positive  direction, this creates a positive vorticity flux,

 z 1 v 
  0,
r r R  t r R

which corresponds to the generation of negative vorticity, as discussed in Section 11.5.1.

Thus, the initial acceleration of the rod generates a finite amount of vorticity within the
fluid adjacent to the rod boundary. This initial vorticity is then redistributed within the
fluid according to our vorticity transport mechanisms [in this example, it eventually
diffuses to infinity].

Does it make any difference how fast the rod is accelerated to R ? If I accelerate the
rod very slowly, or very quickly, will it influence how much total vorticity is generated?
The answer is no. There will always be the same amount of vorticity generated,
regardless of how the cylinder accelerates. If we expand our region of integration for the
circulation outward toward infinity, the value of initial will remain zero for an extended
period, and thus fluid  2R 2 for all modes of acceleration, as long as the terminal
angular velocity of the rod is the same.

11.5.2.2 Acceleration of a Circular Vessel

Now, using the rotating rod example as a guide, what happens if I place a circular vessel
full of quiescent water on a turntable and suddenly start it rotating at an angular velocity
 [here we will again assume two-dimensional behavior]? For this situation, vorticity is
generated at the inner boundary of the vessel, but how much? Consider the circulation
within the boundaries of the vessel (i.e. @ r = R).
   
     dA   V  d s  (R )2R  2R 2 (11.58)

Here, as in our previous example, vorticity is generated adjacent to the bounding surface
during the acceleration of the vessel, yielding a circulation of   2R 2 . After the flux
of new vorticity passes into the fluid adjacent to the outer boundary, the vorticity is
spread by viscous diffusion uniformly throughout the fluid, creating at steady state a solid
body-like velocity field within the fluid, as shown in Figure 11.13.

347
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

 
Initially  Steady State

Figure 11.13 Illustration of the development of the velocity field within the fluid
inside a circular vessel subjected to sudden rotation at a constant
angular velocity .

After reaching a steady state condition, what if we suddenly bring the vessel to rest?
What happens? When the rotation terminates, v  r R , the velocity at the vessel surface,
becomes zero, and the circulation within the vessel becomes:
   
   V  d s  v  2R  0!    dA
 (11.59)

Equation 11.59 indicates that the total vorticity of the fluid within the cylinder is now
zero. What happened to original vorticity? The answer is that it was cancelled out by a
new flux of vorticity of opposite sign, which was generated at the wall surface, and
transferred to the fluid adjacent to the vessel wall when the vessel rotation was
decelerated and stopped. However, note that while the circulation for the total fluid
within the vessel will be zero at the instant of stoppage, the local response of the fluid
will be quite unsteady and protracted, while the original vorticity and the newly generated
vorticity interact through viscous diffusion to bring the velocity field within the vessel to
its new steady state of zero (vorticity, velocity, and circulation). This process is similar
to what happens to the fluid in a cup of coffee or tea after it is stirred. Since the bounding
surface of the cup is stationary, the circulation of the fluid within the cup must always be
zero. Consequently, after stirring and removal of the stirring implement, there must be
equal amounts of positive and negative vorticity contained within the cup. Over a period
of time, viscous diffusion (and some advective transport and stretching/tilting) will cause
the competing vorticities within the cup to cross-cancel, resulting in the eventual decay to
a quiescent (but well mixed) cup of beverage.

Note that while the above examples deal with vorticity generated by the
acceleration/deceleration of a surface adjacent to a fluid, the same process also holds for
the acceleration of a fluid adjacent to a stationary surface. This mechanism of vorticity
generation is most clearly apparent when considering a fluid that accelerates/decelerates
above a flat surface, such as the interaction that occurs when gusts of wind pass in
proximity to the earth. Since wind gusts (acceleration) and lulls (deceleration) occur all
the time, wind interactions with the earth provide continual sources of both positive and
negative vorticity, which subsequently interact in complicated vorticity transport
processes that further affect local wind patterns.

348
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

11.5.3 Vorticity Generation by a Streamwise Pressure Gradient

To assess the generation of vorticity by a streamwise pressure gradient, consider a fully-

developed laminar flow between parallel, bounding surfaces. This is the Poiseuille flow
that we considered in Section 6.3.2 and Section 11.4.2 above.
y

dP
0
dx h
x

In Section 6.3.2, the velocity and vorticity for this flow were determined as:

h 2 dP  y  y  y y
u   1  4Umax 1   (11.60)
2 dx  h  h  hh
y
and
4 U max  y 
z   2  1 (11.61)
h  h 
z(y)
2
h dP 1 dP 8U
where U max      max (11.62)
8 dx  dx h2

Recall that this flow has solid boundaries, is steady, and has a constant pressure gradient
that is decreasing in the x-direction (to give a positive, x-direction flow). Simplifying Eq.
11.48 for a solid boundary with a steady flow, and employing Eq. 11.62, we have for
y = 0:
0 0
z 1 u 1 P v 8Umax
   z  (11.63a)
y y 0
 t y  0  x  y  0 h2

Likewise, for the y = h boundary, we have the same result:

0 0
z 1 u 1 P v 8Umax
   z  (11.63b)
y yh
 t y  h  x  y  h h2

Thus, since the flux of vorticity is equivalent at the respective boundaries (y = 0 and y =
h), equal but opposite amounts of vorticity will be generated. Since Eq. 11.63a indicates
z
that  0 , this indicates that negative vorticity
y y  0
 4 U max 
 z  , from Eq .11.61 is generated at a constant rate from the lower

y 0
h 

349
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

surface. Note that the vorticity gradient at the upper boundary, Eq. 11.63b, also gives
z
 0 , which we would initially anticipate would indicate the generation of
y y  h
negative vorticity. However, Eq. 11.61 indicates that the vorticity generated at the
 4 U max 
boundary  z y  0   is positive. Here we note that the surface normal for the
 h 
upper boundary will be in the negative y-direction. Recall from our discussion in Section
11.5.1 that a negative surface normal will reverse our convention for vorticity generation,
such that a positive vorticity gradient results in the generation of positive vorticity.

As vorticity is generated at each of the opposing surfaces, it will diffuse toward the
midline between the plates, resulting in a mutual cancellation of the respective opposing
sign vorticities. Note that if there were not mutually opposing surfaces, but only one
bounding surface, the presence of a pressure gradient would only generate vorticity at
that bounding surface. Examples of such pressure-gradient flows are the external flow
over a cylinder or an airfoil, as we discussed in Chapter 9 for inviscid potential flows. In
the presence of viscosity, such external flows will result in the generation of vorticity at
the body boundary due to the spatially varying pressure gradients that are created over the
bodies due to their respective shapes. The subsequent transport of the generated vorticity
will of course result in a modification of the associated velocity field, which will modify
the external flow, and thus affect the subsequent generation of vorticity. The process, as
one can imagine, can get quite interactive and complicated.

Consider the external flow over a cylinder. In Section 9.8.1 we showed that inviscid
potential flow around a cylinder is subject to large spatial variations of the surface
pressure (and thus pressure gradient). For a viscous flow, the variation of the pressure
gradient from negative to positive, as the external flow circumvents the cylinder, will
result in the generation of both positive and negative vorticity within the bounding flow.
Since no opposing surface is present to generate counter-balancing vorticity, these types
of flows undergo very strong modifications from the hypothetical inviscid flow, which
further modify the pattern of vorticity generation. Often, a preponderance of one sign of
vorticity or the other will be generated, resulting in flow field adjustments that allow the
flow over the body to release or "dispose" of the continually generated and accumulating
vorticity. Often these flow field adjustments result in very unsteady and often
complicated processes of vorticity “shedding”, often as discrete vortices, which pass into
the departing flow field. We will discuss these types of external flows, and the processes
for shedding vorticity later in Chapter 15.

11.5.4 Vorticity Generation by Surface Transpiration

While transpiration through bounding surfaces is not widely encountered in many

practical fluid applications, transpiration is of importance in such things as turbine blade
cooling, paper making, and flow control (particularly for control of separated flows,
which we discuss later in Chapters 14, 15, and 16). To examine vorticity generation due
to transpiration through a porous surface, we utilize the example that was done in Section

350
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

11.4.1 for a fully developed flow over a flat plate with suction. From Eq. 11.31 and
11.32, the expressions for the velocity and vorticity were determined as:

 
Vy

u  U  1  e   (11.64)
 
Vy
du V 
and z    U e 
(11.65)
dy 

This flow is steady with no streamwise pressure gradient, but with a constant suction
velocity through the bounding surface, v y 0  V . The vorticity flux equation, 11.48,
thus simplifies to:
0 0
z 1 u 1 P v
   z (11.66)
y y 0
 t y  0  x  y  0

Substituting v = -V = constant for this flow and employing Eq. 11.65 @ y=0, we have:

z V V2
 z  U (11.67)
y y0
 y 0 2

z
Since  0 , this implies that vorticity is being generated at the surface (which has
y y 0

V
a positive normal) in the amount z y 0  U , as shown in Figure 11.14. This

negative vorticity will diffuse outward from the surface. However, as we discussed in
Section 11.4.1, this vorticity is simultaneously transported (by advection) back towards
and removed through the surface by the suction. Because of this balance between
viscous diffusion and transpiration, the velocity profile for this flow does not change, and
attains the fully developed condition shown in Figure 11.14.

y
u
z

u, z
UV U

 V
z
 0  " vorticity flux"
y y0

Figure 11.14 Generalized velocity and vorticity curves for flow over a porous flat
 
plate with suction v y 0  V . The positive surface gradient of
vorticity indicates a flux of negative vorticity across the boundary.

351
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

However, what will happen if, as the flow passes along the flat surface, the flow leaves
the porous surface with suction, and passes onto a solid boundary with v y 0  0 ?
z
Equation 11.66 then would give  0 , which indicates that there will be no flux of
y y 0

vorticity from the solid surface. Consequently, no additional vorticity will be generated,
nor will the existing vorticity in the flow field be restrained by the surface suction.
Therefore, the existing vorticity will diffuse outward as the fluid moves along the solid
boundary, resulting in a continual increase in the thickness of the vorticity bearing layer.
This process, as we will discuss in Chapter 13, results in a growing “boundary layer.”
The flow for this situation is no longer fully developed, and (as we will see in Chapter
13) the analysis of the flow becomes more complicated. The general character of this
flow over a solid boundary is represented in Figure 11.15.
y
z u

u, z
U

Figure 11.15 Generalized velocity and vorticity curves for flow over a solid flat
 
plate v y 0  0 . Note that there is no vorticity flux since the gradient
of the vorticity at the surface is zero.

z
Note that although  0 , at the boundary we have z  0 (from the previous
y y 0
y 0

developed vorticity distribution). Since no new vorticity is generated over the solid
boundary, the total vorticity that initially existed within the vorticity-bearing layer will
remain constant and negative, and will continually redistribute as the flow advects
downstream, with a non-zero vorticity value at the boundary.

Now consider what happens if we transition from a porous plate with suction, to a porous
plate with injection of fluid through the surface and into the flow field. Assuming that
the injection is equal, but opposite, to our suction process, this would transition from a
boundary condition of v y 0  V to a boundary condition of v y0   V . Since this flow
would start with an initial distribution of negative vorticity z  0 within the vorticity-

bearing region, and a non-zero vorticity at the boundary z y  0  0 , the vorticity flux
will now be given by:

z V
 z 0 (11.68)
y y 0
 y 0

352
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

Equation 11.68 indicates that injection will result in the generation of positive vorticity at
the surface (recall our discussion in Section 11.5.1). This positive vorticity will be both
advected (by the injected fluid) and diffused away from the surface, initially modifying
the vorticity and velocity profiles as shown in Figure 11.16.

y y
u z u
z

u, z u, z
U U
v=V v=V
(a) Shortly after transition to injection (b) After extended exposure to injection

Figure 11.16 Generalized velocity and vorticity curves for flow over a porous flat
 
plate with injection v y0   V . Here the negative slope of the
vorticity at the surface indicates a flux of positive vorticity across the
boundary.

As shown in Figure 11.16a, the flux of positive vorticity due to injection will begin to
cross-cancel with the original negative vorticity distribution within the vorticity bearing
layer, such that the vorticity distribution will develop a vorticity maximum (in absolute
terms) away from the plate surface. At this maximum, the slope of the vorticity is zero,
z
 0 . This null vorticity gradient roughly reflects a null second derivative of
y
   2u 
streamwise velocity  z   2  0  , which reflects a cross-stream inflection in the
 y y 
streamwise velocity. As we will discuss in Chapters 13 and 17, a cross-stream inflection
in the streamwise velocity is generally an undesirable flow characteristic, since it is a
precursor of unstable flow behavior, and can quickly result in a breakdown of the flow
from laminar to turbulent flow.

Assuming that a stable flow can be maintained, as the flow proceeds along the porous
plate, the fluid injection results in further generation of positive vorticity, and the
vorticity profile will skew even further, with z y  0  0 . When z y  0  0 , as shown in
Figure 11.16b, the vorticity gradient at the bounding surface becomes zero,
z V
 z  0 , and the flux of vorticity across the boundary will cease. Since
y y  0  y  0
z y 0
 0 , this means that the slope of the velocity at the surface will be zero,
u
 0 , such that the flow closely adjacent to the surface will become essentially
y y  0
stationary. Since no new vorticity is generated, the flow again grows only through a

353
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

redistribution of the vorticity within the vorticity bearing layer (assuming that it remains
in a stable, laminar flow--which is unlikely). Thus, as the flow moves along the
boundary, and fluid injection continues, the vorticity-bearing layer will spread farther
from the surface, but there will be no further addition of vorticity.

11.5.5 Vorticity Generation by Baroclinic Effects

As mentioned in Section 11.5, baroclinic effects [P ≠ P()] also can generate vorticity. In
Chapter 10 (Sections 10.2-10.5 in particular), we illustrated that when the gradients of
pressure and density are not aligned, such as in buoyancy-induced flows, there will be
temporal changes in circulation for a defined region of fluid. In this section, we address
the generation of vorticity that gives rise to the circulation changes due to baroclinic
behavior. To assess this behavior, we consider Eq. 11.5, the equation we derived leading
to the vorticity transport equation:

    1
t
  
 
 V        V   2   P   2
 
or rearranging:
 
D    1
    V   2   2   P (11.69)
Dt  

D
Here, we know that reflects the changes of the vorticity field due to material
 Dt

changes,    V models vorticity addition/removal due to stretching/compression

effects, and 2 represents diffusion of existing vorticity due to viscous effects. The
 1 
 2   P term is an interesting term, in that is does not deal with a modification or
 

transport of existing vorticity since it requires no action on  , the vorticity field. Thus,
this is a term that can create vorticity where none existed, and without the assistance of
viscosity. For simplicity, consider a two-dimensional (Cartesian, x-y plane), inviscid
flow. Eq. 11.69 then simplifies to:

D  1 
    P (11.70)
Dt  2 

Equation 11.70 illustrates that if the gradients of pressure and density do not align, the
cross product of the gradients will be non-zero, and thus vorticity will be generated. For
  ˆ  ˆ
example, in an x-y plane the component of vorticity is   z kˆ , with   i j
x y
P P
and P  î  ĵ . Thus, Eq. 11.70 reduces to a single component equation for z ,
x y
given by:

354
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

Dz  1   P  P 
  2    (11.71)
Dt    x y y x 

While Eq. 11.71 looks a little daunting, it is really quite instructive as to how baroclinic
flows generate vorticity. Consider the sea breeze problem we examined previously in
Section 10.2.3. For that example, we made the approximation that isobars of pressure
would align with the earth, but the isopycnals of density would align vertically (due to the
temperature differential between the air over land relative to the air over the ocean). This
is generically illustrated in Figure 11.17 below.
y

x
Land Sea

Figure 11.17 Idealized distributions of air density and pressure from land to sea.

Figure 11.17 is highly idealized, but illustrates the most extreme case. Here we assume P
 P
= P(y) and  =  (x), and  0 , and  0 (since pressure will decrease with altitude).
x y
Since the gradients are orthogonal to each other, Eq. 11.71 reduces to:

Dz  1   P 
  2    0 (11.72)
Dt    x y 
(+) (+) (-)
Equation 11.72 indicates that the conditions shown in figure 11.17 will generate negative
vorticity. Recall in Section 10.2.3 that for these same types of conditions, we determined
that the air would experience a negative change in circulation, reflecting a clockwise air
flow, which is of course consistent with the generation of negative vorticity. One might
think of this as the denser air over the sea, under the influence of the pressure gradient,
moving inland (on shore) to displace the lighter air, which must move upward and
seaward to satisfy continuity.

Of course, in the evening, when the land cools faster than the sea, the density will be
greater over the land than the sea, causing the generation of positive vorticity and a
counter-clockwise circulation, such that the direction of the air flow will reverse to a
seaward (off-shore) breeze.

355
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

Now, consider the movement of a variable density fluid through a duct due to a pressure
difference, as shown in Figure 11.18.

Velocity

Figure 11.18 The flow of a fluid with vertical density variations in a horizontal duct.

Here, the density varies with y and the pressure with x. If we assume that the flow is left
P
to right, then the pressure must decrease in the positive x-direction and  0 . Initially,
x

we’ll assume that the density decreases in the positive y-direction, so  0 . Thus, Eq.
y
11.71 then reduces to:

Dz  1   P 
   2    0 (11.73)
Dt    y x 
(+) (-) (-)
Equation 11.73 indicates that negative vorticity will be generated as the flow proceeds in
the x-direction. Physically, we reason that the pressure gradient will accelerate the less
dense material faster than the more dense material, thus a differential in cross-stream
velocity will develop. Since we assumed the flow to be inviscid, the flow would continue
to accelerate differentially, and thus continually generate negative vorticity, such that the
total vorticity will continually increase (since vorticity does not dissipate in an inviscid
fluid). Note that if the density increased in the y-direction (opposite to Figure 11.18),
then positive vorticity would be generated.

The present discussion illustrates how baroclinic processes are another source of vorticity
within a flow field. Here, the vorticity does not emanate from a boundary, nor is
viscosity necessary. As pointed out in Chapter 10, such baroclinic processes are a key
element in many atmospheric and oceanic flows. The particular development of rotating
flows, such as tornados and hurricanes, depend on strong thermal warming, and thus
density variations, to generate vorticity. Such density-generated vorticity can accumulate
and concentrate into tornadic flow patterns. Taken to an extreme, very strong
temperature/density variations in wild fires have, under the right terrain and conditions,
generated whirling flame concentrations termed “firenados”. This link shows several
images of such phenomena, as well as some scary videos. Strong density gradients within
the oceans can also yield strong rotational behavior, which can create/modify ocean

356
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

currents, which in turn can modify the local weather patterns (e.g. such processes
contribute in complicated ways to the cyclical El Niño/La Niña weather patterns).

11.6 Circulation Changes in a Fixed Reference Frame

In Section 10.1, we examined how circulation can change for a region of fluid that moves
within a viscous flow field. In the present section, we examine the changes in circulation
that can take place within a fixed region as the flow field moves through it. While in
Section 10.1 we focused on a defined material region, which was free to move and
deform, here we focus on a defined spatial region, through which the flow will pass. Our
concern is how the circulation within this fixed region changes as vorticity-bearing fluid
passes into and out of the region. Here we consider a fixed curve in space, as shown in
 
figure 11.19, for which the circulation is given by a line integral as    V  d s .

 
Figure 11.19 A fixed region in space, with circulation    V  d s .

Now, since we consider a fixed curve, and don’t follow a material region, we can write
the change in  with time as a conventional, not a material, derivative:

    D
t t 
 V  ds (note: this is not )

Dt
V 
=  ds (11.74)
t

Note that we can bring the time derivative inside the integral, because the region over
which we integrate is of fixed dimensions (see Liebniz rule on integration limits).

However, using the Navier-Stokes equation (Eq. 5.44), we can solve for the time
derivative of velocity, giving:
neglect

  P 
V
t

  V V 

 
 g   2 V (11.75)

We now substitute Eq. 11.75 into 11.74, and simplify and consolidate the terms, using a
vector identity from Chapter 2 [Eq. 2.16g ---prove to yourself how we get this
relationship], and then Stokes' Theorem, section 2.4.4.

357
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

    

t
 
  V   V  ds  
P 

 d s    2 V  d s
   V 2    

      V     d s  
P 
 d s    2 V  d s
  2  

0, since operating on a scalar, Eq.2.16a

    V 2    

       V        dA  
P 
 d s    2 V  d s (11.76)
A   2  

From Stokes’ Theorem =0

(if P=P()
or constant) C

Thus, for barotropic flows [P=P()], Eq. 11.76 reduces to: A


t
 
 

 
       V  dA    2 V  d s (11.77)
A C

Note that Eq. 11.77 is similar to Eq.10.10, with the exception of the additional first term
on the right hand side of Eq. 11.77. This term, as we will illustrate in the following,
represents the transport of vorticity into and out of the spatial region C.

As an illustration of how Eq. 11.77 can be employed, consider the boundary layer that
develops when a viscous fluid flows over a stationary flat plate, as shown in figure 11.20.
We want to examine how and why the vorticity changes as the flow passes onto and over
the plate surface. To do this, we consider an arbitrary rectangular region C, as shown by
the dotted line in figure 11.20, which is h high, and extends in a streamwise direction
from x1 to x2. Here, we assume that h is high enough that the upper extent of region C is
in a region of uniform flow, such that any deformation of the passing fluid takes place
only within the region C.

U, P U, P Edge of boundary layer

Region C

y h Since this is a steady flow,

x for region C
x1 x2

Figure 11.20 The development of a boundary layer by a viscous fluid impinging on,
and passing over, a stationary flat plate.

If this is a steady flow, then  0 , which means that Eq. 11.77 reduces to:
t

358
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

    

V  dA

  
 
V  ds 2
(11.78)
A C

Now, consider the viscous term on the right of Eq. 11.78. Examining the flow in figure
 
11.20 within a two-dimensional x-y plane for V  uˆi  vˆj and ds  dxˆi  vˆj , we have
that:
 
    2V  2V  
 V  d s   2  2   d s
2 .
  x  y  0 or negligible on y
0 at x boundaries boundaries

  2u  2u   2v 2v 
  2  2 dx   2  2 dy (11.79)
 x y   x y 
 2u
As we will show in Chapter 13, The only significant term in Eq. 11.79 is  dx , since
y 2
changes of u with respect to x are zero along the x-directed boundaries of region C, and
since v << u, changes in v at the y-directed boundaries are either zero, or negligible.
 2u
Additionally, the remaining 2 term will contribute to the integral in Eq. 11.79 only
y
 2u
along the x-boundary on the plate surface at y = 0, since  0 at the y = h boundary,
y 2
where the velocity profile is uniform. Thus, we can reduce the viscous term of Eq. 11.79
to:

xx2
    2u    2u 
C C  y2  x x  y2  dx  negligible terms
2
 V  d s    dx   (11.80)
1 y 0

Now consider the x-direction Navier-Stokes equation at the plate surface, which reduces
to (after dropping terms that are zero at y = 0):

 2u 1 P
  (11.81)
y 2 y 0
 x y  0

Substituting Eq. 11.81 into Eq. 11.80 we get:

  x 2  2u x2
1 P P2  P1
  V  ds  x  y2   x
2
dx  dx  (11.82)
x1

1 y0 y 0

Here P1 and P2 are the pressures at the respective x locations on the plate surface. Thus,
the impact of viscosity on the circulation of region C is reflected by the change in
pressure along the plate surface.

359
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

The first term on the left of Eq. 11.78 is a bit harder to assess. If we expand the integrand
 
for a two-dimensional flow in the x-y plane, where V  uˆi  vˆj ,   z kˆ , and

dA  dx dy k̂ , we can write:

    

V  dA
 
x

y

    u    v  dxdy
z z (11.83)
A A  

Now if we integrate this term over the area encompassed by the region C,
x1  x  x 2 and 0  y  h , we have:

 h x2  x2 h

   
 
A     V  dA  0 x x z u dxdy  x 0 y z vdydx
1 1
0 0
h h x2 x2

  uz x  x 2 dy   uz x  x1 dy   v 

z yh dx   v 
z y 0 dx
0 0 x1 x1

(11.84)
Vorticity Vorticity
advected advected
out right face in left face

Here, we note that z y  h  0 (since we have uniform flow at y = h) and v y0  0 (at the
plate surface), which negates a contribution by the last two integrals of Eq. 11.84. If we
now substitute Eqs. 11.82 and 11.84 back into Eq. 11.78 we obtain:

h h
P2  P1

  u 
0
z xx2 dy   uz x  x1 dy
0
(11.85)

Pressure Flux of vorticity

changes transported across
along plate fixed region C

Equation 11.85 indicates that if P2  P1 , there is no new vorticity created inside of region
C. However, if P2  P1 , vorticity is created inside of region C.

Now, consider two different fixed regions above the plate, C and C', as shown in figure
11.21. Region C encompasses only the immediate leading edge region of the plate,
whereas region C' encompasses a region, that begins just after the leading edge region,
and extends downstream. Using Eq. 11.85, we can assess the pressure and vorticity
behavior within each of these respective regions, and the impact of pressure changes on
the generation of vorticity.

360
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

U, P
C C'
Ustag= 0
P1=Pstag y h
P'2= P
P2=P'1=P
x
x1 x2
x'1 x'2
Figure 11.21 Illustration of two fixed regions encompassing the boundary layer
(dotted line) created by a viscous fluid impinging on, and passing over,
a stationary flat plate. Region C encompasses the leading edge of the
plate; region C' encompasses a region following the leading edge
region. Note that U and P are the free stream velocity and pressure.

For curve C, Eq. 11.85 gives,

0
P  Pstag h h


  u 
0
z xx2 dy   uz x  x1 dy
0
(11.86)

U 2 vorticity vorticity
 transport out transport in
2

P  Pstag
U 2
Here, we note that  comes from applying the Bernoulli equation from
 2
the impinging free stream to the leading edge of the plate (see figure 11.21), which we
surmise is a stagnation point. Since the flow undergoes little deformation prior to
impacting the stagnation point, the use of the Bernoulli equation is a reasonable
assumption. Additionally, since the flow will be uniform until the point of impingement,
we surmise that no vorticity is transported into region C, or z x  x  0 . Thus, rewriting
1

Eq. 11.86, we have:

h
U 2
 uz x  x
0
2
dy  
2
(11.87)

Equation 11.87 implies that vorticity is transported out of region C, so vorticity must
have been generated at or in the vicinity of the leading edge of the plate.

Now, considering curve C', Eq. 11.85 yields:

h h
P  P

  u 
0
z xx2 dy   uz x  x1 dy  0
0
(11.88)

vorticity vorticity
transport out transport in

361
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

In Eq. 11.88 we assume that the pressure at both x1' and x2' is P, since the streamlines
along the plate will be essentially parallel, and the pressure of the outer region of uniform
flow will be “impressed” across the thin boundary layer region (we confirm this at the
end of section 13.2.1). Thus, rearranging Eq. 11.88, we have:

h h

 uz x  x dy   uz x  x dy
0
2
0
1
(11.89)

vorticity = vorticity
transport out transport in

Equation 11.89 indicates that the vorticity transported into and out of C' are identical.
Thus, the vorticity generated at the leading edge of the plate is the source of all the
vorticity for a flat plate flow. Once the vorticity is generated at the leading edge, it will
only redistribute laterally by viscous diffusion, as it is advected along the plate with the
flow. Of course, if the external flow accelerates or decelerates due to geometry changes,
this will result in pressure changes within the uniform flow external to the boundary
layer. And as we discussed in Chapter 9 on potential flows, these inviscid-like pressure
changes within the outer flow will be impressed across the boundary layer, and reflected
as changes in streamwise pressure along the plate surface. As shown in Section 11.5.3,
the presence of such streamwise pressure changes along the surface will give rise to
additional vorticity generation (negative or positive, depending on the change in surface
pressure), and thus modification of the development of the viscous boundary layer.

In Chapter 13, we will examine how this additional vorticity is redistributed within a
developing boundary layer, and how pressure gradients impact the development of
boundary layers. As we will see, additional vorticity generation by surface pressure
changes can either minimize boundary layer growth, or cause very rapid growth,
depending on whether the pressure decreases or increases in the direction of flow.

362
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

11.7 Vorticity Transport Equations: Cartesian and Cylindrical Coordinates

11.7.1 General Vector Equation

   
D    
  (V  )  (  V)   2 
Dt t

11.7.2 Cartesian Equation Components (3-D in x,y)

x-dir.:
         2  2  2 
 x u x v x w x     u   u   u    x  x  x 
 t   x x y y z z   
x y y  x 2 y 2 z 2 
   
y-dir.:
        2  2  2  
 y y y y   v v v   y y y
  t  u x  v y  w y     x x   y y  z z      
   x 2  y 2  z 2 

   
z-dir.:
         2  2  2 
 z u z v z w z    w w w   z z z 
       
 t x y y   x x y y z z 
 x 2 2 2 
  y z
 

11.7.3 Cartesian Equation (2-D)

z  z  z   2 z  2 z 
u v    

t x y  x
2
y 2 

11.7.4 Cylindrical Equation Components (3-D)

r-dir.:
r  r v    r   r
 vr       v z
t r r    z
v   v v   1  2 2


  r r    r  v     z r    rr   12  2 r   2 r  22  
r r    z  r  r r  r  z r  
-dir.:
  v     
 vr    r   v z
t r r    z
v    v   v    1   1     2  r 
2 2
 r   v r   z    r   2 2  2  2 
r r    z  r  r r  r  z r  

363
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

z-dir.:
z  z v   z  z
 vr   vz
t r r  z
v z  v z v z 1    z  1  2  z  2  z 
 r   z   r  2  
r r  z  r r  r  r 
2
z 2 

11.7.5 Cylindrical Equation (2-D in r, )

z z v  z 1   z  1  2 z 

 vr    r  2 2 
t r r   r r  r  r  

364
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

Problems

1. Starting with Eq. 11.2, derive Eq. 11.6, showing all the vector simplifications required
in detail.

2. An inviscid, three-dimensional flow is described by the velocity field:


V  5x  yˆi   3y  zˆj   2z  x kˆ
Determine:

a) The vorticity field for this flow

b) The vortex stretching and tilting that is taking place, by calculating the value of

   V .

c) Separate the    V value of part b into the stretching components and the tilting


components.

3. A circular tank filled with water rotates on its axis such that the water in the tank is in
steady state rotation; the wall of the tank, of radius Ro, moves with a tangential velocity
of Vo (see figure).
V0 î 
îr

 R0 Tank

Do the following:

a) Write an expression (do not derive) for the velocity profile of the water within the tank,
v(r), for the steady state conditions;

At t = to, the rotation of the tank is abruptly stopped. Assume that the tank is deep enough
such that the flow behaves two-dimensionally (r,  dimensions). Do the following:

b) When the tank is stopped, indicate: (1) where vorticity is generated and (2) what causes
the vorticity generation.

c) State the steady-state solution for the velocity, v(r), in the tank when t

d) Determine the total amount of vorticity (i.e. "new" circulation) that is generated for t > to.

365
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

4. In section 11.4.3, we considered the 2-D flow outside a circular rod of radius R, rotating at a
constant angular velocity  in a fluid of infinite extent with kinematic viscosity  and density
. The solution for the steady-state velocity field in the fluid external to the rod using the
vorticity equation was:
R 2
v  .
r
a) Determine the vorticity for r > R and provide a brief physical explanation of what this
result implies in one or two sentences.

b) Determine the circulation for r=2R and r=. Again provide a physical explanation of
what this result implies in one or two sentences.

c) If I suddenly stop the rod rotating, assuming a viscous and laminar 2-D flow, which of the
terms in the vorticity vector equation shown below would be relevant to the solution of the
subsequent unsteady flow behavior? Indicate the terms that would be non-zero by
indicating the identifying number in the general equation below.
  

t

 

 V        V   2 


1 2 3 4
d) Briefly explain why the terms you did not select are not relevant.

e) After the rod rotation is stopped, is there any vorticity in the fluid? If so, where does it come
from? How much is generated?

v u
5. At a given instant of time, the vorticity,  z   , for a two-dimensional,
x y
inviscid (µ=0), constant density flow field is specified at each point in the field
(z  0).

Do the following:
a) Using appropriate simplifications of the vorticity transport equation, give an
argument that shows that if the particle path lines are known (i.e. the velocity

field, Vx, y, t  , is known), one could determine z for each point in the flow
field at a later time.
 2  2
b) Show that for this flow field an equation,    z can be developed
x 2 y 2
for the stream function, .

c) The circulation within a marked circle of fluid in the flow field is 0 at time t=0.
If the marked fluid moves within the flow field, such that at a later time t=1 the
circulation within the marked fluid is 1. What is the relationship between 0 and
1 and why?

366
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

6. A circular tank of water of radius Ro rotates such that the velocity is Vo at Ro . A

stationary rod of radius Ri, located at the center of the tank.(see figure below).
Assume the rod and tank are infinite in the z-direction so the flow in the annulus
between the rod and the tank wall is two-dimensional.

î îr
Vo

r
Tank
Ri Ro

Do the following:

a) Simplify the following 2-D vorticity transport equation to allow you to calculate the
vorticity distribution in the water, if the flow is at steady state conditions. Indicate
the terms that can be neglected in the equation, and list the reasons that they are
assumed/can be show to be negligible

b) For one boundary condition, let the vorticity at r=Ri be z = o. A second
dz
boundary condition is  0 at r=Ro. Explain how this second boundary
dr
condition arises.

c) Using the vorticity boundary conditions in (b), solve the simplified equation from
(a) to determine an expression for r  .

7. A circular tank of water of radius Ro rotates such that for t < to the velocity is Vo at
Ro. A stationary rod of radius Ri, located at the center of the tank.(see figure below).
Assume the rod and tank are infinite in the z-direction so the flow in the annulus
between the rod and the tank wall is two-dimensional. One can show (see problem 5),
what at steady state, the vorticity within the annulus is z  o = constant.

î îr
Vo

Tank
Ri Ro

367
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

At t = to, the rotation of the tank is abruptly stopped (Vo=0 for t > to). Assume that the
flow behaves two-dimensionally (r,  dimensions). Do the following:

a) Simplify, using appropriate assumptions, the 2-D vorticity transport equation to

yield an equation that could be solved for the vorticity and velocity field inside the
tank for t > 0 (DO NOT ATTEMPT TO SOLVE THIS EQUATION!).

b) Determine the total circulation within the tank (inside r = Ro ) for both t < to and t >
to. Briefly explain your result for both cases.

c) Using your result for t < to from part b) of this problem, determine the value of o
in terms of Vo, Ri, and Ro.

8. A circular rod of radius Ri rotates at an angular velocity, , inside of a stationary tank

of water of radius Ro (see figure below). Assume the rod and tank are infinite in the z-
direction so the flow in the annulus between the rod and the tank wall is two-
dimensional.

î îr


r
Tank
Ri
Ro

Do the following:

a) Simplify the two-dimensional vorticity transport equation to allow you to calculate

the vorticity distribution in the water, if the flow is at steady state conditions.
Indicate the terms that can be neglected in the equation, and list the reasons that
they are assumed/can be shown to be negligible

b) For one boundary condition, let the vorticity at r=(Ri+Ro)/2 be z = o. A second
d z
boundary condition is  0 at either r=Ri or Ro. Explain how we get this
dr
second boundary condition from the concept of vorticity generation.

c) Using the vorticity boundary conditions in b), solve the simplified equation from a)
to determine an expression for r  .

368
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

9. A circular rod of radius Ri rotates in steady state at an angular velocity, , inside of a

stationary tank of water of radius Ro (see figure below). Assume the rod and tank are
infinite in the z-direction so the flow in the annulus between the rod and the tank wall is
two-dimensional. One can show (see problem 7), what at steady state, the vorticity
within the annulus is z  o = constant.
î îr

 r
Tank
Ri
Ro

At t = to, the rotation of the rod is abruptly stopped. Assume that the flow behaves two-
dimensionally (r,  dimensions).

Do the following:

a) Simplify, using appropriate assumptions, the 2-D vorticity transport equation to

yield an equation that could be solved for the vorticity and velocity field inside the
tank for t > to (DO NOT ATTEMPT TO SOLVE THIS EQUATION!).

b) Determine the total circulation within the tank (inside r = Ro ) for both t < to and t >
to. Briefly explain your result for both cases.

c) Using the steady state result for t < to that z  o = constant, determine the value
of o in terms of , Ri, and Ro.

10. A circular rod of radius Ri rotates at an angular velocity, , inside of a stationary tank
of water of radius Ro (see figure below). Assume the rod and tank are infinite in the z-
direction so the flow in the annulus between the rod and the tank wall is two-
dimensional.
î îr

 r
Tank
Ri
Ro

Do the following:

a) Simplify the two-dimensional vorticity transport equation to allow you to calculate

the vorticity and velocity distributions in the water, if the flow is at steady state

369
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

conditions. Indicate the terms that can be neglected in the equation, and list the
reasons that they are assumed/can be shown to be negligible

b) Specify the necessary boundary conditions, for both z and v to allow you to solve this
equation (you must specify at least one condition for the vorticity);

c) Solve the O.D.E. to obtain z(r) and v(r) in terms of , r, Ri and Ro;

d) Determine the vorticity, and circulation for r = Ri and r = Ro. Provide a physical
explanation for these results in one or two sentences.

11. A uniform viscous flow approaches a flat plate at uniform velocity, U. After encountering
the leading edge of the plate, the main outer flow velocity accelerates according to
 x
U  U  1   , where U is the velocity approaching the plate, and P is the pressure of the
 L
approaching flow. If the flow has no vorticity, and no vorticity is transported in the flow
before it encounters the plate at x = 0, determine the vorticity transport across the plane at x =
L. Assume a stagnation point at the leading edge of the plate.

U, P UL, PL
Region C

y h Since this is a steady flow,

x for region C
x=0 x=L

12. Consider the 2-D flow outside a rotating circular cylinder, at the surface of which there is an
inward radial velocity, V, at the cylinder surface (r = R) due to suction through the wall of the
cylinder. Again, assume that the cylinder is in a fluid of infinite extent with kinematic
viscosity  and density ; the cylinder is of radius R and rotates at a constant angular velocity
. Determine the solution for the steady-state velocity field in the fluid external to the
cylinder using the vorticity equation for your solution:

  
t
 
  
 V        V   2 

Let one boundary condition be v   0 at r; consider the other boundary conditions
carefully, and perform your solution as follows:

a) Simplify the vorticity transport equation to an ordinary differential equation for

z using the continuity equation to assist the simplification (i.e. find vr first).

370
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 11

b) Consider the appropriate boundary conditions for vorticity and solve the resulting
O.D.E. to for  = (r, R, k, and A), where k = VR/ and A is an unknown
constant.

c) Using the results of b) solve for v = v(r, R, k, ), apply appropriate boundary
conditions, and show that a solution exists over the flow field only if k = VR/>2.

d) Determine the circulation,  = (r), and note its limit at k = 2; recalculate the
value for vorticity, with A known from part c).

e) If  = 2 and R = 1 (both dimensionless), plot z, v, and  for 1 < r < 10, for k =
3 and k = 6, and for the no suction case, V = 0, which is the solution determined
in section 11.4.3. Plot separate graphs of z, v, and , with the three curves for k
= 3, 6, and no suction on each plot. Interpret the results of these plots.

f) Extra credit: Explain why there is the limiting behavior indicated in part c) and
d), and see if you can determine what the limit of k = 2 implies physically.

371
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 12

Chapter 12

More Complicated Navier-Stokes Solutions

Contents

12.1 Fully-Developed Flow in a Rectangular Duct .............................................................. 373

12.2 A Suddenly Accelerated, Infinite Plate ....................................................................... 376

12.3 Viscous Decay of a Line Vortex (Oseen Vortex) .......................................................... 380

12.4 Ekman Drift: A Wind Driven Flow .............................................................................. 385

12.5 A Laminar Jet Issuing from a Narrow Slot ................................................................... 394

In Chapter 6 we explored solutions of the Navier-Stokes equation where the equation was
reduced to only one dependent velocity variable and one independent spatial variable
[e.g. u = u(y)]. These solutions were all fully-developed flows within parallel
constraining boundaries of either planar or radial symmetry. In all cases, the Navier-
Stokes equation reduced to an ordinary differential equation of a boundary value type.

In this chapter, we consider several flows where the velocity field is a function of more
than one space parameter, or one space parameter and time. In some cases, such as fully-
developed flow in a rectangular channel, the result is a Poisson-type equation, which is a
boundary value problem for one velocity component as a function of two spatial variables
[e.g. u = u(y,z) ].

Other types of problems are initial value problems, which are unconstrained in one or
more dimension. These types of problems can involve one dependent velocity variable,
one independent spatial variable, and time, such as the response of a bounding fluid to the
temporal motion of a boundary in one dimension [e.g. u = u(y,t) ]. These problems can
also involve the development of an unconstrained flow in two dimensions, such that u =
u(x,y) and v = v(x,y).

All of these problems require the solution of partial differential equations. The Poisson
type of equation has fairly standard solutions, which make the determination of the result
relatively straight forward. The unbounded initial value problems require some creativity
to develop analytical solutions. One approach that often proves very effective in solution
of such problems is the use of a similarity solution. Similarity solutions generally
require that a flow be unconstrained on one boundary, and that a variable can be
determined that collectively collapses the two independent variables into one independent
variable. Another technique that is employed for solution of flows in which both u and v
vary, is the use of the stream function to reduce the Navier-Stokes equations to one
independent variable, , with the penalty of increasing the order of the equation by one.

372
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 12

In this chapter, we will explore the use of all these approaches, which will prepare us for
an examination of boundary layers flows in Chapter 13.

12.1 Fully-Developed Flow in a Rectangular Duct

Consider the fully developed, steady flow in a rectangular cross-section duct, as shown in
figure 12.1
h y

z
x
b

Figure 12.1 Fully-developed flow in a rectangular duct.

This is a Poiseuille type flow (Section 6.3.2) with two sets of constraining surfaces. Here
we select a Cartesian coordinate system with its origin on the symmetry axis of the duct.
We again invoke the assumptions of:

1. Steady flow
2. Fully developed flow; no variations in x-direction
3.  = constant
4. Neglect body forces

Since the flow is assumed incompressible, the incompressible continuity equation in three
dimensions is:
2
u v w
  0 (12.1)
x y z

u
If the flow is fully developed, this means that u does not change with x, thus  0 , and
x
Eq. 12.1 becomes:

v w
 0 (12.2)
y z

Rearranging, Eq. 12.2, and integrating from y = -b/2 to y = +b/2, we have:

yb / 2 yb / 2
v w
v b / 2  v b / 2   dy    dy (12.3)
yb / 2
y yb / 2
z

Now, since v = 0 at y = -b/2 and +b/2 (since both limits are solid boundaries), then Eq.
12.3 becomes:

373
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 12

y b / 2
w

y b / 2
z
dy  0 (12.4)

w
If we assume that  f z, y  , then:
z

yb / 2
w

yb / 2
z
dy  g(z, y) y   b / 2  g(z, y) y   b / 2  0

or

g ( z, y ) y   b / 2  g ( z, y ) y   b / 2 (12.5)

Since the limits on y are arbitrary, g(z,y) = 0 is the only solution that consistently satisfies
w
Eq.12.5, which in turn means that  0 is the only value that will also consistently
z
v w
satisfy Eq. 12.4. Consequently, from Eq. 12.1, we have   0 , or both v and w
y z
are constants. Since v and w are both zero at solid boundaries, we conclude that:

v = 0, w = 0 everywhere. (12.6)

Additionally, it follows that all derivatives of v and w are also zero everywhere.

Using the above assumptions and our result from Eq. 12.6, the x, y, and z-direction
Navier-Stokes equations in three-dimensions simplify as indicated:

1 2 3 3 4 2
u u u u 1 P   u  2u  2u  2
u v w  x    2  2  2  x-direction (12.7a)
t x y z  x  x y z 
1 2 3 3 4 3 3 3
v v v v 1 P  2v 2v 2v 
u v w  y    2  2  2  y-direction (12.7b)
t x y z  y  x y z 
1 2 3 3 4 3 3 3
w w v w 1 P  2w 2w 2w 
u v w  z    2  2  2  z-direction (12.7c)
t x y z  z  x y z 

Thus, Eqs. 12.7 reduce to:

1 P   2u  2u 
   2  2   0 (12.8a)
 x  y z 

374
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 12

P
0 (12.8b)
y
P
0 (12.8c)
z

P
To assess the character of , we again establish this functionality like we did in Section
x
6.3.2 by taking the derivative of the x-direction differential equation, Eq. 12.8a, with
respect to x, which yields:

   2 u  2 u    1 P 
    0 (12.9)
x  y 2 z 2  x   x 

since u and its derivatives must be a function of y and z only (for a fully-developed flow).

P
Equation 12.9 again indicates that must be a constant, and P=P(x) only. This is
x
reasonable, since the shear stress will be a function of the derivative of the velocity
profile, which will also be invariant with streamwise position. So, noting that u = f(y, z)
only, the x-direction differential equation, with appropriate boundary conditions
simplifies to:

dP  2u  2u
  2  2 0
dx y z
or

 2 u  2 u 1 dP
   constant
y 2 z 2  dx
(12.10)
Poisson type equation

The corresponding boundary conditions are:

b
u  0, y  
2
h
u  0, z  
2

Note Eq. 12.10 and the corresponding boundary conditions result in a fairly standard
solution via conventional analytical solution techniques, which are available in most
graduate mathematics texts on partial differential equations. To see solutions for
Eq.12.10, and for a number of other common duct geometries, see White (1991).

375
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 12

12.2 A Suddenly Accelerated, Infinite Plate

This example deals with a situation where an infinite plate is suddenly accelerated from
u = 0 to u = V instantaneously. This is generally known as Stoke's first problem. We
determine how the velocity adjacent to the plate develops as a function of time. For this
situation, there is no vertical (v) velocity. Vorticity is generated instantaneously at the
surface, which subsequently diffuses outward (y-direction) due to the fluid viscosity.
y

x V

For this flow, the Navier-Stokes equation simplifies to:

u  2u 
 2  z  vorticity "flux" (12.11)
t y y

Thus, only the local x-direction acceleration and shear forces are non-zero. Similar to a
Couette flow, we can show that the pressure gradient for this flow is zero in both the x
and y directions, since the plate is assumed to be infinite in x.

The appropriate initial and boundary conditions for the flow are:

Initial condition: u( y, t  0)  0 (fluid is initially quiescent)

Boundary conditions: u( y  0, t  0)  V (plate has velocity V after t = )
u( y  , t  0)  0 (fluid velocity remains quiescent far
from the plate, for all times > 0)

To solve this type of equation, we employ what is termed a “similarity” solution. For this
solution process, we assume that the velocity profile will expand in the y direction due to
the viscous diffusion processes, and will scale on some yet to be determined universal
parameter that is a function of y and t. The approach is to introduce a similarity variable
that reduces the governing equation from a partial differential equation (P.D.E.) to an
ordinary differential equation (O.D.E.). What we are assuming when we introduce a
similarity variable is that the shape of the velocity profiles will be of the same “shape” as
time progresses. It is as if we “stretch” the velocity profile in the direction normal to the
plate (y) with increasing time. This approach is generally effective for flows that are
unbounded, and thus have no characteristic length (unlike pipe or channel flow
geometries, which have a characteristic diameter or width).

u
For the present solution approach, we assume a velocity form of  f () , where the
V
non-dimensional velocity is assumed to retain the same “shape” when scaled on .

376
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 12

Here,  is a similarity parameter, which we assume to be dependent on y and t, of the

form:
  Ay n t m where A, n, and m are all constants. (12.12)

The procedure is to substitute f() and  into the original P.D.E., simplify the resulting
equation, and by observation determine the appropriate values of n, m, and A that will
reduce the P.D.E. of Eq. 12.11 to an O.D.E. (ordinary differential equation) in terms of f
and . Thus, we first determine the u derivative terms of Eq. 12.11 in terms of , as
df d 2f
follows, using the chain rule, and designating  f  and  f  :
d d2
Term 1:

1 u f () df   m m
   f  f Ay n mt m 1  f ( Ay n t m )  f  (12.13a)
V t t d t t t t

Term 2:

1 u f () df  n n
V y

y

d y
 f

y y

 f Any n 1t m  f  Ay n t m  f 
y



2
1  2 u f 2 ()                 2
   f     f    f     f    f 
V y 2 y 2 y  y   y  y y  y   y 
  y 2
  2
 
 f  Any n 1t m  f  An n  1y n 2 t m (12.13b)
n 2
n n  1 n2
n n  1
 f Ay t 
n m 2
2
 f Ay t 
n m
2
 f  2
2
 f 
y y y y2
2 

Substituting Eqs. 12.13 into Eq. 12.11gives:

0 if n=1
m 2 n
2
n n  1
f   f  2  f 
t y y2

To simplify, we cancel out  terms, and set n = 1, which eliminates one of the terms in
the equation, giving:
m 1
f  f  2
t y

Rearranging gives:

my 2
f   f  0 (12.14)
 t

377
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 12

We note that f   2f   0 is a solvable equation, so we work toward that form by letting:
my 2 my 2
2    2  
 t 2t
Setting this relationship equal to our original assumption for the form of  from Eq.
12.12, we have:
1 1
 my 2  2  m  2  12
n m
  Ay t         yt (12.15)
 2t   2 

By equating the constants on the left hand and right-hand sides of Eq. 12.15, we have:
1
1  1 2 1
n  1from previous assumption , m   , which makes A    
2  4  2 

1 1 y
So, for this case, n  1, m   , and A   
2 2  2 t

Substituting these constant values into Eq. 12.12, reduces Eq. 12.14 to:

f   2f   0 (12.16)

Equation 12.16 is an O.D.E. for f as a function of the similarity (stretching) parameter,

y
 .
2 t

For this similarity parameter, the initial and boundary conditions become:

Initial condition: u( y, t  0)  0  f (  )  0 Note that these

conditions are
identical, and thus
Boundary conditions: u( y  0, t  0)  V  f (  0)  1 represent one
limiting condition
u( y  , t  0)  0  f (  )  0 (12.17)

Thus, we have not only reduced the differential equation from a P.D.E to an O.D.E., but
we have reduced the limiting conditions from three to two as well (which we must, if we
are to achieve a solution).

To solve Eq. 12.16, we let   f  and substitute into Eq. 12.16, such that:
  2

Separating variables and integrating gives:

ln   2  C

378
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 12

df
  f 
d
 C1 exp  2  


Integrating f ' gives:  

f  C1  exp  2 d  C 2  C1
2

erf ()  C 2
0

Where erf() is a tabulated function known as the error function. Since this integral has
no analytical solution, the function is solved numerically, and tabulated values of the
function are available in most standard books of mathematical functions or from the
Wolfram Alpha website. Applying the two limiting conditions from Eq. 12.17, we have:

=0

f (0)  1  C1 erf (0)  C2  C2  1
2
=1
 2
f ()  0  C1 erf ()  1  C1  
2 
u u
This gives a final solution for  f () , and thus ( y, t ) , of:
V V
f ()  1  erf()
u  y  (12.18)
 1  erf  
V 2 t 
 

Figure 12.2 shows plots of u/V as a function of both , and y and t. Note how the use of
, the similarity variable, collapses the velocity profile to only one curve.

Eta vs. u/V y vs. u/V

3 12
t=1
2.5 10 t=4
t=7
2 8
Eta

1.5 6
y

1 4

0.5 2

0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
u/V u/V

Figure 12.2 Velocity profiles for a suddenly accelerated plate using both similarity
scaling and physical scaling (non-dimensional y and t, with  = 1)

Note that the velocity profiles plotted with respect to y migrate outward with time, but are
“similar” in shape. Thus, as time increases the profiles are “stretched” in the y direction

379
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 12

as the vorticity originally generated by the accelerated surface gradually diffuses

outward.

12.3 Viscous Decay of a Line Vortex (Oseen Vortex)

The velocity of an inviscid line vortex, as shown in Section 9.7.3, is described in


cylindrical coordinates by the equation v r  0 , v   , where  is the total circulation,
2r
or strength, of the vortex. The shape of this velocity profile is shown in figure 12.3.

Velocity vs. Radius

Inviscid Vortex
10
9
8
7
6
Velocity

5
4
3
2
1
0
0 2 4 6 8
Radius

Figure 12.3 This graph shows the initial shape of an inviscid line vortex velocity
profile (letting Velocity=2v/) as a function of radius.

Recall that this vortex is considered irrotational, except for a spike of vorticity at the
origin, where the azimuthal velocity is infinite. In a real, viscous fluid, such a
concentration of vorticity would be quickly diffused away from the origin, strongly
modifying the behavior of the velocity profile as time increases.

To assess the behavior of an initially inviscid line vortex in a viscous fluid, we assume
that (1) all the initial circulation, , is retained within a radius of infinite extent for all
time, (2) the initial velocity distribution at t = 0 is that of an irrotational line vortex, given

by v  , and (3) the azimuthal velocity is forced to be zero at r = 0, immediately
2 r
after the initial time zero.

To solve for the velocity of a decaying line vortex, a problem originally solved by Oseen,
we first determine the governing equation. Since the flow will only be a function of the

380
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 12

radius and time, the only relevant velocity component in this problem is the -
component, v, such0that the continuity equation becomes:
1
rvr   1 v  0 since v = f(r) only (12.19)
r r r 

This gives vr = 0, since vr = 0 at r = 0 will be a boundary condition.

Subsequently, since v = f(r) only, and vr = 0, the Navier-Stokes equations reduce to:

v    1 
   rv  (12.20)
t  r  r r 

Thus, we have a balance between temporal changes and viscous diffusion. There is no
momentum transport—only diffusion of vorticity—and there is no pressure change in the
azimuthal direction.

Since this problem has no characteristic dimension, this is an ideal candidate for a
similarity solution. However, to simplify our similarity assessment, we assume a
function:

F  rv (12.21)

Where rv is called the reduced circulation (i.e. the circulation contained within any
radius r), and  is an appropriate similarity parameter. Here we assume that   ar n t m ,
where a, n, and m are constants to be determined. We can determine the values of a, n,
and m by substituting F and  into equation 12.21, and judiciously selecting a, n, and
m so as to render a tractable ordinary differential equation. However, noting that
F  rv (r, t , ) , we can apply dimensional analysis to determine the appropriate non-
r
dimensional groups for this set of parameters. For this case, we can show that   .
t
Thus, for our similarity parameter and velocity function we use:

r F
 , and v  . (12.22)
t r

The initial condition for the velocity profile is assumed to be an irrotational vortex:


@ t = 0, r > 0 v  (12.23a)
2 r

The two boundary conditions that apply here are:

for t > 0, r = 0 v  0 (12.23b)

381
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 12

and

for t > 0, r    = constant (12.23c)

Boundary condition #2 implies that the circulation is fixed from the beginning of the
decay process on, so we can calculate the circulation for t = 0 for any radii, as:


  v  2 r   2 r    .
2 r

Since this circulation will be fixed for an infinite radius (i.e. we assume the vorticity will
diffuse radially, but never quite reach an infinite radius), we can solve for the velocity for


r as v  .
2 r

Using the similarity parameters from Eq. 12.22, we establish the terms for Eq. 12.20 as
follows:
3

v    F  F  F  1 rt 2
1 r 
    where   
t t  r  r t 2rt t 2  2t t 2t

and
 1

rv    F  F   F where 
r r r t r t

So,
 1 
 rv     F    2 F  F    2 F  F
r  r r  r  r t  r t r t r r t rt

Substituting into Eq. 12.20, we have:

F  F F    Fr Fr 2 

    2    3     (12.24)
2rt  r t rt  r  t t 

r
Multiplying Eq. 12.24 through by rt, and noting that   gives:
t

F t  Fr Fr 2  1 F


2
 2  
r 

t  
 

  2  F  2 F    F
t

Solving for F , we have:

382
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 12

 1 
F    F (12.25)
 2
Using the initial/boundary conditions from Eq. 12.23, the appropriate limiting conditions
for Eq. 12.25 are:

 
@ t = 0, r > 0   : v   F= (12.26a)
2 r 2

@ t > 0, r = 0  = 0: v   0  F=0 Same (12.26b)

 
@ t > 0, r     : v    F= (12.26c)
2 r 2

Since Eq. 12.26a and 12.26c are identical, we have reduced the limiting conditions on Eq.
12.25 to two, which satisfies the order of Eq.12.25, thus yielding a tractable problem.

To solve Eq. 12.25 for F, we let   F , separate variables and integrate:

d  1   d  1  
         d
d   2     2 

2   2 
ln  ln   lnC1  lnC1 exp  
4   4 
or
dF  2 
  C1 exp   (12.27)
d  4

Integrating Eq. 12.27 for F, we get:

 2 
F  2C1 exp    C2 (12.28)
 4

Applying the limiting conditions of Eq. 12.26 to Eq. 12.28:

0
 
F   2C1 exp    C2   C2 
1 2 2
C 
F0  2C1 exp 0  C 2  0  C1  2 
2 4

Substituting C1 and C2 into Eq. 12. 28 gives a final expression of:

383
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 12

   2 
F 1  exp  
2   4 
And finally,

F    r 2 
v   1  exp   (12.29)
r 2 r   4 t 

Figure 12.4 shows the behavior of the velocity as a function of radius and time.

Velocity vs. Radius

(0<t<10)
1.2
t=0
1 t=0.1
t=0.5
0.8 t=1
t=4
Velocity

0.6 t=10

0.4

0.2

0
0 2 4 6 8
Radius

Figure 12.4 This graph shows the generic shape of the velocity profiles (letting =1,
and Velocity=2v/) as a function of time and radius.

Note the decay of the maximum velocity with time, and the movement of this maximum
away from the origin due to viscous diffusion. However, the “shape” of the profile is
essentially the same, only “stretched” radially.

The behavior of the vorticity is given by:

1 drv    r2 
z   exp   (12.30)
r dr 4 t  4 t 

The generic shape of the vorticity profiles (letting =1, and Vorticity=4z/) is shown
in Figure 12.5. Note that the vorticity is all concentrated at the origin (an infinite spike)
at t=0. Subsequently, the vorticity diffuses away from the center due to viscosity.

384
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 12

Vorticity vs. Radius

(0<t<10)
1.2
t=0
1 t=0.1
t=0.5
0.8 t=1
t=4
Vorticity

0.6 t=10

0.4

0.2

0
0 2 4 6 8
Radius

Figure 12.5 The general behavior of the vorticity for a decaying line vortex (letting
=1, and Vorticity=4z/) as a function of time and radius.

An additional solution to the decaying vortex problem was determined by G.I. Taylor,
using the assumption that the angular momentum of the vortex is a finite value, H. His
expression:

H r  r2 
v  exp   (12.31)
8  t 2  4t 

is similar, but reasonably different from the Oseen expression of Eq. 12.29.

12.4 Ekman Drift: A Wind-Driven Flow

We know that much of the motion of the oceans and lakes are strongly influenced by the
motion of the air above them. Let us consider the temporal development of the velocity
field of an initially quescent body of water, which is suddenly subjected to a constant
surface wind shear stress of o. we know that the shear stress must be continuous at the
air-water interface, as must the velocities. We could try to solve the coupled problem
where surface wind shear also varies with time, but that is a much more complicated
problem. So, we will presume that o = constant, and solve this as a similarity problem,
similar to how we solved the suddenly accelerated plate problem of section 12.2.

385
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 12

For this wind driven flow, like the suddenly accelerated plate, the Navier-Stokes equation
within the water simplifies to:

u  2u
 2 (12.32)
t y

Thus, only the local x-direction acceleration and shear forces are non-zero. Note that we
assume that our coordinate system is located at the water surface, with y oriented
vertically upward from the surface, as shown.

Air x o = constant

Water

The appropriate initial and boundary conditions for the water are:

Initial condition: u( y  0, t  0)  0
u 
Boundary conditions: ( y  0, t  0)  o  constant
y w
u
( y   , t  0)  0
y
u( y  , t  0)  0

Since we have a second-order boundary condition that is non-zero, this means that we
u
need to do a solution using y, t  as the similarity profile variable, and then determine
y
u(y,t) from that solution. To do this, we take the derivative of our Eq. 12.32 with respect
to y, such that:

  u     2 u    u  2  u 
         2   (12.33)
y  t  y  y 2  t  y  y  y 

u
Note that Eq. 12.33 is an equation for y, t  , which can be solved for using a
y
u y
“similarity” solution of the form  f () , where the similarity parameter is   ,
y 2 t
and using a process similar to what we used in solving for the velocity in section 12.2.

To solve Eq. 12.33, we reduce the equation to an ODE in terms of f(), and solve to
u
determine f(), and thus y, t  . Once we have a solution for f(), we can solve for
y

386
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 12

u
u(y,t) by integrating y, t  , and applying the two first-order initial and boundary
y
conditions. We first determine the derivative terms of Eq.12.33 in terms of , using the
1 1
 
df d 2f y y 2 t 2
chain rule, and designating  f  and  f  with    :
d d 2 1 1
2
2 t 2 2

Term 1:
1 3
   
  u  f () df    y 2 t 2
 y 1 
     f where    1 1  2t   2t
t  y  t d t t t 4  2 2
 2 t 
  f 
 f      
 2t  2t

Term 2:
 
  u  f () df    1  y 1 
     f where    1 1  y  y
y  y  y d y y y 1 1
2 2 t 2  2 2 t 2 

 f 
y

2
 2  u  f 2 ()                 2
    f     f
   f    
f   f
  
y 2  y  y 2 y  y   y  y y  y   y  y 2
2  
   2        1 
 f    f 0 where     0
 y y 2 y  y  y  12 12 
 2 t 
2
 f  2
y

Substituting the reduced Terms 1 and 2into Eq.12.33 gives:

 2
 f    2 f 
2t y

y2
 f   f 
2 t

 2f   f   f   2f   0 (12.34)

387
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 12

Eq.12.34 is an O.D.E. for f as a function of the similarity (stretching) parameter,

y
 .
2 t

For this similarity parameter, the two second order boundary conditions shown
previously become:

u  o
Boundary conditions: ( y  0, t  0)  o  f (   0) 
y w 

u
( y  , t  0)  0  f (   )  0
y

Thus, we have reduced the differential equation from a P.D.E to an O.D.E., with two
appropriate boundary conditions. Note that unlike the suddenly accelerated flow, we did
not have to reduce three boundary conditions to two.

To solve our ODE, we let   f  and substitute into the

O.D.E., such that

  2

Separating variables and integrating gives:

ln    2  C
or
df
  f 
d
 C1 exp   2  
Integrating f ' gives:

 
f  C1  exp   2 d  C 2  C1
2

erf ()  C 2
0

Applying the two boundary conditions we have:

=0
  
f (0)  o  C1 erf (0)  C 2  C 2  o
w 2 =-1
w
 2 2  o 
f (  )  0  C1 erf (  )  C 2  C1  C2   
2     w 

388
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 12

u u
Substituting C1 and C2, gives a solution for  f () , and thus  f ( y, t ) , of:
y y
o
f ()  1  erf()
w

u  o   y 
 1  erf   (12.35)
y  w   2  t 
  

u
Figure 12.6 is a non-dimensional plot of predicted by Eq. 12.35, showing the
y
similarity shape of the shear profile in the water. Note that within the water,   0 .

(u/y)/o
0.0000 0.2000 0.4000 0.6000 0.8000 1.0000
0
-0.2
-0.4
-0.6
-0.8
 -1
-1.2
-1.4
-1.6
-1.8
-2

y
Figure 12.6 Similarity variable,   , vs. the non-dimensional velocity
2 t
gradient in water, (u/y)/o, for a wind-driven flow with constant

applied surface shear stress, o .
w

u
To obtain u(y,t), we integrate Eq.12.35 for with respect to y (I used Wolfram Alpha to
y
do the integration), and obtain the messy:


 t o  y2    y  
u ( y, t ) 
2 exp     y 1  erf     g( t )
w
 


4 t    2  t  

   
 
Applying our original first-order initial and boundary conditions:

389
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 12

Initial condition: u( y  0, t  0)  0

o
u 0exp   y1  erf    g(0) = 0
w

u  o 0  y1  1  g(0)  g(0)  0
w

Boundary condition: u( y  , t  0)  0

 o  t 
u(  , t )  2 exp      1  erf     g( t )  0
w   

 o  t 
u(  , t )  2 0    1  1  g( t )  0
w   
o
u(  , t )  00  0  g( t )  0  g( t )  0
w

  y 
Note, that the term y 1  erf   will go to zero as y-. Prove this to yourself
  2  t 
 
using L'Hospital's rule.

Thus:
o 
 t  y2    y  
u 2 exp    y  1  erf    (12.36)
w   
 4 t    2  t  

   


2 o t
Note that in Eq.12.36 the collection of terms has the units of velocity, so
w 
letting:
2 o t
u0  (12.37)
w 

we can rewrite Eq.12.36 as:

u  y2  y    y 
 exp    1  erf   (12.38)
u0 4  t 2 t  
    2  t 

390
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 12

y
And substituting our similarity parameter,   , into Eq. 12.38 gives:
2 t
u
u0
 exp   2    1  erf   (12.39)

u
Note that Eq. 12.39 is a "similarity" velocity profile for .
u0

Equations 12.38 and 12.39 indicate that u 0 is the velocity at the water surface (y = 0 and
  0 ), which according to Eq. 12.37 increases proportional to t.

Water Surface Velocity vs. time

Vwind = 2 m/s
0.45
0.4
Surface velocity (m/s)

0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
0 500 1000 1500 2000 2500 3000 3500 4000
time (s)

Figure 12.7 Water surface velocity, u 0 (m/s) vs. time (s) for a wind-driven flow with

constant applied surface shear stress, o . Assumed wind velocity
w
Vwind=2 m/s, and 0  0.002air Vwind  u 0  , from Roll (1965).
2

To examine some numbers for the movement of the water at the surface, it was pointed
out by White (1991) that Roll (1965), citing several empirical studies, estimates the wind
stress for airflow over water surface to be given by the relation:

0  0.002air Vwind  u 0  .
2
(12.40)

kg
Assuming typical standard properties for water and air of  w  10 3 ,
ms
m2 kg
 w  10 6 , and  air  1.2 3 , we solve Eq. 12.37 for 0 , and substitute 0 into
s m

391
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 12

Eq. 12.40 to give an equation for u 0 in terms of Vwind and t as:

u0
 Vwind  u 0   0
2
369.3 (12.41)
t

Note that t is in seconds, and velocities in m/s.

Assuming a wind velocity of 2 m/s (about 4.5 mph), we solve Eq. 12.41 using a root
finding program for a series of times. A graph of u 0 vs t is shown in Figure 12.7. Note
that the water surface velocity, u 0 , initially grows quite quickly, but slows with time.
Also, the numbers seem somewhat unrealistic, since the water surface velocity, u 0 , is
predicted to increase to 20% of the wind velocity after one hour of wind exposure. As
White points out, the water surface velocity should only grow to about 3% after an hour
of wind exposure. He also notes that for such flows at higher wind velocities, both the
water and air flows will be turbulent, with much higher effective turbulent viscosities
(known as eddy viscosities--see Chapter 17). However, if the turbulent values of w are
larger, this would suggest that u 0turbulent  u 0laminar , which would result in even higher
predicted values of u 0 .

So, what is the problem? This inconsistency with empirical values is most likely due to
the assumption of constant shear. In reality, the shear will diminish as the water surface
velocity increases. Note that Eq. 12.37 indicates that u 0 will continually increase
proportionally as t . In reality, as time increases u 0 must reach some equilibrium value,
such that u 0 << Vwind. Since we don't specify the wind velocity in the similarity model,
and the boundary conditions that  w , 0   a , 0 , and u w ,0  u a ,0 , our similarity model over
simplifies, and thus over estimates the acceleration of the water. Thus, the Ekman drift
problem is an interesting example of the use of similarity, but is limited in its practical
applicability.

Graphs 12. 8 and 12.9 show the velocity behavior within the water, dimensionally for Eq.
12.38 and non-dimensionally for Eq.12.39.

Figure 12.8 shows the dimensional effect of viscous diffusion of the applied shear within
the water. Note that the thickness of the affected layer of water (that where u > 0) grows
relatively slowly, taking 1000 seconds for the effect of the applied wind stress to be felt
at a depth of 10 cm (roughly 4 inches). Even after 1 hour, the extent of the affected fluid
only reaches 19 cm (about 7.5 inches).

Figure 12.9 shows the shape of the velocity ratio relative to the similarity parameter, .
It is interesting to note that the shape of the velocity appears similar to that of the shear
stress, but the functions for the shear stress and the velocity are quite different.

392
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 12

u/uo
0 0.2 0.4 0.6 0.8 1
0
-1
-2
-3
y (cm)

-4
t=10s
-5
-6 t=100s
-7 t=1000s
-8 t=3600s
-9
-10

Figure 12.8 Water depth, y (cm), vs. the water velocity ratio, u / u 0 for a wind-

driven flow with constant applied surface shear stress, o . Velocity
w
profiles are shown for four increasing time increments, t = 10, 100,
1000, and 3600 seconds

u/uo
0.0000 0.2000 0.4000 0.6000 0.8000 1.0000
0
-0.2
-0.4
-0.6
-0.8
 -1
-1.2
-1.4
-1.6
-1.8
-2

y
Figure 12.9 Similarity variable,   , vs. the water velocity ratio, u / u 0 for a
2 t
o
wind-driven flow with constant applied surface shear stress, .
w

393
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 12

12.5 A Laminar Jet Issuing from a Narrow Slot

As we demonstrated in the examples of Sections 12.2, 12.3, and 12.4, the use of a
similarity solution is an effective method of reducing a P.D.E. to a tractable O.D.E. This
process reduces the number of independent variables by one or more (e.g. from y and t to
 for the suddenly accelerated plate of Section 12.2). However, how do we approach a
problem that involves more than one dependent variable? Note that our two previous
examples dealt with only one non-zero velocity variable (u for the suddenly accelerated
plate, and v for the decaying line vortex).

In the following example, we deal with a flow that requires the determination of two
dependent velocity components, u and v, and thus we must develop a solution that
simultaneously satisfies both the continuity and the Navier-Stokes equations. Thus, we
have u = u(x,y) and v = v(x,y).

Here, we show how this is done making use of the stream function, as well as a similarity
variable.

The problem we address here is a two-dimensional laminar jet, which issues from a slot
in a wall into a region of infinite extent and constant pressure. The solution we seek will
only apply for a downstream region removed from the immediate vicinity of the slot. In
this downstream region, the velocity profiles are expected to exhibit similar behavior,
with the velocity profile stretching laterally in the y-direction due to viscous diffusion,
and correspondingly contracting in the x-direction, as illustrated below.

P = constant y

  constant x
Constant pressure

An additional assumption will be that the total momentum flux in any x-plane of the jet
will remain constant, and that no other external forces act on the flow. In reality, this is
not exactly the case (although the momentum will not change rapidly). However, as we
will see, we require four limiting conditions to solve the problem fully, and we can only
identify three appropriate boundary conditions. Therefore, we use the assumption of
constant momentum flux as our fallback limiting condition.

An assessment of the Navier-Stokes equation indicates that the y-direction component

equation has negligible influence on the solution (we will demonstrate how we prove this
for a boundary layer flow in Chapter 13), so the governing equations for this flow are the
x-direction component of the Navier-Stokes equation and the continuity equation (2-D).

394
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 12

Neglect relative to y-changes

0
u u 1 dP   2u  2u 
Navier-Stokes (x-direction): u v    2  2  (12.42a)
x y  dx  y x 

u v
Continuity equation:  (12.42b)
x y

We further assume that the pressure within the solution region is constant, which makes
the pressure gradient within the flow region negligible. Additionally, we assume the
viscous changes in the x-direction are minimal in comparison to viscous changes relative
to the normal (y) direction.

Finally, we identify three clear boundary conditions, all relative to y.

Boundary Conditions: @ y  0: v0 (12.43a)

u
@ y  0: 0 (12.43b)
y
@ y: u 0 (12.43c)

Because Eqs. 12.42a and 12.42b contain a first-order derivative of u with respect to x, we
technically require a bounding condition for u at an appropriate x location. One
possibility would be u = 0 as x  . However, such a boundary condition does not
prove useful in establishing the coefficients of integration. Thus, we use a fallback
assumption of a constant momentum flux, which will be used after the fact to allow a
(somewhat) complete solution.

To begin the solution, we note that we can reduce Eqs. 12.42a and 12.42b to one equation
by the use of the stream function, since the stream function is an exact solution of the
continuity equation (in two-dimensions). The use of the stream function reduces the two
partial differential equations (P.D.E.) to one P.D.E. equation with one dependent variable
(), since it eliminates the continuity equation from the set of equations.

Substituting the stream function into Eq. 12.42a, where:


u
y

v
x
u  2

x yx
u  2

y y 2

395
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 12

 2 u  3

y 2 y3

yields:

  2    2   3
   (12.44)
y yx x y 2 y 3

Thus, by using the stream function, we have reduced two P.D.Es to one P.D.E. for
  x, y  . We now search for an independent variable,  , which will convert the
partial differential equation of Eq. 12.44 to one ordinary differential equation.
The determination of an appropriate similarity approach is not always straightforward,
and requires a bit of insight, skill, and trial and error. Many approaches that work seem
reasonable after the fact, but are not exactly obvious initially. For this problem, in
addition to a similarity parameter, the stream function must be further scaled on the
streamwise variable, x, such that we seek a functional relationship of the form:

  a x bf () and   Cc xd ye (12.45)

Here, a, b, c, d, e, and C are all constants, to be determined. As we will see, this form of
scaling reduces Eq. 12.44 to a tractable ODE, but the form of the scaling was probably
not obvious to the folks who first solved this problem, and evolved from several different
attempts at an effective scaling. Often, dimensional analysis of the governing equation
may be helpful in the determination of an appropriate scaling approach, but not always.

To seek a general similarity reduction of Eq. 12.44, we could employ the general
functional forms of Eq. 12.45, substitute into Eq. 12.44, and manipulate a, b, c, d, e, and
C to yield a tractable ordinary differential equation. However, this is a hard, tedious
process, so we will short cut it by using the following previously established functional
forms:

1 y
  x1 / 3f () and   2/3
(12.46)
3 x

Thus, we first determine the various components of Eq. 12.44 as:

  f   2 y 2
v  23
  x 1 3f  Noting that  
x 3 x x x 9 x
53
3x

 f  2f 
3x 2 3

  f  1
u   x1 3f   13 Where 
y y 3x y 3 x 2 3

396
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 12

u  2 f   f 
 2  13 
y y 3x y 9x 

 2 u  3 f   f 
2
 3   53
y y 9x  y 27x

u  2 1
   4 3 f   2f 
x xy 9x

Now substituting these relationships back into Eq. 12.44 and simplifying (do this as an
exercise) gives:

f   f f   f 2  0 (12.47)

f f 

f   f f   0  f   f f   0 O.D.E. for f 

And the modified boundary conditions from 12.43 are:

1) y  0  0 : v0  f 0
u
Boundary Conditions: 2) y  0  0 :  0  f   0 (12.48)
y
3) y       : u  0  f  0

Integrating Eq.12.47 for f gives:

f   f f   C1  From Eqs .12.48 for   0  0  0  C1  0

Therefore,

f   f f   0 (12.49)

Note that we can write, (f 2 )  2ff  . This will help simplify Eq.12.49, but the
coefficients are wrong to allow a closed form integration. So, we substitute further,
letting:

f  2F  f   2F  f   2F (12.50)

Thus, substituting Eqs.12.50 into Eq. 12.49, gives

397
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 12

2F  ( 2F)(2F)  0
F  2FF  0


F  (F 2 )  F  F 2  0 (12.51)

Integrating Eq. 12.51 gives:

F  F 2  constant = 2  choose for convenience

dF
or   2  F2 (12.52)
d

Separating variables, and integrating Eq.12.52 gives:

dF
 2
 F2 
 d
0 0
F   tanh   C2  for   0  F(0)   tanh( 0)  C2  C2  0

Thus, f  2F  2 tanh  (12.53)

Now, using f to solve for the velocity components gives:

f 2 2 2 2
u

 1  1 sech2 ( ) 
y 3x 3 3x 3 3 x 13
1  tanh 2    (12.54a)

1 y
where   2/3
3 x
and

v

x 3x

  2 3 f  2f  
2 
3x 23

2 sech2   tanh   (12.54b)

Note that we never used our third boundary condition to arrive at these expressions.
However, we note that Eq.12.54a reflects u = 0 (f '= 0) for   , thus our third
boundary condition is satisfied.

However,  is an undetermined constant from the integration process, and (since we

have no other appropriate boundary condition) we establish  by invoking the physical
condition that the momentum flux for the jet will remain constant (within a reasonable
distance of the orifice). Thus, if we let J be the momentum flux, defined by:

398
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 12

 u dy (12.55)

2
J


Substituting Eq. 12.54a for the x-direction velocity, u, into Eq. 12.55, and solving for 
in terms of J, the result is:

1
 J  3
  0.8255 
  

Figure 12.10a and b illustrate the u velocity predicted by Eqs. 12.54a, illustrating the
"similar" shape of the velocity profile, when scaled on , and the spread of the velocity
profile, when scaled on y, as a function of streamwise distance, x. Note that figure
12.10b also demonstrates the decrease in umax (the center line velocity, at y =  = 0) with
increasing distance, x.

4 20
x = 1.0 x = 1.0
x = 2.0 x = 2.0
2 x = 3.0 10 x = 3.0
eta

0 0
y

-2 -10

-4 -20
0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8
u u
(a) u vs.  (similarity parameter) (b) u vs. y (physical distance)

20
x = 1.0
x = 2.0
10 x = 3.0

0
y

-10

-20
-2 -1 0 1 2
v
(c) v vs. y (physical distance)

Figure 12.10 Streamwise velocity for a laminar jet. Here we let , , and J = 1, for
three progressive x distances, x =1.0, 2.0, and 3.0.

Figure 12.10c shows the behavior of the transverse v velocity, which appears a bit
unusual. However, there is an explanation for the pattern shown. Note the change in the

399
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 12

transverse velocity behavior depending on the y location: when the jet is close to the wall
slot (x=1.0), near the centerline (|y| < 4)the transverse velocity, v, is away from the
centerline, but well away from the centerline (|y| > 10), the velocity is very strongly
toward the centerline. This reflects an induced flow created by the jet, which causes fluid
away from the centerline to flow toward the jet. However, as the jet moves further along
x, this flow pattern mediates, although the same outward/inward pattern is still present,
although much weaker.

To better illustrate the jet flow pattern, Figure 12.11 shows a plot of the velocity vectors
for u and v calculated for the region 1 < x < 6 and -10 < y < 10; overlaid on these velocity
vectors are selected streamlines for the flow. These vectors and streamlines illustrate the
strong inflows near the wall, the decreasing u-velocity away from the wall, and the slow
spreading of the jet as it proceeds outward.

4
x

0
10 8 6 4 2 0 -2 -4 -6 -8 -10
y

Figure 12.11 Velocity vector/streamline plot for a laminar wall jet. Here we let , ,
and J = 1. Velocity vectors are shown in black, and streamlines in blue.
Jet enters upward at x,y =0,0.

Equations 12.54 reflect an idealized solution, since under normal flow conditions the jet
would break down due to the rapid generation of turbulence. However, Eqs. 12.54 will
be a reasonable model of a low density or very viscous jet flow, at very low Reynolds
number.

Note that if we tried to use the boundary condition u = 0 as x  , this would be

indeterminate in establishing . When x  , 0, and Eq. 12.54a will become:
0
2 2 2 2
u0
3 ()

1  tanh (0) 
2

3 ( )


This result is indeterminate for , so  cannot be determined from an assumption of u = 0

as x  .

400
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 12

This approach of using similarity variables permeates fluid mechanics. In the next
chapter, we will continue on and address a very common flow behavior that yields a
similarity solution: the development of laminar boundary layers for flows over external
surfaces.

References

White, F. M. (1991), Viscous Fluid Flow, 2nd ed., McGraw-Hill, New York. (section 3-
3.3, pp. 119-122)
(link: http://ftp.demec.ufpr.br/CFD/bibliografia/viscous_fluid_flow_frank_m_white_second_edition.pdf)

Roll, H. U. (1965), Physics of the Marine Atmosphere, International Geophysics Series,

vol. 7, Academic Press, New York.

Problems

1. Consider the solution for a decaying vortex solved in section 12.3. Using the solution
for the velocity (Eq. 12.29), determine the shear stress for the vortex, and create a
graph showing r vs. r from 0 < r < 6 cm for water ( = 2x10-5 kg/cm-s,  = 10-2
cm2/s). On one graph, show 4 plots of t = 1, 5, and 10 seconds. Briefly explain what
is happening as time increases.

2. Consider the situation where an infinite plate and the flow above it are moving at a
constant velocity U. At t = 0, the plate is suddenly stopped, decelerating from u = U
to u = 0 instantaneously.
y

x V
For this flow, the Navier-Stokes equation simplifies to:

u  2u
 2
t y

Thus, only the local x-direction acceleration and shear forces are non-zero. Solve this
u
equation for u(y,t) using a “similarity” solution of the form  f () , where the
U
y
similarity parameter is   . Indicate clearly the appropriate boundary and
2 t

401
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 12

initial conditions, reduce the equation to an ODE in terms of f(), and solve to
determine f(), and thus u(y,t).

Create two graphs: One of u/U vs.  for 0 <  < 2; another showing three plots of
u/U vs. y between 0 cm < y < 2 cm, for t = 1 s, 10 s, and 100 s. Assume the fluid to
be water with  = 10-2 cm2/s.

3. For the Ekman Drift problem done in section 12.4, using equation 12.41, determine
the value of Vwind (in m/s) for which u 0 will be 3% of Vwind after one hour of
exposure.

4. For the Ekman Drift problem done in section 12.4, make graphs for 0  y  10 cm of:

 w u
a. vs . y , and
0 y

u w cm 2 u w
b. vs . y (let   102 , which will give units of cm)
0 s 0

Comment on what these show about the flow.

5. For the Ekman Drift problem done in section 12.4, use a similarity method similar to
that used for the water to derive the behavior of the shear stress and velocity for the
air flow above the water. You will use the same reduced equation and similarity
parameter,  (only for the kinematic viscosity in air), and will determine solutions for
the region y  0 . Also, your initial condition will be u( y  0, t  0)  U = constant,
u
and the second and third boundary conditions will be ( y  , t  0)  0 and
y
cm 2
u( y  , t  0)  U . From your solutions, plot (let   102 , which will give
s
u  U 
units of cm):
0

 a u
A.  vs . , from 0 <  < 2
 0 y
uU 2  t
B.  vs . , from 0 <  < 2, where u 0  o
u0  
uU
(note that is termed a velocity "deficit")
u0
u  U 
C. y vs . , from 0 < y < 10 cm
o
Comment on the behavior of your graphs, and what they show about this flow.

402
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter13

Chapter 13

Introduction to Boundary Layer Theory

Contents

13.1 Motivation ............................................................................................................... 403

13.2 Development of Boundary Layer Equations ................................................................ 404

13.2.1 An Order of Magnitude Reduction of N-S Equations ........................................ 407
13.2.2 The Boundary Layer Equations ........................................................................ 410

13.3 Solution of the Boundary Layer Equations for a Flat Plate Flow .................................. 412

13.4 Boundary Layer Parameters ...................................................................................... 418

13.4.1 Displacement Thickness, * ............................................................................ 418
13.4.2 Momentum Thickness,  ................................................................................. 419
13.4.3 Friction Coefficient, cf ..................................................................................... 420

13.5 Boundary Layer Characteristics for a Flat Plate Flow .................................................. 420
13.5.1 An Example of Boundary Layer Parameters ..................................................... 425

13.6 Falkner-Skan Solutions for Non-Zero Pressure Gradients ............................................ 427

13.7 Conclusion ................................................................................................................ 436

13.1 Motivation

Almost all practical fluid systems, from pumps to airplanes, have fluid interacting with
solid surfaces. In such real fluid flows, as opposed to “ideal” potential flows, the “no
slip” condition due to the fluid viscosity constrains the fluid to adhere to, and not slip,
when in contact with a solid surface. The interaction of a real fluid with a solid surface
can also generate vorticity, due to the processes we discussed in sections 11.5 and 11.6.
When vorticity is generated, it will diffuse away from the surface, but very slowly in
comparison to its rate of transport parallel to the surface. Consequently, the interaction of
a fluid with a solid boundary results in the creation of a thin layer of vorticity-bearing
fluid adjacent to the solid surface. This thin region provides a transition zone from the
no-slip solid boundary to the outer region fluid, where the fluid behaves in a relatively
inviscid manner, like a potential flow. This transitional region is correspondingly termed
the “boundary layer”, and is the region within which fluid shear stresses are dominant.
Consequently, we generally think of a real fluid flow in proximity to a solid surface as
comprised of a thin, viscously-dominated boundary layer, with an adjoining outer region
that behaves in a relatively inviscid manner. While this model is a little oversimplified,
this two-region approach is a practical way to model such flows, from which to seek
practical engineering solutions. To familiarize yourself with the physical characteristics

403
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter13

of a boundary layer, I recommend you view a video of visualized boundary layers, which
can be accessed here. This is a YouTube video of an old educational film (circa 1961),
but one that is still quite relevant, showing the development and physical behavior of real
boundary layers.

Boundary layers, as we will discuss, are subject to both viscous diffusion and to pressure
gradients imposed by the outer-region flow. While viscous diffusion will cause the
spread of vorticity, the presence of a pressure gradient will cause the generation of
vorticity at an adjacent surface. As discussed in Section 11.5.3, depending on the
direction of the pressure gradient (positive or negative), the additional vorticity generated
by a pressure gradient can cause the cumulative vorticity comprising the boundary layer
to either increase or decrease. For complicated geometries, the generation of vorticity,
and the development of boundary layers, can be quite complex and difficult to model
mathematically. However, to begin our assessment of boundary layers, we will first
confine ourselves to the most elementary type of boundary layer flow—one that develops
over a thin flat plate with no pressure gradient. This flow, first assessed by Prandtl, and
first solved mathematically by Blasius, is the simplest example of a broad field of fluid
mechanics known as boundary layer theory.

In the present chapter, we first examine the development of what are termed the boundary
layer equations, and then apply these equations to assess the behavior of the laminar
boundary layer for a flat plate flow, both without and with a pressure gradient. In the
following Chapter 14, we will illustrate some approximate solution techniques that allow
the rapid assessment of more complicated laminar boundary layers with pressure
gradients. Later, in Chapter 17, we will address some simple models and solutions for
boundary layers when turbulence is present.

It should be recognized that we will only scratch the surface of this very expansive topic
of boundary layer theory. And we will only address two-dimensional flows for which
semi-analytical solutions are possible. In reality, most practical flows are sufficiently
complicated that one must resort to numerical solutions of the complete boundary layer
equations. In this text we will not cover numerical solution techniques, which is a topic
unto itself. However, upon completion of the present material, a student should be well
prepared to take advanced graduate courses on numerical techniques that should allow
one to solve, or at least understand the solution procedures, for more complicated
boundary layer flows.

13.2 Development of the Boundary Layer Equations

To characterize the thickness of the boundary layer, we choose the variable delta, . We
take this thickness to be the distance away from a solid boundary where the velocity
approaches that of the outer region. Since we can envision that the velocity may very
slowly asymptote to the outer region streamwise velocity, U, we arbitrarily designate the
boundary layer thickness as that point where u = 0.99U. As we will see, this is not a
particularly good designation, since if we use this criterion to determine the boundary
layer thickness from experimental velocity measurements, the result will be highly

404
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter13

dependent upon the accuracy of the measurement instrument employed. For example, a
 1% accurate instrument might yield values of  with  10% accuracy or worse,
depending on how the velocity profile within the boundary layer asymptotes to U. We
will discuss alternative ways to characterize the effects and scale of the boundary layer
later in section 13.4.

As figure 13.1 shows, we consider the thickness of the boundary layer, , to grow slowly
as a function of the streamwise distance, x, such that (x) << x. Note that the boundary
layer growth in figure 13.1 is exaggerated for illustration purposes. In reality, a boundary
layer will generally measure in height less than 1/100th of the streamwise distance from
its point of origin. In figure 13.1, we assume that the boundary layer will start at zero
thickness at x = 0, and grow from that point onward. However, that is not necessarily the
case, since  could have some non-zero initial value, due to vorticity generated on a
previous surface. Additionally, we must generally assume that U, the velocity of the
outer region (generally known as the free stream), may also be a function of the
streamwise distance.
U(x)
u = 0.99U
y

(x) = boundary layer

x

Figure 13.1 A generic model of boundary layer development (not to scale)

Generally, laminar boundary layers are computed using either a combination of analysis
and numerical computation, for simple flows, or purely numerically, for more
complicated flows. On the other hand, because of the tremendous complexity of the flow
behavior, exact boundary layer solutions of a turbulent boundary layer are done
employing significant simplifications and numerical computations. As we will discuss in
Chapter 17, most practical engineering solutions of turbulent boundary layer
characteristics are essentially empirically-derived, requiring significant use of
approximations or empirical curve fits. We will discuss the approaches for assessment of
turbulent boundary layers later in Chapter 17. Here, we develop the appropriate
equations to determine the characteristics of laminar boundary layers.

U U(x)

y 

 1
(x) *  
x L 100

Figure 13.2 A parametric representation of boundary layer growth.

405
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter13

We first consider a model of the boundary layer, as shown in figure 13.2, which shows
the relevant parameters at a point L in the direction of flow over a solid surface.

At any given distance along the surface, the relevant parameters are:

L, the position along the surface;

U, the local outer region velocity;
, the kinematic viscosity of the fluid (the ratio of absolute viscosity to density); and
, the local boundary layer thickness.

If we apply a process of dimensional analysis to this set of parameters, we can determine

a set of two non-dimensional parameters:


(13.1a)
L

and

UL U
 Re L or  Re  (13.1b)
 

The first parameter, Eq.13.1a, is the thickness of the boundary layer relative to the local
streamwise position, L (which represents a characteristic dimension of the flow at that
point). The second parameter, Eq.13.1b, is termed the Reynolds number (after Osborn
Reynolds, who first characterized turbulence), which represents a ratio of the local inertia
forces to viscous forces. The Reynolds number can be based on either the local position,
or the local boundary layer thickness, depending upon what one chooses as the repeating
parameter in the non-dimensionalization process. Using the parameters of Eq. 13.1, we
can indicate a functionality of the form:

 U L   U 
 f    or  f   (13.2)
L    L   

Equation 13.2 indicates that the growth and behavior of the boundary layer is contingent
upon the magnitude of the local Reynolds number.

Note that we can expand the definition of the Reynolds number, based on the boundary
layer thickness, as follows:

U   U   U   U2 u 2 inertia forces

Re       (13.3)
   U  u viscous forces
     
    y

As Eq. 13.3 illustrates, the Reynolds number represents the relative magnitude of inertia
forces within the boundary layer to the comparable viscous forces (i.e. shear stresses).

406
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter13

As we will see, when the Reynolds number is quite large this allows us to simplify the
governing equations quite significantly, and makes the computation of boundary layers
more tractable (although the solution of the resulting equations still requires the use of
sophisticated analytical and numerical procedures).

13.2.1 An Order of Magnitude Reduction of the N-S Equations

To begin our assessment of the governing equations, we model the flow situation as
shown in figure 13.2, and use this model to appropriately non-dimensionalize the Navier-
Stokes equations. We then simplify the equations using an order-of-magnitude
assessment, which lets us establish which terms within the governing equations can be
considered negligible or at least very small relative to the dominant terms, and thus to a
good approximation can be neglected .

Note that the terms comprising the governing equations will be non-dimensionalized and

compared relative to the non-dimensional boundary layer thickness, *  , which we
L
pointed out is of the order (size) of 1/100. Using this scaling on * allows us to establish
the order of magnitude of each term within the equations (i.e. how dominant each term
is). This order of magnitude hierarchy is given by:

1
*  1  , etc.
*

We begin the non-dimensionalization process by defining the following non-dimensional

ratios,

x y u v P
x*  , y*  , u*  , v*  , P*  (13.4)
L L U U U 2

as our non-dimensional variables. Considering our model of the boundary layer in figure
13.2, we use physical reasoning to establish the order (size) of the dimensionless
variables in Eq.13.4 as:

x*  0(1), y*  0(*), u*  0(1), v*  0(*), P*  0(1) (13.5)

Here, 0(1) means this is, or could be, a dominant variable, and 0(*) indicates a variable
which is an order of magnitude less dominant (roughly 100 times smaller). The rationale
for this ordering is that large changes can take place in the streamwise (x) direction and
the (u) velocity component, whereas much smaller changes take place in the normal (y)
direction and (v) velocity component. Since the pressure may or may not change
significantly, we assume that could be a potentially dominant term of 0(1).

The starting governing equations are the continuity and the steady-state Navier-Stokes
equations in two dimensions (Note that the same process can also be applied for the more
complicated, three-dimensional case), which for an incompressible flow are:

407
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter13

u v
 0 continuity (13.6a)
x y
u u 1 P   2u  2u 
u v    2  2  x-direction N.S. (13.6b)
x y  x  x y 
v v 1 P  2v 2v 
u v    2  2  y-direction N.S. (13.6c)
x y  y  x y 

To see how this order of magnitude analysis works, consider the continuity equation,
Eq.13.6a. Using the definitions in Eq. 13.4, we can write:

x  Lx*, y  Ly*, u  U  u*, v  U  v * (13.7)

Substituting the variables in Eq.13.7 into Eq.13.6a and simplifying, we have:

u v U  u * U  v * u * v *
     0 (13.8)
x y Lx * Ly * x * y *

Now, assessing the order of magnitude of each term in Eq. 13.8, using the relative
magnitudes of each variable that were assumed in Eq. 13.5 (we assign the magnitude of
each term as the ratio of the magnitude of the variables comprising the derivative), we
have:

u * v *
 0 (13.9)
x * y *

0(1) 0(*)
  0(1)  0(1)
0(1) 0(*)

The order of magnitude assessment of Eq.13.9 indicates that both terms in the continuity
equation can be of comparable magnitude, and thus both must be retained. At first this
may seem a bit strange, since both the v* and y* terms are small. However, the ratio of
these terms indicates that while the values of v* and y* may be small, the ratio of the
changes will be comparable to the ratio of the streamwise changes. This is logical, since
for two-dimensional incompressible flow any changes in the velocity in one dimension
must be balanced by comparable changes in the velocity in the other dimension. As we
will see in the following assessment of the Navier-Stokes equation, all terms in the
Navier-Stokes equations are not comparable, which allows us to neglect selected terms as
essentially non-contributory.

Next, we non-dimensionalize the x-direction Navier-Stokes equation, Eq.13.6b. For

example, since u = Uu* and x = Lx*, then for the first term of Eq.13.6b we can write,

408
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter13

u U u *  U 2  u * 
u  U  u *    u * 
x L x *  L  x * 
Performing similar non-dimensionalizations of each term of Eq. 13.6b, and substituting
those back into Eq.13.6b yields:

 U 2  u *  U 2  u *  U 2  P *  U     2 u *  2 u * 
 u *  v *  
 L  x *   L  y *    L  x *   L2  x * 2  y * 2 
         

 U2 
Dividing through by    , and simplifying yields:
 L 

u * u * P * 1   2 u *  2 u *  UL
u*  v*      where Re  (13.10)
x * y * x * Re  x * 2 y * 2  

Applying an order of magnitude analysis to Eq. 13.10 using Eq. 13.5 yields:

0(1) 0(1)  0(1) 0(1)  1   2u * 

0(1)  0(*)  0(1)  0(?)    
0(1) 0(*)  0(1)
2
0( * 2 )  Re  y *2 

neglect 0(1) also  Re 

Both terms 0(1) relative to
Must be because viscous terms can’t be zero!

Note that on the right side of Eq. 13.10, our order of magnitude analysis indicates that the
two derivative terms arising from the viscous shear are of 0(1) and 0(1/*2). Now, since
 2u *
*<<1, this means that 0(1) << 0(1/*2), and we can effectively neglect the term
x *2
 2u *
relative to the term. Note that we do not initially know the order of the reduced
y *2
1   2u * 
term,   , relative to the other terms in the equation. However, we reason that
Re  y * 2 
the viscous term must play an important role in the development of the boundary layer (it
cannot be negligible, or we would have basically an inviscid flow). Thus, we further
reason that the right-hand viscous term must be of 0(1), or of the same magnitude as the
momentum terms on the left-hand side of Eq.13.10. In order for this to be the case, this
 1 
requires that the Reynolds number of the flow must be very large, or ReL  0 2  .
 * 
That the Reynolds number must be large was one of the key assumptions made by
 2u
Prandtl, who derived the original boundary layer equations. Thus, neglecting the
x 2
term, the x-direction Navier-Stokes equation reduces to:

409
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter13

u u 1 P   2u 
u v    2  (13.11)
x y  x  y 

Now, non-dimensionalizing the y-direction N-S equation, Eq. 13.6c, and assessing the
order of magnitude of the terms gives:

v * v * P * 1   2 v *  2 v * 
u*  v*      (13.12)
x * y * y * Re  x *2 y *2 

0(*) 0(*)  0(*) 0(*) 

0(1)  0(*)  0(?)  0( *2 )  
0(1) 0(*)  0(1)
2
0( *2 ) 

All 0(*) or less

Note that our order of magnitude analysis indicates that all terms in Eq. 13.12 must be of
0(*) or less, with the possible exception of the pressure term. This is particularly
interesting, since it implies that all the inertia and viscous terms in the y-direction
 2u
Eq.13.12 are much, much smaller than any of the terms (except the term) in the x-
x 2
direction Eq. 13.10. This suggests that we can neglect all the terms in Eq.13.12 relative
to the terms in Eq. 13.10. But this also means that the y-direction pressure gradient term,
P *
, must also be of order 0(*) as well (since it must be in balance with all the other
y *
terms of Eq. 13.12).

Comparison of the order of magnitude results for Eq.13.10 to Eq.13.12 indicates that
P * P *
 0(* ) relative to  0(1) . Thus, to a good approximation we can say that at
y * x *
any streamwise location the pressure change across the boundary layer is essentially
negligible, or P  P(x ) , which means that the pressure at the edge of the boundary layer is
impressed across the boundary (i.e. the pressure within the boundary layer must be P(x),
the pressure at the interface of the boundary layer with the outer region flow).

13.2.2 The Boundary Layer Equations

Since all the terms comprising the y-direction equation are of 0(*) or less, this means
that to a good approximation we need only consider the solution of the reduced x-
direction boundary layer equation (Eq.13.11) and the continuity equation (Eq.13.9) in
order to determine a mathematical solution for the velocity field. Thus, the governing
equations for a two-dimensional boundary layer are:

410
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter13

u u 1 P  2u
u v   2 (13.13a)
x y  x y
And

u v
 0 (13.13b)
x y

The Two-Dimensional
Boundary Layer Equations

Note that Eqs.13.13 require specification of four boundary conditions to allow a complete
solution: Three for u (two on y, and one on x), and one for v (one on y). These are
respectively (see figure 13.3):

Boundary Conditions: 1) u=0 @ y=0

2) v=0 @ y=0
3) u  U  (x) @ y   or 
4) u = u(y) @ x=0
3
U(x)
y

1 2
4

Figure 13.3 Locations of the four boundary conditions required for solution of the
boundary layer equations.

Note that the boundary layer equations are parabolic. This means that the fluid can only
evolve based on the imposed boundary conditions as it moves downstream, and is not
influenced by downstream conditions (similar to a rock falling toward the ground; the
rock doesn’t know the ground is approaching until it hits). Here, the only way that the
downstream geometry can affect the boundary layer development is through changes in
the external pressure gradient, which will change with geometric conditions.

Now that we have derived the governing boundary layer equations, these can be solved
exactly for a few cases (the solutions, however, are not exactly simple). In addition, we
can obtain some approximate solutions for other cases (we will discuss how one does this
in Chapter 14). Additionally, the boundary layer equations can be solved numerically,
using relatively standard numerical integration techniques. However, one caveat on
numerical solutions is that under extreme changes in geometry, and thus pressure

411
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter13

gradient behavior, the assumptions that were invoked in the order of magnitude analysis
may break down (e.g. that v*  0(*) ), which requires that one revert to the solution of
the full Navier-Stokes solution, which makes solution procedures much more
complicated.

13.3 Solution of the Boundary Layer Equations for a Flat Plate Flow

Let’s now use the boundary layer equations (Eqs.13.13) to examine the simplest flow
application: a steady, laminar flow over a flat plate with a uniform approach flow (i.e.
U   constant) to the plate. This is essentially the type of flow shown in figure 13.3.

To establish the pressure gradient for this flow, we employ the Bernoulli equation for an
inviscid flow, and apply it to the flow just outside of the boundary layer. We assume this
external flow will be uniform (the flows outside the boundary layer are generally
relatively uniform, and thus behave basically like an inviscid flow). Applying the
1
Bernoulli equation, we have P  U 2  constant along a streamline external to the
2
boundary layer. If we differentiate this equation with respect to x, we get:

dP dU  dP
 U  0 0 (13.14)
dx dx dx
0
Since U   constant, there is no pressure gradient along the flat plate. Thus, substituting
Eq. 13.14 into Eq.13.13a eliminates the pressure gradient term, such that the boundary
layer equations for this flat plate flow become:

u u  2u
u v  2 (13.15a)
x y y
And
u v
 0 (13.15b)
x y

With the corresponding boundary conditions:

@ y0 @ y  @ x0 (13.16)

u0 u  U u  U
v0
Now, since we have no fixed dimension for this problem, we will assume a similarity-
type solution (that we explored in Chapter 12), with  assumed as a similarity (i.e.
“stretching”) variable that scales all the boundary layer velocity profiles along the plate,
or:

412
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter13

u  y
i.e.  f 
U 
Using physical reasoning, we assume that  will be a function of (i.e. depends on) the
free stream velocity, U, the distance along the plate, x, and the kinematic viscosity, :

i.e.  = f(U, x, )

Applying dimensional analysis, we can show that an appropriate functional relationship

should be:

 U x
 f  
x   

It was shown by Blasius that the appropriate dimensional relationship that preserves
dimensional correctness, and appropriately reduces the Eqs.13.15 is:

x

U

To scale changes in the y direction on changes in , we define a similarity variable, ,

where:

y U
 , so we let   y (13.17)
 x

This gives a velocity function of the form:

u
 f () (13.18)
U

However, before we can apply our similarity variable, we must reduce the two partial
differential boundary layer equations to one partial differential equation. To do this, we
make use of the stream function, like we did in section 12.5 for the laminar wall jet.
Why? Remember that the steam function is already an exact solution of the continuity
equation, and that the 2-D velocity components (in the Cartesian system) that satisfy
continuity are given by:
 
u and v (13.19)
y x

Substituting the functional relations of Eq.13.19 into the boundary layer equations of
Eqs.13.15 gives:

413
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter13

  2   2  3
   x-direction momentum (13.20a)
y xy x y 2 y3

and
 2  2
 0 continuity (13.20b)
xy yx

Thus, Eq.13.20b is satisfied exactly, leaving us with only Eq.13.20a, which is a non-
linear, partial differential equation for only one dependent variable (), rather than u and
v in the original Eqs.13.15.

While we now must solve only Eq.13.20a, the equation is a bit daunting since we have
raised the order of the equation from second order to third order. To obtain a similarity
solution that will reduce Eq.13.20a to an ordinary differential equation, we must relate
our dependent variable in Eq.13.20a, , to the velocity function, f(), and thus the
similarity variable, , from Eq.13.17 and 13.18. To do this, we note from Eq.13.19 that

u . We then use this relationship to integrate the stream function with respect to y
y
at a fixed streamwise (x) location, i.e.:
y
 
d x  fixed  dx  dy  udy x  fixed     o   udy (13.21)
x y 0

0 Set = 0

Equation 13.21 represents the planar volumetric flow rate of fluid over the indefinite
region between the surface (y = 0) and an indefinite height y. Substituting 13.18 into Eq.
13.21 and utilizing Eq. 13.17, we obtain:

y y  y
x
   udy   U  f ()dy   U  f () d 
(note that udy reflects the planar flow rate)
0 0 0
U 0

or

U
  U  x  f ()d  U  x F() where  y (13.22)
0
x

Define this integral

as F()

Note that in Eq.13.22 we set the integral of f() equal to F(), which is just another
indefinite function of , since  is an indefinite integration limit.

Now, using the expression for  from Eq. 13.22, we determine the respective terms in
Eq. 13.20a, and substitute these to obtain an ordinary differential equation for F()

414
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter13

 1 U     1 U 1 
e.g.  F  U  x F  y 
x 2 x x x 2 x x 2x

and thus,

 1 U 
 ( F  F) = -v
x 2 x
Likewise, we obtain:

  2 U  3  U 2
 U  F = u  U F =u/y  F =2u/y2
y y 2
x y 3 x

 2

    
   U  F  U  F   U  F      U  F =u/x
xy x  y  x x  2x  2x

Substituting these relationships into Eq. 13.21a,

  2   2  3
  
y xy x y 2 y3

gives,
  1  U   2    U  12 
1
 U U2
UF  F      F  F U    F    F ,
 2x   2  x     x   x
or
U 2 1 U 2 U 2 U 2
 FF  FF  FF  F ,
2x 2 x 2x x

which simplifies to what is known as the Blasius' equation:

1
F  FF  0 (13.23)
2
1
  1  U   2 U
Now, since u   U F , v      F  F , and   y ,
y x 2  x  x

we can derive the boundary conditions for F that apply for the Blasius equation as
follows:
y  0: u0   0: F  0
v0   0: F0
3 B.C.
y   : u  U     : F  1 same
x  0: u  U     : F  1

415
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter13

Note that the last two of the four boundary conditions for u and v in x and y reduce to one
common bounding condition for F' with respect to . Thus, we reduce the third order
P.D.E. with four boundary conditions for (x,y) to a third order O.D.E. for F() with
three boundary conditions. Equation 13.23 is non-linear, and doesn’t lend itself to a
closed form solution. The solution of this equation is obtained using standard numerical
solution techniques, generally a Runge-Kutta numerical technique. I chose to solve the
equation using the MATLAB function bvp4C, which works quite quickly. The result is a
set of tabulated data for F, F, and F , as shown in Table 13.1.

Table 13.1: Blasius table for F, F'and F'' vs. 

 F F' F''
0 0.0000 0.0000 0.3321
0.2 0.0066 0.0664 0.3320
0.4 0.0266 0.1328 0.3315
0.6 0.0597 0.1989 0.3301
0.8 0.1061 0.2647 0.3274
1 0.1656 0.3298 0.3230
1.2 0.2380 0.3938 0.3166
1.4 0.3230 0.4563 0.3079
1.6 0.4203 0.5168 0.2967
1.8 0.5295 0.5748 0.2829
2 0.6500 0.6298 0.2668
2.2 0.7812 0.6813 0.2484
2.4 0.9223 0.7290 0.2281
2.6 1.0725 0.7725 0.2065
2.8 1.2310 0.8115 0.1840
3 1.3968 0.8461 0.1614
4 2.3058 0.9555 0.0642
5 3.2833 0.9916 0.0159
6 4.2797 0.9990 0.0024
7 5.2793 0.9999 0.0002
8 6.2793 1.0000 0.0000
9 7.2793 1.0000 0.0000

The non-dimensional velocity profiles developed from this numerical solution are shown
in figure 13.4, on the following page.

416
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter13

6 6

5 5

4 4

 3  3

2 2

1 1

0 0
0 0.2 0.4 0.6 0.8 1 1.2 0 0.2 0.4 0.6 0.8 1
u v
 F Re x
U U

Figure 13.4 The behavior of the velocity components u and v within a Blasius flat
U
plate boundary layer, as a function of   y .
x

For the Blasius velocity components, note that:

u v 0.865
 U  F(0)  constant @   
 U Re x
 2u 1 v 1
 U  F(0)  ?   F(0)F(0)  0 For Re x  104   0.00865 
 2
2 U 100
this indicates an inflection
From D.E.
point occurs at the surface
Satisfies assumption employed in
(this means the velocity
simplifying the boundary layer equations.
profile is just stable)
i.e. that u >> v

 2u
Since  0 at  = 0, the streamwise velocity profile is inflectional at the solid surface,
 2
which makes the profile barely stable. As we will discuss in Chapter 17, destabilization
of a laminar boundary layer leads to transition from a well-behaved laminar flow (i.e. that
moves in laminas, or layers) to a much more chaotically mixed flow, which is termed
turbulence. Additionally, the above result that at a Reynolds number of 104 the normal
velocity at the edge of the boundary layer and beyond is on the order of 1/100 of the
magnitude of the mean outer flow supports our previous assumption in Section 13.2 that
streamwise velocity changes are the only significant velocity changes within the
boundary layer. Clearly, at larger Reynolds numbers, this ratio will be even smaller.
However, at lower Reynolds numbers the terms we neglected in Eqs. 13.6 will play a
more significant role, requiring the solution of the entire Navier-Stokes equation. The
good news is that the region within which such a solution is warranted is confined to very

417
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter13

near the initiation of the boundary layer, and neglecting this region has a minimal effect
on our Blasius solution. For example, for water flow over a flat plate at 1 m/s, a Reynolds
number of 104 will be reached after only 1cm.

13.4 Boundary Layer Parameters

As we pointed out in section 13.2, the boundary layer thickness parameter, , which
characterizes the thickness of the boundary layer, is a rather arbitrary parameter. It is
defined as the location above the surface where the velocity approaches 99% of the free
stream velocity, which may be difficult to establish with precision. We now examine
several parameters that can be derived based on more physical properties, and can more
reliably characterize the physical characteristics of the boundary layer.

13.4.1 Displacement Thickness, *

The displacement thickness, illustrated in figure 13.5, represents an imaginary

displacement of fluid from the surface to account for mass flow “lost” by the formation
of the boundary layer, when compared to the mass flow that would exist without the
boundary layer present (i.e. if the flow all the way to the boundary moved at the free
stream velocity, U).

U U
“lost” mass flow

*

Uniform Flow Boundary Layer Flow

Figure 13.5 The characterization of the displacement thickness as the lost mass
displaced from the surface by a distance *, modeled as the
displacement of a flow of uniform outer region velocity from the
surface by a thickness *.

To establish an equation for the displacement thickness, we use our model in figure 13.5
and equate the mass flow within the actual boundary layer to an equivalent, displaced
mass moving at U as:

   *
m
 tot   udy
0
  U  dy   U dy   U  dy
* 0 0
(13.24)

Total Equivalent
-U*
flowrate flowrate
within within
boundary displaced
layer uniform
flow

418
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter13

Reorganizing Eq. 13.24 gives:

 
U    U dy   udy
*

0 0

and solving for * gives:


 u 
   1 
*
dy for  = constant (13.25)
0
U  

Note that Eq. 13.25 is a very reliable measure of the degree to which the outer flow is
"displaced" from the proximity of the bounding surface. In practice, such a displacement
will make the effective geometry of the solid boundary appear slightly "thicker" to the
outer flow, which results in a modification of the outer flow field, and consequently the
associated pressure gradient. Comprehensive solution techniques generally take this
boundary layer displacement into account by using an iterative procedure to converge on
a combined boundary layer and outer region solution.

13.4.2 Momentum Thickness, 

Another useful physical characterization of the boundary layer is known as the

momentum thickness. The momentum thickness is a hypothetical displacement of fluid
of uniform velocity, U  , away from a bounding surface to account for the momentum
“lost” due to the formation of the boundary layer velocity profile (and dissipated by the
action of the applied shear stress). Figure 13.6 illustrates this hypothetical displacement
of momentum, .
U
U

Displaced Boundary Layer Flow

Uniform Flow

Figure 13.6 The characterization of the momentum thickness, , as a thickness of

fluid moving at the uniform outer region velocity, U  , that would have
to be removed to account for the momentum “lost” due to the formation
of the boundary layer.

To establish an equation to calculate the momentum thickness, we multiply the

momentum flux of a uniform flow by the lost momentum “thickness,” represented by .
We then equate this lost momentum equal to the difference between the maximum
possible momentum (reflected by U) that could be carried by the actual boundary layer
mass flow, less the actual momentum within the boundary layer, i.e.:

419
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter13

 
U 2    (udy )U    (udy )u (13.26)
0 0

Hypothetical Possible Actual

Thickness of Momentum Momentum
“Lost” Momentum
“Lost”
Momentum

Solving Eq. 13.26 for , and assuming an incompressible flow, we have:


u  u 
 1  dy (13.27)
0
U   U  

As we will see, because the momentum thickness reflects the momentum “lost” by the
formation of the boundary layer due to viscous effects, it is directly related to the shear
stress and the cumulative shear forces (i.e. drag) exerted by the fluid on a solid boundary.

13.4.3 Friction Coefficient, cf

The friction coefficient is a convenient, non-dimensional parameter reflecting the shear

stress at the surface divided by the dynamic pressure equivalent of the free stream flow
outside the boundary layer, defined as:

du

 wall dy
y 0
cf   (13.28)
1 1
U 2 U 2
2 2

The friction coefficient characterizes the non-dimensional shear stress at a specified

location along the boundary, and as such allows the determination of engineering shear
stress data for any similar laminar flow, regardless of the flow conditions. As we will
see, for a flat plate flow this coefficient is only a function of the local Reynolds number,
U x
Re x   , which allows effective analytical and empirical engineering correlations to

be developed where c f  f (Re x ) . As we will discuss in Chapter 17, this functional
dependence of cf on Reynolds number also applies to turbulent flows.

13.5 Boundary Layer Characteristics for a Flat Plate Flow

Using our solution for the velocity behavior that we developed in section 13.3, we can
now assess the various boundary layer properties for a laminar, flat plate flow.
Examining Table 13.1, we note that u/U  0.99 when  ≈ 5, so we choose  = 5 as the

420
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter13

hypothetical edge of the boundary layer, in similarity units. Thus, letting y =  and  = 5
in Eq. 13.17, we obtain:
u U  U x 
@ 5,  F(5)  0.992   @   5     Re x  5
U x x  x

 5
  (13.29)
x Re x

Note that Eq.13.29 indicates that the ratio of the boundary layer to its streamwise location
is only a function of the local Reynolds number. The use of the approximately equal
symbol () in Eq.13.29 reflects the arbitrary selection of u/U  0.99 as the edge of the
boundary layer, although Eq.13.29 is generally taken as a hard value in practice.

To solve for the displacement thickness, we use Eq.13.17 to allow the integration of
Eq.13.25 as a function of  (at a fixed x location):


 u  U U x
*   1  dy note:   y  d  dy  dy  d
0
U   x x U

Substituting for dy, and integrating from 0    5 gives,

5 5
 u  x x x
   F50
U  0
*  1   d  (1  F)d 
0  U  U U x



x
5  3.283  x 1.717
Re x Re x

Solving for the ratio of displacement thickness to the streamwise location, we have:

 * 1.717
 (13.30)
x Re x

Similarly, we can solve for the momentum thickness to give:


u  u  x
 1  dy  0.660
0
U  U  Re x

 0.660
 (13.31)
x Re x

421
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter13

Note that the solution for Eq.13.31 is a bit tricky, and requires the use of the original
boundary layer equation, Eq.13.13, and Table 13.1 to determine the integral. This is left
as an exercise for the reader.

One might wonder what the impact is of selecting  = 5 as the “edge” of the boundary
layer when integrating for the coefficients in Eq.13.30 and 13.31. To illustrate this, we
can calculate that the coefficient in Eq.13.30 changes from 1.7167 for integration over
0 <  < 5, to 1.7207 for integration over 0 <  < 9, or a 0.23% change. Correspondingly,
the coefficient in Eq.13.31 changes from 0.6600 for 0 <  < 5, to 0.6642 for 0 <  < 9, or
a 0.63% change. Clearly, the extension of the integration beyond  = 5 has only a
minimal effect on the displacement and momentum thickness results.



*
 *


x
Figure 13.7 The generic behavior of the boundary layer, displacement thickness, and
momentum thickness relative to the streamwise development distance, x.
The above analysis illustrates that the boundary layer, the displacement thickness, and the
momentum thickness all demonstrate the same functional relationship with regard to the
streamwise Reynolds number. As figure 13.7 shows, each of these properties grows
proportionally to the square root of the streamwise distance. When considered as a ratio
of each boundary layer property to the streamwise distance, these ratios all scale with the
inverse of the streamwise Reynolds number.
  
Note that Eqns. 13.29, 13.30, and 13.31 all suggest that , , and become infinite for
x x x
small Reynolds numbers. As discussed previously in Section 13.3, within the region very
near the leading edge of the flat plate, the assumptions employed to develop the boundary
layer equations fail, since the changes in y and x will become comparable. This leading
edge region requires the use of the full Navier-Stokes equations to develop a properly
exact solution. However, the amount of correction is so small, that the Blasius solution is
considered essentially exact for regions immediately removed from the leading edge.
To calculate the shear stress for the flat plate boundary layer, we substitute into the wall
shear stress equation, and use Table 13.1 to determine the shear stress as:

du u  U
w     where u  U  F,   y
dyy 0
 y 0 x
U 2
U  0.332U 
2
(13.32)
 U  F(0)  0.332 
x    U x Re x
0.332  
 

422
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter13

Substituting Eq. 13.32 into Eq. 13.28 for the friction coefficient, cf, we obtain:

w 0.664
cf  2
 (13.33)
1
2 U  Re x

cf

Figure 13.8 The generic behavior of the friction coefficient vs. streamwise distance,
x, for a laminar, flat plate flow.

Equation 13.33 indicates that the friction coefficient behaves similar to the ratios of the
boundary layer parameters (Eqns. 13.29, 13.30, and 13.31) with streamwise distance, i.e.
decreasing with increasing Reynolds number. This behavior illustrates that as the
Reynolds number increases, due to either increased outer region velocity or streamwise
distance, or decreased kinematic viscosity, viscous forces will decrease relative to inertia
forces, resulting in a consequent decrease in the friction coefficient. This behavior of cf
as a function of streamwise distance is illustrated in figure 13.8. Note that this decrease in
cf, does not mean that the actual shear stress will decrease, since if U  increases, this will
also increase the dynamic pressure  U , such that the shear stress will increase
1
2
2

3
(proportional to U 2 if x and  remain constant).

We should also note that a comparison of Eq. 13.33 for the friction coefficient and Eq.
13.31 for the momentum thickness shows that:
w 0.664 
cf  2
  (13.34)
1
2 U  Re x x

Note that in establishing Eq. 13.34, we have used the value of 0.664 for the constant in
Eq.13.31, which is the constant one gets for an upper limit of  = 9. This is where U =
U∞, which sets the equivalency between Eqs. 13.33 and 13.31. What Eq. 13.34 illustrates
is that the friction coefficient, cf, is directly proportional to the momentum thickness, θ.

This direct relationship of the momentum thickness to shear stress is even more clearly
illustrated by the calculation of the total plate drag due to shear stress, which follows.
To calculate the total drag on a flat plate due to a developing laminar boundary layer, we
integrate the shear stress, w(x), over the total streamwise length of the plate, L (and
width, W, although the two-dimensional boundary layer doesn’t vary with width):

x L

423
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter13

L
Drag    w Wdx (13.35)
0

Substituting Eq.13.32 into Eq.13.35, and simplifying yields:

L
1
Drag  0.332U 2 W  1
dx
0  U x 
2
 
  
L
U 2 W  x 2 
1 1
U 2 WL 2
 0.332 1
 1   0.664 1
 U    2  0
2 
 U  2
   
     

U 2 WL 0.664
Drag  0.664 1
 U 2 WL (13.36)
 UL  2 Re L
 
  
To non-dimensionalize the drag force of Eq.13.36, we divide through by the dynamic
force, 12 U2 (WL) , yielding:

Drag 1.328
CD   (13.37)
1
2 U  WL)
2
Re L

In Eq.13.37, CD is termed the drag coefficient, indicating the relative amount of viscous
drag due to the formation of a laminar boundary layer on a flat plate. By referencing the
drag to the outer region dynamic pressure ( 12 U2 ), and the area of the plate over which
the drag is generated (WL), Eq. 13.37 indicates that the relative drag decreases with
increasing Reynolds number (based on the total streamwise plate length). This behavior
is similar to the dependence of the friction coefficient, defined by Eq.13.33, on the local
Reynolds number. However, the friction coefficient is a local property, whereas the drag
coefficient is a cumulative property, which takes into account the integrated effect of
friction over the entire area of boundary layer development.

CD

Figure 13.9 The generic behavior of the drag coefficient as a function of total plate
length, L, for a laminar, flat plate flow.

424
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter13

Figure 13.9 shows the generic behavior for the drag coefficient as a function of total
streamwise plate length. It is instructive that increasing the length of the plate results in a
decrease in the drag coefficient. However, because CD is based on the total plate area,
note that actual drag will increase proportional to the square root of the plate length, as
illustrated by Eq. 13.36.

As discussed above, the friction coefficient, cf, is proportionally related to the momentum
thickness along the plate. However, if we compare Eq. 13.37 for the drag coefficient and
Eq. 13.31 for the momentum thickness at the end of the plate, x = L (again using a
constant of 0.664 for an upper limit of  = 9), we have:

Drag 1.328  x  L 0.664

CD  2
 and 
1
2 U  WL Re L L Re L

Where  x L is the value of θ at x = L. Thus, we can write:

Drag 2
CD  2
 x L
1
2 U  WL L

Or


Drag  U2 WxL  (13.38)

What Eq.13.38 indicates is that the total drag on the plate is directly proportional to the
momentum thickness at the end of the plate. This shows the characterization of the
momentum thickness as the momentum “lost” due to the viscous shear stresses acting on
the fluid as it passes over the plate. This is important, since this means that if the velocity
profile at the trailing edge of the plate is known, the integrated momentum thickness for
that profile is directly proportional to the viscous drag on the plate. And by extension, if
the velocity profile is known at any x location along the plate, Eq. 13.34 can be used to
determine the shear stress at that location.

13.5.1 An Example of Boundary Layer Parameters

Consider a thin flat plate 1 meter wide and 2 meters long that is towed through 20 °C
water at 0.1 m/s. We will compute the boundary layer thickness, displacement thickness,
and momentum thickness at the trailing edge of the plate in cm, and then compute the
total drag due to surface shear stress on the two surfaces (top and bottom surfaces) in
Newtons.

We assume that the flow is a Blasius-type flow and uniform in width.

m2
The kinematic viscosity of water at 20 °C is roughly   106 , and L = 2 m.
sec
Thus, the Reynolds number based on the plate length is:

425
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter13

U L
0.1 m 2m
Re L    sec  2x105 .
2
m

106 
sec

Note that flat plate Reynolds numbers less than 5x105 are generally considered laminar,
so this flow is well below that level. As we will discuss in Chapter 17, on smooth
surfaces with little external flow disturbance, a laminar flow boundary layer can often be
sustained for Reynolds numbers up to and slightly in excess of 106.

The boundary layer, displacement, and momentum thicknesses for this flow are:

 x L 5 5L 52m
   x L    0.0224 m  2.24 cm
L Re L Re L 2x105

1.717L 1.7172 m
*x L    0.00768 m  0.768 cm
Re L 2x105

0.660L 0.6602m
*x L    0.00295 m  0.295 cm
Re L 2x105

At this moderate Reynolds number, these thicknesses are quite small. Note that
 x L
 0.0112  1.12% .
L

The total drag on one surface is given by Eq. 13.36:

0.664
Drag  U 2 WL
Re L
kg
Here, W = 1 meter, U = 2 m/sec, and the density is roughly   1000 . Equation
m3
13.36 predicts a drag of:

0.664 kg m 2
kg  m
Drag  1000 3 4 2 1m2m  11.88 2
 11.88N
2x10 5 m s s

Alternatively, if we use Eq. 13.38 to calculate the drag using the momentum thickness,
we get:

kg m2
 
Drag  U 2 W x L  1000 3
4  2
1m0.00295m  11.8 kg 2 m  11.8N
m s s

426
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter13

Note that this value is slightly less than the calculation obtained using Eq.13.36. This is
due to the different coefficient value used in calculating the momentum thickness (0.66
vs. 0.664, as discussed previously in Section 13.5).

Using our first calculated values of drag, the total drag for both sides of the plate is:
Total Drag  2x11.84 N  23.68 N .

13.6 Falkner-Skan Solutions for Non-Zero Pressure Gradients

One might ask, what happens when the pressure gradient is non-zero for a laminar
boundary layer? Well, an external pressure gradient can have quite an effect on the
behavior of the boundary layer. As we will discuss qualitatively in section 14.5, a
favorable pressure gradient (an accelerating flow, with a decreasing stream-wise
pressure) will result in a boundary layer that grows slowly and remains quite stable.
However, the presence of an adverse pressure gradient (a decelerating flow, with
increasing stream-wise pressure) will cause a rapid growth of the boundary layer, often
leading to flow separation (i.e. stagnation of flow very near the surface, causing the
boundary-layer fluid to detach and move away from the bounding surface) or rapid
degeneration of the boundary layer to turbulence. In this section, we will address the
simplest boundary layer flow with a pressure gradient, known as a Falkner-Skan solution.
We will again assume that the boundary layer velocity profile has a similarity solution,
and assess what type of pressure gradient would still allow a similarity approach to be
feasible.

To start our solution, we expand the solution of the boundary layer equations to flows
which have a free stream velocity that varies as a function of x, or U∞ = U∞(x). Using the
same similarity assumption that was employed by Blasius, we again assume a stream
function and similarity variable as defined by Eq. 13.22:

U
  Ux F() where  y
x

However, we now assume the free stream velocity to be of the form U   U  ( x ) .

Since the pressure gradient for these flows will be non-zero, the boundary layer equations
to be addressed are Eqs. 13.13:

u u 1 P  2u
u v   2
x y  x y
And

u v
 0
x y

427
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter13

Using the stream function relations for u and v developed in section 13.3, Eqs. 13.13 can
again be reduced to a single partial differential equation (since Eq. 13.13b, continuity, is
identically satisfied by the stream function) of the form:

  2   2 dU   3
  U  3 (13.39)
y xy x y 2 dx y

Here, we have assumed that the pressure gradient within the boundary layer is again
equivalent to that imposed by the adjacent free stream flow (since dP  0) , and have
dy
applied the Bernoulli equation to establish the effect of the pressure gradient using:

dP dU  1 dP dU 
 U  0    U
dx dx  dx dx

Now, using the similarity variables  and  from Eq. 13.22, we again determine the
respective derivative terms in Eq. 13.39 and simplify the resulting equation. The result is
dU 
an ordinary differential equation for F() and U  . Note that this substitution
dx
process is a bit messier than the Blasius formulation, since we must consider the x-
derivatives of U∞ in Eq. 13.22. The respective derivatives of  are:

 1/ 2  1 1 dU  1 1 d 
 U  x   F F  F  (13.40)
x  2 U  dx 2 x dx 

and

 y  U   1 1 dU     1 1 dU  
           (13.41)
x 2  x  x U  dx  2  x U  dx 

Substituting Eq.13.41 into Eq.13.40 and simplifying, we obtain:

1/ 2
 1  U      x dU   
   F  F   F  F = - v
x 2  x    U  dx  

Likewise, we can show:

1/ 2
 d   U  
 U  x  F
1/ 2
 U  F = u where  
y dy y  x 

and

428
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter13

1/ 2
 2  U 
2
 U  F  U     F = u/y
y y  x 

and

 3 U 2
3
 F = 2u/y2
y x

and

1/ 2
 2     1  U     x dU    x dU    x dU    d
      1  F    1  F    1  F
xy y  x  2  x   U  dx   U  dx   U  dx   dy
1 U   x dU  
  2F  F  F
2 x  U  dx 
= u/x

Substituting these relationships into Eq. 13.39:

  2   2 dU   3
  U   
y xy x y 2 dx y3

gives,

U F 1 U  x dU 
2F  F  F
2 x  U  dx 
1/ 2
1 U    x dU     U  1 / 2
    F  F   F  F U    F
2 x    U  dx    x 
dU  U2
 U    F
dx x

Simplifying gives:

U 2  x dU    x dU   2x dU  2 
2x
  
 2F2  FF  FF  FF  FF   FF  FF    F
 U  dx   U  dx  U  dx  
 x dU   2  x dU   2x dU 
 2F  FF   FF   2F
 U  dx   U  dx  U  dx

And, rearranging we finally have:

429
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter13

1  x dU   2 1  x dU    x dU  
F  FF   F   FF     0
2  U  dx  2  U  dx   U  dx 

1 x dU   1 
F  FF  1  F 2
 FF  0 (13.42)
2 U dx  2 

Equation 13.42 is a generalized equation for an assumed similarity velocity profile within
the boundary layer, which takes into account the effect of the pressure gradient on
boundary layer development. However, in order for this equation to be solvable, all
terms in the equation must be either be derivatives of F, functions of , or constants.
Thus, in order to for Eq. 13.42 to be solvable, the term related to the pressure gradient
 x dU  
effects,   , must be a constant. Thus, we set:
 U  dx 

x dU 
 m  constant (13.43)
U  dx

Assuming a relationship according to Eq. 13.43, we separate variables and integrate,

which gives:

U   x 
ln    m ln 
 U0   x0 
or
m
 x 
U   U 0   (13.44)
 x0 

While it is not initially obvious, this velocity function represents the velocity behavior for
inviscid flow over a wedge or corner. Recall from Section 9.7.5 that the complex
potential function for a wedge/corner flow could be represented by (Eq. 9.41 and 9.42):

A n A rn
( z )  z  cos n  i sinn
n n

From which we could determine the velocity along a radius r (Eq. 9.46) as:

1 
vr   A r n 1 cos n
r 
(13.45)

The generic configurations of the bounding shapes giving such an inviscid flow are
shown below for n >1 and n <1.

430
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter13

r &x
U 


U
u=0 r &x u=
(a) Concave corner (b) Convex corner

n > 1 (concave corner flow) n < 1 (convex corner flow)

Here, we consider that the boundary develops along a surface in the r-direction, which we
equate with the x-direction in the boundary layer equations. Thus, we can rewrite
equation 13.43, in terms of x, as:

n 1 m
 x   x 
vr  U  A x n 1
cos n  Ax n 1
0 cos n   U 0  
 x0   x0 

Where Ax 0n 1 cos n  U 0 , and m = n-1. While the types of free stream velocities that
Eq. 13.44 can represent are limited, it can represent the general behavior for an
accelerated flow (m > 0) and a decelerated flow (m < 0) for a flow satisfying our
similarity constraints.

Thus, for a free stream velocity described by Eq.13.44, Eq. 13.42 becomes:

1  1 
F  FF  m 1  F 2  FF  0
2  2 
or

1
F 
2

1  m FF  m 1  F 2  0  (13.46)

Equation 13.46 is similar to the classic Falkner-Skan equation, first derived by Falkner
and Skan (1931). In their derivation, they used a modified similarity variable:

m 1 U
FS  y
2 x

This makes the reduced equation, like 13.44, a little bit simpler (but not much), and it
makes the boundary layer merge with the outer region for FS  5 , like the Blasius
boundary layer does. However, by retaining the Blasius similarity variable, this shows
the reflective variation of the outer boundary layer  limit. As we will see, as m varies,
not only will the shape of the velocity profile change, but the extent of the boundary layer
in  units. The choice of  does not change the results [since (m+1)/2 is a constant for
any selected flow], but only how they scale in  units.

431
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter13

To set the boundary conditions, we again note that:

1/ 2
  1 U 
u  U F , v     F  F  mF  F
y x 2 x 
and
U
 y .
x

Although the velocity expressions are a bit more complicated, the boundary conditions on
F and  for the Falkner-Skan equation are the same as for the Blasius equation, as
follows:

y  0: u, v  0    0: F  F  0
y: u  U  : F  1

Note that we do not include an initial condition at x = 0, since this would over-determine
the problem. Unlike the flat plate flow, the initial flow field at x = 0 is unrealistic, since
for m > 0 the velocity must be zero, and for m < 0 the velocity is undetermined at x = 0.
So, the free stream velocity behavior defined by Eq. 13.44 can only apply for values of
x > 0. And for x > 0, the three boundary conditions listed suffice. Thus, we have again
reduced the third-order P.D.E. with four boundary conditions for (x,y) to a third-order
O.D.E. for F() with three boundary conditions.

Like the Blasius equation, Eq. 13.46 is non-linear, and doesn’t lend itself to a closed-form
solution. The solution of this equation is obtained numerically using a Runge-Kutta
differential equation solver (I again used the MATLAB function bvp4C). Note that while
solutions for accelerating flows exist for all values of m > 0, the corresponding solutions
for decelerating flows can only be obtained for -0.0905 < m < 0. This latter result
suggests how sensitive laminar boundary layers are to positive (adverse) pressure
gradients. Figures illustrating u (streamwise) and v (normal) velocities and shear stress
profiles for select m values from 1 to -0.0905 are shown in figure 13.10a and 13.10b.

Note that for m > 0, the u-velocity profiles are much flatter, with a thinner boundary layer
than the Blasius flat plate flow (represented by m = 0). This is a result of the assistance
of the decresing pressure gradient, which acts in the direction of flow, and helps counter
the viscous forces. Correspondingly, the v-velocity profiles for m > 0 show the
development of an increasingly negative velocity normal to the wall (i.e. toward the
wall) as m increases. Note that for m = 1, the v-veloctiy moves very strongly toward the
surface. To understand what is happening, note that m = 1 defines a free-stream flow that
is impinging normal to the bounding surface (i.e. a “stagnation” flow), as represented for
an inviscid flow in section 9.7.5.1, where n = 2  m = 1 in Eq. 13.44. For a stagnation
flow, the v velocity will actually be the velocity of the inviscid free stream toward the
surface, and thus the reason that v increases so rapidly away from the surface.

432
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter13

(a) Nondimensional x and y-direction velocities, u and v, vs. 

(b) Nondimensional shear stress, , vs. 

Figure 13.10 Falkner-Skan similarity velocity and shear-stress profiles vs. the
m
 x 
similarity parameter, , for a free-stream velocity U   U 0   .
 x0 

433
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter13

As illustrated by figure 13.10b, the shear stress within an accelerated boundary layer flow
increases quite rapidly with increasing m, which is a reflection of the increased
streamwise velocity gradient at the wall, as reflected in figure 13.10a. Additionally, the
slope of the shear stress (the second derivative of u) is positive for m > 0 flows, which is
indicative of greater flow stability (we will discuss this further in Section 14.5.1). Note
that solutions to the Falkner-Skan equation exist for m > 1. However , these flows would
represent flows into a strongly concave corner (reversing the flow back on itself), which
is physically unrealistic, and so of little practical interest.

If we consider the boundary layer behavior for decelerating flows, m < 0, the range over
which flow solutions are available is limited to -0.0905 < m < 0. As shown by the u-
velocity and shear profiles of figures 13.10a and 13.10b, as m becomes more negative the
u-velocity profile develops an obvious inflection, indicated by a maximum in the shear
stress developing away from the wall. When m = -0.0905, the velocity near the wall
essentially comes to a stop, and the wall shear stress becomes zero. While zero shear
stress might seem like a good thing, since this would indicate no boundary drag, this
condition is problematic. Zero shear stress implies that a layer of stagnant fluid is present
adjacent to the wall, around which the flow moving downstream along the wall must
divert, causing what is termed a flow separation (a separation of the boundary layer from
the surface). This also means that the scaling assumptions that were employed to derive
the boundary layer equations will no longer be be valid. Thus, beyond a point of
separation, the boundary layer equations become invalid, and the flow must be assessed
using the full Navier- Stokes equations---making the solution procedure much more
complicated. We will discuss the onset and consequences of flow separation in detail in
Section 14.6 and in Chapter 15.

For the Falkner-Skan solutions we can derive appropriate boundary layer parameters for
displacement and momentum thickness. The displacement thickness is:


 u  x
*   1  dy where dy  d
0
U  U

8 8
 u  x x xx
   Fm 80
U  0
*  1   d  (1  Fm )d 
0  U  U U x

x
*m  8  Fm (8) (13.47)
Re x

In Eq. 13.47, Fm indicates the function of F() as determined from integrating Eq. 13.46
for a specified m value, and *m is the corresponding displacement thickness for the
specified m value. Note that we take  = 8 as the outer integration limit, since when m <
0 the point at which u  U  moves beyond the  = 5 we assumed for the Blasius
solution (m=0).
The momentum thickness takes a bit more work to determine, but is:

434
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter13

2x  Fm 0  mFm 8  8

 n 8
u  u  x
m  
0
1 
U   U  
dy 
U  F 1  F d 
0
m m 
Re x  1  3m
 (13.48)


To show the relative changes in the displacement and momentum thickness, we

normalize Eqs. 13.47 and 13.48 on the respective Blasius expressions for a zero pressure
gradient from Eqs. 13.30 and 13.31 to give:
x
8  Fm (8)
m *  *m Re x
 *'     0.58248  Fm (8) (13.49)
 * Blasius  *m 0 1.717
x
Re x
and

2x  Fm 0  mFm 8  8

 
m m Re x  1  3m   F 0  mFm 8  8
'     3.03 m  (13.50)
 Blasius  m 0 x  1  3m 
0.660
Re x

2.5 4.5
(delta*)' 4
2.0 (theta)' 3.5
H 3
(*=delta*)' 1.5 2.5
H
2
(=theta)' 1.0
1.5

0.5 1
0.5
0.0 0
-0.2 0 0.2 0.4 0.6 0.8 1
m

Figure 13.11 The behavior of normalized * and  (as per Eqs. 13.49 and 13.50), and
shape factor H, as a function of the power law exponent m in Eq. 13.44.

The normalized displacement thickness and momentum thickness from Eqs. 13.49 and
13.50 are plotted for a range -0.0905 < m < 1 in figure 13.11. In addition, the ratio of
*
these properties, H  , known as the "shape factor" of the boundary layer is also

plotted in figure 13.11. This latter property is a quantitative indication of how broadly
the boundary layer extends, relative to the momentum lost to viscous friction. As figure

435
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter13

13.11 shows, for accelerating flows (m > 0), the normalized displacement and momentum
thicknesses, and the shape factor, all decrease with increasing m. However, the shape
factor decreases markedly slower, since the relative changes in the displacement and
momentum thicknesses are quite similar. For decelerating flows (m < 0), all three
properties grow quite rapidly with decreasing m, and essentially terminate for
m = -0.0905, where the boundary layer essentially reaches separation.

It should be noted that the when H grows beyond the Blasius value of 2.6 at m = 0, the
boundary layer becomes quite unstable, and will usually breakdown into turbulence
before it would undergo laminar separation. The process of breakdown to turbulence and
the role of instabilities on fluid behavior is discussed in detail in Chapter 17.

13.7 Conclusion

In the present chapter, we developed the equations governing boundary layer

development, and touched on the basics of boundary layer theory. The boundary layer
equations developed in Section 13.2 can be applied to a broad variety of flow geometries,
as long as the flow remains laminar, and the order of magnitude assumptions employed in
Section 13.2 remain valid (/x < 1/100). The laminar flat-plate solution of Blasius,
developed in Section 13.3, is accepted as essentially an exact solution, and widely used to
reference laminar flat-plate boundary layer growth, along with the boundary layer
parameters developed in Section 13.5.

However, for more complicated flows, the solution of the boundary layer equations can
be much more involved, requiring significant numerical computations, as illustrated by
the Falkner-Skan solutions of section 13.6. However, as we will demonstrate in the
following chapter, we can often obtain good engineering results for somewhat
complicated laminar flows using approximate solution techniques, which are much more
tractable than the solution of the full equations, and provide results of acceptable
engineering accuracy.

References

V. M. Falkner and S. W. Skan, (1931), "Some Approximate Solutions of the Boundary

Layer Equations," Phil. Mag., vol.12, no. 7, pp 865-896. [also in Aero. Res. Coun. Rep.
and Mem. no 1314, 1930]

Problems

1. A thin flat plate 1 ft. wide and 4 ft. long is towed through 68 °F water at 2 ft./s
Compute the drag due to surface resistance on the two surfaces (in lbf.)

2. For the flow described in problem 1, determine the boundary layer, displacement, and
momentum thicknesses at the end of the plate.

436
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter13

3. A laminar, uniform air flow at 20 °C passes over a thin flat plate 1 meter wide and 4
meters long. If the Reynolds number based on the plate length is 5x105, determine the
flow velocity, and the boundary layer, displacement, and momentum thicknesses at the
end of the plate, compute the drag due to surface resistance on the two surfaces (in
Newtons.)

4. Use the tabular data from table 13.1 to calculate the shear stress, yx, vs.  in the
Blasius boundary layer for water ( = 10-3 N-s/m2,  = 10-2 cm2/s) for a flat plate flow
with U=10 cm/s, at a location x = 100 cm. Plot a graph of yx vs.  for 0 <  < 8.
Hint: copy the table from this chapter to an Excel sheet to do the plotting. Note that
 u v  v
 yx     , and you will have to determine from  U  x F() , similar to
 y x  x
u v
how is determined on page 13 in section 13.3. How much does the term
y x
contribute to the shear stress?

5. Using the Blasius results from section13.3 and table 13.1, determine if the boundary
U x
layer equations should give reasonable results for Re x    1000 . If not, what is

the minimum value for Re x that would satisfy our initial assumption for using the
v 
boundary layer equations? Hint: consider the ratio of and at the edge of the
u x
boundary layer.

6. Consider a uniform flow, u = U entering a 2-D duct with sharp leading edges as shown
below. The duct height is h and the flow remains laminar at all times through the duct.
Assume the flow may be broken into viscous and inviscid regions as shown. Make use
of the Blasius solution and determine:

a) h = h(x) such that dp/dx = 0 (let ho be the initial duct height).

b) the length, L, where your expression for h(x) ceases to be valid.

7. For the Blasius solution, prove that /x = 0.664/(Rex)0.5 (hint: make use of the original
ODE to assist the integration).

437
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter13

2x  Fm 0  mFm 8  8

8. For the Falkner-Skan solution, prove that  m    , where
Re x  1  3m 
 m is the momentum thickness, and Fm and Fm are functions at a specified value of m.
(hint: make use of the original ODE, Eq. 13.44, to assist the integration). Take the
upper limit of the boundary layer where @ =8 (where u  U  ). Note that the values
of the Fm function, and its derivative value are:

at   0, Fm  0, Fm  0, Fm  a constant depending on m;

at   8, Fm  a constant depending on m, Fm  1, Fm  0

9. The following table is a Falkner-Skan solution for m = -0.05, a moderate adverse

pressure gradient. Copy this table to an Excel sheet, and determine and plot y vs. u (y
vertical axis, y horizontal axis) at x=10, 20, and 30 cm, for Uo = 10 cm/s at x = 10 cm,
and  = 10-2 cm2/s. All three plots should be on one graph. Hint: use the expression for
u on page 26, and note that U   f (x ) given by Eq. 13.44.

Falkner-Skan table
for m = - 0.05

 F F' F"
0 0 0 0.2135
0.2 0.0043 0.0437 0.2235
0.4 0.0176 0.0894 0.2332
0.6 0.0402 0.137 0.2424
0.8 0.0725 0.1863 0.2509
1 0.1148 0.2372 0.2582
1.2 0.1675 0.2895 0.264
1.4 0.2307 0.3427 0.268
1.6 0.3046 0.3965 0.2698
1.8 0.3893 0.4504 0.2691
2 0.4848 0.504 0.2658
2.2 0.5908 0.5566 0.2595
2.4 0.7073 0.6076 0.2504
2.6 0.8337 0.6566 0.2386
2.8 0.9697 0.7029 0.2241
3 1.1147 0.7461 0.2075
3.2 1.2679 0.7858 0.1892
3.4 1.4287 0.8217 0.1698
3.6 1.5963 0.8536 0.1498
3.8 1.7699 0.8816 0.13

438
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter13

4 1.9487 0.9057 0.1108

4.2 2.1319 0.926 0.0927
4.4 2.3189 0.9429 0.0762
4.6 2.5089 0.9566 0.0614
4.8 2.7013 0.9676 0.0486
5 2.8957 0.9762 0.0377
5.2 3.0916 0.9828 0.0287
5.4 3.2887 0.9878 0.0215
5.6 3.4867 0.9915 0.0157
5.8 3.6852 0.9942 0.0113
6 3.8843 0.9961 0.008
6.2 4.0836 0.9974 0.0055
6.4 4.2832 0.9983 0.0037
6.6 4.4829 0.9989 0.0025
6.8 4.6828 0.9993 0.0016
7 4.8827 0.9996 0.001
7.2 5.0826 0.9998 0.0007
7.4 5.2826 0.9999 0.0004
7.6 5.4826 0.9999 0.0002
7.8 5.6826 1 0.0001
8 5.8826 1 0.0001

10. The table shown in Problem 9 is a Falkner-Skan solution for m = -0.05, a moderate
adverse pressure gradient. Using appropriate information from that table, determine
general expressions for *, , and shape factor H = */. On a single graph, plot U  ,
* ,and  vs. x (all on one graph) from x = 10 to 100 cm, where Uo = 10 cm/s and  =
10-2 cm2/s. Note that U   f (x ) (given by Eq. 13.44) and must be taken into account
when calculating Re x . Plot * and  on the primary vertical axis, and U  using a
secondary vertical axis.

439
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 14

Chapter 14

Approximate Solutions of the Boundary Layer Equations

Contents

14.1 The Approximation Concept....................................................................................... 440

14.2 The Momentum Integral Equation for a Boundary Layer ............................................. 441
14.2.1 The Simplified Momentum Integral Equation .................................................. 445
14.2.2 Using Similarity Velocity Profiles with the Momentum Integral Equation .......... 446

14.3 Approximate Solution with a Pressure Gradient: Thwaites' Method ........................... 451

14.4 Predicting Boundary Layer Characteristics with Thwaites' Method .............................. 456
14.4.1 Flat Plate Flow ................................................................................................ 456
14.4.2 Linearly Accelerating Flow .............................................................................. 457
14.4.3 Decelerating Flow ........................................................................................... 459

14.5 The Relationship of Shear, Drag, and Momentum Thickness when dP/dx = 0 ............. 461

14.6 The Effect of a Pressure Gradient on Boundary Layers ................................................ 462

14.6.1 Favorable Pressure Gradient—Accelerating Flow ............................................. 464
14.6.2 Zero Pressure Gradient—Constant Velocity Flow ............................................. 465
14.6.3 Adverse Pressure Gradient—Decelerating Flow ............................................... 466

14.7 Boundary Layer Separation ....................................................................................... 467

14.1 The Approximation Concept

The exact solution of the boundary layer equations can, as we saw in Chapter 13, be
mathematically quite complicated. Because of this complexity, exact solutions are
possible for only a selected number of cases, such as the generic flat plate and Falkner-
Skan solutions we did previously in Sections 13.3 and 13.6. However, by relaxing the
need for exactness, and using some creative approximations, we can develop some rather
useful approaches that will yield good, approximate (but relatively accurate) solutions of
the boundary layer equations. These approximation approaches are fairly versatile, and
allow consideration of a much broader variety of flow geometries and conditions.

In this chapter, we illustrate two ways to do an approximate solution:

1) Solve the exact equations using an approximate solution technique, or

2) Solve the approximate equations exactly.

Both of the approaches we discuss are relatively old techniques, but they illustrate the
kind of creative thinking that can be employed to simplify the solution of the boundary

440
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 14

layer equations, and develop relatively accurate results. As discussed previously, we will
not be covering numerical solution techniques in this book, but an understanding of these
approximate approaches can give an appreciation of how more complicated numerical
approaches can be applied to either the full equations or the approximate approaches we
develop in this chapter.

14.2 The Momentum Integral Equation for a Boundary Layer

This first approach was originated by Theodore von Karman, and is often called the von
Karman Integral Equation. This approach is based on a similarity concept, assuming that
most boundary layer velocity profiles will look more or less similar at any position along
a boundary (e.g. the Falkner-Skan solutions of section 13.6). Thus, we might be able to
assume a mathematically simple similarity velocity profile, which satisfies the main
boundary conditions, and then use this assumed velocity profile to derive the boundary
layer parameters for a particular flow through the use of the integrated boundary layer
equations.

The Idea:

Develop an equation that can accept “approximate” velocity profiles as input and
yield accurate (close, but approximate) shear stress, , * , and  as outputs.

The Approach:

Integrate the boundary layer equations across the boundary layer, between
0  y   , and use assumed similarity velocity profiles to solve the resulting
integral equation, and establish the resultant boundary layer properties.

We start with the two-dimensional, steady boundary layer equations, and the appropriate
boundary conditions from Eqs. 13.13. We let U be the velocity at the edge of the
boundary layer, which is assumed to be a function of x and behaves in an essentially
inviscid manner. Thus, to a good approximation the pressure gradient is given by the
dP dU  dP
Bernoulli equation as  U  . Substituting for in Eq. 13.13a, we have:
dx dx dx

u u dU   2u
u v  U  2 (14.1a)
x y dx y
u v
 0 (14.1b)
x y

B.C. y  0 : u, v  0 (14.1c)
y   : u  U  x 

Note that in Eq. 14.1c we do not include the initial condition [u(y) at x = 0], since this
condition will be incorporated into the integral equation that will be derived here. For
our purposes of integration, all we require are the three spatial boundary conditions listed.

441
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 14

To begin our integration process, we first separate the v-momentum transport term
(second term on left) in Eq. 14.1a by parts, and use the boundary layer continuity
equation, Eq. 14.1b, to rewrite one of the resulting terms, to give:

u   v  ( uv )  u 
v  ( uv )  u     u  (14.2)
y y  y  y  x 

(from continuity)

We now substitute Eq. 14.2 into Eq. 14.1a, collect like terms, and integrate across the
boundary layer from the wall (y = 0) to the edge of the boundary layer (y = ):

   
u ( uv ) dU   2u
 2u
0
x
dy  0 y dy   U
0
dx
dy  
0 y
2
dy (14.3)

1 2 3 4

To facilitate the integration of Eq. 14.3, we have labeled each term of the integrated
equation with a number. We now consider each of these terms in the following, and
determine what they reduce to when integrated.

Consider term (2) of Eq. 14.3 first: 0 (at y=0)

  
( uv )  v u
0 y dy  uv 0  U  v   0; Note :  dy  v   v 0    dy
0
y 0
x
v u

y x (from continuity)
 
( uv) u
0 y dy  U 0 x dy (14.4)

To obtain Eq. 14.4, we again make use of the continuity equation, Eq. 14.1b. By
integrating the v-derivative across the boundary layer, and noting that v = 0 at y = 0, we
develop a term for v  in terms of the integrated u-derivative, which is substituted to yield
Eq. 14.4.

Term (4) in Eq. 14.3 can be integrated easily, yielding first derivative limits at the wall
and the edge of the boundary layer. Since the u velocity asymptotes to the outer region
u
velocity at the edge of the boundary layer, this means that the derivative
y y  
u
asymptotes to zero, leaving only the term.
y y0

442
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 14

0 (because u=Uf(y)at y=)

  
 2u   u  u u u
0 y 2 dy  0 y  y dy  y  y  y
0 y  y0


 2u u
0 y2 dy   y (14.5)
y0

Finally, note that we can rewrite term (1) as:

 
u ( u 2 )
 2u
0
x
dy  
0
 x
dy (14.6)

Substituting Eqs. 14.4, 14.5, and 14.6 back into Eq. 14.3 gives:

  
( u 2 ) u dU   u 
0 x dy  U 
0
x
dy  U
0

dx
dy   
 y  y 0
(14.7)

Since U is not a function of y, we can move U in the second term of Eq. 14.7 inside the
integral, and then expand the resulting integral using integration by parts:

 uU  
  
u u dU 
U  dy   U  dy     u  dy (14.8)
0
x 0
x 0
x dx 

Substituting Eq. 14.8 into Eq. 14.7, and rearranging terms, we get:

uU 
   
( u 2 ) dU  dU  u
0 x dy  0 x  dy  u
0
dx
dy  U
0

dx
dy  
y y  0
(14.9)



 x u  
2
 uU  dy dU 0
0  u  U  dx
0


dy 

by Leibnitz Rule 0 0


 
u 2  uU  dy   y  x 
f y, x dy  
 x  f
dy  f x , x 
x 
 f 0, x 
0
x 0 x y  0 0 x x x
Note that here,
f y, x  f((x))=0 and f(0)=0

Note that to simplify Eq. 14.9, we invoke Leibnitz rule, which considers the functional
change in the limits of the integration. As the above generic Leibnitz relationship shows,
while the upper integration limit, , will change with x, the value of the integrand at both
the limits (y =  and y = 0) will be zero, thus negating any effect of moving the derivative
function from outside to inside the integral. We also identify the last integrated term in

443
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 14

Eq. 14.9 as proportional to the wall shear stress,  w , which illustrates its relationship to
the momentum and pressure gradient changes reflected by the terms on the left side of
Eq. 14.9.

Rearranging the terms on the left side of Eq. 14.9, we can rewrite the equation as:

 
2  u u   u  dU 
2
 

x 0
U 

U
 
2

U 
 dy 


  U
0 
 1 U 
 dx
dy   w

(14.10)

To simplify further, we multiply Eq. 14.10 by –1 and factor the U  terms out of the
integrals, since U  x  does not depend on y. We then identify a set of terms in the
resulting equation that represent the boundary layer properties of displacement thickness
(*) and momentum thickness (), which we defined previously in Section 13.4 (Eqs.
13.25 and 13.27):

  2 u  
 
 u dU   u  
x  0 U   1  U
U  1  dy   U  dy  w
 U   dx 0   

 *

Rewriting Eq.14.10 in terms of these boundary layer properties, we have:

dU  w

x

U 2    * U 
dx



 dU 
U 2  2U 
x dx

Dividing through by U 2 yields,

d 1 dU  
 (*  2)  w2 (14.11)
dx U  dx U 

Momentum Pressure gradient Shear

changes /momentum effects
effects

In Eq. 14.11, we have identified the physical process that each term represents. Changes
in the momentum thickness obviously reflect changes in the total x-direction momentum
(i.e. the integrated momentum within the boundary layer). The last term reflects the
non-dimensional shear stress on the bounding surface. The middle term is a bit trickier,

444
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 14

since it reflects both pressure gradient effects and momentum changes. You will note that
back in Eq. 14.9 we combined a momentum term and the pressure gradient term, which
mixes the effects. However, this mixed property term is only significant when the outer
flow varies with x, which results in both pressure gradient effects, and
acceleration/deceleration of the boundary layer. This results in a change in the boundary
layer momentum, since the outer region velocity constitutes the outer boundary condition
for u. We will discuss the physical interrelationship of the terms in Eq. 14.11 later in
Section 14.4.

14.2.1 The Simplified Momentum Integral Equation

We now define another boundary layer property, H, termed the boundary layer “shape
*
factor”, defined as H  . This shape factor indicates the degree of distortion of the

boundary layer velocity profile, and is a good indicator of the type of behavior reflected
by the boundary layer. For example, laminar boundary layer shape factors are typically
larger than 2.2, whereas shape factors for turbulent boundary layers (which we will
discuss in Chapter 17), are generally less than 1.6. Variations of H also reflect whether a
flow is accelerating or decelerating: for laminar flows, values of H greater than 2.6 are
indicative of a decelerating outer flow (dU/dx < 0); H values less than 2.6 are indicative
of an accelerating outer flow (dU/dx > 0). The Blasius flat plate boundary layer has an
H value of 2.6.

Inclusion of the shape factor, and noting that c f  1 w 2 , allows Eq. 14.11 to be written
2
U 
as:

d  dU  c f
 ( H  2)  (14.12)
dx U  dx 2

The final result of the integration of the boundary layer equations across the boundary
layer is a differential equation for (x ) . This classic equation is called the von Karman
Momentum Integral Equation, after Theodore von Karman, who first derived it.

Note that the integral equation (i.e. Eqs. 14.11 and 14.12) is exact. That is, we have not
made any approximations to bring it to its present form. However, the approximation
now comes from the assumption of a particular similarity function for the velocity profile
within the boundary layer. The approach is to assume a simple profile that has the proper
velocity “shape” and satisfies the proper boundary conditions. We do this by writing
possible similarity velocity profiles in terms of a stretching (i.e. similarity) variable,
  y /  . We then apply the appropriate boundary conditions to the assumed general
velocity profile in order to determine the particular form for the assumed velocity profile.
We can do this effectively for laminar flows, which we demonstrate in the following
section. We can also apply this integral approach for turbulent flow behavior, although
this requires the use of some additional assumptions and empirical curve fits, as we will
demonstrate in Section 17.7.

445
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 14

14.2.2 Using Similarity Velocity Profiles with the Moment Integral Equation

For an approximate boundary layer velocity profile, we could choose any number of
possible velocity profile shapes, such as a linear profile, a sinusoidal profile, high-order
polynomials, etc. Interestingly, almost any assumed velocity profile works reasonably
well, as long it matches the boundary conditions. For our first example, we consider a
second-order polynomial velocity profile, where:
u
 a  b  c2 (14.13a)
U

In Eq.14.13a,  = y/ and a, b, c are undetermined constants.

The applicable boundary conditions for the flow are:
1) u  0 @ y  0 (  0)
2) u  U @ y   (  1) (14.13b)
du
3)   0 @ y   (  1)   0 @  1
d
Note that we only require as many boundary conditions for the assumed profile as there
are undetermined constants in the profile. So, for this case, we require three: (1) no slip
at the surface, (2) matching the outer region velocity at the boundary layer edge, and (3)
asymptoting to zero shear stress at the boundary layer edge. If we assumed a higher
order polynomial, we would have to match higher order derivatives (e.g. d2u/dy2 = 0 at 
= 1, etc.).

1
0.9
2
0.8 u  y  y
0.7  2    
U  
0.6
y/ 0.5
0.4
0.3
0.2
0.1
0
0 0.2 0.4 0.6 0.8 1 1.2

u/U

Figure 14.1 The shape of a second-order polynomial similarity velocity profile for a
laminar boundary layer.

To determine the particular form of this polynomial velocity profile, we solve for the
undetermined constants in Eq. 14.13a using the boundary conditions from Eq. 14.13b.

446
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 14

Thus:
B.C. #1  0  a  0  0  a  0
B.C. #2  1  b  c b=2
B.C. #3  0  b  2c c = -1
2
u  y  y
 2  2  2     (14.14)
U  
The shape of this assumed polynomial velocity profile from Eq. 14.14 is shown in figure
14.1.

Now, we use this approximate velocity profile described by Eq. 14.14 to determine the
functional relationship of the boundary layer parameters *, , and w in terms of the
boundary layer thickness, .

 u  dy
*   1   dy Note that d  , so dy   d
0
U   

 1 1
 u 
  1 
U

d   1  2  2 d 
0    0
1
 3  
*     2    (14.15a)
 3 0 3
and
  1
u  u  u  u 
 1  dy    1  d
0
U  U  0
U  U 
1
(14.15b)
2
   (2   )(1  2   )d 
2 2

0
15
and,
du du d 1 du 2U 
0      (14.15c)
dy y 0
d dy 0  d 0 

U  2  2 
0
We now substitute the boundary layer parameters from Eqs. 14.15 back into the
momentum integral equation, Eq. 14.11, to obtain a differential equation for (x):

d  2    4  1 dU  2U  2
     
dx  15   3 15  U  dx U  U 
2

or
2 d 3 dU  2
  (14.16)
15 dx 5 U  dx U 

447
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 14

dU 
However, recall that for a flat plate flow, U   const.  0
dx
2 d 2
  for a flat plate flow
15 dx U 

To solve for   (x ), we separate variables and integrate,

15
 d  U  dx
2
 15x
  C1 (14.17)
2 U
To determine C1 in Eq. 14.17, we need a bounding condition on  and x for the flow. For
a flat plate flow, with the boundary layer initiating at the leading edge, we have:

  0 @ x  0  C1  0

Note that if the boundary layer had some prior initial value (e.g.  0 at x0), we would have
used those values as our starting conditions, and then determine the constant C1 in Eq.
14.17. However, for C1 = 0 in Eq. 14.17, we establish the relationship for boundary layer
growth with streamwise distance as:

30 x x
  5.477
U U
or
 5.477 U x
 where Re x  (14.18)
x Re x 

Now using Eq. 14.18, and our previous Eqs. 14.15 -- the relationship of *, , and  w as
functions of  -- we can write:

 x 1.826 x
*   1.826  (14.19a)
3 U Re x

2 x 0.730 x
  0.730  (14.19b)
15 U Re x

w 
2U 

2U 

2U 2


0.730 12 U 2  (14.19c)
 x U x Re x
5.477 5.477
U 

448
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 14

Comparing the results of the integral solution, Eqs. 14.18 and 14.19, to the Blasius
solution of section 13.5 (Eqs. 13.29, 13.30, 13.31, and 13.32), we have:

 5.477
  1.095
 Blasius 5

* 1.826
  1.063
Blasius 1.717

 0.730
  1.106
Blasius 0.660

cf w 0.730
   1.106
c f Blasius  w Blasius 0.660

Therefore, the results from using our “approximate” velocity profile end up being about
10% high. Note that these results, although higher, are still relatively close to the exact
Blasius solution (and a lot easier to obtain). Such results are generally quite adequate for
most engineering applications.

In general, one can obtain better comparisons to the Blasius' solution by employing
velocity profiles that better approximate the Blasius velocity profile. Table 14.1 shows
the results obtained for the second-order polynomial profile of Eq.14.14, compared with
results for six other approximate velocity profiles, and the Blasius profile (Table 13.1).
Fig. 14.2 is a graphical comparison of these approximate profiles to the Blasius profile.
As Fig. 14.2 shows, the variation of the approximate profiles from the Blasius can be
quite significant. However, several of the results of Table 14.1 show remarkable
similarity to the Blasius results.

Table 14.1 illustrates that increasing the order of the polynomial profile does not
necessarily assure a closer match to the Blasius result; for example, the four-term, fourth-
order profile gives the worst results of all the approximate velocity profiles shown—even
worse than the linear profile. One might also assume that a velocity profile that most
closely compares to the Blasius profile would yield the best comparative boundary layer
parameters. However, this is also not necessarily the case, since a comparison of the
results of table 14.1 with the shape of the profiles in Fig. 14.2 illustrates that the closest
u
approximate profile shape is the second-order polynomial (i.e.  2  2 ), but this
U
profile does not give boundary layer parameters that are the closest to the Blasius results.
The best comparisons of boundary layer parameters are given by the three-term, fourth-
order polynomial and the sine-shaped profile, both of which diverge further from the
Blasius velocity profile than the second-order profile. Thus, while establishing the precise
“shape” of the velocity profile is desirable, it is not essential in order to obtain good
results for boundary layer parameters, and in particular the friction factor, cf.

449
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 14

Table 14.1
Comparison of predicted boundary layer characteristics
for selected approximate velocity profiles:
Flat plate flow with no pressure gradient

Assumed Velocity Profile

u  y  *  *
 f    f   Re x Re x Re x H
c f Re x
U  x x x 
f     3.464 1.732 0.577 3.00 0.577
f   2  2 5.447 1.826 0.730 2.50 0.730
f   1.5  0.53 4.641 1.740 0.646 2.69 0.646
f   3  32  3 7.483 1.871 0.801 2.34 0.801
f   4  62  43  4 9.486 1.897 0.843 2.25 0.843
f   2  23  4 5.836 1.751 0.685 2.55 0.686
 
f    sin   4.795 1.742 0.655 2.66 0.655
2 
Blasius Solution 5 1.717 0.660 2.60 0.660

Assumed Velocity Profile

u  y  *  H cf
 f    f  
U   Blasius  * Blasius  Blasius H Blasius c f Blasius
f     0.693 1.009 0.874 1.154 0.874
f   2  2 1.089 1.063 1.106 0.962 1.106
f   1.5  0.53 0.928 1.013 0.979 1.035 0.979
f   3  32  3 1.497 1.090 1.214 0.900 1.214
f   4  62  43  4 1.897 1.105 1.277 0.865 1.277
f   2  23  4 1.167 1.020 1.039 0.981 1.039
 
f    sin   0.959 1.014 0.992 1.023 0.992
2 
Blasius Solution 5 1.717 0.660 2.60 0.660

Note that Table 14.1 lists two different fourth-order velocity profiles. These result from
the use of the assumption of a different fifth boundary condition in developing the
assumed profile. For the four-term profile, the fifth boundary condition was
d 3u
 0 @ y   (  1) . For the three-term profile, the fifth boundary condition was
dy 3
d2u
 0 @ y  0 (  0) . See problem number 5 at the end of this chapter to see how
dy 2
we determine this latter boundary condition using the original boundary layer equations.

450
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 14

1.2
u/U = 
2
u/U = 2-
1.0 u/U = 1.5-0.5
3

2 3
u/U = 3-3 +
2 3 4
0.8 u/U = 4-6 +4 -
3 4
u/U = 2-2 +
u/U = sin( /2)
0.6 Blasius
y/
0.4

0.2

0.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2
u/U

Figure 14.2 Comparison of six different approximate velocity profiles with the
Blasius solution.

14.3 Approximate Solution with a Pressure Gradient: Thwaites’ Method

The use of similarity velocity profiles, as was done in Section 14.2, is one method of
solving the momentum integral equation exactly, using approximate velocity profiles.
However, the resulting equation can only be solved analytically for selected functional
variations of the free stream velocity with x. Additionally, the assumed velocity profile
may not be appropriate for various types of free stream flows where the velocity profiles
are not "similar". In this section, we will present another approximate method, which
does not require an assumed velocity profile, and can be used for any type of free stream
velocity, whether accelerating or decelerating.

We now consider the development of an exact solution of the approximated integral

equation using curve fits of existing analytical and experimental data. This technique has
no restrictions, and yields good, approximate answers within roughly 10%. The method
is based on the evaluation of all existing solutions of laminar flows available at the time it
was developed, and employs some judicious tabulation and curve fitting of the data to
create a method that establishes key boundary layer parameters by reducing the boundary
layer equations to a one-dimensional integral of the free-steam velocity behavior. Other
boundary layer characteristics are then obtained by correlation with a set of tabulated
data. The limitation of the method is that it is only useful for laminar flows. In honor of

451
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 14

the gentleman who developed this technique, Thwaites (1949), this approach is termed
Thwaites’ Method.

As our starting point, consider again the momentum integral equation, Eq. 14.12. Rather than employ
an approximate velocity profile, let us see how we can further manipulate the terms in the equation,
to yield terms that can be related to physical aspects of the flow, which we might be able to model
approximately:

Starting with the momentum integral equation, Eq. 14.12,

d 1 dU  
 ( 2   * )  w2 ,
dx U  dx U 
u
we note that at the bounding solid surface (y = 0),  w   . Additionally, we
y 0
multiply the equation through by 2U /  to give:
u

2 d 1 dU  2U  2U  y 0
U  ( 2  * )  (14.20)
 dx U  dx   U 2

Simplifying and rearranging Eq. 14.20 yields,

2 d  2 dU   *   u 
U  2  2     (14.21)
 dx   dx    U  y 0 

Now, we introduce the following dimensionless variables,

  u 
   (14.22a)
U  y 0
*
H (14.22b)

2 dU 
n (14.22c)
 dx

Note that Eq. 14.22a reflects a shear stress parameter, Eq. 14.22b is the “shape”
parameter discussed in Section 4.2.1 (reflecting the degree of deformation of the
boundary layer velocity profile), and Eq. 14.22c characterizes a pressure gradient
parameter. Substituting Eqs. 14.22 into Eq. 14.21 and rearranging a bit yields (prove this
yourself):

d  n 
 U    2n( H  2)   (14.23)
dx  dU  / dx 

452
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 14

Since U(x) is assumed known, Eq. 14.23 is a single equation for the three unknowns n,  , and H.
Thwaites collected all available analytical and experimental results for laminar boundary layers, and
computed the corresponding values of the n,  , and H parameters for each set of data. For example,
for a flat plate (using the Blasius solution of Section 13.5):

0.664x 1 U2 0.664 0.664

2
  u 
      0.220
U  y o U Re x 2  Re x 2

1.726x Re x
H  2.60
Re x 0.664x

2 dU 
n 0
 dx

Table 14.2
Parametric boundary layer
data developed by Thwaites

n  H
0.090 0.000 3.70
0.088 0.015 3.62 F(n) = 2[n(H+2)+L] vs. n
0.086 0.027 3.54 1.5
0.084 0.038 3.47
0.080 0.056 3.35
0.076 0.072 3.24 1
0.072 0.085 3.16
0.068 0.095 3.09 F(n)= 6.00n + 0.47
0.064 0.104 3.04
R² = 1.00 0.5
0.060 0.113 2.99
0.056 0.122 2.94 F(n)
0.048 0.138 2.87 0
0.040 0.153 2.81 -0.3 -0.2 -0.1 0 0.1 0.2
0.032 0.168 2.75
0.016 0.195 2.67 -0.5
0.000 0.220 2.60
-0.016 0.244 2.55
-0.032 0.268 2.49 -1
-0.048 0.291 2.44
-0.064 0.313 2.39
-0.080 0.333 2.34 -1.5
-0.100 0.359 2.28 n
-0.120 0.382 2.23
-0.140 0.404 2.18
-0.200 0.463 2.07
Figure 14.3 Linear regression of F(n) vs. n,
-0.25 0.5 2.00 reflected in Eq. 14.24.

In this way, Thwaites was able to develop a comprehensive table of n,  , and H for all
the available laminar boundary layer data, which is shown in Table 14.2. Using this
collected data, he determined that a linear curve fit, shown in Fig. 14.3, could be
developed for the parameter function on the right side of Eq. 14.23. Thwaites determined

453
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 14

that the parametric fit, Eq. 14.24, is essentially only a function of the parameter n (Note
that it did take Thwaites a number of tries to determine this proper collection of terms for
F(n)).

F(n )  2[n(H  2)  ]  0.47  6n (14.24)

Note that Eq. 14.24 is an Excel curve fit I did of Thwaites data, which yields a slightly
different equation than Thwaites determined, who utilized a less accurate regression
analysis. Substituting Eq.14.24 into Eq. 14.23, we obtain:

d  n 
 U    0.47  6n  F( n ) (14.25)
dx  dU  / dx 
2 dU 
Further substituting n   from Eq. 14.22c into Eq.14.25 gives:
 dx

d  2  2 dU 
U    0.47  6
dx     dx
or
d(  2 ) dU 
U  62  0.47 (14.26)
dx dx

We make the left-hand side of Eq. 14.26 a total differential by multiplying through by Um , and
establish an integrating factor, m:

d(  2 ) dU 
U m 1  62 U m  0.47U m (14.27)
dx dx
Noting that we can write,

d( U m 12 ) d(  2 ) dU 
 U m 1  2 ( m  1)U m ,
dx dx dx

we choose m+1 = 6, or m = 5, which allows us to rewrite the left side of Eq. 14.27 as:

d(  2 ) dU  d( U 6  2 )
U 6  6 2 U 5   0.47U 5 (14.28)
dx dx dx
Integrating Eq. 14.28 gives:
x

U  6

2
 0.47  U 5 dx (14.29)
0
x0

If we assume a starting value of 0  0 at x 0  0 in Eq. 14.29 we have,

454
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 14

x
U   0.47 U 5 dx
6 2
 (14.30)
0

Rearranging Eq. 14.30 gives an expression for the momentum thickness, :

x
2
 0.47U  6  U 5 dx (14.31)
 0

Substituting Eq. 14.31 into Eq. 14.22c gives an expression for n:

x
dU 

6
n  0.47U U 5 dx  (14.32)
dx 0

Thus, for a known U(x), we can compute the momentum thickness, , from Eq. 14.31,
and the pressure gradient parameter, n, from Eq. 14.32. Knowing n, the values of  and
H can then be obtained by either interpolation from Table 14.2, or a curve fit of H vs. n
and  vs. n. And knowing n,  , and H, all the boundary layer parameters (and w) can
be computed.

However, be careful using Eqs. 14.31 and 14.32. As we point out above, these are
derived assuming an initial value of  = 0 at x = 0. If the boundary layer has had some
initial development (e.g. development over a flat plate followed by an acceleration), then
you will need to integrate Eq.14.29 using an initial value of o = init at xo = 0, with the
corresponding initial velocity, U = Uinit at xo = 0.

Thus, reintegrating Eq.14.29 with  = init and U = Uinit at xo = 0, will give:

x

U  6 2
 U  U  6 2

6 2
init init  0.47 U5dx (14.33)
   init
0

or

6 x
 2 U init 2
 6 init  0.47U 6  U 5 dx
 U  0
0.5
 U6 2 x

. 0
6 5
   init
6
 init  0 47U  U  dx  (14.34)
 U 

with the value of n being given by:

dU  U init
6 x

 
n   6 init2
 0.47U 6
U 5
dx  (14.35)
dx U  
  0 

455
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 14

Equations 14.34 and 14.35 will then yield results for the behavior of a boundary layer
which had some prior development. Note that if we have o = init at xo = xinit, then the
computation gets a bit messier, but is still a straight forward integration, starting from x =
xinit, not x = 0.

Note that Thwaite's technique is essentially a curve fit of available data via the boundary
layer equation. This is not truly a predictive technique in that all the data shown in Table
14.2 was either measured experimentally or developed from analytical data, so what we
really have is a sophisticated curve fit of parametric data. That being said, Thwaites’
method is a very effective technique for prediction of a wide range of laminar flows, and
is still the technique of choice for engineering prediction of laminar flows, giving results
that are good to 10% or better.

14.4 Predicting Laminar Boundary Layer Characteristics Using Thwaites’ Method

To illustrate the use of Thwaites’ method, we will examine three different flows: the
conventional flat plate flow, an accelerating flow, and a decelerating flow.

14.4.1 Flat Plate Flow

A flat plate flow has a constant velocity outer region flow such that U  = constant. Thus,
dU 
 0 , and the pressure gradient parameter, n , is calculated from Eq. 14.32 as:
dx
x
dU 
dx 0
n  0.47U  6 U5dx

n0

Note that n = 0 for all x locations, which also fixes the values of  and H, as obtained
from Table 14.2.
x n  H
0 0.220 2.60

The value of the momentum thickness, , is obtained from Eq. 14.31 as:

x
2
 0.47U  6  U 5 dx
 0

 0.47U 6 U 5 ( x  0)

0.47x

U

456
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 14

x  0.685
  0.685 or  (14.36)
U x Re x

Note that the value of  determined from Eq. 14.36 is about 3% greater than that
determined from the Blasius solution, Eq. 13.31. This may seem curious, since the
Blasius values were incorporated into the data used to determine the function F(n) in
Eq.14.25. However, F(n) was a linear curve fit of the cumulative data, and as one can see
in Fig. 14.3, the curve fit is slightly higher than the actual data for n = 0, the flat plate
flow. Other versions of Thwaites' method have employed linear curve fits of the data in
Table 14.2 that purposely force the curve fit to pass through the flat plate data exactly,
since it is the most reliable data value in Table 14.2. However, the present curve fit is
really a more accurate reflection of the cumulative data. And a 3% variation is well
within the accuracy one might expect for this technique (roughly  10%).

Now, if the boundary layer had some previous development, such that  = i at x = 0, we
would use Eqs. 14.34 and 14.35 to determine the boundary layer characteristics. So, for
the momentum thickness, we would have from Eq. 14.34:

x
 2  i2
  0.47U 6  U 5 dx
  0

Or solving for :

2 2
 2 i2 0.47x    0.47    0.47
     i    i 
  U x  x   U x   x  Re x
 
  
dU 
And since  0 , n is still 0. Note that if this was an accelerating or decelerating
dx
dU 
flow,  0 , and n would also be non-zero.
dx

14.4.2 Linearly Accelerating Flow

x
Here we examine a linearly accelerating flow given by U   U 0 (1  x ) , where x   .
L
dU  U 0
Thus,  , and n is determined using Eq. 14.32 (where we note that dx  Ldx ):
dx L

x
U0
n  0.47U 0 6 (1  x' ) 6
L 0 U50 (1  x)5 Ldx
Uo

x
 0.47  (1  x)6 

1  x6  6  0 x
L

457
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 14

n
 0.078
1  x
 
1  x6  1  0.078 1  1 6  (14.37)
 (1  x ) 
6

Equation 14.37 indicates that n varies significantly with x', which requires that one
interpolate the corresponding values of  and H at any x' location from Table 14.2, or by
means of a curve fit of  and H vs. n. The following table shows the respective values of
n,  , and H at x' = 0 and x' = 1. Note that as the flow accelerates, the shear (reflected by
 ) increases, and the shape factor, H, decreases. A decrease in H means that the velocity
profile displays a stronger gradient near the wall (and thus a higher shear stress).

x' n  H
0 0 0.220 2.60
1 -0.077 0.329 2.35

Using Eq. 14.31, the momentum thickness for this accelerating flow is determined as:

 2 0.078L  1 
 1  
 U 0  (1  x ) 6 

1
L  1  2
  0.279 1  6 
(14.38)
U0  (1  x ) 

Equation 14.38 also shows a strong initial dependence on x'. However, note that for
x' > 1,  becomes essentially constant. Additionally, Eq. 14.37 indicates that for
x' > 1, n becomes a constant, which in turn means that  and H also become constants.
This means that the boundary layer basically stops growing and maintains a constant
shape!

If  and H become constant, this means that o is a result of pressure gradient effects
only. This can be shown from the original integral equation, Eq. 14.12, as follows:
0
d 1 dU  
 ( 2   * )  0 2 (14.39)
dx U  dx U 

Pressure Gradient Shear Stress

Equation 14.38 indicates that after roughly x' = 1, as the flow continues to accelerate, no
d
further momentum is lost from the flow (since  ceases growing) and  0 . Thus, as
dx
shown by Eq. 14.39, the changes in the pressure exactly balance the shear stress. Note
d
that if we rewrite Eq. 14.39 (for  0 ), noting that from Bernoulli,
dx

458
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 14

dP dU  *
 U   0 , and H  , we get:
dx dx 

dU  dP dP
0  U  (2  * )    (2  H)  (2  H)  (14.40)
dx dx dx

Thus, since  and H will be constants, Eq. 14.40 indicates that the shear will be directly
proportional to the pressure gradient (note that since this is an accelerating flow, the
dP
pressure is decreasing, and   0 , such that 0  0 ).
dx

14.4.3 Decelerating Flow

U0 x
Now let's examine a decelerating flow given by U   , where again x   .
(1  x ) L
Here, we first determine  using Eq. 14.31:

x
2
 0.47U  6  U 5 dx
 U
0

x
2
 0.47U 0 6 (1  x' )6  U50 (1  x' ) 5 Ldx ' x
 0
L
x'
2  (1  x' ) 4 
 0.47U 01 (1  x' )6 L  
  4 0

2

 0.1175
L
U0

(1  x' )6  (1  x' )2 
1/ 2
 
  0.1175
L
U0

(1  x' ) 2 (1  x' ) 4  1   (14.41)
 

dU  U0
To get n, we note that  , and using Eq. 14.22c we have:
dx L1  x'
2

n
 2 dU 
 0.1175
L

(1  x' ) 2 (1  x' ) 4  1   U0 
2
 dx U0  L1  x' 

n
2 dU 
 dx

 0.1175 (1  x' )4  1  (14.42)

459
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 14

Equations 14.41 and 14.42 suggest that  and n will grow quite rapidly with x'.
However, for a boundary layer type flow, the limiting condition will be when the shear
stress becomes zero, since at that point the flow near the wall will cease to move
downstream, and the boundary layer assumptions fail. We note that for this decelerating
flow the solution will fail when:

u  u
0   0   0
y 0 U1 y 0

When   0  n = 0.090 (from Table 14.2). Thus, the limiting condition for Eq. 14.42 is:

 
n  0.1175 1  x '  1  0.090
4

Solving for x' gives:

x'   00..1175  14  1  x'  0.1528

1
090

The following table shows the initial and limiting conditions for this decelerating flow.

x' n  H
0 0 0.220 2.60
0.1528 0.090 0 3.70

Using this limiting value of x', we calculate the corresponding limiting value of  from Eq. 14.41:

  0.1175
L
U0

1.15286  1.15282 
L 0.346L
  0.346 
U0 U0L


As an example, consider an air flow for which L=1 meter, U0=1 m/s, =1.6x10-5 m2/s

Separation starts at 0.1528 meters

The corresponding momentum thickness:  = 1.38 mm

Thus, when n = 0.090, H = 3.70  * = 5.11mm (0.201 inches)

460
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 14

dP
14.5 The Relationship of shear, drag, and momentum thickness when 0
dx

A boundary layer flow for which the external velocity, U  , remains a constant can be
used to illustrate a unique relationship between shear and drag on a plate and the
momentum thickness,  . Consider the momentum integral equation, Eq. 14.11.

d 1 dU  
 (*  2)  w2
dx U  dx U 
dU 
If external flow is constant, that means that  0 , and the momentum integral
dx
equation reduces to:

d 
 w2
dx U 
Which can be rewritten as:

d
w  U 2 14.43
dx

Equation 14.43 is remarkable, since it indicates that the wall shear stress is a direct
function of the change in the momentum thickness. Additionally, if we assume a plate of
length L, and arbitrary width W, we can calculate the drag on that plate as:

x L x L 
d L

 w Wdx  U2 W  dx  d
2
Drag  dx  U  W
x 0 x 0 0

Drag  U2 W  L  0  14.44

Here, 0 and L are the values of the momentum thickness at x = 0 and x = L

respectively. Equation 14.44 is again remarkable, since it shows that the drag on a given
plate of a length L is directly proportional to the difference in the momentum thickness at
the end of the plate less the momentum thickness at the beginning of the plate. And this
does not just apply to a laminar flow, but also to (1) a turbulent boundary layer as well, as
we will discuss in Chapter 17, and/or (2) any type of boundary layer conditions, such as
transition from laminar to turbulent flow, or the effects of surface roughness and
irregularities.
dP
So, as long as the external flow is a constant (i.e.  0 ), a measure of the drag over a
dx
specified length of a surface (usually a flat surface) can be established by Eq. 14.44 by
measurement of the velocity profile and calculation of  at the beginning and end
locations. And if one of those locations is the leading edge of a plate where the boundary
layer initiates (where  is zero), measurement of  at any downstream location will
establish the total drag to that location. Pretty nifty.

461
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 14

14.6 The Effect of a Pressure Gradient on Boundary Layers

As the Thwaites solutions shown in Section14.4 illustrate, the external pressure gradient
can have a particularly strong effect on the boundary layer. Since shear effects are
always working against the fluid inertia (particularly near the bounding surface), the
effect of the pressure gradient can either work with inertia and against shear effects (for
an accelerating flow, with decreasing pressure in the flow direction), or it can work
against inertia, and with the shear effects (for a decelerating flow, with an increasing
pressure in the flow direction). This latter situation, called an adverse pressure gradient,
leads quickly to the termination of forward flow near the surface, resulting in reverse
flow near the wall. As we have discussed before, this flow reversal causes a “separation”
of the boundary layer from the surface, and the development of large energy losses in the
flow. The Thwaites example done in section 14.4.3 is a particularly good illustration of
the problem created by an adverse pressure gradient. For that example, the shear stress
went to zero (the precursor to a flow reversal at the surface) by 15.3 cm (roughly 6
inches) from the initiation of the pressure gradient—a relatively short distance along the
boundary!

Insight into this interplay between shear, inertia, and pressure can be obtained by a
further manipulation of the integral equation to separate out the effects of inertia and
pressure changes. Recall that Eq. 14.11 was written as:

d 1 dU  
 (*  2)  w2
dx U  dx U 

Multiplying Eq. 14.11 by U 2 and rearranging yields:

 d dU   * dU 
 U 2  2U     U   w
 dx dx  dx

Combining the left two terms, we have:


d U 2  
  * U 
dU 
 w
dx dx

dP dU 
And noting that  U  , and substituting, we get:
dx dx


d U 2   dP
 *    w (14.45)
dx dx

Momentum Pressure Shear

changes gradient effects
effects

462
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 14

Equation 14.45 now clearly illustrates the interplay between shear, pressure gradient, and
momentum changes. The interesting aspect of Eq. 14.45 is the impact of the pressure
gradient.
dU  dP
Recall that for an accelerating flow,  0 and thus  0 . Thus, for an accelerating
dx dx
flow, both the pressure and the change in momentum contribute to an increased wall
shear stress. Interestingly, even if the momentum thickness ceases growing (as was
illustrated in the example in Section 14.4.2), the outer region velocity (U) will continue
to increase, such there will still be a positive momentum change in the boundary layer.
Thus, for an accelerating flow the absolute shear will continue to increase, with the
velocity profile "stretching" in the streamwise direction, since our boundary conditions
require that u = U at the edge of the boundary layer.

dU  dP
Conversely, for a decelerating flow  0 and thus  0 . Consequently, pressure
dx dx
in a decelerating flow works in opposition to the momentum changes, and the shear will
decrease. And, since U is decreasing, the momentum changes will also decrease (even
as the momentum thickness grows rapidly), which further reduces the shear. As Eq.
14.45 illustrates, despite continued growth in the momentum thickness, the momentum
will grow more slowly (since U 2 is decreasing), and the pressure gradient will act more
strongly in opposition to the flow as the displacement thickness grows. Ultimately, the
momentum cannot withstand the opposing pressure forces within the boundary layer, and
the flow nearest the boundary will cease to move downstream, and the boundary layer
flow will separate from the surface (more on this in section 14.7).

To qualitatively examine the effect of a pressure gradient, dP/dx, on the boundary layer
velocity profile, consider the two-dimensional boundary layer equation (Eq. 13.13a)
evaluated at the bounding surface (i.e. at y = 0). By the no slip condition we have u = 0
at y = 0; and for a solid surface, we must have v = 0 at y = 0, thus:
0 0
u u 1 dP  2u
u v   2 at y = 0.
x y  dx y

Thus, at the bounding surface we have:

 2u 1 dP
 (14.46)
y 2 y0
 dx

Equation 14.46 indicates that the pressure gradient imposes a second-order boundary
condition on the velocity profile at y = 0. Recall that in section 11.5.1 and 11.5.3 we
showed that a streamwise pressure gradient is the source of vorticity at a bounding
u
surface. Since at the wall boundary  z   , then Eq. 14.46 can be rewritten as:
y

463
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 14

 2u 1 dP 
  z (14.47)
y 2 y 0
 dx y

Equation 14.47 indicates that the second order boundary condition is a vorticity source
term, controlled by the pressure gradient, and which will be seen to have a significant
modifying effect on the boundary layer velocity profile.

To illustrate the impact of this pressure gradient boundary condition on the behavior of
the boundary layer velocity profile, we qualitatively examine three different generic
pressure gradient cases: a favorable pressure gradient (dP/dx < 0), no pressure gradient
(dP/dx = 0), and an adverse pressure gradient (dP/dx > 0).
dP
14.6.1 Favorable pressure gradient  accelerating flow (  0)
dx

y y y
Slope=0 U Slope=0
   

Decreasing
Slope<0
positive
slope

 2u u u
y 2 y

 2 u 1 dP u
@ y=0  0   w  0 u0
y 2  dx y

 2u u
@ y= 0 0 uU
y 2 y

Figure 14.4 Effect of a favorable pressure gradient (dP/dx<0) on the shape of the
boundary layer velocity profile.

A favorable pressure gradient, as discussed above, acts in concert with the boundary layer
momentum, and thus aids the flow very near the boundary to continue moving
downstream. This is like having the wind at your back when running. As Fig. 14.4
illustrates, the second-order boundary condition is negative at the bounding surface (y =
0), and will of course be zero at the edge of the boundary layer (y = ). Since the second-
order boundary conditions reflect the local slope of the shear stress, the shear has a
positive value, but with a negative slope at y = 0, and asymptotes to zero shear at y = ,
with zero slope. The resulting velocity profile will maintain a full, non-inflectional shape
while the pressure gradient remains favorable. Such flows are quite stable, with the
boundary layer remaining attached to the surface (no reversed surface flow). Because
this is a very stable type of flow, a favorable pressure gradient will help a boundary layer

464
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 14

maintain laminar behavior, and resist transition to turbulence (we will discuss this further
in Chapter 17)

dP
14.6.2 Zero pressure gradient  constant velocity flow (  0)
dx
y y y
Curve Slope=0 U Slope=0
   
Must be cannot be 0,
therefore
negative
shear curve
must be Slope=0
inflectional

 2u u u
y 2 y

 2 u 1 dP u
@ y=0  0   w  0 u0
y 2  dx y

 2u u
@ y= 0 0 uU
y 2 y

Figure 14.5 Effect of a zero pressure gradient (dP/dx = 0) on the shape of the
boundary layer velocity profile.

For a zero pressure gradient flow, or flat plate flow, the pressure gradient has no impact
on the boundary layer development. Such a flow, as was shown in the Blasius solution of
Section 13.3, is a balance between inertia effects and shear effects. As is shown in Fig.
14.5, the second-order boundary condition is zero at the bounding surface (y = 0), and
must also be zero at the edge of the boundary layer (y = ). Since the second-order
boundary conditions reflect the respective local slope of the shear stress, the shear will
have a positive value at y = 0, but with a zero slope. The shear again asymptotes to zero
at y =  with zero slope. Since the shear must experience a negative slope to connect
between y = 0 and y = , the shape of the shear profile will be as shown in Figure 14.5,
 2u
and must be negative between y = 0 and y = , as shown. The resulting velocity
y 2
profile will be less full than for a favorable pressure gradient, and will experience a point
of inflection at y = 0. Since a functional inflection signals the possibility of unstable
behavior, a flat plate laminar boundary layer is just barely stable. As such, small
perturbations, such as velocity or pressure fluctuations, or surface irregularities, can cause
a breakdown of the boundary layer to turbulence. In general, beyond a Reynolds number
of roughly 5x105, very good flow conditions (i.e. smooth surface and minimal
pressure/velocity fluctuations in the outer flow) are required to maintain a laminar

465
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 14

boundary layer over a flat plate with a zero pressure gradient. However, with ideal flow
conditions laminar flows can be maintained in excess of Rex = 106.

dP
14.6.3 Adverse pressure gradient  decelerating flow (  0)
dx
y y y
Slope=0 U Slope=0
   
 2u Slope<0
0
y 2 Inflection
Therefore, velocity Slope>0 point
profile is inflectional

 2u u u
y 2 y

 2 u 1 dP u
@ y=0  0   0  0 u0
y 2  dx y

 2u u
@ y= 0 0 uU
y 2 y

Figure 14.6 Effect of an adverse pressure gradient (dP/dx>0) on the shape of the
boundary layer velocity profile.

An adverse pressure gradient, as opposed to a favorable pressure gradient, will act in

opposition to the boundary layer momentum, and thus further restrain the flow very near
the boundary. The is like having the wind in your face while running, which makes
moving forward that much harder, for both you and the boundary layer fluid As Fig. 14.6
illustrates, the second-order boundary condition is now positive at the bounding surface
(y = 0), while it still asymptotes to zero at the edge of the boundary layer (y = ). The
shear stress still (up to a point) has a positive value, but with a positive slope at y =0, and
it again must asymptote to zero at y = , with zero slope. As shown in Figure 14.6, the
shear stress must reach a maximum above the surface, and then develop a negative slope,
 2u
which will allow it to decrease to zero at y = . To do that, must pass through zero,
y 2
which indicates that a point of inflection must occur above the boundary surface. As
discussed in section 14.6.2, a point of inflection reflects a point of potential instability,
which makes these adverse pressure gradient flows highly likely to a break down to
turbulence. Additionally, the dual penalty of the shear stress and the pressure gradient
acting in opposition to the fluid inertia can quickly reduce the velocity of the fluid very
near the surface, strongly distorting the velocity profile (as shown in Fig. 14.6), which
can rapidly lead to boundary layer separation, as described in the following Section 14.7.

466
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 14

14.7 Boundary layer separation

When a boundary undergoes a sustained adverse pressure gradient, the flow very near the
bounding surface can quickly be slowed to a halt, such that the wall shear stress becomes
zero. Note that zero wall shear stress was the condition limiting the flow development
for the decelerating flow example done using Thwaite's method in section 14.4.3. As
shown in Fig. 14.7, downstream of this point of zero wall shear stress, the flow adjacent
to the bounding surface will actually reverse direction, forming a stagnant or recirculating
region of fluid. Downstream of the point of zero wall shear stress, the impinging
boundary layer flow will divert away from the surface and around this stagnant or
reversed flow region.

Figure 14.7 Illustration of the development of a separation point of zero shear

stress, and subsequent flow separation, due to a sustained adverse
pressure gradient (dP/dx > 0).

This formation of a reversed flow region, and the subsequent diversion of the impinging
flow around this region of reversed flow, is termed a "flow separation." When flow
separation occurs, the boundary layer ceases to be a boundary layer, and becomes a much
more complicated flow, which can no longer be properly modeled using the boundary
layer equations. Such separated flows must be addressed using the full Navier-Stokes
equations and sophisticated numerical solution techniques.

Physically, the consequence of boundary layer separation is that the displacement

thickness, which we envisioned and defined in section 13.4.1, becomes quite large,
effectively displacing the outer free stream flow around the region of stagnant or reversed
flow. This diversion of the outer flow will result in a significant modification of the
pressure distribution within the flow. As we will discuss in Chapter 15, the creation of
these separated flow regions, and the associated modified pressure field, result in a
significant net pressure force acting in the flow direction, which is termed pressure drag.

As we discussed in section 14.4.3, an additional effect of a sustained adverse pressure

gradient is to move the point of inflection in the boundary layer velocity profile away
from the surface, which makes the boundary layer much more sensitive to a transition to
turbulence. When this happens, the inertia forces overwhelm the viscous forces in the
boundary layer, and the velocity profile cannot keep its shape, subsequently breaking
down into rapid, local mixing behavior known as turbulence, which in turn causes
increased surface shear stress. The good news is that the mixing caused by the turbulence
will feed more energy from the outer region flow into the flow very near the bounding

467
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 14

surface, which helps maintain the downstream movement of the boundary layer.
However, while a transition to turbulence may initially keep the flow attached to the
bounding surface, under a continued adverse pressure gradient, the now turbulent
boundary layer will eventually undergo flow separation as well.

In general, adverse pressure gradients, if strong or sustained, will always lead to flow
separation, and increased drag or flow resistance. In the following Chapter 15, we will
examine the impact of flow separation on both the drag on external bodies, and the
development of large temporal pressure fluctuations on the bodies. And in Chapter 17,
we will examine the basics of turbulence, and its effect on flow separation.

References

Thwaites, B. (1949). Approximate calculation of the laminar boundary layer.

Aeronautical Quarterly, vol. 1, pp. 245-254. Also see. Thwaites, B (1960).
Incompressible Aerodynamics. Oxford University Press, London. pp.62-64.
ISBN 0-486-65465-6.

Problems

1. Using the momentum integral equation, assume a velocity profile of the form
u y
 a  b , where   . Determine the constants a and b, and then the boundary
U 
layer parameters , *, , and c f for a flat plate flow. Compare you results to table
14.1 in section 14.2.2.

2. Using the momentum integral equation, assume a velocity profile of the form
u y
 a  b  c2  d3 , where   . Determine the constants a, b, c, and d, and
U 
then the boundary layer parameters , *, , and c f for a flat plate flow. Assume that
all derivatives are zero at   1 . Compare you results to table 14.1 in section 14.2.2.

3. Using the momentum integral equation, assume a velocity profile of the form
u y
 a  b  c2  d3 , where   . Determine the constants a, b, c, and d, and
U 
then the boundary layer parameters , *, , and c f for a flat plate flow. Assume that
the shear stress is zero at   1 , but for a fourth boundary condition, assume that
d 2u
 0 at   0 . Note that this boundary condition is obtained from the original
dy 2
boundary layer equation when   0 , assuming that U=constant (see section 14.5,
Eq. 14.42). Compare you results to table 14.1 in section 14.2.2.

468
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 14

4. Using the momentum integral equation, assume a velocity profile of the form
u y
 a  b  c2  d3  e4 , where   . Determine the constants a, b, c, d and
U 
e, and then the boundary layer parameters , *, , and c f for a flat plate flow.
Assume that all derivatives are zero at   1 . Compare you results to table 14.1 in
section 14.2.2.

5. Using the momentum integral equation, assume a velocity profile of the form
u y
 a  b  c2  d3  e4 , where   . Determine the constants a, b, c, d and
U 
e, and then the boundary layer parameters , *, , and c f for a flat plate flow.
d 2u
Assume that the shear stress and are zero at   1 , but for a fifth boundary
dy 2
d 2u
condition, assume that  0 at   0 . Note that this boundary condition is
dy 2
obtained from the original boundary layer equation when   0 , assuming that U =
constant (see section 14.5, Eq. 14.42). Compare you results to table 14.1 in section
14.2.2.

6. Using the momentum integral equation, assume a velocity profile of the form
u y
 a sin( b ) , where   . Determine constants a and b, and then the boundary
U 
layer parameters , *, , and c f for a flat plate flow. Assume that all derivatives are
zero at   1 . Compare you results to table 14.1 in section 14.2.2. What are the
problems with this assumed velocity profile? [hint: consider the boundary condition
d 2u
for at   1 ]
dy 2

7. Water enters a two-dimensional nozzle at free-stream velocity U0. The free-stream

velocity accelerates exponentially according to the relationship, U(x)= U0exp(kx),
where x is the distance from the start of the nozzle, and k is a constant. Assume that
the kinematic viscosity, , is constant and the boundary layer thickness is   0  0
at x = 0. Assuming a boundary layer profile of the form, u/U(x)=A+By+Cy2+Dy3,
use the momentum integral equation and, for a bounding surface of the nozzle,
determine:

a) Expressions for *, , and 0 as functions of .

b) A differential equation for  as a function of x, with U0, , and k as constants.

c) A solution for  in terms of U0, 0, , k, and x

469
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 14

d) A plot of (x) (in inches) for 0x5 ft, for both 0=0 inches and 0=0.5 inches,
letting k=0.2 ft-1,  = 10-5 ft2/sec., and U0=1 ft/sec.

e) For the same values of k, , and U0 as in part d, determine the x location (in feet)
and the free-stream velocity U (in ft/sec.), where  becomes a maximum.
Determine the location of the maximum for both 0 = 0 inches and 0 = 0.5 inches.
Physically explain what is happening to give such significantly different maxima.

8. Assuming the same conditions as problem 7, where U(x)= U0exp(kx), determine an

expression for (x) using Thwaites method, and plot (x) vs. x for 0  x  5 ft, for 0
= 0 inches, letting k = 0.2 ft-1,  = 10-5 ft2/sec., and U0 = 1 ft/sec. On the same graph,
plot for comparison the (x) expression you derived in problem 7 using a direct
solution of the momentum integral equation.

Extra Credit: Again assuming the same conditions as problem 7, use Thwaites
method to
determine (x), and plot (x) vs. x for 0  x  5 ft, for 0 = 0.5 inches, letting k=0.2 ft-
1
, =10-5
2
ft /sec., and U0 = 1 ft/sec. On the same graph, plot for comparison the (x) expression
you
derived in problem 7 using a direct solution of the momentum integral equation.

Note: You cannot do this directly using the original Thwaites equation, but need to
employ Eqs. 14.34 and 14.35 for a non-zero momentum thickness.

9. A fluid entering a diffuser is decelerated according to the velocity U(x) = U0/(1+ax),

where a is a constant indicating how rapidly the diffuser decelerates the flow.
Assume that U0 and  (kinematic viscosity) are known constants. Using Thwaites
method, do the following:

a) Determine an expression for  as a function of U0, , a, and x.

b) Determine an expression for the parameter n as a function of U0, , a, and x.

c) Assuming that a is inverse units to x, and determine values of  at x = 0.5 for both
a = 0.1 and a = 0.2 in terms of U0 and .

d) Determine the values of x where U = 0.9U0 for both a = 0.1 and a = 0.2; compare
the respective values of  that occur at these points, and comment on the reason
for any variations.

e) For a = 0.1, determine the maximum value of  and the respective x value at
which this maxima occurs. Briefly explain the physical processes which result in
the development of this maxima.

470
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 14

10. A fluid entering a diffuser is decelerated according to the velocity U(x)= U0(1- ax),
where a is a constant indicating how rapidly the diffuser decelerates the flow.
Assume that U0 and  (kinematic viscosity) are known constants. Using Thwaites
method, do the following:

a) Determine an expression for  as a function of U0, , a, and x.

b) Determine an expression for the parameter n as a function of a, and x.

c) Assuming that a is in inverse units to x, determine values of  at x = 1 for both a =

0.05 ft-1 and a = 0.1 ft -1 in terms of U0 and .

d) Determine the values of x where U = 0.90 U0 for both a = 0.05 and a = 0.1;
compare the respective values of  that occur at these points, and comment on the
reason for any variations.

e) For a = 0.1, determine the maximum value of  and the respective x value at
which this maxima occurs. Briefly explain the physical processes that result in
the development of these maxima.

11. A fluid entering a nozzle is accelerated according to the velocity U(x) = U0(1+ax)
where a is a constant indicating how rapidly the nozzle accelerates the flow. Assume
that U0 and  (kinematic viscosity) are known constants. Using Thwaites method, do
the following:

a) Determine an expression for  as a function of U0, , a, and x.

b) Determine an expression for the parameter n as a function of a, and x.

c) Assuming that a is in inverse units to x, determine values of  at x = 1 for both a =

0.05 ft-1 and a = 0.1 ft -1 in terms of U0 and .

d) Determine the values of x where U = 1.10 U0 for both a = 0.05 and a = 0.1;
compare the respective values of  that occur at these points, and comment on the
reason for any variations.

e) For a = 0.1, determine the maximum value of  and the respective x value at
which this maxima occurs. Briefly explain the physical processes that result in
the development of these maxima.

12. Consider the flow of air over a cylinder. Assume that the impinging velocity is U=1
cm/s, and the cylinder radius is R=10 cm. The kinematic viscosity for air is  = 16
mm2/s. As the flow passes around the cylinder, a laminar boundary layer develops on
the cylinder surface. Assume that the velocity at the edge of the boundary layer can
be approximated by the potential flow solution of U = 2Usin(), where  is the
angle measured from the cylinder stagnation point around the cylinder.

471
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 14

a. Using Thwaites method, neglect curvature effects and determine the development
of the momentum thickness () as a function of x, where x = R, when  is in
radians. You can integrate the Thwaites equation analytically (you need to
review your integral tables)—do it in terms of , and then express in terms of x.

b. Determine the angle max (in degrees) where the solution is no longer valid, using
a parallel calculation of the parameter n. Why does the solution fail at this point?

c. Plot  and n on the same graph (using different axis scales) vs. x, for 0 cm < x <
20 cm. Use appropriate scales for  and n, such that the full extent of the changes
in  and n can be observed and compared to each other. I suggest you use smaller
increments of  near  = 0 to effectively resolve the curves.

13. Consider the flow of water with a Falkner-Skan type of outer region velocity
m
 x 
U   U 0   , where U0, xo, and m are a constants, with Uo = 10 cm/s and xo =
 x0 
100cm. The kinematic viscosity for water is is  = 0.01 cm2/s. The initial momentum
thickness at xo is o = 0.2 cm.

a. Using Thwaites method, determine expressions for the momentum thickness ()
and the Thwaites pressure gradient parameter (n) as a function of x, in terms of o,
, U0, xo, x, and m. Note that you must start with your integration for the
momentum thickness using Eq. 14.29 since o0.


b. Create a graph of (where o is the initial momentum thickness, 0.2 cm) vs.
0
Show plots for m = 0 (Blasius), = 0.25 (accelerating), and = - 0.05 (decelerating).
x x
For your graph, use increments of of 0.1 from 1   4.
x0 x0
x
c. Create a second corresponding graph of n vs. showing plots for m = 0
x0
(Blasius), =0.25 (accelerating), and =-0.05 (decelerating). For your graph, again
x x
use increments of of 0.1 from 1   4 . Is the decelerating flow likely to
x0 x0
separate? What is your basis for judging whether or not it separates?


d. Create a third graph of (where B is the Blasius value for m = 0) vs. m, for
B
x
 4 . Plot your graph for  0.1  m  1 (use increments of 0.05). Compare
x0
your graph to the results for the Falkner-Skan solution of Figure 13.11 in Chapter
13.

472
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 14

14. Consider the flow of water with a Falkner-Skan type of outer region velocity
m
 x 
U   U 0   , where U0, xo, and m are a constants, with Uo = 10 cm/s and xo =
 x0 
100 cm. The kinematic viscosity for water is is  = 0.01 cm2/s. The initial
momentum thickness at xo is o = 0 cm.

a. Using Thwaites method, determine expressions for the momentum thickness () and
the Thwaites pressure gradient parameter (n) as a function of x, in terms of o, , U0,
xo, x, and m.

x
b. Create a graph of  (in cm) vs. . Show plots for m = 0 (Blasius), = 0.25
x0
x
(accelerating), and = - 0.05 (decelerating). For your graph, use increments of of
x0
x
0.1 from 1   4.
x0
x
c. Create a second corresponding graph of n vs. showing plots for m = 0 (Blasius),
x0
=0.25 (accelerating), and = - 0.05 (decelerating). For your graph, again use
x x
increments of of 0.1 from 1   4 . Is the decelerating flow likely to separate?
x0 x0
What is your basis for judging whether or not it separates?


d. Create a third graph of (where B is the Blasius value for m = 0) vs. m, for
B
x
 4 . Plot your graph for  0.1  m  1 (use increments of 0.05). Compare your
x0
graph to the results for the Falkner-Skan solution of Figure 13.11 in Chapter 13.

x
e. Using your expression for n vs. , use a root-finding program to determine the
x0
value of m that the Thwaites method would indicate that the flow would reach
separation at x = 400 cm. How does this compare with the actual Falkner-Skan
solution?

473
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 15

Chapter 15

Flow Separation and Drag Over Real Bodies

Contents
15.1 Causes of Separation ................................................................................................. 474

15.2 Effects of Separation .................................................................................................. 477

15.3 Drag ......................................................................................................................... 478

15.4 Unsteady Separation/Drag Effects ............................................................................. 489

15.4.1 Vortex Shedding by a Circular Cylinder ............................................................ 490
15.4.2 Vortex Shedding by Other Shapes ................................................................... 492

15.5 Control of Separation and Drag ................................................................................. 494

15.5.1 Passive Control ............................................................................................... 494
15.5.1.1 Surface Effects .................................................................................. 494
15.5.1.2 Splitter Plates .................................................................................... 495
15.5.2 Active Control ................................................................................................ 496

15.1 Causes of Separation

As discussed previously in Chapter 14, when a boundary layer flow comes under the
influence of an adverse pressure gradient (i.e. with pressure increasing in the flow
direction), both the local wall shear stress and the increasing pressure will act to retard
the flow immediately adjacent to the bounding surface. In sometimes a very short
distance, the fluid adjacent to the surface will be become stagnant (i.e. be brought to rest),
and often begin to move in opposition to the main flow. This condition results in what is
known as flow separation, where the initially stagnated fluid will cause the following
flow to divert outward, away from the surface, and around the stagnant or rearward
flowing wall fluid. This process will cause a substantial change of the flow field,
resulting in significant losses of flow energy and pressure due to the intense mixing
between the stagnant fluid and the downstream flow bounding this stagnant fluid.

The conditions necessary for such a flow separation behavior to develop are:

1) A local, adverse pressure gradient, and/or

2) Strong curvature or sharp corners

Figure 15.1 shows several examples of separated flows. Figure 15.1a is typical of flow
separation occurring over bluff bodies. Here we show a cylinder, but the same behavior
is characteristic of flows where the width and length of the body are comparable, such as
rectangular or elliptical bodies. As shown in Chapter 9, flows passing around these types
of bodies will experience significant surface pressure variations. If the boundary layer

474
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 15

that develops on these bodies encounters an adverse pressure gradient, it will be subject
to possible (and often rapid) boundary layer separation. As discussed earlier in Section
14.5, once a boundary layer separates, a region of reversed flow, or a recirculating wake,
is created downstream of the body. This wake, as shown by the pressure sketch of Figure
15.1a, will be of relatively constant pressure that is significantly lower than the leading
edge pressures (this relatively constant trailing-edge pressure is generally characteristic of
separated flow regions or wakes). The cumulative differential in pressure around the
body will result in a cumulative force in the flow direction (determined by integrating the
surface pressure over the body), which is termed a "drag" force. As we will discuss in
section 15.3, the total drag is the sum of both pressure imbalances and surface shear.

85
a) Cylinder
Laminar
separation

Flow region

Surface
Inviscid 120-140
Pressure
Turbulent
Turbulent
Laminar separation
0   region

b) Airfoil c) Flat Plate

No separation No separation
Flow

Separation at sharp leading edge

Separation point
Flow
Separated wake
Separated wake

d) Smoke flow visualization of separated flows

1. Separation for an airfoil: 2. Laminar separation 3. Turbulent separation

high angle of attack behind a sphere behind a sphere

Figure 15.1 Some typical examples of separated flows: a) a circular cylinder; b) an

airfoil; c) a flat plate; d) smoke visualizations of separation behind an
airfoil and a sphere.

475
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 15

However, note that when a body is of significant thickness (i.e. not a flat plate or thin
airfoil), the drag due to differences in surface pressure around the body greatly exceeds
the drag due to the surface shear stresses.

Additionally, the type of boundary layer on the body has a significant effect on the
magnitude of the pressure drag. As we discuss in Chapter 17, a transition to turbulence
results in an increase in boundary shear stresses; however, the mixing associated with
turbulence will keep the boundary layer energized, so it will generally remain attached to
the surface much farther around a body. As shown on the right of figure 15.1a, a laminar
boundary layer will separate from a bluff body surface much sooner than a turbulent
boundary layer. The result is a trailing surface pressure that is lower for a laminar
separation than a turbulent separation, and consequently the pressure drag for laminar
separation will exceed that for a turbulent separation, for the same impinging flow.

Figure 15.1b illustrates flow around a symmetric airfoil. When a symmetric airfoil (i.e.
having the same shape above and below the centerline) is aligned with the flow, as in the
upper example of Figure 15.1b, the flow will (ideally) pass around the airfoil without
undergoing flow separation. Even though there will be an adverse pressure applied to the
trailing portion of the airfoil (i.e. that following the thickest part of the airfoil), airfoils for
aircraft applications are generally designed such that the region of adverse pressure will
not result in a significant separation. Generally, to avoid a trailing edge separation, the
thickness (i.e. thickest portion of the airfoil) is kept thin relative to the length of the
airfoil. For a low thickness to length ratio, the flow will generally pass around the airfoil
without separating. Note that there will still be an unbalanced pressure force on the
airfoil, since the pressure at the leading edge will be the stagnation pressure of the flow,
and the pressure at the trailing edge will generally be that of the local static pressure. The
pressure distribution will result in a nominal amount of pressure induced drag, but much
less than for a comparable body without an extended trailing edge (e.g. a circular
cylinder). Additionally, surface shear stresses will more significantly contribute to the
overall drag on an airfoil.

However, when an airfoil is placed at an angle of attack (the angle of the airfoil centerline
to the flow direction), as shown in the lower portion of figure 15.1b, the pressure
distribution over the airfoil will change significantly, creating an unbalanced pressure
distribution between the top to the bottom of the airfoil. This unbalanced pressure
distribution results in a lift force on the airfoil (necessary for flight), but also increases the
pressure imbalance from front-to-back of the airfoil, with a subsequent increase in the
pressure-induced drag on the airfoil. If the angle of attack is increased too severely, the
adverse pressure gradient over the trailing portion of the airfoil will cause the flow to
separate from the upper surface, with an accompanying reduction in surface pressure over
the upper surface; the result is a marked increase in pressure-induced drag. If separation
occurs, it will develop quite rapidly, creating a large region of separated fluid, or wake. It
also results in a significant loss in lift, which can have disastrous consequences on an
aircraft, in that the aircraft will not be able to sustain flight. Such situations are known as
"stall" and can develop quite rapidly; the occurrence of stall is most often the cause of
aircraft accidents and associated fatalities.

476
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 15

Figure 15.1c represents flow over a flat plate. As we know, when a flat plate is oriented
with its surface aligned with the flow direction, as in the upper portion of figure 15.1c, a
boundary layer develops over the flat surfaces of the plate, resulting in frictional drag.
Since the plate is aligned with the flow, the outer flow remains uniform and essentially
parallel, with no variation in the bounding pressure. And since no pressure gradient
develops, there is no flow separation. Consequently, the plate only experiences drag due
to viscous shear.

However, when a flat plate is placed at an angle of attack, as shown in the lower portion
of figure 15.1c, the local pressure gradient at the sharp leading edge is so high (even at
low angles of attack) that a flow separation occurs immediately, developing a large
region of separated flow/wake on the upper surface of the plate, with a consequent low
pressure. The result is very large pressure drag, which increases essentially
proportionally to the sine of the angle of attack.

Figure 15.1d shows flow visualizations of separated flows using smoke streamers in a
wind tunnel. The first visualization shows the separation developing on the upper surface
of an airfoil at a high angle of attack. This degree of separation is severe, and would
result in significant loss of lift and increased drag; typically such a flow would not sustain
the flight of an aircraft, resulting in the "stall" referred to above. The second and third
images show the difference between laminar and turbulent separation behind a sphere.
Note that similar to the sketches shown on the right of figure 15.1a, the wake is smaller
when the flow over the body is turbulent, and will thus have less pressure drag. Two
interesting YouTube videos show a variety of flow separation behavior for real flows:
here using smoke visualization in air for separation over an airfoil, and here using particle
visualization in water for separation over a cylinder. Both of these are videos of
relatively old research movies, and a bit grainy, but they show some excellent sequences
of separated flow behavior for real flows.

15.2 Effects of Separation

Flow separation can have a number of deleterious effects, including:

1) Form or Pressure drag

2) Unsteady flow behavior

In Section 15.1 we discussed pressure drag (often called form drag, since it is a strong
function of the form of the body). In Section 15.3 we will develop methods for predicting
and calculating the forces due to drag over various types of bodies, and discuss the trade-
off between pressure drag and drag due to shear stresses.

Unsteady flow behavior occurs when the separated flow developing over a body changes
temporally, creating a time-varying pressure field, with associated time-varying pressure
drag. As we will discuss in Section 15.4, such time-varying behavior can create
unwanted, and often destructive, periodic loading of a body.

477
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 15

Finally, in Section 15.5, we will discuss several methods for controlling, minimizing, and
stabilizing separation and the resultant pressure drag, and the tradeoffs between using
passive or active drag control techniques.

15.3 Drag

Drag is defined as the total force exerted on a body in the direction of the fluid motion by
the effects of the fluid, and is a combination of (a) unbalanced pressure forces and (b)
friction forces (i.e. fluid shear stress), both acting on the body surface. Note, however,
that pressure forces act normal to the surfaces, whereas friction forces act parallel to the
surfaces. The total resultant force acting in the direction of the fluid flow comprises the
total drag on the body. The resultant force acting normal to the direction of flow is
termed the lift on the body.

V D Frontal Area of a Cylinder, A=(D)(L)

Figure 15.2 Characterization of the frontal area of a bluff body.

To quantify the drag force, we establish an equation that relates the actual drag force to the
dynamic pressure (i.e. the pressure equivalent to the fluid momentum flux, 12 V 2 ), and
the frontal area of the body (i.e. the cross-sectional area of the body, as viewed from the
direction of the impinging flow), as illustrated in figure 15.2.

So, we write the drag equation as:

Total Drag Force Fluid Density Fluid Velocity

D  CD 1
 V2 A
2 (15.1)

Drag Coefficient Projected Frontal Area of Body

In Eq. 15.1, C D is the drag coefficient, and is proportional to the non-dimensional total
drag on a body, the combination of the shear stress drag and the form or pressure drag.
We equate the total drag as proportional to the dynamic pressure ( 12 V 2 ) and the frontal
area of the body. Note that the dynamic pressure is the pressure difference that would
result from bringing the impinging velocity to zero, or stagnation. A drag coefficient of
one or greater is considered high, with drag coefficients of 0.2 or less considered low
drag values.

Analytical or computational determination of accurate drag coefficients for any but the
simplest geometric body shapes has generally not been possible. Thus, we generally rely

478
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 15

on experimental data. However, the experiments must be done judiciously by employing

dynamic similitude techniques. First we determine the non-dimensional parameters upon
which drag depends. Secondly, we determine the functional relationship of the drag
coefficient to these parameters. Thirdly, we conduct experiments that cover the range of
the appropriate parameters.

From dimensional analysis and similitude we can establish that:

C D  C D (geometry, Re, Fr, M), where

geometry  scaled similarity (i.e. same geometric shape)

VL inertia forces
Re  Reynolds number =  (L is a characteristic length)
 viscous forces

V 2 inertia forces
Fr  Froude number =  (for air-water interactions)
gL gravity forces

V fluid velocity
Mach  Mach number =  (for compressible flows)
C speed of sound

For our purposes (i.e incompressible, single fluid), we need only consider the geometry
and Reynolds number, giving:

C D  C D (geometry, Re)

However, even this reduction in influencing parameters still requires a number of

carefully controlled experiments to determine CD for a range of geometries and Reynolds
numbers. As an example, figure 15.3 shows the general behavior of the drag coefficient
for a circular cylinder.

The drag on a cylinder is a classic example of bluff body drag. Note that at very low
Reynolds numbers (ReD < 10), the flow over the cylinder is laminar, with limited flow
separation, such that CD is basically due to viscous shear, and roughly proportional to
1/ReD. As ReD increases beyond 10, pressure drag starts to dominate, with CD becoming
essentially constant at ≈ 1.2 for 103 < ReD < 105 - 106, while the boundary layer over the
cylinder remains laminar.

Note that the broad upper range of ReD for laminar behavior depends on the character of
the boundary layer. For a smooth cylinder, without upstream disturbances, the flow will
remain laminar to a much higher ReD (up to 106). However, if the cylinder surface is
rough, or the external flow contains disturbances, the boundary layer on the cylinder can
transition earlier to a turbulent boundary layer, which increases local mixing, that in turn
reduces the separation behind the cylinder, as shown in figure 15.1a. When the boundary
layer over the cylinder is turbulent, CD decreases sharply to roughly 0.3, or roughly 25%
of the drag with a laminar boundary layer.

479
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 15

1000

100

CD
10

1 ≈ 1.2

Roughened
surface ≈ 0.3
0.1
2 3 4 5 6 7 8
0.01 0.1 1 10 10 10 10 10 10 10 10
ReD
Figure 15.3 The generic behavior of the drag coefficient for a smooth circular
cylinder (in blue) [after Hoerner, 1965]

Thus, in figure 15.3, the CD vs. ReD curve shows a region of 105 < ReD < 106 over which
the boundary layer may become turbulent, with a consequent decrease in CD. Note that
surface roughness, while causing early transition to turbulence, and thus a reduced CD,
also has a parasitic effect on the surface drag due to increased shear effects, which
generally results in a CD higher than that for a smooth cylinder (≈ 0.3) after transition.

1000

100

CD
10

1
≈ 0.47

0.1
Roughened
≈ 0.10
surface
0.01
2 3 4 5 6 7 8
0.01 0.1 1 10 10 10 10 10 10 10 10
ReD
Figure 15.4 The generic behavior of the drag coefficient for a smooth sphere (in
blue) [after Hoerner, 1965].

480
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 15

For drag on axisymmetric and three-dimensional bodies, the drag coefficient is still based
on the frontal area as observed in the flow direction. For example, the frontal area of a
sphere will be D2/4. In general, the drag coefficients for axisymmetric bodies tends to
be lower than those for two-dimensional bodies, such as cylinders. Figure 15.4 shows the
behavior of CD for a sphere as a function of the Reynolds number based on sphere
diameter. Note that the behavior is quite similar to that for a cylinder, with a plateau in
CD of ≈ 0.47 over a range from 103 > ReD > 105, and transition to turbulence in the range
105 < ReD < 106, with a subsequent drop in CD to ≈ 0.10 (an almost 80% reduction!).
Again, roughening the sphere surface will cause an earlier transition, and a consequent
decrease in CD.

Note that both figure 15.3 and 15.4 illustrate that roughening the surface of a body will
result in an earlier transition to a turbulent boundary layer over the body, and a significant
reduction in CD. Many practical flow situations take advantage of this reduction in CD by
purposely causing the boundary layer to transition to turbulence. This is done by
roughening of the bounding surface, or placing irregularities (e.g. a cable or wire) on and
around the surface. These surface treatments generate turbulence, which reduces the
separated region, and thus the hydrodynamic drag. Another quite common technique is
the roughening of the surface of a ball used in a sport. Examples are tennis and golf balls.
The "fuzz" on the exterior of the tennis ball greatly reduces both the drag and
unsteadiness (more on this in Section 15.5.1), so the ball will fly faster and truer. Golfers
found in the 1800's that an old, scratched-up ball would fly farther and truer than a new
one. Consequently, manufacturers started putting "dimples" on golf balls, to again reduce
drag and unsteadiness. Even today, golf ball manufacturers are forever tweaking the
dimple pattern on their golf balls, to reduce the drag and thus increase the distance the
ball will travel, assuming the same initial velocity. This link discusses the history of golf
balls and drag, and this link shows an interesting simulation of the flow due to the
dimpling of a golf ball.

1.2
Friction (3%)

CD
Pressure
0.3
Pressure

Laminar Turbulent

Figure 15.5 The relative contribution of fluid friction and pressure forces to
cylinder drag.

Figure 15.5 shows the general difference between laminar and turbulent drag on a smooth
cylinder. Note that drag due to shear stresses (i.e. friction) is a minimal contribution for
both laminar and turbulent flow (although friction drag will be higher for turbulent
flow—see Section 17.7.5.3). The decrease in total drag is quite remarkable, and
illustrates why many practical applications want to promote turbulent flow over bluff
bodies.

481
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 15

C
(a) Flow t = thickness
t
C = chord

=shape parameter

100% Flat plate

t
 0.25
C
(b)
Cylinder (laminar flow)

3%
0
t 1
C

Figure 15.6 The generic effect of geometry on the contribution of surface shear drag
vs. pressure drag to the total drag on a smooth body: (a) definition of a
body shape parameter, (b) relative contribution of surface shear drag to
total drag versus the body shape parameter, t/C.

Figure 15.6 shows a generic characterization of the relative contribution of shear stress to
the total drag on a body, using the ratio of the body thickness, t, to its length, C (known
generally as the chord of the body). For a flat plate, of negligible t/C, essentially all the
drag is due to surface shear, whereas for a smooth cylinder, with t/C = 1, the surface
shear contributes roughly 3% of the total drag (note: this would increase for roughened
bodies). The friction drag and pressure drag are about equal for t/C = 0.25, so this is not
a linear change with t/C. Clearly, for t/C > 1, essentially all the drag will be due to
pressure drag.

Generally, increasing t/C is accompanied by increased variations in the local pressure,

resulting in local adverse pressure gradients. As discussed in Chapter 13 and 14, the
presence of adverse pressure gradients almost assures the possibility/probability of a
separation of the boundary layer, an increased imbalance of pressure forces, and thus
more drag on the body.

Figure 15.7 shows CD values for a number of common two-dimensional shapes of infinite
depth into the page (e.g. long rods of different cross-sectional shapes). These values
assume a laminar approach flow with that Re  104. The Re and cross-sectional area are
based on the vertical dimension of the body normal to the flow direction. The CD values
for the smooth, more rounded bluff bodies will be reduced significantly when transition
to turbulence occurs, so one needs to understand the nature of the impinging flow. The
sharp-edge bodies, such as flat plates or bodies with sharp edges, will be relatively
insensitive to turbulent transition, since separation will normally occur at those sharp
edges, regardless of the type of boundary layer or impinging flow.

482
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 15

Plate
Plate perpendicular
to a surface
CD = 1.98 CD = 1.40

Square rod Square rod:

Rotated 45°
CD = 2.05 CD = 1.55

Isosceles triangle: Isosceles triangle:

flat side corner
CD = 2.00 CD = 1.55

Half cylinder: Half cylinder:

flat side curved side
CD = 1.7 CD = 1.16

Half tube: Half tube:

inner side outer side
CD = 2.30 CD = 1.20

Hexagon: Hexagon:
flat side at corner
CD = 0.70 CD = 1.00

Figure 15.7 Drag coefficients for selected two-dimensional bodies of infinite extent
(into the page) at Re  104 (Reynolds number based on vertical
dimension of the body). [after Hoerner, 1965]

You might note that most of these bodies have drag coefficients that are larger than that
for a cylinder, and often much larger. This demonstrates the drag penalty for a body with
sharp corners. As we showed in Section 9.7.5.2, potential flow around a sharp corner
results in an infinite local velocity at the corner, which would also reflect an infinite
adverse pressure gradient. And while such infinite flow singularities cannot exist in a
real flow, very large pressure differences will develop at sharp edges, such that flow
separation takes place immediately. Flow separation at sharp corners creates large

483
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 15

pressure differences between the front and the back of the body, and thus large drag
forces and large CD.

Notice in figure 15.7 that bodies with more pronounced corners have higher CD. For
example, CD = 2.05 for a square rod with a flat leading edge, whereas rotating the rod 45°
to give a pointed leading edge, and thus moving the corners toward the center of the body
reduces the CD to 1.55, a 25% reduction; a similar reduction in CD is noted by changing
the orientation of the hexagonal-shaped rod.

It also should be noted that CD for a flat plate is reduced by 30% when it is located on and
normal to a bounding surface, as opposed to being far removed from bounding surfaces.
This is due to the presence of a boundary layer on the bounding surface, which reduces
the local pressure at the leading edge of the plate, and may act to stabilize the trailing
wake. This reduction in CD is common for any bluff body mounted on a surface parallel
to the flow direction. In all cases, the CD for the bluff body will be reduced due to the
presence of the bounding surface.

Figure 15.8 shows the effect of bluff body thickness and boundary layer characteristics
on the CD for two-dimensional elliptical rods. As pointed out above, the thicker a body is
relative to its length, the larger the pressure gradients affecting the boundary layer on the
body, the greater the flow separation, with correspondingly greater form drag. This is
clearly demonstrated in figure 15.8, which shows that as t/L (termed the thickness ratio)
decreases from 1 (a circular cylinder), the CD correspondingly decreases. Note that all the
bodies shown in figure 15.8 will have the same frontal area, so this figure shows that as
the bodies become more extended the subsequent modification of the pressure gradient
will strongly reduce the form drag on the body. As shown, for laminar flows, extending a
circular body by only one thickness (i.e. from L = t to L = 2t), reduces the drag by 50%!
And extending it by seven thicknesses (to L = 8t) reduces the drag by 80%. This shows
the benefits of a little "streamlining". Additionally, reducing t/L generally makes the
flow more stable, such that time-varying separation effects are reduced as well. We will
discuss this further in section 15.4.

t/L = 1 0.5 0.25 0.125

t
L
Laminar CD  1.20 0.60 0.35 0.25

Turbulent CD  0.30 0.20 0.15 0.10

Figure 15.8 The laminar (Ret<105) and turbulent (Ret>106) drag coefficients for a
series of two-dimensional oval-shaped bodies of infinite extent (into the
page). [after Hoerner, 1965]

484
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 15

Figure 15.8 also shows the sensitivity of CD to the nature of the boundary layer. We noted
in Figure 15.3 the significant decrease (75%) in CD that takes place when the flow over a
smooth cylinder becomes turbulent in the Reynolds number range 105 < ReD < 106. As
one would expect, this sensitivity to the type of boundary layer extends to t/L < 1, with
turbulent boundary layers giving markedly lower CD values (roughly 60% lower). One
caveat is that extending the length of a body adds additional surface area, such that shear
drag will begin to make more significant contributions to the total drag.

(a) Flat leading edge rod (2-D)

L/H 0.1 0.5 1.0 2.0 4.0 6.0

H
CD 1.97 2.02 2.05 1.82 0.93 0.90

(b) Rounded leading edge rod (2-D)

L/H 0.1 0.5 1.0 2.0 4.0 6.0

H
CD - 1.16 0.90 0.70 0.58 0.54

(c) Flat-faced cylinder (3-D)

L/D 0.1 0.5 1.0 2.0 4.0 6.0

D
CD 1.17 1.15 0.90 0.85 0.84 0.82

Figure 15.9 The drag coefficients at Re  104 (based on H or D) as a function of

body length (L): (a) a two-dimensional rod of infinite extent (into the
page) with a flat leading edge; (b) a two-dimensional rod of infinite
extent with a semi-circular leading edge; (c) a flat-faced cylindrical rod.
[after Hoerner, 1965]

Figure 15.9 further illustrates the impact of added body length on CD, for some generic
types of smooth bodies with laminar boundary layers. Figure 15.9a is a rectangular rod
with a flat leading edge. For this body, a very short body length (e.g. L/H = 0.1) is
essentially a flat plate with the flow normal to the plate, as shown in figure 15.7. As
body length is increased, the CD increases slightly to 2.05 at L/H = 1 (the square rod of
figure 15.7). Further increases in length cause CD to decrease markedly to roughly 0.9 by
L/H = 4, where the CD appears to stabilize. Apparently, increasing length beyond L/H =
1 must have a significant impact on the separation characteristics in the wake of the body

485
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 15

for 1 < L/H < 4, with a corresponding reduction of CD. However, increasing the length
beyond L/H = 4 appears to have minimal impact on the wake behavior, or CD.
The body shown in figure 15.9b is a rectangular rod with a semi-circular leading edge.
Note that rounding the leading edge reduces the drag by roughly 50% from the
rectangular rod with a flat leading edge (figure15.9a). Obviously, elimination of sharp
corners, which reduces the separated wake of the rod, has a marked effect on CD. Again,
there is a corresponding decrease in CD with increased length, with increases beyond
L/H=4 having a limited effect on CD.
The third body shown in figure 15.9c is a flat-faced circular rod, with its axis in line with
the flow direction. This is considered a three-dimensional body, since it is radially
symmetric in the flow direction. The behavior of CD for this body with increases in
length is similar to that for the flat leading edge rectangular rod (figure 15.9a), but with
less dramatic changes in CD. For example, for the rectangular rod, CD decreases 53%
from L/H=0.1 to L/H=4; for the cylindrical rod, CD decreases only 28% from L/D=0.1 to
L/D=4. This suggests that the wake of the cylinder is less affected by the length
increases, most likely because of the closer exposure of the wake to the main flow due to
the axisymmetry. This closer exposure of the wake to the main flow is also why the CD
for the cylinder does not change significantly after L/D = 2
Example: Drag Force on a Suspended Telephone Cable/Smoke Stack

Consider a telephone cable of 10 mm diameter, with a wind blowing perpendicular to

the cable at 10 m/s. Determine the force per unit length on the cable due to drag.

Assuming air at 20 C, the kinematic viscosity is 1.5x10-5 m2/s, and the density is 1.2
kg/m3. Thus, the Reynolds number for the cable would be:

m mm
10 10mm10  3 
VD s m
Re D    6.66x10 3
 2
1.5x10  m
5

s
From figure 15.3, at this Reynolds number the CD for a smooth cylinder is roughly 1.
The frontal area of the cable (per meter of length) is A = (0.01)m x (1)m = 0.01m2.
Thus, we calculate the drag/meter as:
kg m2 kg  m
Drag / m  C D  12 V 2 A  11.2 3 0.5100 2 0.01m 2  0.6  0.6 N
m s s2
Not a particularly large force. For a suspended cable 30 meters long, the total force
on the cable would be 18 Newtons, or about 2.4 lbf.

Now, what if the same wind blew on a power plant smoke stack, 10 meters in
diameter and 80 meters high? Here, the Reynolds number would be:

VD
10 m 10m
Re D   s  6.66x107
2

1.5x10  ms
5

486
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 15

From figure 15.3, this would place the flow in the turbulent region, with a CD of
roughly 0.4. So, the drag on the smoke stack would be:
kg m2 kg  m
Drag  C D  12 V 2 A  0.4 1.2  3 0.5100 2 800m 2  19,200  19,200 N
m s s2
A substantial force (roughly 4,300 lbf), but easily handled by such a large structure.
Note that if the wind were to blow twice as hard (i.e. 20 m/s), the force would
quadruple to 76,800 N (17,300 lbf), which could begin to be worrisome for the
operators. Note that 20 m/s is roughly in the range of what would be considered a
gale force wind.

Circular Disk Square Disk

CD = 1.17 CD = 1.18

Cube Cube:
Rotated 45°
CD = 1.07 CD = 0.81

60° Cone: 60° Cone:

flat side point side
CD = 1.10 CD = 0.50

Hemisphere: Hemisphere:
flat side curved side
CD = 1.17 CD = 0.42

Hemispherical Cup: Hemispherical Cup:

inner side outer side
CD = 1.42 CD = 0.38

Parachute

CD = 1.20

Figure 15.10 Drag coefficients for selected three-dimensional bodies at Re  104

with laminar approach flow (Reynolds number based on diameter or
vertical dimension of the body). [after Hoerner, 1965]

487
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 15

Figure 15.10 shows CD values for selected smooth three-dimensional bodies with laminar
approach flow and Re  104. For these bodies, the Reynolds number is based on the
frontal vertical dimension (i.e. diameter for bodies of revolution, like a cone, and frontal
height for rectangular shapes). The behavior of these bodies is similar to that of the two-
dimensional bodies shown in figure 15.7, with a significant difference depending on the
orientation of the body. Again, sharp corners at the leading edge create much higher CD
values. Note particularly the hemispherical cup, for which CD varies by a factor of 375%
depending on orientation. Of course this elevated drag effect is put to good use for such
devices as parachutes (the last shape in figure 15.10), where high drag provides lowered
terminal velocities for falling bodies (such as paratroopers), or when employed as air
brakes (e.g. on aircraft or drag racers).

Example: Terminal Velocity with a Parachute

Consider an 80 kg man jumping from an airplane, and using a parachute to descend.

Assume that the air (near the ground) is at 20 C, the kinematic viscosity is 1.5x10-5
m2/s, and the density is 1.2 kg/m3. What should the frontal area of the parachute be to
assure landing at a safe speed of 6 m/s?

Let the weight of the parachute and pack be 5 kg, and neglect the drag coefficient of
the man. Then the drag force required at landing will be that to balance the
gravitational force on the combined man + parachute:

m kg  m
Drag required  Mg  85kg 9.81 2
 833.8  833.8 N
s s2
The drag equation is:

Drag  C D  12 V 2 A

Solving for the frontal area gives:

2 Drag
A
C D V 2

From figure 15.10 CD = 1.2 for a parachute, so the required frontal area is:

kg  m
2833.8
A s2  32.2m 2
kg m2
1.21.2 3 36 2
m s
D2
And since the frontal area will be A   , the parachute diameter will be 6.4
4
meters, or 21 feet. The amount of cloth to make the necessary parachute (assuming a
hemispherical shape) would be about 64 m2 or 210 ft2—a lot of cloth. Additionally,
The Reynolds number for this chute would be 2.5x106, so it is easily exceeds the
ReD > 104 criteria listed in figure 15.10.

488
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 15

15.4 Unsteady Separation/Drag Effects

As we discussed in Section 15.3, when an impinging flow encounters a bluff body, the
flow will initially undergo strong acceleration toward the thickest portion of the body,
followed by strong deceleration after passing the point of maximum thickness. For
example, in Section 9.8.1 we showed that inviscid flow over a circular cylinder will
initially accelerate to the point of maximum thickness, and then decelerate from that point
to the trailing edge. Correspondingly, the pressure along the cylinder surface will
decrease to the point of maximum thickness, yielding a favorable pressure gradient,
followed by a pressure increase to the trailing edge of the body, reflecting an adverse
pressure gradient. Figure 15.11 shows the general streamline pattern and non-
dimensional pressure behavior for inviscid flow over a cylinder, from Section 9.8.1.

(a)

Cp vs 
1
0
-1 Region of favorable Region of adverse
pressure gradient pressure gradient
CP

-2
-3
0 30 60 90 120 150 180

 (degreees)

(b)

Figure 15.11 Inviscid, potential flow over a circular cylinder: (a) Streamline
behavior [from figure 9.7]; (b) Surface pressure coefficient for inviscid
P  P 2
flow over a cylinder, where Cp 
1 U 2
 1  4 sin  [from figure 9.9].
2

As was illustrated by the Falkner-Skan solutions in Section 13.6, and discussed in general
in section 14.6, the boundary layer of a viscous fluid cannot be maintained under even a
mild adverse pressure gradient, and will quickly separate from a cylinder, or other
similarly bluff bodies, when an adverse pressure gradient is encountered. However, as we
showed in Section 11.5.3, a region of favorable pressure gradient (as encountered in the
leading portion of bluff bodies) will cause vorticity to be generated at the bounding
surface. So, what happens to the vorticity generated by the favorable pressure gradient
over the leading side of a cylinder or other bluff body when a boundary layer separates?
The answer is that the vorticity is transported into the wake of the cylinder when the
boundary layer separates from the cylinder/body surface (generally, shortly after reaching
the maximum body thickness). And because a sheet of vorticity is very unstable, the
separating vorticity will generally concentrate into discrete vortices, which form on or

489
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 15

near the body; these vortices then periodically detach from the wake, and are carried
downstream. This detachment of discrete vortices is termed "vortex shedding" since the
vortices are formed within the separated boundary layer, and then periodically "shed" and
carried downstream by the flow field. A YouTube video, shown here, uses laser
illuminated dye in a water flow to visualize the type of unsteady, yet periodic vortex
shedding behavior that can develop in the wake of bluff bodies.

As you might expect, the process of periodic vortex shedding causes an often strong
oscillation of the wake of the bluff body. This wake oscillation results in a change in the
trailing edge pressure on the body, which correspondingly causes a temporal change in
both the drag and lift forces on the body. Depending on the size, frequency, and
regularity of this vortex shedding process, these oscillations of the body wake and the
associated pressure field can act as a strong forcing function on the body. A good
example of periodic forcing by vortex shedding on a cylindrical pendulum can be viewed
here.

When this vortex-induced forcing is near the natural frequency of a solid body, it can
cause the development of significant periodic movement of the body within the flow
field, and under extreme cases failure and collapse of the solid structure comprising the
body. The YouTube video shown here shows the forced oscillation of a street light due to
vortex shedding.

15.4.1 Vortex Shedding by a Circular Cylinder

The process of vortex shedding from a circular cylinder can display very periodic
behavior, with the formation of very discrete vortices that are carried downstream by the
main flow, as shown schematically in figure 15.12.

U
D


Shedding frequency = N

Figure 15.12 Schematic of vortex shedding from a circular cylinder, with a spacing
of L, and a frequency of N.

We can characterize the frequency of the shedding of the separated vortices by a non-
dimensional frequency number termed the Strouhal number, given by Eq. 15.2:

ND
Strouhal Number = St = (15.2)
V

490
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 15

Note that the Strouhal number (Strouhal, 1878) depends on the Reynolds number of the
VD
cylinder (based on the cylinder diameter, D), Re D  . Figure 15.13 shows the

general behavior of the Strouhal number for a cylinder as a function of Reynolds number,
as determined from a number of experimental studies. The grey region in this figure
indicates the spread of the data for a series of studies, with the red line indicating the
general consensus for St = f(ReD).

Generally, the separated wake behind a cylinder is relatively steady until ReD  60. Once
ReD > 60, vortex shedding initiates, and the Strouhal number increases with ReD [i.e. St =
St (Re D ) ] up to about Re D 103 , where St becomes constant at  0.21, until ReD  105.
For ReD > 105, the vortex shedding becomes less consistent and irregular, with different
studies suggesting a range of behavioral changes. The red line in figure 15.13 represents
a compromise. Most studies show that when ReD > 107 the vortex shedding becomes
disorganized, and Strouhal numbers may no longer be determined.

0.4

0.3
Spread in data

ND
St 
U 0.2

0.1

0 2 3 4 5 6 7
10 10 10 10 10 10 10

Figure 15.13 The behavior of the Strouhal number as a function of Reynolds

number. The red line is the consensus behavior for a series of
experimental studies, and the grey region indicates the spread in the
empirical data. [after Hoerner, 1965]

Note that the region of constant St can be put to good use as a velocity-sensing device, by
noting that for a circular cylinder we can approximate that St  0.21 within the region
103 < ReD < 5x105. Thus:

ND ND
St   0.21  V  4.76D  N V  N
V 0.21

So, within the region of constant Strouhal number the velocity is directly proportional to
the shedding frequency. Thus, by correlating the local flow velocity with the shedding
frequency of a small bluff body, the shedding frequency can be used to calibrate velocity
measurement devices, such as hot-wire anemometers.

491
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 15

Example: Since each shed vortex is a source of low pressure (at the center of the
vortex), the passage of a vortex will create a temporal pressure change. Since sound is
created by time-varying pressure, if the vortex passing frequency is high enough,
there will be an audible sound. So, consider air at 20 m/sec (about 40 mi/hr) passing
across two different circular rods: a) 2mm coat hanger, and b) a 2 cm rod.

The vortex shedding frequency will be different, due to the differing diameters, which
will result in different audible frequencies. The shed vortices are sources of sound
pressure, and thus we can “hear” the difference in the shedding behavior.

Here we estimate the shedding frequencies:

a) Coat hanger  D  2 mm, V  20 m/s,   1.5 105 m2/sec

m m
( 20) ( 2)mm(103 )
VD s mm  2.66x103  St = 0.21 (from figure 15.13)
Re D   2
 m
1.5 105
s
ND
St  0.21 
V
m
(0.21)(20)
N s  2100 Hz
m
( 2)mm(10 ) 3

mm

b) 2 cm rod  D  2 cm, V  20 m/s,   1.5  105 m2/sec

Re D  2.66x104  St = 0.21 (from figure 15.13)

N  210 HZ

Human hearing can detect sounds in the range of 20 to 20,000 Hz (check here to see how
well your hearing is). So, the shedding from a coat hanger will be in the mid-higher
frequency range of our hearing, and the 2 cm rod (roughly the diameter of a broom stick)
will shed near the low end of our hearing. Try this as an experiment. You should be able
to swing these types of rods by hand at roughly the 20 m/s speed. Note that the sounds
will also be of different amplitudes, because the different diameters generate different
strength and size vortices (the coat hanger sound will be harder to hear).

15.4.2 Vortex Shedding by Other Shapes

Generally, any bluff body will display similar periodic shedding, with some bodies
shedding more periodically than others. Hoerner (1965) has shown that within
103 < ReH < 105 (where H is the vertical dimension of a body), the Strouhal frequency is
roughly inversely proportional to the drag coefficient of the body, with an empirical
relationship according to Eq. 15.3:

492
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 15

NH 0.21
St   0.75 (H is vertical dimension of the body) (15.3)
U CD

So, generally a lower drag body will have a higher Strouhal number, and higher drag
body a lower Strouhal number. However, the more streamlined a body is, the less likely
it is to have organized vortex shedding. Additionally, three-dimensional bodies will
generally not shed in an organized manner. Organized, periodic vortex shedding
primarily occurs for flow over two-dimensional cross-section bluff bodies (e.g. long rods,
wires, or poles).

So, what are common bodies that shed vortices? Almost anything that has a bluff-shaped
body will shed vortices. And the bodies, and the vortices they shed, can often be quite
large. For example, all the following can undergo some form of vortex shedding---
periodic, or otherwise:

 girders
 buildings
 trucks & cars
 airfoils
 big ships
 dock pilings and bridge supports
 suspended pipe lines and cables

What are the consequences? Periodic pressure loading of the body, which can cause
fatigue and failure of a solid body, or can act as an external forcing function, that can
excite large oscillations at the system natural frequency. Examples are:

 Tacoma Narrows bridge collapse (1932)

(http://www.youtube.com/watch?v=j-zczJXSxnw)
 swaying flagpoles (https://www.youtube.com/watch?v=ptYrbQGk6DQ)
 buffeting of cars by shedding from large trucks
 swaying buildings
(one shows the skyline moving, another shows a pendulum in a room swaying)

On a windy day, the Sears Tower in Chicago will shed large vortices, which cause it to
sway slowly, with a movement of up to three feet on a windy day. Of course the higher
the wind speed, the greater the movement, and the higher the frequency of the swaying
motion. For example, assuming a rectangular-shaped building of the dimensions of the
Sear Tower, we get following approximate frequencies

@ 20 mph – 0.021 HZ  48 second oscillation period.

@ 60 mph – 0.063 HZ  16 second oscillation period

People occupying the upper floors of the tower have been known to get "seasick" from
this motion. See this video of building sway caused by vortex shedding.

493
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 15

15.5 Control of Separation and Drag

Since pressure/form drag is significant on bluff bodies, it is of practical advantage to

employ methods to reduce pressure drag (i.e. a process generally termed “control”), and
or reduce or eliminate the pressure fluctuations due to vortex shedding. There are many
different approaches to accomplish this reduction, most of which fall into two categories:
(1) passive control, and (2) active control.

15.5.1 Passive Control

15.5.1.1 Surface Effects

The most common passive control method is to promote mixing of the fluid passing over
the body to keep the flow attached, and thus delay separation, which correspondingly
reduces pressure drag and vortex shedding. Some general methods are shown in Figure
15.14. The simplest passive methods attempt to cause the boundary layer on the body to
transition to turbulence as soon as possible. For these techniques, the body is either
purposely roughened (e.g. the dimples on a golf ball), or fitted with a "tripping" device,
such as a wire, rod, or saw-toothed blade, which generate localized disturbances. In all of
these cases, artificial disturbances are generated, which promote the development of a
turbulent boundary layer (see Chapter 17), which subsequently delays flow separation
(see figure 15.1a, and the associated visualizations in 15.1d, pictures 2 and 3).

Roughen
Turbulence Increased
a) rough surfaces generator
mixing

b) trip wires

c) turbulence generators
Trip wire

Figure 15.14 Passive methods for delaying separation by creating increased mixing.
Methods causing an early transition to turbulence: (a) roughening
surfaces, and (b) use of a trip wire. Methods generating vortices near
the surface: (c) turbulence or vortex generators projecting from the
surface.

More complicated passive methods utilize larger protrusions from the body surface (on
the order of the boundary layer thickness, or larger), which generate large-scale vortices
to promote strong mixing near the region of natural separation, and again delay
separation. These are generally termed "vortex" or "turbulence" generators", since they
promote turbulent-like behavior by generating streamwise vortices near a surface.
However, they normally create stream-wise vortices of larger scale than naturally
occurring turbulence. These devices range from small protruding tabs at an angle of
attack, which generate trailing vortices, to fabrication of bounding surfaces with span-
wise waviness, which also promotes the development of stream-wise vortices. Regardless
of how they are generated, the creation of stream-wise vortices promotes downstream

494
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 15

mixing, and can be very effective in delaying separation and reducing drag. Additionally,
streamwise vortices will interact with the separating vorticity layer, and interfere with the
vortex shedding process, and thus reducing or eliminating the periodic pressure variations
due to vortex shedding.

A variation on trip wires and vortex generators that has proven relatively successful is
wrapping a wire spirally around a cylinder. This is relatively easy to do, and has the
combined effect of tripping the boundary layer, to make it turbulent, while generating
streamwise vortices, which interfere with the periodic vortex shedding process. An
additional benefit of this approach is that unlike the trip wire, it has no preferred
directional orientation, and will be effective regardless of the flow direction. This spiral
wrapping approach has been used successfully in a wide variety of applications, from
power plant chimneys (see images here) to large suspended cables and pipelines.

Passive mixing methods are generally easy to apply and implement, and do not require an
additional control system to operate. However, such methods have varying degrees of
effectiveness in the control of separation, depending on the type of body. The placement
of trip wires and turbulence generators is sometimes a bit tricky, and may require a bit of
trial and error to optimize the best location. Additionally, turbulence generators extract
energy from the mean flow to generate stream-wise vortices. This energy extraction
process creates additional drag on the body. However, these devices normally delay
separation sufficiently, and reduce pressure drag enough, to override the added device
drag of the turbulence generators.

15.5.1.2 Splitter Plates

Another passive method of drag and vortex shedding control is the use of splitter plates,
several variations of which are shown in figure 15.15. The type of effect they can have
on flow behavior is illustrated here. Splitter plates are flat plates located in the wake of a
bluff body, and extending the span of two-dimensional geometry rods. Figure 15.15

a) Thwaites flap CD = 0.94 (1.11)

b) CD = 0.64 (1.11)

V
c) CD = 0.67 (1.11)

d) CD = 1.63 (1.98)

e) CD = 1.67 (1.98)

Figure 15.15 The use of splitter plates for reduction of drag and vortex shedding
effects. CD values are for the respectivesplitter arrangements shown
(unmodified CD values are shown in parentheses) [after Hoerner (1965) ]

495
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 15

shows several types of plate sizes and relative locations in the wake of the body, from a
very short plate (about ¼ D in length, known as a Thwaites flap), shown in figure 15.15a,
to a long plate (about 4D in length), shown in figure 15.15b.

The objective of these plates is to stabilize the wake by reducing the unsteady interaction
of the boundary layers separating from the top and bottom of the body. This stabilization
also causes an increase in the trailing edge pressure, and thus reduces the drag on the
body. It is interesting that the location of a splitter plate of roughly 1D length at a
distance of 3D downstream of a cylinder, as shown in figure 15.15c, has almost the same
effect on drag and unsteadiness as that of the 4D plate in figure 15.15b.

Generally, the complete elimination of vortex shedding requires a splitter plate of length
4D or greater. Shorter plates will cause reductions in the Strouhal number, but will not
eliminate the process. The 1D plate was found to reduce the Strouhal number from St ≈
0.19 when the plate is in contact with the cylinder, to roughly St ≈ 0.13 when the plate is
3D removed from the cylinder (figure 15.15c). However, locating the plate farther from
the cylinder ( > 3D) results in a jump in Strouhal number back to St ≈ 0.21, the value
typical of a cylinder with no splitter plate (and a return to CD = 1.11, the value with no
plate).

While splitter plates can reduce drag from 10 to 40%, their chief impact is to reduce or
eliminate the vortex shedding. So the primary application of these splitter plates would
be where unsteady loading on a structure is a problem. However, fabrication of such
control plates is often awkward, particularly if they are to be located away from the body,
such as those in figure 15.15c and d. If the plates are not sufficiently rigid, they can be
subject to forcing by the alternating vortex interactions, which can result in harmonic
flexing of the plate, which can potentially exacerbate the vortex shedding. Additionally,
to be effective, the direction of the flow over the body must be from a consistent
direction. If a flow is an atmospheric or oceanic flow, and varies significantly in
direction, the use of flaps can be quite problematic and is not recommended.

15.5.2 Active Control

Active control of flow separation and drag is done using some form of auxiliary system
to remove the stagnant wake fluid downstream of a flow separation or to reenergize the
boundary layer fluid. As opposed to passive methods, which extract energy from the
mean flow, and reenergize the boundary layer through mixing processes, active control
requires the addition of energy to the local boundary layer from some external source.
These approaches are generally more complicated than passive devices, but can be
directed at specific regions of the flow, and can often be very effective in minimizing
separation and pressure drag for flows that cannot be controlled by passive means.

Figure 15.16 shows several general approaches to active control of separation. There are
others, but those shown illustrate the most common approaches. One particularly
effective technique is the use of suction (figure 15.16a) to remove boundary layer fluid
that has been slowed by an adverse pressure gradient, or the stagnant wake fluid
downstream of a separated region. Suctioning off of the low-speed or stagnant fluid will

496
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 15

bring the higher-speed outer region flow into closer proximity of the bounding surface,
and thus keep the boundary layer energized. Such techniques are not easy to implement,
since they require a separate pump or suction device to create the suction, and a locally
perforated (i.e. porous) bounding surface. In such suction systems, contamination of the
porous surface, creation of the proper flow though the porous surface, and proper location
of the porous surface are all problems that must be addressed, usually by trial and error.

a) suction  remove separated fluid/wake

b) blowing  keep boundary layer moving/energize wake

c) rotating surfaces  rudders Separated flow Attached flow

No rotation With assisting

rotation

Figure 15.16 Active methods for delaying separation by removal of regions of

separated fluid or reenergizing the boundary layer: (a) fluid removal
by surface suction; (b) reenergizing the boundary layer/wake by
injection of streamwise fluid through the bounding surface; (c)
reenergizing boundary layer fluid by use of a moving surface.

To properly account for the drag reduction using active fluid removal, one must also
account for the additional energy required to run the suction device. Often the reduction
of flow energy losses due to the drag reduction achieved is exceeded by the energy
required to run the fluid removal system, resulting in a net increase in "lost" flow energy,
rather than a decrease in lost flow energy. Often suction systems are utilized when drag
reduction is not the objective, but when one wants to reduce separation, and keep a flow
attached and steady. If vortex shedding or unstable behavior is threatening the integrity
of a structure, the penalty paid to operate the suction or injection system may be justified.

A second active technique for controlling separation/drag is fluid injection through the
bounding surface (figure 15.16b), generally termed "blowing". This is usually done
through surface holes or slots angled to inject the fluid in the direction of the downstream
flow. Again the objective is generally to either (1) reenergize the boundary layer fluid,
but in this case by the introduction of "new" high-velocity fluid, or (2) create vortices
similar to a vortex generator (see an example here) to help control separation. This
technique can be effective, if done correctly. However, it has drawbacks similar to the
use of suction, such as contamination of the injection openings, proper placement of the
injection ports, creation of an appropriate injection pattern, and the requirement of a
separate pump or pressure device to create the flow. An additional concern is the angle
of the injection openings, which cannot, due to practical considerations, be parallel to the
bounding surface, but must be at some angle to the flow direction. Of course, one must
again account for the added energy required to create the injection process vs. any
improvement in separation/drag. In general, injection systems are most effective for
airfoil-shaped blades used in gas turbine environments, and are generally more effective
in reducing heat transfer to turbine blades and reducing flow separation, than reducing

497
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 15

overall drag. In such gas turbine environments, the turbine blades are essentially hollow,
and receive the pressurized fluid by bleeding off air from the compressor of the gas
turbine system (see "Cooling" here).

A third approach for active separation/drag control is to employ a moving boundary

(figure15.16c). Such an approach is only possible using a moving belt, or a circular
rotor, built into the boundary of a body. The objective is to use a moving boundary to
reenergize the boundary layer fluid, such that the boundary layer does not have to work
against both an adverse pressure gradient and surface shear. I am not aware of any
practical applications of moving belt flows for separation control; however, rotors have
been put to practical use at the junction of rudders on large ships, where the flow
velocities are not particularly high, and where the objective is to reduce separation
downstream of the ruder junction, or pivot. This process helps the rudder retain
aerodynamic "lift", and thus its effectiveness as a control surface.

The schematic in Figure 15.15c shows the general placement of a rotor at a rudder
junction. As shown, such a configuration will keep the local boundary fluid energized
such that it can remain attached under significant changes in direction. Such rotor
devices have been successfully utilized on the rudders of large cargo ships, and allow
them to have a much tighter turning radius when maneuvering in tight spaces, such as
harbors. However, such rotor devices are not practical in aerodynamic flight systems,
due to the velocities involved, and the weight of the rotor and accompanying powering
system. It should be noted that such rotational devices applied in practice will reduce
separation, and potentially the drag, although that is generally not their primary objective.

References

Hoerner, S. F. (1965). Fluid-Dynamic Drag, Hoerner Fluid Dynamics, Bricktown New

Jersey.

Problems

1. Consider a telephone pole of circular cross section of 28 cm diameter, with a gale

force wind blowing perpendicular to the pole at 20 m/s. Determine the force per unit
length on the pole due to drag if the pole surface is (a) smooth, or (b) rough. Assume
air at 20 C, the kinematic viscosity is 1.5x10-5 m2/s, and the density is 1.2 kg/m3
(neglect any effect of the ground on the drag).

2. A square sign post of 28 cm on a side, with a gale force wind blowing perpendicular
to the post at 20 m/s. Determine the force per unit length on the post due to drag.
Assume air at 20 C, the kinematic viscosity is 1.5x10-5 m2/s, and the density is 1.2
kg/m3. Determine the highest drag to be expected, and the lowest drag
for these conditions, and why?

498
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 15

3. A square sign post of 20 cm on a side, supports a thin, circular metal sign of 70 cm in

diameter, with the sign attached to the post starting at 2 meters above the ground.
Hurricane force winds blow perpendicular to the post at 50 m/s. Determine the Total
force on the post and on the sign due to drag. Assume air at 20 C, the kinematic
viscosity is 1.5x10-5 m2/s, and the density is 1.2 kg/m3. What is the highest drag to be
expected, and what is the lowest for these conditions, and why?

4. A circular cross-section leg of an off-shore oil platform is of 3 meters in diameter,

and extends down 100 meters. An ocean current of 4 meters/s flows past the leg.
Determine the total force on the leg due to drag. Assume the flow velocity is uniform
over the entire leg, sea water is at 10 C, the kinematic viscosity is 1.3x10-6 m2/s, and
the density is 1020 kg/m3.

5. A two smooth spherical pieces of metal of density 3050 kg/m3 are dropped over the
side of a boat. The smaller piece is 2 cm diameter, and the larger piece is 20 cm
diameter. Assume sea water is at 10 C, the kinematic viscosity is 1.3x10-6 m2/s, and
the density is 1020 kg/m3. When the spheres reach terminal velocity (no longer
accelerating), what will their descent velocities be (in m/s) and which will descend
faster? What are the results if the spheres are roughened?

6. A car top carrier for a kayak has a front cross-strut that is circular in cross-section,
with diameter 3 cm and 1.3 meters long. Assume the cylinder is smooth. When the
car is driven, the air flows perpendicular to the strut. Assuming the flow is uniform
approaching the strut, determine the total force on the strut due to drag, when the car
travels at 30 m/s without the kayak on the rack. Assume air at 20 C, the kinematic
viscosity is 1.5x10-5 m2/s, and the density is 1.2 kg/m3. How much power (in Watts)
from the engine is required to maintain this force? If we change the cross-section of
the rod to an oval with t/L=0.25, where t = 3 cm, how much is the force and power
requirement reduced?

7. The drag coefficient for a runner is roughly 1.2. The world record holding sprinter,
Usain Bolt, can run at an average speed of about 10.6 m/s for 100 meters. The world
record holder in the 1500 m run is Hicham El Guerrouj, who covered the distance in 3
minutes and 26 seconds. Your typical recreational jogger runs at a speed of about 3
m/s in running 5000 m. Assuming a cross sectional area of 0.75 m2 for all three
runners, an air density of 1.2 kg/m3, and that they run at a constant speed for their
respective distances, determine the drag force that acts on each runner during their
run, the power they must expend, and the total work they do due to drag in covering
their respective distances.

499
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 15

8. For the conditions given in problem 1, determine the approximate vortex shedding
frequency for the telephone pole.

9. For the conditions given in problem 4, determine the approximate vortex shedding
frequency for the oil platform leg.

10. For the conditions given in problem 6, determine the approximate vortex shedding
frequency for the circular strut. Would this be audible? How might you improve the
design of this strut?

11. For the conditions given in problem 2, determine the approximate vortex shedding
frequency for the square post, for both the highest drag and lowest drag configuration.

12. We want to calibrate a hot-wire anemometer (an electrically-heated wire that can
measure high frequency velocity variations) by using vortex shedding from a small
cylindrical rod. We want to do this in air, by measuring the vortex shedding
frequency with the hot-wire anemometer, and correlating it to the respective velocity.
If we need to calibrate the anemometer over a range of velocities from 1 and 20 m/s,
what is the range of rod diameters that we could use to assure that the velocity will be
a linear function of the shedding frequency? Assume air at 20 C with kinematic
viscosity of 1.5x10-5 m2/s. What is the lowest shedding frequency we might measure
with this range of rods?

13. Consider the flow of air over a cylinder. Assume that the impinging velocity is U = 1
cm/s, and the cylinder radius is R = 10 cm. The kinematic viscosity for air is  = 16
mm2/s. As the flow passes around the cylinder, a laminar boundary layer develops on
the cylinder surface. Assume that the velocity at the edge of the boundary layer can
be approximated by the potential flow solution of U = 2Usin(), where  is the
angle measured from the cylinder stagnation point around the cylinder.

a. Using Thwaites method (from Chapter 14), neglect curvature effects and determine
the development of the momentum thickness () as a function of x, where x = R,
when  is in radians. You can integrate the Thwaites equation analytically (you
need to review your integral tables)—do it in terms of , and then express in terms
of x.

b. Determine the angle max (in degrees) where the solution is no longer valid, using a
parallel calculation of the parameter n. Why does the solution fail at this point?
[hint: consider the point of separation]

c. Using the pressure distribution given by the potential flow solution over a cylinder
(from Chapter 9), determine the drag per unit depth on the cylinder. For this real

500
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 15

fluid flow, assume that after max the flow is separated, and that the pressure will
remain constant at the P(max) value until =180 is reached. Neglect the shear
stress on the cylinder surface. Express your results in terms of the drag coefficient:

Drag
Cd  1
2
U 2 ( 2R )

How does the value you obtain compare to what you might expect for this flow
(i.e. in figure 15.3)?

501
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

Chapter 16

Introduction to Non-Newtonian Fluid Behavior

Contents

16.1 Introduction............................................................................................................. 503

16.1.1 Shear Thinning Fluids (Pseudoplastics) ........................................................... 503
16.1.2 Shear Thickening Fluids (Dilatants) ................................................................. 505
16.1.3 Viscoplastics (Bingham plastic) ....................................................................... 506

16.2 Power Law and Bingham Plastic Fluid Models ………………………………………………………….. 506
16.2.1 Ostwald-de Waele Power Law ....................................................................... 507
16.2.2 Ideal Bingham Plastic …………………………………………………………………………………..…508

16.3 Simple Non-Newtonian Flows: Couette and Poiseuille .............................................. 508

16.4 Power Law Fluid Solutions......................................................................................... 509

16.4.1 Parallel Plate Flows ....................................................................................... 510
16.4.1.1 Couette Flow between Parallel Plates ……………………………………………… 510
16.4.1.2 Poiseuille Flow between Parallel Plates ……………………………………………. 512
16.4.2 Flows with Circular Symmetry ....................................................................... 516
16.4.2.1 Couette Flow between Concentric, Rotating Cylinders ……………………… 516
Outer cylinder rotating at velocity Vo, Inner cylinder fixed……………………… 517
Inner cylinder rotating at velocity Vi, Outer cylinder fixed………………………. 519
16.4.2.2 Poiseuille Flow in a Circular Duct………………………………………………………. 524

16.5 Ideal Bingham Plastic Solutions ................................................................................ 529

16.5.1 Parallel Plate Flows ....................................................................................... 529
16.5.1.1 Couette Flow between Parallel Plates………………………………………………. 530
16.5.1.2 Poiseuille Flow between Parallel Plates …………………………………………… 531
16.5.2 Flows with Circular Symmetry ...................................................................... 541
16.5.2.1 Couette Flow between Concentric, Rotating Cylinders……………………… 541
Outer cylinder rotating at velocity Vo, Inner cylinder fixed……………………… 542
Inner cylinder rotating at velocity Vi, Outer cylinder fixed ……………………… 547
16.5.2.2 Poiseuille Flow in a Circular Duct ……………………………………………………… 552

16.6 Conclusion ............................................................................................................... 561

16.7 Governing Equations for Non-Newtonian Fluids ........................................................ 564

16.7.1 Stresses on a Non-Newtonian Fluid: Cartesian and Cylindrical Coordinates .... 564
16.7.2 Continuity and Momentum Differential Equations in Cartesian Coordinates ... 564
16.7.3 Continuity and Momentum Differential Equations in Cylindrical Coordinates . 564

16.8 Table 1: Select properties of power law and ideal Bingham plastic fluids ……………..566

502
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

16.1 Introduction

To this point, we have assumed that fluids are Newtonian, and deform linearly under an applied
stress. Newtonian fluids are the most commonplace fluids, representing air and most other
gasses, water, most oils, most liquid fuels, and mercury, to name a few. However, many other
common materials qualify as fluids, but do not deform linearly under an applied stress. Some
examples are many foodstuffs, such as peanut butter and ketchup, cosmetics such as nail polish
and toothpaste, industrial materials such as freshly mixed cement and drilling mud, and
biological materials like blood and mucus. Therefore, while we will not cover such non-
Newtonian fluids in depth, it is appropriate that we examine some common types of non-
Newtonian fluids, and understand how these behave differently than Newtonian fluids.

So, how do non-Newtonian fluids behave differently than Newtonian fluids? The key is how the
shear stress is a function of the strain rate,  :

  f  16.1
We know that for a Newtonian fluid the shear stress changes linearly with the shear strain rate,  ,
such that:
  du  
    e.g.  yx      ,
  dy  
where the coefficient relating shear stress and strain rate is the viscosity of the fluid, .

Thus, non-Newtonian fluids are those that demonstrate non-linear changes of shear stress with
strain rate. Figure 16.1 shows a generic graph of several common types of non-Newtonian fluid
behavior. Non-Newtonian fluids that deform continuously under an applied shear stress (the
definition of a fluid, from Section 1.1), are generally categorized as either shear thinning fluids,
shear thickening fluids, or viscoplastics.

16.1.1 Shear Thinning Fluids (Pseudoplastics)

A shear thinning fluid, often called a pseudoplastic fluid, will become less viscous with
increasing shear stress. It turns out that many common materials display shear thinning behavior.
Two examples are peanut butter and paint. At room temperature, peanut butter is highly viscous
at low shear rates. However, when one applies a high shear stress, as you do when spreading it
on bread, the peanut putter flows, and can be distributed evenly (pretty much) over a piece of
bread. However, once the shear is reduced, the peanut butter will recover to its more viscous
nature, and will move very little, if at all. If we were to perform the same process with a
Newtonian fluid, the fluid would continue to flow under very little stress (such as that applied by
gravity), and flow off the bread. Note that if we were to heat the peanut butter, it would change
its characteristics. Although it would still behave as a shear thinning fluid, it would spread, and
be less stable at much lower shear stresses. This explains why peanut butter sandwiches, when
they are subjected to a hot environment (like leaving one in a car on a warm summer day), may
result in peanut butter oozing out of the sandwich.

503
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

Shear Thinning
(Pseudoplastics)

Shear Thickening
(Dilatant)

Figure 16.1 The generic behavior of power-law and Bingham plastic non-Newtonian fluids
relative to Newtonian fluid behavior. The n values refer to the consistency index
for Ostwald-de Waele power law model, Eq. 16.2.

Another example of a shear thinning fluid is paint. Paint is very viscous when held within the
bristles of a paintbrush. However, when we brush paint on a wall, the shear stress applied
between the brush and the wall causes the paint to become less viscous, which allows it to spread
evenly across the wall. Moreover, when that paintbrush shear stress is removed, the paint will
return to its more viscous state. In this relaxed state, the thin paint layer will become more
viscous, and behave more like a solid, maintaining an even layer on the wall while drying
(however, if the paint layer is too thick, it may move under the effect of gravity).

Did you ever have that age-old problem of getting ketchup out of a bottle? You open the bottle,
and it just will not come out. So, you shake the bottle toward your food, and whoosh! Out comes
the ketchup, often quite rapidly. That is because ketchup is a shear thinning, pseudoplastic fluid.
It is very viscous under low stresses (when you try to pour it), but becomes less viscous (and
flows more quickly) when you shake the bottle and apply a higher stress.

An interesting, and lethal, pseudoplastic is quicksand. Quicksand is a suspension of loose sand

saturated by water. The suspension will appear solid, but a sudden small (1 to 2%) change in
pressure, or application of a shock, will cause the suspension to transform rapidly to a low
viscosity fluid. If a human or animal walking on the quicksand causes the shock, they will start
to sink. The good news is that due to the higher density of the quicksand, a human will only sink

504
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

to about their waist. The bad news is that further agitation of the quicksand by rapid, struggling
movements will further reduce the quicksand viscosity, causing one to sink further. The solution
to escaping (if you are careless enough to step into quicksand) is to move your legs slowly,
which causes the quicksand to remain more viscous, and slowly rotate your body to a face up
floating position. Then you do a really, really slow backstroke to move your body over and out
of the quicksand.

16.1.2 Shear Thickening Fluids (Dilatants)

In contrast to shear thinning fluids, shear thickening fluids, many of which are termed dilatant
fluids, become more viscous as the shear stress is increased (see this link). These types of fluids
are generally suspensions of particulate matter in a surrounding liquid, and are less common than
pseudoplastic fluids. Examples of such fluids are wet cement and synovial fluid. The
explanation of why this thickening occurs with increased shear stress is speculative; it is
presumed that when the fluid is at rest, or under low shear stress, the supporting liquid, which
inhibits the suspended material from interacting strongly, lubricates the suspended material.
However, when the dilatant material is under high shear stress, the suspension expands, reducing
the liquid lubrication, and allows the suspended material to interact strongly, increasing the
apparent viscous behavior. There are other speculations as to the physical reasons for shear
thickening, but there is no definitive answer.

This shear thickening process within particulate suspensions, such as the cited wet cement,
clearly explains our experience with these types of suspensions. For example, if one steps into
wet cement, the material will allow you to move slowly, with low shear motions, to extract your
foot. However, quick, high shear motions will make the material behave in a much more viscous
manner, and you will have little chance of easily extracting your foot, most probably losing your
shoe.

Synovial fluid, which is a suspension of biological material, is secreted by the cartilage

protecting our body joints, and has a consistency like egg whites. It acts as a lubricant, becoming
more viscous under strong shearing movements, such as those generated by running. This more
viscous behavior acts to protect the higher loading on the joint surfaces. However, synovial
fluid becomes less viscous when shear and movement are reduced, such as slow walking, where
joints are under lower loading. This lowered viscosity of course allows easier movement of the
joint.

A truly interesting aspect of certain dilatant fluids is they will essentially solidify under a sharp
impact. With certain fluids, (a highly concentrated mixture of cornstarch and water is one
example), this rapid thickening allows a person to run or walk briskly across a pool of the
mixture, without sinking in. However, if one just stands on the mixture, the viscous effect is
diminished, and a person will slowly sink in, and have real trouble extracting themselves (see
this link for a video example).

An unexpected practical application of shear thickening fluids is in body armor (see this link).
Tests have shown that filling Kevlar body armor with a small layer of certain shear thickening
fluids can increase the effectiveness of the armor to bullet impacts, while allowing the armor to
remain relatively mobile under normal body motion. However, use of the fluid increases the
weight of the armor, which may be a significant disadvantage.

505
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

16.1.3 Viscoplastics (Bingham plastic)

Fluids that do not continuously deform until a finite shearing stress is exceeded are termed a
viscoplastic. These materials behave as a rigid body under low shear stresses, but flow as a fluid
under high shear stresses. The fluid behavior of a viscoplastic, after reaching a minimum yield
stress, can behave like either a Newtonian or a shear thinning fluid. The simplest of these types
of non-Newtonian materials is an ideal Bingham plastic, named after E. C. Bingham who
originally proposed the model. As shown in Figure 16.1, an ideal Bingham plastic will deform
similar to a Newtonian fluid, after a critical yield stress is exceeded. Often these materials are
composed of suspensions of particles or large molecules, which in a quiescent state interact with
one another, creating a weak solid structure. However, under a critical level of applied stress the
solid structure breaks down and begins to flow like a fluid. If the stress drops below the critical
level, the particles/molecules reform as a weak solid. An interesting aspect of a Bingham plastic
is that when it is at rest in an open container, its free surface will not even out like true fluids, but
will retain peaks and valleys on the surface. For example, honey, a very viscous Newtonian fluid,
will slowly relax to give a smooth free surface, but mayonnaise, a Bingham plastic, will not relax
to a smooth free surface, regardless of the time allowed.

Examples of a Bingham plastic are drilling mud and toothpaste. Drilling mud is a suspension of
special clays in either water or a petroleum fluid, generally resembling thick chocolate milk, or a
milk shake. The mud is pumped through oil field drilling systems to lubricate and cool the
drilling bit, and remove the drill cuttings. Under high pumping pressure, the fluid flows in a
Newtonian-like, free-flowing fashion. However, reducing the pumping pressure below a certain
level causes the mud material to transform into a “gel-like” structure, which resists flow. This
solidification holds the drill cuttings in place until the flow is resumed. When the pressure is
raised above a critical level, the mud will again start to flow as a fluid, carrying along the
suspended drill cuttings. However, the pressures required to cause the mud to flow are quite
significant, and we will see later in Section 16.5.2.2.

You probably never paid much attention to toothpaste as a fluid, but it also behaves as a
Bingham plastic. If you uncap a tube of toothpaste, the paste will remain in the tube, as a gel-
like solid—even if you invert the tube with the opening facing downward. However, when you
squeeze the tube, which raises the internal pressure, and thus the shear stress at the wall of the
outlet of the tube, the paste will flow out of the tube like a very viscous liquid; and when you
stop squeezing, the tooth paste returns to a solid, or gel-like state.

16.2 Power-Law and Bingham Plastic Fluid Models

Many mathematical models have been developed to describe (or attempt to describe) non-
Newtonian fluid behavior. A study of these models, and their various fluid behaviors, occupies
thousands of technical papers, and numerous books on the topic (see for example, Bird, Stewart,
and Lightfoot, 2002, Bird, Armstrong, and Hassager,1987,and Chhabra and Richardson, 1999).
However, to give you some understanding of how non-Newtonian fluids differ from Newtonian
fluids, we will examine only two of these models: a simple power-law model of Ostwald and de
Waele, (which models both pseudoplastic and dilatant fluids) and the viscoelastic ideal Bingham
plastic model. Both of these models are approximations of actual fluid behavior, but allow us to
examine the generic behavior of a broad range of non-Newtonian type fluids.

506
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

Both of these models are based on curve fits of empirical data, and were hypothesized by their
namesakes as reasonable curve fits of observed and measured fluid behavior. Both of these
models recognize that, like a Newtonian fluid (Eq. 16.1), the fluid shear stress is a function of the
shear strain rate,  , where again   f    .

16.2.1 Ostwald-de Waele Power Law

Whereas the shear stress for a Newtonian fluid is given by:

  du  
    e.g.  yx      ,
  dy  
the equation for an Ostwald-de Waele power-law fluid is a variation on Newton’s law, and given
by:
  K 
n
16.2
Here K is termed the flow consistency index (with SI units of Pa sn), which is similar to (but not
identical to) viscosity, and n is termed the flow behavior index (n > 0). Values of K and n for a
range of pseudoplastic and dilatant fluids are shown at the end of this chapter in Table 1.

Note in Figure 16.1 that fluids with 0 < n < 1 are considered pseudoplastic fluids (shear
thinning); the closer the value of n is to 0, the more the fluid behaves plastically. For values of n
< 0.5, the shape of the curve will start to approximate the behavior of a Bingham plastic, with
such fluids often modeled reasonably well as either a power-law or Bingham plastic fluid.

Figure 16.1 also shows that fluids with n > 1 are dilatant fluids (shear thickening). Such fluids
are less common and often only display thickening effects over only a certain range of shear
rates. Most shear thickening fluids are suspensions, and often don’t display consistent behavior
over their entire shear stress range, often displaying pseudoplastic behavior at low shear rates,
and transforming to dilatant behavior at higher shear rates.

While a nice model of a number of non-Newtonian fluids, there are a couple of issues with the
Ostwald-de Waele power law equation of Eq. 16.2. First, since the flow behavior index, n, is
generally a fractional number, the equation only works for positive shear rates, since we cannot
take a fractional power of a negative value. This is often circumvented by assuming absolute
values of strain rate. However, for the examples shown in this chapter, coordinates and limits of
integration are used that will yield positive shear rates.

Secondly, the flow consistency index K for specific fluids has units that depend upon the value
of n, to yield a shear stress that is in force per area. As we point out above, the units of K will be
SI units of Pa sn. Thus, the time units are unique to the respective value of n for a specific fluid.
This is quite awkward, and makes effective comparisons of one fluid to another difficult, as we
will see.

Finally, if we define a relative shear viscosity for an Ostwald-de Waele fluid, we have:

  K      app    app  K   
n n 1
,

507
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

where  app is the apparent viscosity of a fluid. For pseudoplastic fluids (n < 1), this apparent
viscosity will become infinite as the shear rate approaches zero, which cannot physically happen.
Other fluid models address this inconsistency, but generate expressions that are more
complicated.

Despite the limitations, the Ostwald-de Waele model does a nice job of approximating the
characteristics of fluids that are both shear thinning (n < 1) and shear thickening (n >1).

16.2.2 Ideal Bingham Plastic

The equation for an ideal Bingham plastic fluid is a two-region model, and is referenced to the
strain rate as:

0 when   0
and
  0  B    when   0 16.3

Here 0 is termed the yield stress, and  B is the apparent viscosity once the material starts to
flow as a fluid. In regions where   0 , the material will not deform    0 , and simply remain
static, or move in uniform motion as a solid. Values of 0 and  B are also listed in Table 1 for a
number of ideal Bingham Plastic materials.

Notice that this section is titled “ideal Bingham Plastic.” As shown in Figure 16.1, an ideal
Bingham Plastic deforms linearly, similar to a Newtonian fluid, after reaching the critical yield
stress. In reality, Bingham fluids more often behave as pseudoplastics after reaching the yield
stress. However, to make life easier, we will use the ideal plastic case.

I must note that the properties listed in Table 1 come from a wide variety of sources, and may
vary widely, depending on the constituency of the material, which is often vague. For example,
there are several listing for the properties of mayonnaise, which vary quite broadly, as do the
listings for latex paint, the constituents of may or may not be listed. Also note that the cited
properties are often shear dependent (i.e. they may vary with the shear value of the material). So,
Table 1 should be taken as exemplary, but not definitive.

16.3 Simple Non-Newtonian Flows: Couette and Poiseuille

In this chapter, we examine the behavior of power-law and Bingham plastic fluids in two simple
types of flow situations that we examined previously in Chapter 6 for Newtonian fluids: Couette
and Poiseuille flows. Recall that these Newtonian flows were the simplest types of flows that we
modeled, and resulted in relatively simple equations modeling the respective flow behaviors.
Moreover, once the controlling parameters were specified (e.g. pressure gradients, pipe
diameters, viscosity, etc.), the non-dimensional flow behavior was identical for any set of
parameters (e.g. the velocity profile for a Newtonian Poiseuille flow always has a parabolic
shape). However, as we will discover in this chapter, flows of non-Newtonian fluids can

508
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

undergo quite different flow behaviors, depending on the magnitude of the controlling
parameters and the type of non-Newtonian behavior.

We will consider Couette and Poiseuille flows first in Cartesian coordinates for flows between
parallel plates, then in cylindrical coordinates for flows between rotating, concentric cylinders
and flows along circular ducts. We will only consider fully-developed flows, which will
simplify the analysis and reduce all flows to one-dimensional flows. However, to deal with non-
Newtonian flows we will have to step back from the conventional Navier-Stokes momentum
equations, and consider the momentum equations in terms of stress terms, rather than the
previous rate deformation equations shown in Section 5.8. The governing momentum equations
in terms of stress terms are listed at the end of the chapter in section 16.7, for both Cartesian and
cylindrical coordinates. Which directional equation is used to assess a particular flow depends
on the geometry and the direction of motion, as we will show in the following sections.

16.4 Power Law Fluid Solutions

The Ostwald-de Waele model was indicated in Eq. 16.2 as   K    , where  is the
n

deformation rate. However, the relationship of  to the fluid behavior will vary depending on
the geometry and coordinates considered, and the flow orientation of interest. The three model
variations we will use here are:

Non-Newtonian, Cartesian:
n n 1
 u   u   u 
x-direction only: yx  app  yx  app    K   , and  app  K  16.4
 y   y   y 
Non-Newtonian, Cylindrical
n
 v     v 
θ direction only: r  app r  app r     K r    
r  r   r  r  
n 1
  v 
where  app  K r    16.5
 r  r 

n
 v   v 
z-direction only: rz  app rz  app  z   K  z 
 r   r 

n 1
 v 
where  app  K z  16.6
 r 

In the above equations,  app is the apparent viscosity relative to the linear deformation rate  .

509
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

16.4.1 Parallel Plate Flows

For fully-developed Couette and Poiseuille flows between infinite parallel flat plates, there is no
motion in the lateral directions, but only in the streamwise, x-direction, and the governing
equations reduce to the x-direction momentum equation (Eq. 16.108b)

u u u u 1 p 1    yx  zx 
u v w   g x   xx    16.7
t x y z  x   x y z 
Simplifying Eq.16.7 for steady, fully-developed flow in the x-direction only, we get:

n
p  yx  u 
   0 where yx  K  
x y  y 
Thus, for power-law flow between parallel plates, the governing equation is:

   u 
n
 p
K    16.8
y   y   x
 p   p 
Equation 16.8 applies for both a Couette flow   0  , and a Poiseuille flow   constant  .
 x   x 
The velocity behavior for these flows are derived in the following Sections 16.4.1.1 and 16.4.1.2
respectively.

16.4.1.1 Couette Flow between Parallel Plates

Using the coordinate system in figure 16.2, we have boundary conditions of u = 0 at y = 0, and
u = U at y = h, where we consider a moving upper plate and a fixed lower plate.

y U
h
x

Figure 16.2 Couette flow between a lower stationary plate and an upper moving plate.

Since there is no pressure gradient for a Couette flow (see the detailed discussion of this in
section 6.3.1), Eq. 16.8 reduces to:

   u 
n

K   0 16.9
y   y 

Integrating Eq. 16.9 once we have:

1
u  C1  n
 
y  K 

510
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

Integrating again:

1
 C n
u   1  y  C2
K
Applying the boundary condition u = 0 at y = 0, gives:
1
 C n
0   1  0  C 2  C2  0
K
1
 C n
u  1 y
K
And applying the second boundary condition u = U at y = h:

1
n
 C n U
U   1  h  C1  K  U
K h y
Thus,
1 x
 U  n n
 K  
1
 C1  h Uy
u   y    y
n
16.10
K  K  h
 
 
Equation 16.10 is the same result we determined for a Newtonian fluid in Section 6.3.1, Eq. 6.8,
despite the non-linear relationship of shear stress to deformation rate for a power-law fluid. To
examine this non-intuitive result, consider the shear stress distribution.

The shear stress is given by Eq.16.4 as:

n
 u 
yx  K  
 y 
u U
Differentiating Eq. 16.10, gives:  .
y h
Thus,
n
U
yx  K    constant 16.11
h
Thus, for a given plate gap, h, and plate velocity, U, the shear stress, Eq. 16.11, is constant across
the gap (the same as for a Newtonian fluid). Since the shear stress is constant, the deformation
rate,  , will also be constant, and thus the velocity profile has to be linear. However, note that if
the velocity changes (or the gap changes), the shear stress will not change linearly, as it does for
a Newtonian fluid.

It would be nice if all our other examples were so straightforward. However, as you will see in
Section 16.4.2.1, Couette flow between rotating, concentric cylinders can get quite messy.

511
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

16.4.1.2 Poiseuille Flow between Parallel Plates

It might first seem that we should treat this solution the same way we did a Newtonian flow in
Section 6.3.2, where we selected the origin at the lower plate. However, since the integration
process deals with values of n that are generally non-integer, this creates problems with the
integration process, and in developing velocity equations that can be displayed effectively.
Since the flow will be symmetric about the midplane between the plates, it is the most effective
to locate the coordinate system as shown in figure 16.3, with the origin on the midplane between
the plates. This allows the assessment of the flow in the upper half of the gap, recognizing that
the flow in the lower half of the gap will be a mirror reflection. Using this coordinate system,
u
the boundary conditions for the upper half of the flow are  0 at y = 0 (assuming flow
y
h
symmetry), and u = 0 at y  .
2

Decreasing pressure,
y

h/2
x h

Figure 16.3 Poiseuille flow between parallel flat plates, power-law fluid.

p
For a Poiseuille flow,  constant (as shown in section 6.3.2), and the governing equation
x
from section 16.4.1 is Eq. 16.8

   u   p
n

K      constant ,
y   y   x
h u
with boundary conditions u = 0 at y  , and  0 at y = 0.
2 y
Integrating once gives:

n
 u  p
K   y  C1

 y x
u
Applying boundary condition  0 at y = 0:
y
p
0  0   C1  C1  0
x
Thus,

512
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

n
 u  p
K   y
 y  x
or
1
u  p y  n
 
y  x K 
Integrating gives:
1
n 1
 n   p 1  n n
u    y  C2
 n  1   x K 
h
Applying the second boundary condition, u = 0 at y  .
2
1 n 1 1 n 1
 n  p 1  n  h  n  n  p 1  n  h  n
u       C2  0  C2       
 n  1  x K   2   n  1  x 2K   2 
Thus,
1
 n 1

 n  p h  n  h   2y  n 
u         1
 n  1  x 2K   2   h  
 
2y
Defining y*  , and designating u n  u  n  , since the u velocity profile will be a function of
h
n, we have:
1
 n  p h  n  h   n 1

un        y *  n 1
 16.12
 n  1  x 2K   2  
At y* = 0  u n  u n,max  u n,CL , so from Eq. 16.12 we have:
1
 n   p h  n  h 
u n,CL       16.13
 n  1   x 2K   2 
p
Note in Eq. 16.12 and 16.13, that the pressure gradient, , must be positive to allow
x
1
calculation of the power of , which makes u n and u n,CL negative velocities (like a Newtonian
n
fluid, a positive pressure gradient yields a negative velocity). However, we can deal with
u
positive behavior by defining u *n  n . Thus, dividing Eq. 16.12 by 16.13 gives:
u n,CL
n 1
u *n  1   y * n 16.14
Equation 16.14 gives the non-dimensional velocity variation relative to the maximum centerline
velocity as a function of n, the flow behavior index. A graph of Eq. 16.14 is shown for a series
of n values in figure 16.4(a).

A limitation of Eq. 16.14 is that it does not allow a comparison of the relative velocity behavior
of a power law fluid to that of a Newtonian fluid, since the K and n values will not be consistent.
However, there is a way to do this for flows at equal volume flow rates.

513
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

We calculate the volume flowrate per unit depth, Q, in the upper half of the gap by integrating
h
the velocity from y = 0 to y  (note that the total flowrate between plates would be twice Q):
2
h
y
y*1 y*1
2
h h  n 1

Q  u n dy   u n,CL u *n dy*  u n,CL     n  dy*
1  y*
0 0
2 2 0  
h   n  h  n 1 
Q  u n,CL 1     u n,CL  
2   2n  1   2  2n  1 
Substituting for u CL from Eq. 16.13
1
2
 n   p h  n h
Q     16.15
 2n  1   x 2K  4
1
 p h  n h
Notice that the parameter   in Eq. 16.15 is also common to the same parameter in
 x 2K  2
the velocity equation, Eq. 16.12. Solving for this parameter from Eq. 16.15 gives:

1
 p h  n h Q
   16.16
 x 2K  2 h  n 
 
2  2n  1 
Substituting Eq. 16.16 into Eq. 16.12, we have:

2Q  2n  1   n 1

un   1   y *  n
 16.17
h  n 1  
Using Eq. 16.17, we can now do a comparison of the velocity behavior for both pseudoplastic
(n < 1) and dilatant (n > 1) fluids to a Newtonian fluid (n = 1) at the same volume flowrates.

Note that for Newtonian flows, n = 1:

3Q 
1   y *  ,
2
u Newt 
h  
with the maximum Newtonian velocity at the centerline, y* = 0:
3Q
u Newt, CL  16.18
h
un
Dividing Eq. 16.17 by Eq.16.18, we calculate u *n,rel  , the comparative velocity
u Newt, CL
profiles, for the same flowrate, Q, through a channel of width h as:

un 2  2n  1   n 1

u *n,rel     1   y* n
 16.19
u Newt, CL 3  n  1   
Figure 16.4(b) shows comparative velocity profiles, u *n,rel , Eq. 16.19, for a series of n values.

514
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

Figure 16.4 shows the comparative effect of the fluid behavior index, n, on the velocity behavior
relative to a Newtonian fluid (n = 1). Note in figure 16.4(a) that shear thinning pseudoplastics
(n < 1) display much fuller profiles, while shear thickening dilatants (n > 1) display narrower
profiles. Figure 16.4(b), for equivalent flowrates, Q, shows that the centerline velocities for
pseudoplastics will be less than Newtonian, and dilatants in excess of Newtonian. What is of
real interest for pseudoplastics is that for smaller n values (n  0.5 or less) the behavior can
appear similar to that of a Bingham plastic, which we consider in Section 16.5.2.2, and discuss in
detail in Section 16.6.

(a) (b)

Figure 16.4 Power law velocity profiles for Poiseuille flow between parallel flat plates.
(a) Normalized on the centerline velocity of each respective n value fluid;
(b) Normalized on Newtonian centerline velocity; constant volume flowrate.

What we can’t show effectively is the general effect a power law fluid has on the pressure
gradient. We might consider a relationship derived from Eq. 16.16. Solving Eq. 16.16 for the
pressure gradient in terms of Q and other parameters, we have:

n
 
p 2K  2Q 
   16.20
x h  2  n 
h
  2n  1  
As Eq. 16.20 shows, if Q is assumed positive, we can’t determine a value of. However, if we
assume Q is negative (i.e. flowing right to left in figure 16.3), or just assume a positive value of
p
the numerator, a positive value of can be determined. In addition, since the K values of
x
power-law fluids have differing units (depending on n), the only way to determine relative
pressure gradients is to calculate them individually for specific fluid properties.

515
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

16.4.2 Flows with Circular Symmetry

In this section, we address Couette flow and Poiseuille flow of a power-law fluid with
geometries of radial symmetry. Again, these flows will reduce to one-dimensional, fully-
developed flows with no lateral cross flow. However, although each of these flows is a function
of radius only, the governing momentum equation, and consequent velocity, for each of these
two flows will be different. The Couette flow will depend on the azimuthal (θ) direction
equation and yield v  f (r) , whereas the Poiseuille flow will depend on the axial (z) direction
equation and yield vz  f (r) . They will also depend on different constitutive equations for the
shear stress, as was indicated in Section 16.4.

16.4.2.1 Couette Flow between Concentric, Rotating Cylinders

Vo

Vi
r

Ro
Ri

Figure 16.5 Geometry for Couette flow between concentric rotating cylinders

As shown in Section 6.4.2, the governing equation for this type of cylindrical flow reduces to the
θ-direction momentum equation, 16.109c of Section 16.7.3.

v  v v v v vv
 vr      vz   r 
t r r  z r
16.109c
1 p 11  2 1    z  r   r 

r 
 g   2
  r r
r  r 
r 
 
z
 r 

Since this is a steady, fully-developed flow, the velocity field is given by v  f (r) , and Eq.
16.109c reduces to:

p 1  2
0 
 r r
r  r ,   16.21
where the power-law shear stress is given by Eq. 16.5:

n
   v 
r  K  r    
 r  r  
p
We showed in section 6.3.1 that  0 for Couette flow, so Eq. 16.21 reduces to:

 2
r
 
r  r  0

516
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

Integrating and substituting from Eq. 16.5 for the shear stress:
n
   v  C
 r  K r     21
 r  r  r
or
1 1
  v   1  C1  n 1    n  1 C 1 2

    2    2    n r n ,where   1
r  r  r  Kr  rr  K
Integrating again:
1
n
v    n
2 n
 C2 r 16.22
2r n

Equation 16.22 is a general solution, requiring boundary conditions at Ro and Ri to solve for the
integration constants  and C2. In Section 6.4.2 we performed a general solution given
independent rotations of both cylinders. However, in this case the mathematical manipulation
would be quite messy, so here we will examine two simpler cases:
(a) Rotation of the outer cylinder with the inner cylinder stationary, and
(b) Rotation of the inner cylinder with the outer cylinder stationary.

Boundary Conditions (a): Outer cylinder rotating at velocity Vo, Inner cylinder fixed

Here the boundary conditions are:

(1) v  Vo at r  R o , and (2) v   0 at r  R i

Applying boundary condition v  Vo at r  R o to Eq. 16.22 gives:

1
n
Vo   n 2 n
 C2R o 16.23
2R o n

Applying boundary condition v   0 at r  R i to Eq. 16.22:

1 1
n n
0   n 2 n
 C2R i  C2   n 2
2R i n 2R i n
Substituting for C2 in Eq. 16.23 yields:

 n1     n1 
 n   1 1   n  Vo
Vo   R o  2  2    2    
 2   R o n R i n     1 1


    Ro 2  2
R n R n 
 o i 
 n1 
 n 
Substituting C2, and   back into Eq. 16.22, and simplifying gives
 2 
 

517
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

   
Vo  1 r  Vo r  1 1 
v    
   n
2  n
R
2   2 
  n R n2
2 
 1 1   Ro   Ro  r
Ro 2  2   r 
n n
i i
2 
1    
R n R n 
 o i  Ro   Ri
n   


 r 
Vo    n2 2

 R o   R R n
o 
v  2 
o
 2 16.24
  2 
 R i n 
1   R o    r
n n

  R i  
 
r R
Defining r*  and ri *  i , and substituting into Eq. 16.25 we get:
Ro Ro

   2 2

Vo  r *  1 1  Vo  i   
r * n  r * n

v  2 

  2 2   2
  2n 
n   r * n  
1   1    ri * n
   ri * n  1   r * n 
  ri *    
 
Rearranging terms gives:
 2

 1  ri *  n

  r *  
v   Vo  2 r*
 1  ri * n
 
 
v
Defining a non-dimensional velocity as v  *  , our final expression is:
Vo
 2

 r
 1   
i *  n

v  r* 
v *      2  r * 16.25
Vo
 1  ri *n 
 
 
An alternative way of writing Eq. 16.25, which is often cited, is to let Ri  aR o where
R
ri *  i  a  1 , and ro *  1 then
Ro
 2

 1  a  n

  r *  
v *  r *  2  16.26
 1 a n 
 
 
Note that Eq. 16.5 gives the shear stress as (after a little work):

518
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

n
 2V 
o
n   2
   v    nR   a 
r  K  r     K o
2    16.27
 r  r      r*
 1  a  
n

   

Using Eq. 16.27, we calculate the torque on any surface between ri *  a and ro *  1 as:

T  r  2 Lr  (r)  r  2R o2 r *2  where L is the length of the cylinders

n
 2V 
o
 
 nR o 
2 2
T  2R o a LK 16.28
 2 

 1  a  
n

   
Eq. 16.28 indicates that for a given Vo, Ro, and a, the torque for a given power law fluid is
constant across the circular gap, and is not a function of the radius. Physically, this is as it
should be since if the torque varied, there would be an acceleration of fluid layers relative to one
another, which would violate the premise of a steady-state flow. Note that constant torque across
a Couette flow between concentric cylinders will hold true regardless of the type of fluid within
the cylinders. This constancy of torque will prove very helpful in Section 16.5.2.1 where we
derive the behavior of a Couette flow for a Bingham plastic fluid.

Boundary Conditions (b): Inner cylinder rotating at velocity Vi, Inner cylinder fixed

Here the boundary conditions are reversed, with:

(1) v  0 at r  R o , and (2) v  Vi at r  R i

The integrated equation is again Eq. 16.22:
1
n C1
v    n
2 n
 C2 r , where  
K
2r n

Boundary condition v  0 at r  R o gives:

1 1
n n
0   n
2 n
 C2R o  C2   n
2
2R o n 2R o n
And boundary condition v  Vi at r  R i gives:

1
n
Vi   n 2 n
 C2 R i 16.29
2R i n

Substituting C2 into Eq. 16.29 and rearranging gives:

519
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

 n1 
 n  Vi
 2   
  1 1
  Ri  2  2 
 
R n R n 
 i o 
 n1 
 n 
Substituting into Eq. 16.22 for C2 and   , and simplifying gives:
 2 
 
 r 
Vi    n2 2

 Ro   Ro Ro 
n
v   2
 2
  n2 
n 
 Ri    Ro  n
 r R o 
    1
R R
 o   i  
 
r R
Again defining r*  and ri *  i , substituting and rearranging gives:
Ro Ro
 2
2 
  r *  n

i
 r * n
  r *  i
 r*
v  Vi  2 r *
 1  ri *n  i
 
 
v
With a non-dimensional velocity in terms of Vi as v  *  , our final expression is:
Vi
 2
2 

 i r *  n

 ri * n
v   r *   r*
v *   16.30
Vi  2 r *
 i
1  ri *n
 
 
Ri
Again, an alternative way of writing Eq. 16.30, is to let Ri  aR o , where ri *   a  1 , and
Ro
ro *  1 giving:
 2
2 
  a  n

a n
v   r *   r*
v *   16.31
Vi  2  a
 1 a n 
 
 

520
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

(a) (b)

Figure 16.6 Power-law Couette flow between concentric, rotating cylinders.

(a) Outer cylinder velocity = Vo, inner cylinder fixed, ri * = 0.2, 0.5, 0.8
(b) Outer cylinder fixed, inner cylinder velocity = Vi, ri * = 0.2, 0.5, 0.8
n = 0.2 and 0.5 are shear thinning; n = 2 and 4 are shear thickening.

Figure 16.6 is a graph showing the comparative behavior of: (a) only the outer cylinder rotating
(Eq. 16.25) and; (b) only the inner cylinder rotating (Eq. 16.30). The material behavior for

521
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

three different inner cylinder radii, ri *  0.2, 0.5, and 0.8 , and for a series of flow behavior
indices, n, from 0.2 (shear thinning) to 4 (shear thickening). The behavior of a Newtonian
fluid, n = 1, is shown for comparison. The cross-hatched bar on each figure represents the inner
cylinder wall.

Figure 16.6(a) for the rotating outer cylinder is pretty much as one would assume. As Eq.16.27
indicates, the highest shear stress occurs at the inner wall. For n < 1, shear thinning fluids, the
dv *
fluid will be less viscous near the inner wall, and thus the velocity gradient, , will be
dr
larger, and the velocity profiles will lie above the Newtonian, as shown. For n > 1, shear
thickening fluids, the fluid will more viscous near the inner wall, with correspondingly lower
velocity gradients, with velocity profiles that lag the Newtonian.

Figure 16.6(b) for the rotating inner cylinder shows that the velocity profile behavior is
reversed from figure 16.6(a). Since the shear thinning fluids are less viscous near the inner
surface, they again have correspondingly larger velocity gradients. However, since the inner
cylinder it rotating, the lower viscosity results in the velocity near the inner wall lagging
Newtonian behavior, yielding velocity profiles that also lag the Newtonian. The shear
thickening fluids are again more viscous near the inner surface due to higher shear stress, with
this more viscous region “pulling” the fluid ahead of the Newtonian behavior. This is most
clearly shown for ri *  0.2 and 0.5 , with the respective velocity profiles leading the comparable
Newtonian profile.

However, note that for the rotating inner cylinder with ri *  0.2 and n = 4 (figure 16.6(b), top-
right), v * exceeds 1 at r* = 0.25. This is unusual, since this implies that the fluid will move
faster than the wall rotation! The most likely explanation is that it is a failure of the power-law
model. Remember, back in sections 16.1.2 and 16.2.1 we mentioned that most shear
thickening fluids often don’t display consistent behavior over the entire shear stress range.
Therefore, some relaxation of the shear thickening behavior at the high stress end could change
this apparent unusual behavior. However, if the model holds, let’s consider if such behavior is
physically possible, and why.

Figure 16.7 shows a comparison of profiles of the velocity, v * , and the corresponding velocity
gradient, dv */dr * , for inner cylinder radii ri * = 0.5 and 0.2. Note that adjacent to the inner
cylinder wall, the gradients for the shear thinning fluids are strongly negative. However, the
gradients for the shear thickening fluids, are only weakly negative, and less so near the inner
cylinder wall. In addition, the velocity gradient for the case of ri *  0.2 and n = 4 becomes
positive for r* < 0.25, indicating that there is a maximum at r* = 0.25, where v *  Vi . If one
examines the relative magnitude of the shear stress across the fluid layer for n = 4, the ratio of
 
the shear stresses from the inner wall to the outer wall are inner  4 for ri *  0.5 , but inner  25
outer outer
for ri *  0.2 . Therefore, it appears that the much greater shear stress differential due to shear

522
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

thickening for the smaller inner radius would result in the fluid layers very adjacent to the inner
wall moving in a somewhat solid body manner, which produces the v *  Vi anomaly.

(a) (b)
Figure 16.7 Comparative power law Couette flows with rotating inner cylinder = Vi.
(a) Velocity profiles, v * vs. r * for ri * = 0.5 and 0.2.
(b) Corresponding velocity gradients, dv */dr * vs. r * for ri * = 0.5 and 0.2.
n = 0.2 and 0.5 are shear thinning; n = 2 and 4 are shear thickening.

This same process, although not as accentuated for ri *  0.5 and 0.8 in figure 16.6b is also what
produces velocity profiles in excess of the Newtonian as the higher shear inner region fluid pulls
along the lower shear outer region fluid.

To explore this anomaly a bit more, note that the derivative of the velocity from Eq. 16.31 is:

523
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

 2
2 
  n  2  a  n

 a n
dv * 1   n   r * 
 
  2  16.32
dr * a
 1 a n 
 
 
Setting Eq. 16.32 to zero, and solving for r*, we have:
n
 n  2 2
r*     r *max 16.33
 n 
Eq. 16.33 gives the location where v * will be a maximum, according to Eq. 16.31. Note that
for values of n < 2, there is no solution for r *max . For n >2, there is a solution, but the solutions
only become relevant when r *max  ri * . Note that for ri *  0.2 , r *max  0.25 , which what is
observed in figure 16.6b and 16.7.

Having made this observation, I must remind the reader that this is most likely an anomaly of the
power-law assumption, which may not hold at high shear rates. This also should only occur for
n values greater than 2, which are not all that common.

16.4.2.2 Poiseuille Flow in a Circular Duct

Decreasing pressure,
r

R
z


Figure 16.8 Geometry for Poiseuille flow of a power-law fluid in a circular duct

For a fully-developed Poiseuille flows in a circular duct, as shown in figure 16.8, there is no
motion in the r and θ-directions, but only in the streamwise z-direction, and the governing
equations reduce to the z-direction momentum equation (Eq. 16.109d of Section 16.7.3):

v z v v v v
 vr z   z  vz z
t r r  z
16.109d
1 p 1 1 
  gz   r rz   1 z   zz 
 z   r r r  z 
Since this is a steady, fully-developed flow, the velocity field is given by vz  f (r) , and
Eq. 16.109d reduces to:
p 1 
  r rz   0 16.34
z r r

524
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

p
For a Poiseuille flow,  constant (as shown in section 6.4.1), which allows Eq. 16.34 to be
z
rewritten as:
 p
 rrz   r 16.35
r z
Integrating Eq. 16.35, and equating it to the power-law shear stress from Section 16.4, Eq. 16.6,
gives:
n n 2
p r C1  v   v  p r
rz    K z   C1  Kr  z  
z 2 r  r   r  z 2
v z
One boundary condition will be symmetry at the centerline of the tube,  rz  0 , so  0 at r =
r
0, which renders C1 = 0, and:
n
p r  v 
 K z 
z 2  r 
v z
Solving for :
r
1
v z  p r  n
 
r  z 2K 
Integrating again,
1
n  p 1  n nn1
vz    r  C2
n  1  z 2K 
The second boundary condition is v z  0 at r = R, thus,
1 1
n  p 1  n nn1 n  p 1  n nn1
0   R  C2  C2     R
n  1  z 2K  n  1  z 2K 
Thus, vz becomes:

n 1 1
 n 1

nR n  p 1  n r
   1
n
vz    16.36
n 1  z 2K   R  
 
r
Defining r*  , and designating vz,n  vz  n  , since v z will be a function of n, we have:
R
1
nR  p R  n  nn1 
v z,n     r * 1 16.37a
n  1  z 2K   

From Eq. 16.37a at r* = 0  vz,n  vz,n,max  vz,n,CL :

1
nR  p R  n
v z,n,CL    16.37b
n  1  z 2K 

525
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

p
As I pointed out in Section 16.4.1.2, the pressure gradient, , must be positive to allow
z
1
the calculation of the power of , which makes vz,n and v z,n,CL negative velocities. We
n
v
again deal with this by defining v z,n *  z,n
v z,n,CL

v z,n
 n 1

v z *n   1  r * n  16.38
v z,n,CL  
We obtain the shear stress by differentiating Eq. 16.36 and substituting into Eq. 16.6:

1
 n 1 n
n 1
r 1 1
v z nR  p 1   n 
n
 p r  n
n
   n 1
 
r n  1  z 2K   z 2K 
R n

Thus, Eq. 16.6 gives:

n
 v   p r  p r
rz  K  z   K  
 r   z 2K  z 2
At the tube wall, r = R, the shear stress is:

p R
rz,wall 
z 2
So, the wall shear stress is only a function of the tube radius and the applied pressure gradient.
This is the same result we obtained in section 6.4.1, Eq. 6.64, for a Newtonian fluid. Thus, this
same relationship also applies for a non-Newtonian fluid at r = R.

Like a Poiseuille flow between flat plates, Section 16.4.1.2, we cannot compare the relative
velocities for different n values based on Eqs. 16.36 or 16.37a due to the functionality of the
power law. However, we can compare different power-law fluids at the same volume flowrate,
Q, as I did in section 16.4.1.2. We determine the flowrate by integrating Eq. 16.36 (substituting
for vz, CL) for 0  r  R :
r R
r R  r R
n 1
  2 3n 1

 r  n
 r n r n

Q   v z  2rdr   2v z,n,CL  1   
  rdr  2v z,n,CL 
  R    2 3n  1 n 1 
r 0 r 0
   R n  r 0

 n 1 
  R  v z,n,CL
2
Q 16.39
 3n  1 
Note that for a Newtonian fluid (n = 1), that:

1
Q Newt 
2
 R 2  v z,CL ,

526
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

This is the same relationship of flowrate to v z,CL that we determined in Chapter 6 (section 6.41,
Eqn. 6.62).

To allow a comparison between different power law fluids, we will use the above relationships
for velocity and flowrate, to establish how a particular power law fluid behaves relative to a
Newtonian behavior.

Substituting Eq.16.37b into Eq. 16.39, we have:

1
 n   p R  n
  R  R 
2
Q    16.40
 3n  1   z 2K 
1
 p R  n
Notice that the parameter R   in Eq. 16.40 is also common to the same parameter in the
 z 2K 
velocity equation, Eq. 16.37a. Solving for that parameter from Eq. 16.40 gives:

1
 p R  n Q
R  
 z 2K   n 
  R 
2

 3n  1 
Substituting into Eq. 16.37a for vz,n, gives:

 n  Q  nn1  Q  3n  1   n 1

v z,n     r * 1  2   1 r * n  16.41
 n  1   n  R 2  R  n  1  
   

 3n  1 
Using Eq. 16.41, we can now do a comparison of the velocity behavior for both pseudoplastic
(n < 1) and dilatant (n > 1) fluids to a Newtonian fluid (n = 1) at the same volume flowrates.

For Newtonian flows, n = 1:

2Q
v z, Newt  1  r *2 
2 
R
On the centerline, r* = 0, we have the maximum Newtonian velocity:

2Q
v z, Newt, CL  16.42
R 2
This expression is, of course, identical to the relationship we determined in Section 6.4.1,
Eq. 6.62.
vz,n
Dividing Eq. 16.41 by Eq.16.42, we calculate vz *n,rel  , the comparative velocity
vz,Newt, CL
profiles, for the same flowrate, Q, through a circular duct of radius R as:

vz, n 1  3n  1   n 1

vz *n,rel     1  r * n
 16.43
vz, Newt,CL 2  n 1   

527
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

(a) (b)

Figure 16.8 Power law velocity profiles for Poiseuille flow in a cylindrical duct.
(a) Normalized on the centerline velocity of each respective n value fluid;
(b) Normalized on Newtonian fluid centerline velocity for Q = constant.

Figure 16.8 shows the comparative effect of the fluid behavior index, n, on the velocity behavior
relative to a Newtonian fluid (n = 1). Figure 16.8(a) is plots of Eq. 16.38, with each n value
velocity profile normalized on its maximum centerline value. Figure 16.8(b) is plots of Eq.
16.43, with each n value velocity profile normalized on the Newtonian centerline velocity for the
same flowrate. Like the Poiseuille flows between parallel plates examined in Section 16.4.1.2,
figure 16.8(a) shows that shear-thinning pseudoplastics (n < 1) display much fuller profiles than
a Newtonian flow, while shear-thickening dilatants (n > 1) display narrower profiles. For
equivalent flowrates, Q, figure 16.8(b) shows that the centerline velocities for pseudoplastics will
lag a Newtonian fluid, and dilatants will exceed a Newtonian fluid. Moreover, as I commented
on in Section 16.4.1.2 on parallel flat plate flows, the velocity behavior for smaller n values (n 
0.5 or less) often appear similar to that of a Bingham plastic.

A comparison of figure 16.8 with the similar profiles from figure 16.4 show that figures 16.8a
and 16.4a are identical, since they are both are functionally the same. The relative velocity
profiles of figures 16.8b and 16.4b, normalized on the Newtonian centerline velocity, are similar,
but show a wider spread in the various n-profiles. This is due to a more significant difference in
the centerline velocity to average velocity for Poiseuille circular duct flows than for Poiseuille
parallel plate flows, due to the radial symmetry of the duct flows.

p
Again, like in section 16.4.1.2, we can’t do a relative comparison of for different power law
z
fluids, without calculating the values for each specific fluid.

528
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

16.5 Ideal Bingham Plastic Solutions

An ideal Bingham Plastic, as modeled by Eq. 16.3, requires that a critical stress, termed the yield
stress, be reached before the material will start to flow as a fluid. Prior to reaching that yield
stress, the material will act as a solid, and remain fixed, or move in unison with the surrounding
material. Thus, an ideal Bingham Plastic is modeled by a two-region equation. As was indicated
in Eq. 16.3, the two regions are generically related to the strain rate as:

0 when   0 , and

  0  B    when   0 16.3

Where  is the generic strain rate, 0 is the yield shear stress, and  B is the apparent viscosity
when the material flows as a fluid. The three particular model variations we will use in this
section are:

Non-Newtonian, Cartesian
u
x-direction only:  yx  0, when  yx   0 (assuming 0 >0) 16.44a
y
and
 u 
 yx   B    0 , when  yx   0 16.44b
 y 
Non-Newtonian, Cylindrical
 v   v 
θ direction only:  r  r     0      0 , when  r  0 16.45a
r  r  r  r 
and
   v 
r   B  r      0 , when r  0 16.45b
 r  r  

 v  v z
z-direction only:  rz   z   0  0, when rz  0 16.46a
 r  r
and
 v 
rz   B  z   o , when rz  0 16.46b
 r 

16.5.1 Parallel Plate Flows

For fully-developed Couette and Poiseuille flows between infinite parallel flat plates, there is no
motion in the lateral directions, but only in the streamwise direction, and (like Section 16.4.1) the
governing equations again reduce to the x-direction momentum equation (Eq. 16.108b of Section
16.7.2).

529
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

u u u u 1 p 1    yx  zx 
u v w   g x   xx    16.108b
t x y z  x   x y z 
Again, for steady, fully-developed flow in x-direction, Eq. 16.108b simplifies to:

 yx p
 16.47
y x

16.5.1.1 Couette Flow between Parallel Plates

Using the coordinate system shown in figure 16.9, we again specify boundary conditions of u = 0
at y = 0, and u = U at y = h.

y U
h
x

Figure 16.9 Couette flow between a lower stationary plate and an upper moving plate.

Since there is no pressure gradient for a Couette flow (see the detailed discussion of this in
section 6.3.1), Eq. 16.47 further reduces to:

 yx
p
0
 16.48
y x
Integrating Eq. 16.48 once:

 yx  C1
Now if  yx  0 (assuming 0 > 0), τ yx  constant  τ 0 , and thus Eq. 16.44a applies:

u
 0  u  C2
y
But since u = 0 at y  0 , then u = 0 across the entire gap when  yx  0 for 0  y  h .

Note that when  yx  0 across the entire gap, the upper plate cannot move. So, the boundary
condition of u = U at y = h cannot be met, and u = 0 across the entire channel, and the material
remains fixed until yx  0 .
 u 
When yx  0 , Eq. 16.44b applies,  yx   B    0 . So, integrating Eq.16.48 again, and
 y 
equating  yx to Eq. 16.44b:

530
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

 u 
 yx   B    0  C1
 y 
Rearranging,

u 1
  C1  0  16.49
y  B
Integrating Eq. 16.49,

1
u  C1  0  y  C2 16.50
B
Applying the boundary conditions that u = 0 at y = 0, and u = U at y = h:

1
0  C1  0  0  C2  C2  0
B

1
U  C1  0  h
B

B U
C1   0
h
Substituting C1 and C2 back into Eq. 16.50:
U
1  B U y

u   0  0  y
B  h  x

Uy
u 16.51
h
Thus, as Eq.16.51 shows, the velocity profile is again linear across the gap between the plates,
which is the same as for a Newtonian flow and for a power-law fluid, as shown in Section
6.4.1.1, Eq. 16.10. Note that the shear stress across the gap is:

 u   U
 yx   B    0  B  0  constant
 y  h
Thus, the shear stress across the gap is constant, as is the strain rate. Therefore, in parallel plate
Couette flow of an indeal Bingham plastic, the material will remain static until the yield shear
stress is reached, and then all the material will begin to flow with a linear velocity profile, with a
constant shear stress across the material.

16.5.1.2 Poiseuille Flow between Parallel Plates

Similar to the manner we approached the solution for a power law fluid in Section 16.4.1.2, here
we again employ a coordinate system with the origin located on the midplane of the channel, and
again recognize that the velocity profile will be symmetric about the midplane, as depicted in
figure 16.10. Employing this midplane coordinate system, the boundary conditions for the flow

531
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

u
in the upper half of the flow are  0 for 0  y  y0 (the central plug flow), and u = 0 at
y
h
y . I discuss how we establish the first boundary condition below.
2

Decreasing pressure,
y

y0 h/2
x h

Figure 16.10 Poiseuille flow between parallel flat plates, ideal Bingham plastic.

p
For a Poiseuille flow,  constant (as shown in section 6.3.2). So, we rearrange Eq. 16.47 to
x
integrate for the shear stress.

 yx
p
= constant
 16.52
y x
Integrating Eq. 16.52 once gives:

p
 yx  y  C1 16.53
x
p
Note that Eq.16.53 indicates that the shear stress will vary linearly (since  constant ) , with
x
the highest shear at the plate surface, and decreasing linearly toward the midplane. When the
h  
shear on the upper plate ( y  ) exceeds the yield shear stress  i.e.  h  0  , the fluid
2  yx,y 
2 
adjacent to the wall will behave as a fluid, as long as yx  0 . However, Eq. 16.53 indicates the
shear stress within the fluid region decreases as we move toward the midplane. Potentially, we
will reach a location within the material where  yx   0 (assuming 0 > 0) and fluid behavior will
cease. Since we don’t know where that location is, we arbitrarily designate that location as the
h
yield location, y0 , where 0  y 0  , which I show as a red dotted line in Figure 16.10. Within
2
h
the region adjacent to the wall, y 0  y  , we assume that yx  0 , so the material will behave
2
as a fluid. However, within the region adjacent to the midplane, 0  y  y0 , yx  0 , and the
material will behave as solid. In this latter region, the pressure gradient will cause the material to
move as a solid plug flow, with the material moving at a constant velocity equivalent to the
velocity at y  y0 . To asses this solid region, I apply a force balance over the solid portion of the

532
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

material of arbitrary length, L, as shown in figure 6.11. Note that since the solid plug will move
uniformly, the shear stress within the plug, and at the centerline, y = 0, will be zero, as indicated.

Figure 16.11 Forces acting on the solid region material, 0  y  y0 , of arbitrary length L.

Performing a force balance on the solid material (assuming a unit depth), we have a balance
between the shear stresses and pressure forces acting on the material:

0L   0 L  p1 (y0  0)  p2 (y0  0)  0

0 
 p1  p2  y
p
y0 
0 16.54
L x
In Eq. 16.54, we note that the pressure differences,  p1  p2  , divided by the arbitrary length, L,
is simply the pressure gradient, and the shear stresses will act in a negative direction, with
yx  0 at y  y0 , and yx  0 at y  0 .

Solving Eq. 16.54 for y 0 gives:

0
y0  16.55
 p 
 
 x 
Equation 16.55 indicates that the yield location y0 , which is the bounding location between fluid
and solid behavior, is controlled by two parameters: the yield shear stress and the pressure
h
gradient. Since y0 cannot be greater than , the pressure gradient must exceed the yield stress
2
in order for the material to move. After the pressure gradient exceeds the yield stress, continued
pressure gradient increases will cause y0 to diminish, with the region of fluid behavior
h
y 0  y  expanding outward from the plate surface toward the midplane. We will see the
2
impact of an increasing pressure gradient, and a consequent reduction in y0 , later in this section.

Therefore, a Poiseuille flow of an ideal Bingham plastic will have two regions of behavior:

(1) for 0  y  y0 , yx  0 , the material will behave as a solid plug flow at u = constant,
h
(2) for y 0  y  ,   0 , the material will behave as a fluid with u = u(y).
2 yx

533
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

h
Within the fluid region y 0  y  , and  yx  0 , Eq. 16.53 applies:
2
p
 yx y  C1
x
Applying a boundary condition that yx  0 at y  y0 , we get:
p p
0  y0  C1  C1  0  y0
x x
So,
p
 yx   y  y 0   0
x
Substituting the governing shear stress relationship for the fluid region, Eq. 16.44b:

 u  p
 yx   B    0   y  y0   0
 y  x
Which rearranges to:
u 1 p
  y  y0 
y  B x
Integrating, gives:

1 p  y 2 
u   y0 y   C2 16.56
 B x  2 
h
Applying the boundary condition that u = 0 at y  , we determine C2 as:
2
1 p  h 2 h
C2     y0 
 B x  8 2
Substituting C2 into Eq.16.56 and rearranging gives:

1 p  2 h 2  y0 p h
u y    h  2y  for y 0  y  16.57
2 B x  4  2 B x 2
2y 20
Defining a non-dimensional y variable as y*  , and y0 *  , then u y0 *  u  y0 * , since
h  p 
h 
 x 
the u velocity profile will be a function of y0 * , as well as y*, we rewrite Eq. 16.57 as:
h 2 p
u yo *   y *2 1  2y 0 * 1  y *  for y0 *  y*  1 16.58a
8 B x  
To determine the maximum velocity, which will be the velocity of the material “plug” in the
center of the channel, I let u yo *  u yo *,max at y*  y0 * in Eq. 16.58a:
h 2 p
u yo *,max   y0 *2 1  2y0 * 1  y 0 * 
8 B x  
Which reduces to:

534
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

h 2 p
 y0 * 1
2
u yo *,max   for 0  y*  y0 * 16.58b
8 B x
p
Recall that  0 for a flow in the positive x direction, thus Eq.16.58 will yield positive
x
velocities for a pressure decreasing in the x direction.
u y0 *
Defining u yo * *  gives:
u y0 *,max

u y0 * 1  y*2   2y*0  y* 1

u *y0 *    for y0 *  y*  1 16.59a
 y0 * 1
2
u y0 *,max
and
u *y0 *,max  1 for 0  y*  y0 * 16.59

Figure 16.12 Ideal Bingham plastic Poiseuille flow between parallel flat plates. Normalized
20
on maximum velocity for respective values of the yield location, y0 *  .
 p 
h 
 x 
Equation 16.59 reflects how an ideal Bingham fluid velocity profile varies relative to its
maximum centerline (or plug velocity). Figure 16.12 is a series of velocity profiles of u yo * * vs.
y* calculated from Eq. 16.59 for a range of y0 * values, 0  y0 *  1 .

535
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

As figure 16.12 shows, for a yield location near the wall ( y0 *  1 ), the material profile is
essentially a flat plug flow. However, as the yield location moves toward the midplane
(i.e. the pressure gradient increases), the profile will display a wider region of fluid-like
behavior, and the region of plug flow will decrease. Potentially, if the pressure gradient gets
p 20
quite large, , the yield location will become quite small, and the flow will approach
x h
Newtonian behavior ( y0 *  0 ).

While Figure 16.12 is instructive regarding the shape of the velocity profiles, it does not reveal
the impact of the yield location on the behavior of the material relative to a comparable
Newtonian fluid. Unlike the power-law fluid we examined in section 16.4.1.2, we can compare
the behavior of an ideal Bingham fluid to Newtonian fluid, assuming the same pressure gradient
and fluid viscosity.
h 2 p
A Newtonian fluid will have 0 = 0, and thus y 0 *  0 . Equation 16.58b gives u Newt,max   ,
8 x
which is the same as Eq. 6.19 we derived in Section 6.3.2. If we assume equivalent pressure
gradients and fluid viscosities, B   (not probable, but instructive), we can determine velocity
profiles for a range of Bingham fluids relative to a comparable Newtonian fluid. Thus, dividing
Eq. 16.58a for u yo * by u Newt,max , we obtain:

u y0 *
u *rel   1  y*2   2y0 *  y* 1  for y0 *  y*  1 16.60a
u Newt,max
and
u y0 *,max
u *rel,max   1  y0 *2  for 0  y*  y0 * 16.60b
u Newt,max

Equation 16.60 gives the velocity profile behavior relative to a pure Newtonian fluid, with the
same pressure gradient and fluid viscosity. Figure 16.13 shows series of velocity profiles of
u *rel vs. y* plotted for a range of y0 * values, 0  y0 *  1 .

Figure 16.13 illustrates that the flow of material does not initiate if the pressure gradient is less
p 20
than the yield stress,   y0 *  1 ( y0 *  1 ,the green line, reflects all cases for y0 *  1 ).
x h
As the pressure gradient increases relative to the yield stress, 0 , the yield location, y0 * , shifts
toward the midplane. This results in the expansion of the region of fluid-like behavior, with a
commensurate increase in the maximum (plug) velocity. This correspondingly reduces the extent
of the solid, plug-flow region, and increases the volume flowrate.

536
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

Figure 16.13 Ideal Bingham plastic Poiseuille flow between parallel flat plates for a series
of yield locations, y0 * ; normalized on the maximum (centerline) velocity for
a Newtonian fluid with the same pressure gradient and fluid viscosity.
20
y0 * 
 p 
h 
 x 

To understand the effect of y0 * on both the volume flowrate and the pressure gradient, I integrate
Eq. 16.58 across the duct to determine the volume flowrate, Q.

h h
Since the flow is symmetric about the x-axis, the volume flowrate for   y  will just be
2 2
h h
double the flowrate for 0  y  . Using y  y *   , and an arbitrary plate width W, we have,
2 2
using Eq. 16.58 (this requires a two part integral of both the fluid and solid material regions):

h
y
y*1 y* y0 * y*1
2
h
Q yo *  2 
0
u yo * Wdy  2W 
0
u yo * dy*  hW
2 
y* 0
u yo *,max dy*  hW 
y* y0 *
u yo *dy*

537
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

y* y0 * y*1
Wh 3 p Wh 3 p
Q yo *     y0 * 1 dy* 
2
  y*2 1  2y0 * 1  y* dy*
8B x y*0
8B x y* y0 *  

y*1
Wh 3 p  y* y0 * Wh 3 p  y*3   y*2  
 0   y*0
2
Q yo *   y * 1 y*    y*   2y *
0  y*  
8B x  8B x  3   2   y* y *
0

Which reduces to:

Wh 3 p
Q yo *  
24 B x
 y0 *3 3y0 * 2  16.61

As a check on Eq. 16.61, we note that when y 0 *  0 , the fluid would behave as a Newtonian
fluid, B   , so:

Wh 3 p
Q Newt   16.62
12 x
Equation 16.62 is the same as Eqn. 6.23 for the volume flowrate of a Poiseuille flow derived in
Section 6.3.2.

To compare the respective flow rates, we take the ratio of Q yo * to QNewt :

Q y0 * 1
Q y0 *,rel 
Q Newt

2
 y0 *3 3y 0 * 2  16.63

Equation 16.63 gives the volume flowrate for an ideal Bingham plastic relative to a Newtonian
fluid with identical pressure gradient and fluid viscosity, B   . Figure 16.14(a) is a plot of
Q y0 *,rel vs. y0 * for 0  y0 *  1 .

Note that unlike the power-law solutions, the Bingham plastic solutions are linear with the
pressure gradient. So, solving Eq. 16.61 for the pressure gradient, and letting Q  Q y0 * , we have:

p 24BQ  1 
   16.64
x y0 * Wh   y0 *3 3y0 * 2  
3
 
For a Newtonian fluid, y 0 *  0 , Eq. 16.64 gives:

p 12Q
 16.65
x Newt Wh 3
To compare the respective pressure gradients at the same volume flowrates, we again assume
identical fluid viscosities, B   , and take the ratio of Eq. 16.64 to Eq. 16.65:

p
p x y0 * 2
  16.66
x rel p  y0 * 3y0 * 2 
3

x Newt

538
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

(a) (b)
Figure 16.14 Comparisons of ideal Bingham plastic volume flowrate and pressure gradient
as a function of y0 * -- Poiseuille flat plate flow.
(a) Q y0 *,rel , with identical pressure gradient and fluid viscosity;
(b)  p x  rel , with identical volume flowrate and fluid viscosity.
20
y0 * 
h  p x 

p
In Eq. 16.66, is the relative increase in pressure gradient required to maintain the same
x rel
volume flowrate as a comparable Newtonian flow with the same fluid viscosity. Figure 16.14(b)
p
is a plot of vs. y0 * for 0  y0 *  1 .
x rel

Figure 16.14(a) shows that the flowrate of an ideal Bingham plastic will always be less than that
of a comparable Newtonian fluid with the same pressure gradient and fluid viscosity. The thinner
the fluid region (larger y0 * ), the lower the flowrate. Why is this so? If we do a force balance for
h h
a control volume of length L and 0  y  we can show that the wall shear stress at y  is:
2 2

539
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

p h
wall 
x 2
The governing shear stress equation for an ideal Bingham plastic, Eq. 16.44b, is:

u
 yx   B  0 , when  yx   0
y
h
So, at y 
2
p h u
wall   B  0
x 2 y wall
Rearranging we have:

u 1  p h  p h
   0   1  y0 * 16.67
y wall B  x 2  x 2 B
p h
(Note: this only holds if  0 and y0 *  1 )
x 2
h
What Eq. 16.67 shows is that the slope of the velocity profile at the wall ( y  ) will be
2
u p h
 0 (no flow) until  0 or y0 *  1 . If the pressure gradient increases beyond 0 ,
y wall
x 2
u
thus reducing y0 * , the slope of the velocity at the plate surface will increase,  0 , and
y wall
the material (solid and fluid portions) will begin to move. The higher the pressure gradient
relative to 0 , the larger the region of fluid behavior, and the faster the collective material
p h
moves. The key is the that shear stress at the wall is always the same, wall  . So, the
x 2
larger the yield stress, 0 , the more the pressure gradient has to work to overcome 0 , and the
less work it can do to move the material.

Likewise, figure 16.14(b) illustrates that the thinner the fluid region (larger y0 * ), the higher the
1
relative pressure gradient must be to maintain the same flow rate. For example, if y0 *  , such
3
that 66% of the material behaves as a fluid, the pressure gradient is double that of a Newtonian
fluid. By y0 *  0.6 , where only 40% of the material behaves as a fluid, the required pressure
gradient grows to 5 times the Newtonian. As figure 16.14(b) shows, when y0 *  0.7 the relative
pressure gradient required to maintain the same flowrate is literally off the chart.

Also, since the power required to move the material will be roughly the pressure gradient times
 p 
the volume flowrate  i.e. Power  Q  , the pumping power required to maintain a constant
 x 
flowrate will behave essentially the same as the pressure gradient in figure 16.14(b).

540
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

16.5.2 Flows with Circular Symmetry

In this section, I address Couette flow and Poiseuille flow of an ideal Bingham plastic fluid in
geometries with circular symmetry. Like the parallel plate flows of section 16.5.1, these flows
will reduce to one-dimensional, fully-developed flows with no lateral cross flow. Also, like the
power-law flows I covered in section 16.4.2, different governing momentum equations will
apply for each of these flows. The Couette flow will depend on the azimuthal (θ) direction
equation and yield v  f (r) , whereas the Poiseuille flow will depend on the axial (z) direction
equation and yield vz  f (r) . The two flows will also depend on different constitutive equations
for the shear stress, as indicated in Section 16.5.

16.5.2.1 Couette Flow between Concentric, Rotating Cylinders

Vo
Vi
r

Ro
Ri

Figure 6.15 Geometry for Couette flow between concentric rotating cylinders

As discussed in Section 6.4.2, the governing equation for this type of cylindrical flow reduces to
the θ-direction momentum equation, 16.109c of Section 16.7.3.

v  v v v v vv
 vr      vz   r 
t r r  z r
16.109c
1 p 11  2 1    z  r   r 

r 
 g   2
  r r
r  r  
r 

z
 
r 

Since this is a steady, fully developed flow, the velocity field is given by v  f (r) , and Eq.
16.109c reduces to:

1 2 p
r r
 r r  

16.68
p
I showed in section 6.4.4 that for Couette flow,  0 , so Eq. 16.68 reduces to:

 2
r

r  r  0 
Integrating, we have:

C1
 r  16.69
r2

541
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

However, in order to integrate Eq. 16.69 further, we need to assess and specify the appropriate
boundary conditions.

Boundary Conditions (a): Outer Cylinder Rotating at velocity Vo, Inner Cylinder Fixed

In the case of rotating, concentric cylinders, with an ideal Bingham flow, we cannot apriori
specify the velocity of rotation, Vo, since when  r  0 across the entire gap between the
cylinders, R i  r  R o , there can be no motion, whereas when r  0 over some portion of the
gap, the outer cylinder can rotate. So, how do we set a boundary condition for this problem?

What we do know, from our previous examples of Couette-type flows is that the torque across
the gap will always be a constant (see Sections 6.4.2, Eq. 6.68 and 16.4.2.1, Eq. 16.29). So, if
we let toque = T= constant, then at any r location the torque for a depth of L is given by:

T  r  2 rL r  2Lr 2r

Such that
T
 r  16.70
2Lr 2
For an ideal Bingham plastic, the θ-direction, two-region equations (Eq. 16.45) for the shear
stress are:
 v   v 
 r  r     0      0 , when  r  0 16.45a
r  r  r  r 
and
   v 
r   B  r      0 , when r  0 16.45b
 r  r  
When  r  0 across the entire cylinder gap, the equation that applies is 16.45a, which integrates
to:
v
 C1
r
However, the boundary condition for the fixed inner cylinder at r  R i is v  0 , which will
make v  0 across the gap, since the torque will be insufficient to deform the fluid, and the
fluid will remain stationary.

Equation 16.70 indicates that the highest shear stress will occur at R i (the smallest radius), so
T
the first location that the material will reach r  0 is r  0  at r  R i . However, as
2LR i2
the torque is further increased, the shear stress at R i will increase, and the location where
r  0 will move away from the inner surface and out into the material. We will call this
location where r  0 the yield radius, and designate it by r0 . From Eq. 16.70, we can write:

542
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

T
 r  2
 0
2Lr0
Solving for r0 in terms of the torque we have:

T
r0  16.71
2L 0
Equation 16.71 will be the outer limit at which the Bingham plastic will behave like a fluid, with
fluid behavior spanning a region R i  r  r0 .

Now, within this region R i  r  r0 , the shear stress abides by the Bingham fluid equation,
Eq.16.45b, such that:

   v 
r   B  r      0
 r  r  
or
  v 
Br   r  0
r  r 
From Eqs. 16.70 and 16.71, we substitute for the shear stress values to give:

  v  T T
Br   
r  r  2Lr 2Lr02
2

or
  v  T  r02 1 
     16.72
r  r  2 B Lr02  r 3 r 
Integrating Eq. 16.72, gives:

v T  r02 
 2 
 2  ln  r    C2
r 2 B Lr0  2r 
Employing the boundary condition that v   0 at r  R i , we have:

T  r02 
C2  2  2
 ln  R i  
2 B Lr0  2R i 
Substituting C2 and simplifying:

v T 1  1 1   1   Ri 
   2  2    2  ln   16.73
r 2 B L  2  R i r   r0   r 
To non-dimensionalize Eq. 16.73, we first let R o  R and R i  aR , were a < 1, then we define
r
r*  . Thus, substituting r  r *R , R i  aR , and r0  r0 * R , into Eq. 16.73 and simplifying,
R
we get:

543
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

T 1  1 1  1  a 
v    2  2   2 ln    r * for a  r*  r0 * 16.74
2 B LR  2  a r *  r0 *  r *  
What about the region that is between r0 and Ro = R, where  r  0 ? From Eq. 16.45a we have:
  v 
  =0
r  r 
Integrating gives:

v  C3r
For the case where at least some portion of the material behaves as a fluid adjacent to R i , the
outer cylinder can rotate, and we have a boundary condition of v   Vo at r = Ro = R (even
V
though we don’t know the value of Vo ). Thus, C3  o , and
R
Vr
v   o  Vo r * for r0 *  r*  1 16.75a
R
We note that Vo , since it cannot be specified apriori (since we have specified the torque, T), is
determined by equating the two velocities from the outer and inner regions (Eqs. 16.75a and
16.74) at r*  r0 * :

T 1  1 1  1  a 
Vo r0 *    2  2  2
ln    r0 *
2 B LR  2  a r0 *  r0 *  r0 *  

T 1  1 1  1  a 
Vo    2  2  ln   16.75b
2 B LR  2  a r0 *  r0 *2  r0 *  

Substituting for Vo in Eqs. 16.74 and 16.75a, the two-region velocity equations are:

T 1  1 1  1  a 
v  Vo r*    2   ln   r * for r0 *  r*  1 16.76a
2 B LR  2  a r0 *2  r0 *2  r0 *  
and
T 1  1 1  1  a 
v    2  2   2 ln    r * for a  r*  r0 * 16.76b
2 B LR  2  a r *  r0 *  r *  

Notice that Eq. 16.76a is simply a solid body rotation of the solid portion of the material.

However, what if r  0 all across the entire gap to r = R, or r* = 1? Then r0 *  1 , and only
Eq. 16.76b applies across the whole gap from aR  r  R , or a  r*  1 , so that:

T 1  1 1   a 
v    2  2   ln    r * for a  r*  1 , and r0 *  1 16.77a
2 B LR  2  a r*   r * 
and

544
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

T 1  1  
Vo    2  1  ln  a   at r*  1 16.77b
2 B LR  2  a  

T
Now, what if r0 *  2
 1 , which can certainly happen if T  2LR 2  0 ? For this case,
2LR  0
we need to re-examine our determination of Vo. Here, we note that r* cannot exceed 1 (r = R),
but r0 * can exceed 1. So, revisiting our boundary conditions for determining Vo, we have
v  Vo at r* = 1, which Eq. 16.74 gives as:

T 1  1  1 
Vo    2  1  2 ln  a   16.78a
2 B LR  2  a  r0 * 

Note that the value of r0 * is a direct function of the torque, T, so Vo will increase for r0 *  1 , and
the equation for the velocity distribution for a  r*  1 will be:

T 1  1 1  1  a 
v    2  2   2 ln    r * for a  r*  1 , r0 *  1 16.78b
2 B LR  2  a r *  r0 *  r *  

Since we can solve for Vo from Eqs. 16.75b, 16.77b, and 16.78a, we could non-dimensionalize
v
v  as v  *   , like we did in Section 16.4.2.1. However, here Vo is a function of r0 * , and is
Vo
not a consistent value as it was for the power-law solution. So, we reexamine our expression for
r0 * noting that from Eq. 16.71:

T
r0 * 2 
2LR 2  0
Substituting r0 * 2 into Eq. 16.74, we get:

T 1 1   R0   a 
v   2  2 r*  ln   r * 16.79
4 B LR  a r *   B   r * 
If 0  0 in Eq. 16.79 we have a Newtonian flow, with velocity:

T  1 1 
v , Newt   2  2 r * 16.80a
4LR  a r* 
At r* = 1, Eq.16.80a gives v, Newt 1  Vo, Newt of:

T 1 a 2 
Vo , Newt    16.80b
4LR  a 2 

545
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

So, to keep perspective on the influence of 0 on the outer wall velocity, I normalize our v 
v
velocities for Eqs. 16.76, 16.77, and 16.78 on Vo , Newt , i.e. v  *  . This gives us velocity
Vo , Newt
profiles that compare ideal Bingham Couette behavior to comparable Newtonian behavior
(assuming Bingham   Newtonian for the fluid behavior) at the same applied torque. The three
possible flow cases for a rotating outer cylinder are:

Case 1(a):

When  r  0 across the entire Couette gap, then the material behavior across the whole gap is
governed by the inner radius, r = aR, r* = a, which is where the material will initially begin to
T
flow. Thus, if the yield radius is less than the inner radius, i.e. r0   aR , the applied
2L 0
torque will be less than that required to initiate fluid behavior, T  2La 2 R 2 0 , and the material
will remain stationary under the applied torque, and the outer cylinder cannot move. So,

r  0 , for a  r*  1 v *  0 and Vo *  0

Case 2(a):

When T  2La 2 R 2  0 and  r  0 occurs within the Couette gap at r0 , where aR  r0  R

 a  r0 *  1 , the regions of behavior will be:
a2  1 1  2  a 
r  0 , for a  r*  r0 * v *   2  2   ln  r * 16.81a

1 a 2   a r *  r0 *  r * 
2

and
a 2  1 1  2  a 
r  0 , for r0 *  r*  1 : v *   2  ln  r* 16.81b
1  a   a r0 *  r0 *  r0 *  
2  2 2

a2  1 1  2  a 
with Vo *   2   ln  16.81c

1 a 2   a r0 *  r0 *  r0 * 
2  2

Note that when r0 *  1 , only Eq.16.81a is required across the Couette gap.

Case 3(a):

T
When r0 *  2
 1 , which occurs when T  2LR 2  0 , r* cannot exceed 1
2LR  0

546
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

(r = R), but r0 * may exceed 1. Thus, Vo * is the value of v * at r* = 1, rather than at r0 * = 1,

which yields the equations:

a2  1 1  2  a 
v *   2  2   ln  r * for a  r*  1 , and 1  r0 * 16.82a

1 a 2   a r *  r0 *  r * 
2

with
2 a2
Vo *  1  lna  16.82b

r0 * 2 1  a 2 
Note that when r0 * 1 , then Eq. 16.82 collapses to the Newtonian solution, as it should.

Velocity profile behavior calculated from Eqs. 16.81 and 16.82 appear in figure 16.16(a).

Boundary Conditions (b): Outer Cylinder Fixed, Inner Cylinder Rotating at velocity Vi

For the case of inner cylinder rotation, the derivation of the velocity behavior of an ideal
Bingham plastic is similar to the case for the outer cylinder rotation.

When  r  0 across the entire cylinder gap, the material cannot flow and v  0 across the
Couette gap. The material will remain fixed, until r  0 , which will again initiate fluid
behavior at r  R i .

When r  0 , and a portion of the material begins to behave as a fluid, the equation derivation
is the same through Eq. 16.72.

  v  T  r02 1 
      16.72
r  r  2 B Lr02  r 3 r 
Eq. 16.72 again integrates to:

v T  r02 
 2 
 2  ln  r    C2
r 2Lr0  2r 
However, here the appropriate boundary condition is again at r  R i , where the material first
behaves as a fluid, and where v  Vi . Note that we cannot use a boundary condition of v  0 at
r  R o , since that condition will lie within the region where the material behaves as a solid. To
help simplify the derivation process, we will again let R o  R and R i  aR , where a < 1.
Employing the boundary condition that v  Vi at r  R i  aR , we have:

T  r02  V
C2  2  2 2
 ln  aR    i
2 B Lr0  2a R  aR
Substituting C2 and simplifying:

547
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

v T  1  r02 r02   aR   Vi
   2 2  2 
 ln    16.83
r 2 B Lr02
 2  a R r   r   aR
r
I non-dimensionalize Eq. 16.83, again letting r*  . Thus, substituting r  r * R and
R
r0  r0 * R into Eq. 16.83 and simplifying:

T 1  1 1  1  a  Vi r *
v    2  2   2 ln    r *  for a  r*  r0 * 16.84
2 B LR  2  a r *  r0 *  r *   a
Again, we cannot specify Vi apriori (since we have again specified the torque, T). We
determine the value of Vi by noting that within the region r0 *  r*  1 the material behaves as a
solid, with v   0 . Therefore, letting v  0 at r*  r0 * in Eq. 16.84 and solving for Vi :

T 1  1 1  1  a 
Vi     2 2 
 2
ln   a
2 B LR  2  a r0 *  r0 *  r0 *  
Substituting Vi back into Eq. 16.84 and simplifying, we have the two-region velocity equations:

T 1  1 1  1  r0 *  
v    2  2  ln   r * for a  r*  r0 * 16.85a
2 B LR  2  r0 * r *  r0 *2  r *  
and
v  0 for r0 *  r*  1 16.85b
with
T 1  1 1 1  r * 
Vi    2  2  2
ln  0   a at r* = a 16.85c
2 B LR  2  r0 * a  r0 *  a  

T
When r0 *   1 , we need to again re-examine the determination of Vi. Again we note
2LR 2 0
that r* cannot exceed 1 , but r0 * can exceed 1. Revisiting our boundary conditions for
determining Vi, we have v   0 at r* = 1, rather than at r0 *  1 . Substituting into Eq. 16.84 and
solving for Vi :

T 1  1 1  1 
Vi   1  2   2 ln    a 16.86a
2 B LR  2  a  r0 *  a  
Substituting Vi back into Eq. 16.84 and simplifying, we have the velocity equation for r0 *  1 .
T 1  1  1  1 
v   1  2   2 ln    r * for a  r*  1 , r0 *  1 16.86b
2 B LR  2  r *  r0 *  r *  

548
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

I will again normalize v on the inner wall velocity for a Newtonian flow. If we let 0  0 , in
Eq. 16.86b this gives the equation for a Newtonian flow. When 0  0 , then r0 *   , and Eq.
16.86 reduces to:
T  1  
v , Newt   1  2 r * for a  r*  1 16.87a
4LR  r *  
with
T  a 2 1 
Vi, Newt    16.87b
4LR  a 
Similar to the rotating outer cylinder, here I normalize our v  velocity from Eq. 16.85 and 16.86
v
on Vi , Newt , i.e. v  *  . Again, this compares ideal Bingham Couette behavior to the
Vi , Newt
comparable Newtonian behavior (again letting Bingham   Newtonian ) at the same applied torque.
The possible flow cases for a rotating inner cylinder are:

Case 1(b):

When  r  0 across the entire Couette gap, the material behavior across the whole gap is again
governed by the inner radius, r* = a, which is again where the material will initially begin to
T
behave as a fluid. Thus, if the yield radius is less than the inner radius, i.e. r0   aR ,
2L 0
the applied torque will again be less than that required to initiate fluid behavior, T  2La 2 R 2 0 ,
and the material will remain stationary under the applied torque, and the inner cylinder cannot
move. So,

r  0 , for a  r*  1 v *  0 and Vi *  0

Case 2(b):

When T  2La 2 R 2  0 and  r  0 occurs within the Couette gap at r0 , where aR  r0  R

a  r0 *  1 , the regions of behavior will be:
 a   1 1  2  r0 * 
r  0 , for a  r*  r0 * v  *   2      ln  r * 16.88a
 a  1   r0 *
2
r * 2  r0 * 2  r * 
and
r  0 , for r *0  r*  1 v *  0 16.88b
 a   1 1  2  r * 
with Vi *   2    2   ln 0  a 16.88c
 a  1   r0 * a  r0 *  a 
2 2

549
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

(a) (b)

Figure 16.16 Ideal Bingham plastic Couette flow between concentric, rotating cylinders.
(a) Outer cylinder velocity = Vo, inner cylinder fixed. ri * = a = 0.2, 0.5, 0.8
(b) Outer cylinder fixed, inner cylinder velocity = Vi. ri * = a = 0.2, 0.5, 0.8
T
r0 * 
2LR 2 0

550
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

Case 3(b):

T
When r0 *  2
 1 , which happens if T  2LR 2 0 . In this case, r* cannot exceed 1
2LR 0
(r = R), but r0 * can exceed 1. Therefore, we require that v *  0 at r* = 1, rather than at r0 * =1,
which yields the equations:

 a   1  2  1 
v  *   2  1  2   ln  r * for a  r*  1 16.89a
 a  1   r *  r0 *  r * 
2

with
2  a2 
V *i  1    ln  a  16.89b
r0 *2  1  a 2 

Figure 16.16 shows the comparative Couette velocity profile behavior for an ideal Bingham
plastic with: (a) only the outer cylinder rotating (Eqs. 16.81 and16.82) and; (b) only the inner
cylinder rotating (Eqs. 16.88 and 16.89). The material behavior for three different inner cylinder
radii, ri *  a  0.2, 0.5, and 0.8 , and for a series of yield radii, r0 * , starting at the inner cylinder
radius, and extending outward to r0 *  1.4 or 1.5 .

Note that for all cases, when r0 *  ri *  a (or when r0 *  ri * ) there is no movement of the
material (blue lines). However, once r0 *  ri * the fluid-like material flow begins at ri * and
extends out to r0 * . Figure 16.16(a) shows that if r0 *  1 the region r0 *  r*  1 will behave as a
solid, in solid body rotation. This can best be seen for ri *  0.2 and 0.5 , where the solid portion is
reflected by a linear velocity profile. However, for the case of inner cylinder rotation, figure
16.16(b), the solid portion will remain fixed for the region r0 *  r*  1 . What is clear from
scaling the velocity relative to Vo, Newt and Vi, Newt is that for the same torque the velocity of the
rotating cylinder with an ideal Bingham plastic will always be less than that for a Newtonian
fluid, because of the added resistance due to the yield stress, 0 .

When r0 *  1 , all of the material behaves as a fluid. Note that the larger the gap between
cylinders is, the more nonlinear the velocity profiles are. And as r0 * increases beyond 1, the
more closely the velocity profiles approach the behavior of a Newtonian fluid (which would be
when r0 * 1 ).

Notice that Eq. 16.89b for V *i is identical to Eq. 16.82b for V*o , which may seem unusual.
However, note that the Newtonian values for Vo , Eq. 16.80b, and Vi , Eq. 16.87b, which are
used to normalize the velocities are quite different for the same applied torque. This is a function
of the moment over which the torque is applied ( R o  R for Eq. 16.80b vs. R i  aR for Eq.
16.87b). Thus, for the same applied torque, when the outer cylinder rotates (Case a), the outer
wall shear stress will be less than inner wall shear stress when the inner cylinder rotates (Case b).

551
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

This allows the outer cylinder to rotate at a higher velocity than the inner cylinder, under the
same applied torque. Also, if we examine the angular velocity of the cylinders for a Newtonian
fluid from Eq. 16.80b and 16.87b, we get:

T  1 a2  T 1 a2 
Vo, Newt  o,Newt R     o,Newt    16.90a
4LR  a 2  4LR 2  a 2 

T  a 2 1  T  1 a2 
Vi, Newt  i,Newt aR      i,Newt     16.90b
4LR  a  4LR 2  a 2 

Equation 16.90 shows that under the same applied torque, the angular velocity for both the outer
rotating cylinder and the inner rotating cylinder will be equal in magnitude. The difference in the
sign is a result of our assumption of a positively oriented shear stress on the material surface at
the inner radius. In reality, the inner material surface has a negatively directed normal relative to
r, thus a positive orientation for the shear would point in the negative θ direction, and be
considered a negative torque. Thus, the angular velocities of Eq. 16.90 will be the same.

16.5.2.2 Poiseuille Flow in Circular Duct

Decreasing pressure,
r
r0 R
z


Figure 16.17 Geometry for Poiseuille flow of ideal Bingham plastic in a circular duct

For a fully-developed Poiseuille flow of an ideal Bingham fluid in a circular duct, as shown in
figure 16.17, there is no motion in the r and θ-directions, but only in the streamwise z-direction,
and the governing equations reduce to the z-direction momentum equation (Eq. 16.109d of
Section 16.7.3):

v z v v v v
 vr z   z  vz z
t r r  z
16.109d
1 p 1 1 
  gz   r rz   1 z   zz 
 z   r r r  z 
Since this is a steady, fully-developed flow, the velocity field is given by vz  f (r) , and Eq.
16.109d reduces to:

1 p
 rrz   16.91
r r z

552
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

p
For a Poiseuille flow,  constant (as discussed in section 6.4.1), which allows Eq. 16.91 to be
z
rewritten as:

 p
 rrz   r 16.92
r z
Integrating Eq. 16.92 once yields:

p r C1
 rz   16.93
z 2 r
To determine the constant C1 in Eq. 16.93, we need to consider how an ideal Bingham plastic
will behave for Poiseuille flow in a duct. Like the parallel plate Poiseuille flow of Section
16.5.1.2, (and Newtonian Poiseuille flow, Section 6.4.2) the location of the highest shear
stresses, and thus the location where an ideal Bingham plastic first begins to behave as a fluid
will be adjacent to the cylinder wall, r = R, with the shear stresses decreasing away from the duct
wall. The criteria for fluid behavior to occur is rz  0 . Therefore, if the wall shear stress, rz , is
greater than the yield stress, 0 , the material in some region adjacent to the wall will behave as a
fluid. The extent of this fluid region will be r0  r  R , where we designate r0 as the yield radius,
similar to the yield location, y0 , that we employed for parallel plate flow in Section 16.5.1.2. We
hypothesize r0 as the location where rz  0 , and where the material ceases to behave as a fluid,
and behaves as a solid .

For a Bingham fluid behaving as a solid, Eq. 16.46a specifies that:

 v 
 rz   z   0  , when rz  0 (assuming 0 > 0)
 r 
Thus, the solid behavior region where rz  0 , must behave as a solid, moving uniformly at a
constant velocity as a plug flow, as shown in the central portion of figure 16.17. The extent of
this plug flow region will be 0  r  r0 .

To determine where r0 occurs, we perform a force balance on the region 0  r  r0 , similar to

what we did in Section 16.5.1.2, recognizing that rz  0 at r  r0 , and rz  0 at r = 0. The
result of that force balance is:
p r0
0 
z 2
Solving for r0 gives:

20
r0  16.95
 p 
 
 z 

553
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

Equation 16.95 indicates that the location of r0 is a direct function of the ratio of the yield stress
 p 20 
to the applied pressure gradient. In particular, note that r0 must be less than R  i.e.  
 z R 
in order for any of the material to behave as a fluid.

Therefore, a Poiseuille flow of an ideal Bingham plastic will have two regions of behavior:

(1) for 0  r  r0 , rx  0 , the material will move as a solid at v z = constant as a plug flow,
and
(2) for r0  r  R , rx  0 , the material will move as a fluid with vz  vz  r  .

Within the fluid region r0  r  R , and rx  0 , Eq. 16.93 applies:

p r C1
 rz  
z 2 r
To determine the constant C1, we note that Eq. 16.46a requires that within the solid plug flow
v v z
region that z  0 when rz  0 . And since the derivative of v z , i.e. , must be
r r
v
continuous within the ideal Bingham plastic, the condition that z  0 for the solid material in
r
the region 0  r  r0 , must also apply for the fluid behavior material at the interface r  r0 . To
 v 
apply this condition, we substitute Eq. 16.46b,  rz   z    o , into Eq. 16.93 for the fluid
 r 
region behavior, then substitute from Eq. 16.95 for o :

p r C1  v   v z  p r0
rz    B  z   o   B  
z 2 r  r   r  z 2
v
Solving for z :
r

v z 1 p C
  r  r0   1
r 2 B z Br
v z
Applying the boundary condition of  0 at r  r0 , we have that C1  0 , which gives;
r
v z 1 p
  r  r0 
r 2 B z
Integrating gives:

1 p  r 2 
vz    r0 r   C2
2 z  2 
Applying the boundary condition v z = 0 at r  R

554
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

1 p  R 2 
C2     r0 R 
2 B z  2 
Substituting for C2 and simplifying gives:

1 p  r 2 R 2  1 p
vz      r0  R  r  16.96
2 B z  2 2  2 B z
r
Defining r*  , and letting vz,ro *  vz  ro * , since the vz velocity profile will be a function of
R
ro * , as well as r*, we rewrite Eq. 16.96 as:

R 2 p
v z,r0 *   r *2 1  2r0 * 1  r *  16.97a
4 B z  

The material velocity will reach a maximum at the interface between the fluid and solid
behavior, with vz  vz,r0 *,max at r*  r0 * , which will also be the velocity of the material “plug” in
the central portion of the circular duct:

R 2 p
v z,r0 *,max   r0 *2 1  2r0 * 1  r0 * 
4 B z  
which reduces to:

R 2 p
 r0 * 1
2
v z,r0 *,max  16.97b
4 B z
p
Note that  0 will give a flow in the positive z direction, thus Eq. 16.97 will yield positive
z
velocities for pressure decreasing in the z direction.

v z,r0 *
Defining v z,r0 * *  gives:
v z,r0 *,max

vz,r0 * 1  r *2   2r *0  r * 1 

vz,r0 * *    for r0 *  r*  1 16.98a
 r0 * 1
2
vz,r0 *,max
and

vz,r0 * *max  1 for 0  r*  r0 * 16.98b

Equation 16.98 shows how an ideal Bingham fluid velocity profile varies relative to the
maximum centerline (or plug velocity) for the respective r0 * value. Figure 16.18 is a series of
velocity profiles of vz,r0 * * vs. r* for a range of r0 * values, 0  r0 *  1 .

555
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

Figure 16.18 Ideal Bingham plastic flow in a circular duct. Normalized on the maximum
2 0
velocity for respective values of the yield radius, r0 *  .
 p 
R 
 x 

As figure 16.18 reveals, for a yield location very near the wall ( r0 *  1 ), the material profile is
essentially a flat plug flow. However, as the yield radius decreases (i.e. the pressure gradient
increases), the profile will display a wider region of fluid-like behavior, and the extent of the plug
p 20
flow region will decrease. Potentially, if the pressure gradient gets quite large, , the yield
x R
location will become quite small, and the flow will approach a Newtonian flow behavior ( r0 *  0 ).

While Figure 16.18 is instructive regarding the shape of the velocity profiles, it does not show
how the yield radius affects the behavior of the material relative to a comparable Newtonian
fluid. As we did in Section 16.4.2.2, we can compare the behavior of an ideal Bingham fluid to a
Newtonian fluid with the same pressure gradient and fluid viscosity.

For a Newtonian fluid, 0  0 , which gives r0 *  0 . Equation 16.97b gives:

R 2 p
vz,Newt,max   16.99
4 r

556
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

If we again assume equivalent pressure gradients and fluid viscosities ( B   ), we can compare
velocity profiles for a range of ideal Bingham fluids relative to a Newtonian fluid. Thus, dividing
Eq. 16.97a by Eq. 16.99, we obtain:
vz, r0 *
vz *rel   1  r *2   2r0 *  r * 1  for r0 *  r*  1 16.100a
v z,Newt,max

and
vz, r0 *,max
  r0 * 1
2
vz *rel,max  for 0  r*  r0 * 16.100b
vz,Newt,max
Equation 16.100 gives the velocity profile behavior relative to a pure Newtonian fluid, with the
same pressure gradient and fluid viscosity. Figure 16.19 shows a series of velocity profiles of
vz *rel vs. r * for a range of r0 * values, 0  r0 *  1 .

Figure 16.19 Ideal Bingham plastic Poiseuille flow in a circular duct for a series of yield
radii, r0 * ; normalized on the maximum (centerline) velocity for a Newtonian
2 0
fluid with the same pressure gradient and fluid viscosity. r0 * 
 p 
R 
 x 

Figure 16.19 illustrates that fluid behavior does not initiate if the pressure gradient is less than
the yield stress, r0 *  1 ( r0 *  1 , the green line, represents all cases for r0 *  1 ). As the pressure

557
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

gradient increases relative to the yield stress, 0 , r0 * is reduced, expanding the region of fluid
behavior. This results in an increase in the maximum velocity, a reduction in the extent of the
solid plug-flow region, and a commensurate increase in the volume flowrate.

As a third comparison, we examine how the volume flowrate and pressure gradient of an ideal
Bingham plastic fluid behave relative to a Newtonian flow. To do this, we calculate the volume
flowrate, Q ro * , by integration of Eq. 16.97 across the tube 0  r*  1 .

r R r* r0 * r*1

 v 2rdr  2R  
2 2
Qro *  z vz,r0 *,max r *dr *  2R vz,r0 *r *dr *
r 0 r*0 r* r0 *

R 4 p  0 
r* r * r*1
Qro *     0 r * 1
2
r *dr *   1  r *2   2r0 *  r * 1  r *dr *
2B x  r*0  
r* r0 * 

R 4 p   
r*1
2 r* r0 *
2 r*   r *2 r *4   r *3 r *2  
Qro *     r0 * 1       2r *     
2  r*0  2
0
2B x   4   3 2   r*r0 * 

Which yields a final solution for the volume flowrate, Q ro * , as:

R 4 p
Q ro *    r0 *4 4r0 * 3 16.101
24 B x 
As a check on Eq. 16.101, we note that when r0 *  0 , the fluid would behave as a Newtonian
fluid, so:

R 4 p R 4 p
Q Newt   
3   16.102
24 B x 8 B x

Equation 16.102 is the same as Eqn. 6.62 in Section 6.4.1.

To compare the respective flow rates, I take the ratio of Q ro * to QNewt :

Q r0 * 1
Q r0 *,rel 
Q Newt

3
 r0 *4 4r0 * 3 16.103

Equation 16.103 gives the volume flowrate for an ideal Bingham plastic relative to a Newtonian
fluid with identical pressure gradient and fluid viscosity. Figure 16.20(a) shows Qr0 *,rel vs. r0 *
for 0  r0 *  1 .

To examine pressure gradient variations, we solve Eq. 16.101 for the pressure gradient, and let
Q  Qr0 * :

p 24 B Q
 16.104
z r0 * R  r0 * 4r0 * 3
4 4

558
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

For a Newtonian fluid, r0 *  0 , Eq. 16.104 gives:

p 8Q
 16.105
z Newt R 4
To compare the respective pressure gradient at identical volume flowrates, we again assume
identical fluid viscosities, and take the ratio of Eq. 16.104 to Eq. 16.105:
p
p z r0 * 3Q
  16.106
z rel p  r0 * 4r0 * 3
4

z Newt

p
Figure 16.19b shows vs. r0 * for 0  r0 *  1 .
z rel

(a) (b)
Figure 16.20 Comparisons of ideal Bingham plastic volume flowrate and pressure gradient
to a comparable Newtonian fluid versus r0 * -- Poiseuille circular duct flow.
(a) Qr0 *,rel , assuming identical pressure gradient and fluid viscosity;
(b)  p z  rel , assuming identical volume flowrate and fluid viscosity.
2 0
r0 * 
R  p z 

559
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

Figure 16.20(a) shows that the flowrate of an ideal Bingham plastic will always be less than that
of a comparable Newtonian fluid with the same pressure gradient and fluid viscosity. The thinner
the fluid region (larger r0 * , and larger yield stress, 0 ), the lower the flowrate. The explanation
for this behavior is the same as for the parallel plate flow of section 16.5.1.2. Applying a force
balance at the wall of the duct, we can show that:

vz 1  p R  p R
   0   1  r0 * for r0 *  1 16.107
r wall  B  z 2  x 2 B

Equation 16.107 shows that the slope of the velocity profile at the wall ( r  R ) will be
v z p R
 0 (no flow) until  0 or r0 *  1 . When the pressure gradient exceeds the yield
r wall z 2
stress, this makes r0 *  1 , and a region of material adjacent to the cylinder wall , r0 *  r*  1 , will
behave as a fluid. Eq. 16.107 indicates that the slope of the velocity at the wall will become non-
v z
zero (the sign depending on the direction of the pressure gradient),  0 , and the material
r wall
(solid and fluid portions) will begin to move, collectively, as shown in figure 16.19. As the
p
pressure gradient, , increases further, the yield radius, r0 * , will reduce accordingly. This
z
reduction in the yield radius results in an expanded region of fluid behavior – which allows the
collective material to move more rapidly, as shown by the velocity profiles of figure 16.19, and
the relative flowrate of figure 16.20a.

Of course, the larger the yield stress, 0 , the more the pressure gradient has to work to overcome
0 , and the less work it can do to move the fluid. This is shown clearly in, figure 16.20b, which
illustrates that the thinner the fluid region (i.e. larger r0 * value), the larger the relative pressure
gradient must be to maintain the same flow rate. For example, if r0 *  0.6 , for which 64% of the
material behaves as a fluid, the comparative pressure gradient is 4 times the Newtonian. And by
r0 *  0.8 , where only 36% of the material behaves as a fluid, the comparative pressure gradient
is more than 10 times the Newtonian. Of course, as the yield radius approaches the duct wall,
the pressure gradient required to maintain the same flowrate grows dramatically.

Again, like for Poiseuille flow between parallel plates, the power required to move the material
 p 
will be roughly equivalent to  i.e. Power  Q  , so the pumping power required to maintain a
 x 
constant flowrate will behave essentially the same as the pressure gradient in figure 16.20b.

560
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

16.6 Conclusion

As illustrated by the examples in sections 16.4 and 16.5, non-Newtonian fluids behave markedly
different that Newtonian fluids. In addition, the modeling of non-Newtonian fluids can give rise
to anomalies, such as the flow consistency index, K, for power-law fluids, which have fractional
property units (as a result of fractional values of the fluid behavior index n). This makes it hard
to compare power-law fluids, to either a Newtonian fluid or other non-Newtonian fluids,
particularly pressure gradient variations for Poiseuille flows, as was discussed at the end of
Sections 16.4.1.2 and 16.4.2.2.

Relative comparisons of non-Newtonian fluids to a Newtonian fluid is important, if one is to

understand differences in flow behavior. What constitutes an effective presentation of relative
effects is reflected by figure 16.16 in Section 16.5.2.1. In that figure, I chose to normalize the
relative velocity behavior of Bingham plastic Couette flows for rotating, concentric cylinders to
the Newtonian wall velocity, rather than the wall velocity for the particular Bingham plastic
material. Recall that since we could not specify a common wall velocity, by necessity we had to
assume a value of torque, from which we educed the corresponding wall velocities. Since the
torque was the common parameter for these flows, it seemed most logical to show how the
velocity behavior varies relative to a Newtonian fluid under the same applied torque. This form
of presentation in figure 16.16 illustrates both the variation in the shape of the velocity profiles
and the retardation of the wall velocity due to the presence of a Bingham plastic yield stress (
0  0 ).

So, what if we were to normalize the velocity behavior of the Bingham plastic materials on the
respective wall velocities for those materials, instead of normalizing using the wall velocity for a
comparable Newtonian fluid? Would that form of presentation be as instructive? Figure 16.21
shows a comparison for ri *  0.5 of normalizing Bingham plastic velocity behavior on each
 
material’s moving wall velocity i.e. v  / Vwall,r0 * versus normalizing on the moving wall velocity
for a Newtonian fluid  i.e. v / Vwall, Newtonian  .

The top row of figure 16.21 shows the normalization of the velocity profiles with respect to the
comparable Newtonian wall velocity, as was done in figure 16.16. Using this form of
normalization, it is easy to discriminate between the two regions of the flow process—fluid
versus solid behavior—and to understand the degree of velocity retardation the moving wall
experiences due to the material yield stress of the Bingham plastic. In contrast, scaling on the
actual material wall velocity, as done for the bottom row of figure 16.21, does not clearly reveal
the changes in the velocity behavior between solid and fluid behavior, and does not show the
degree of retardation of the wall velocity under the same applied torque. Also, this latter scaling
approach cannot effectively represent the stationary material behavior when r0 *  ri * (for which
Vwall,r0 *  0 ). This velocity scaling issue illustrates the importance of understanding the method
of presentation of relative flow behaviors in order to reveal important physical behavioral
changes.

561
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

Figure 16.21 Alternative scaling for Couette flow of an ideal Bingham plastic, ri *  0.5 .
Both moving outer wall (left plots) and inner wall (right plots) are shown.
Upper plots scaled on Vwall,Newton for a Newtonian fluid.
Lower plots scaled on Vwall,r0 * for the particular Bingham plastic.

In Sections 16.4.1.2 and 16.4.2.2, I mentioned that certain non-Newtonian fluids may
demonstrate behavior that could be reasonably described by either a power-law or an ideal
Bingham plastic. For example, shear-thinning fluids often display Poiseuille behavior that could
be approximated by either model. Figure 16.22 shows such a comparison for Poiseuille flow in a
circular tube. Note how the red lines (n = 0.5 and r0 *  0.2 ) and the blue lines (n = 0.2 and
r0 *  0.5 ) display similar velocity profiles, suggesting that either model might be applicable.

However, while the profiles appear similar, a shear thinning fluid (n < 1) will always behave as a
fluid under any pressure gradient, whereas if the pressure gradient for a Bingham plastic is
reduced such that:

 p  2
r0 *  1     0
 x  R
the Bingham plastic will behave as a solid. So, one must be careful to assess the type of
behavior the fluid will assume under all conditions.

562
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

Figure 16.22 Comparison of similar Poiseuille velocity profiles for a power-law fluid and
an ideal Bingham plastic flow in a circular duct. Velocities are normalized on
the respective centerline velocities for selected values of the power-law fluid
2 0
behavior index, n, and the Bingham plastic yield radius, r0 *  .
 p 
R 
 x 

I will close this chapter by noting that we have only touched on two types of rather simple non-
Newtonian fluid models, and have examined only two types of steady state flows. You should
be struck by how much more involved the “simple” solutions for a non-Newtonian fluid are than
for a Newtonian fluid. Clearly, if we consider more complicated flows, such as boundary layer
behavior, time-dependent flows, drag, and flows in multi-dimensions, solutions are generally
possible only using numerical approaches. The literature is full of such solutions, as well as other
manifold non-Newtonian fluid models. Hopefully, understanding the simpler examples covered
in this chapter will prepare you to examine, and appreciate, more complicated non-Newtonian
flows.

Speaking of complicated flows, we will now transition back to Newtonian fluids, and examine
what is the most complicated type of flow behavior, turbulence, which along with the process of
transition from laminar to turbulent flow will be the topic of the next, and last, chapter of this
book.

563
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

16.7 The Governing Equations for Non-Newtonian Fluids

16.7.1 Stresses on a Non-Newtonian Fluid: Cartesian and Cylindrical Coordinates

Note: stresses are listed in terms of the generic deformation rate  , and the effective
viscosity, e , which is dependent on the non-Newtonian model employed.

xx  p  2exx xy  e  xy   yx

 yy  p  2e  yy xz  exz  zx
zz  p  2ezz yz  e  yz  zy

rr  p  2err r  e r  r

  p  2e z  ez  z
zz  p  2ezz zr  ezr  rz

16.7.2 Continuity and Momentum Differential Equations in Cartesian Coordinates

u v w
  0 (16.108a)
x y z

u u u u 1 p 1    yx  zx 
u v w   g x   xx    (16.108b)
t x y z  x   x y z 

v v v v 1 p 1   xy  yy  zy 
u v w   gy      (16.108c)
t x y z  y   x y z 

w w w w 1 p 1    yz  zz 
u v w   g z   xz    (16.108d)
t x y z  z   x y z 

16.7.3 Continuity and Momentum Differential Equations in Cylindrical Coordinates

1
rvr   1 v  v z  0 (16.109a)
r r r  z
v r v r v  v r v r v 2
 vr   vz 
t r r  z r
(16.109b)
1 p 1 1 
  gr   r rr   1 r   zr   
 r   r r r  z r 

564
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

v  v v v v vv
 vr      vz   r 
t r r  z r
(16.109c)
1 p 11  2 1    z  r   r 

r 
 g   2
  r r

r  r 
r 
 
z

r 


v z v v v v
 vr z   z  vz z
t r r  z
(16.109d)
1 p 1 1 
  gz   r rz   1 z   zz 
 z   r r r  z 

565
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

Table 1
Select properties of power law and ideal Bingham plastic non-Newtonian fluids

Power Law Bingham Plastic

Temperature
Material (C) n K (Pa sn) 0 (Pa) B (Pa s)

Applesauce 27 0.45 7.3

Chicken (minced) 23 0.1 900
Chocolate 30 0.5 0.7 35 1
Cornstarch in water (17%) 25 1.42 6.9E-04
Cornstarch in water (45%) 25 0.35 0.3
Human Blood 27 0.9 0.004
Lubricating grease 25 0.1 1000
Mayonnaise 25 0.131 100 30 0.13
Nail Polish 25 0.86 750
Paint (latex) 25 0.6 21.5 15 0.065
Peanut Butter 30 0.07 500
Raspberry Jam 25 0.34 12.6
Shaving Cream 25 17.5 0.015
Sunscreen Lotions 25 0.28 75
Synovial Fluid 37 0.4 0.5
Tomato Juice (6% solids) 20 0.56 0.19
Tomato Ketchup 30 0.24 33 13 0.15
Tomato Paste (25% solids) 25 0.5 15 104 0.3
Toothpaste 25 0.28 120 200 10
Water 25 1 0.0089
Wheat Batter 30 0.75 110
Whipped Butter 30 0.057 312

49 %Fine coal slurry in water - 1 0.005

7.5% Kaolin slurry in water - 7.5 0.005
32% Kaolin slurry in water - 20 0.005
14% Sewage sludge - 3.1 0.025
39% Red mud slurry - 23 0.030
40% Drilling mud - 0.51 1.3 11 0.030
Fresh water mud - 0.8 0.319
72% Pulverized fuel ash in water - 0.46 9.3
Cement Slurry - 0.36 3.0 5.4 0.029
20% Sand/Water mixture - 1.48 0.000313
40% Sand/Water mixture - 1.21 0.0269
Power law: PL   app     K    ,  app  K    Ideal Bingham Plastic: BP  B     0
n n 1

566
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

Problems
(note: obtain all material/fluid properties from Table 1)

1. Consider a Couette flow between parallel plates, the lower plate fixed and the upper plate
moving at U=1 m/s. If the gap between the plates is 1 mm, what are the values of shear
stress for the following power law fluids: applesauce, tomato paste, and synovial fluid?

2. Consider a Couette flow between parallel plates, the lower plate fixed and the upper plate
moving at velocity U. If the gap between the plates is 1 mm and a shear stress of 35 Pa is
applied to the upper plate, what will the velocity U be for the following Bingham plastic
materials: mayonnaise, tomato ketchup, paint, and 40% drilling mud?

3. Consider a Poiseuille flow between parallel plates spaced 4 cm apart. If a pressure gradient
of 30 Pa/cm is applied, what are the values of flowrate per unit depth in cm2/s for the
following power law fluids: applesauce, tomato ketchup, and synovial fluid? Note: the
flowrate will be negative for a positive pressure gradient.

4. Consider a Poiseuille flow between parallel plates spaced 1 cm apart. If a pressure gradient
of 100 Pa/cm is applied, what are the values of flowrate per unit depth in cm2/s for the
following Bingham plastic fluids: mayonnaise, tomato ketchup, paint, and 40% drilling mud?

5. Consider a Couette flow between concentric circular cylinders, with diameters of 20 cm and
16 cm, and one meter long. If the inner cylinder is fixed, and the outer cylinder rotates at a
velocity of 10 cm/s, determine the torque in N-cm required to rotate the outer cylinder for
the following power law fluids: applesauce, tomato paste, and synovial fluid.

6. Consider a Couette flow between concentric circular cylinders, both 2 m long with
diameters of 20 cm and 12 cm. The inner cylinder is fixed, and a torque of 200 N-cm is
applied to the outer cylinder. Determine the steady state velocity of the outer cylinder for the
following Bingham plastic materials: mayonnaise and 40% drilling mud.

7. Consider a Couette flow between concentric circular cylinders, both 2 m long with
diameters of 20 cm and 12 cm. The inner cylinder is fixed, and a torque of 1000 N-cm is
applied to the outer cylinder. Determine the steady state velocity of the outer cylinder in m/s
for the following Bingham plastic materials: tomato paste and toothpaste.

8. Consider a Poiseuille flow in a 4 cm diameter tube. If a pressure gradient of 10 Pa/cm is

applied, what are the values of flowrate in cm3/s for the following power law fluids:
applesauce, tomato paste, and 40% drilling mud? Note: the flowrate will be negative for a
positive pressure gradient.

9. Consider a Poiseuille flow in a 2 cm diameter tube. If a pressure gradient of 100 Pa/cm is

applied, what are the values of flowrate in m3/s for the following Bingham plastic materials:
mayonnaise, tomato paste, and paint?

567
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

10. Mayonnaise and paint have properties that can be modeled as either a power-law fluid or a
Bingham plastic. Consider a Poiseuille flow between parallel plates spaced 1 cm apart. If a
pressure gradient of 100 Pa/cm is applied, what are the comparative power-law vs. Bingham
plastic values of flowrate per unit depth in m2/s for each of the fluids (i.e. what are Qpowerlaw
vs. QBingham for each fluid?).

11. A cement slurry can be modeled as either a power law fluid or a Bingham plastic. Consider
a Poiseuille flow in a 20 cm diameter tube. If a pressure gradient of 10 Pa/cm is applied,
what are the comparative power-law vs. Bingham plastic values of flowrate in m3/s (i.e.
what are Qpowerlaw vs. QBingham?).

12. 40% drilling mud can be modeled as either a power-law fluid or a Bingham plastic.
Consider a Poiseuille flow in a 20 cm diameter tube. If a flowrate of 1 m3/s is be achieved,
what are the comparative power law vs. Bingham plastic values of pressure gradient in Pa/m
p p
required to achieve this flow rate (i.e. what are vs. ?). Note that you will
z powerlaw z Bingham
20
need to assume that Q < 0 for the power law model. And since r0  for a Bingham
 p 
 
 z 
p
plastic, you will have to iterate to determine and the appropriate r0 * value.
z

13. Consider a Couette flow of peanut butter between concentric circular cylinders, with
diameters of 20 cm and 16 cm and lengths of 1 m. If the inner cylinder is fixed, and the
outer cylinder rotates, determine the respective torques in N-m required to rotate the outer
cylinder at velocities of 10 cm/s and 50 cm/s.

14. Peanut butter is pumped 100 m through a 12 cm diameter pipe from the processing area to
the bottling area of a plant. If a pressure gradient of 200 Pa/cm is required, and Poiseuille
flow is assumed, determine the flowrate in cm3/s through the pipe, and the total power
required for the pumping process in kWatts. What is the change in flowrate and power
requirements if the pressure gradient is increased to 250 Pa/cm?

15. A 49% fine coal slurry in water is pumped 400 m from a mine site to a railroad tank car
through a 30 cm diameter pipe as a Poiseuille flow. If we want a flowrate of 3 m3/s, determine
2
the pressure gradient required, in Pa/m, and the power required in kWatts. Since r0  p0 for a
z

p
Bingham plastic, you will have to iterate to determine and the appropriate r0 value.
z

16. A 40% sand in water mixture is pumped 1000 m from a mixing site to a processing plant
through a 40 cm diameter pipe as a Poiseuille flow. If a flowrate of 1 m3/s is be achieved,
what pressure gradient in Pa/m is required to achieve this flowrate? Note that you will need
to assume that Q < 0 for the power law model. Also, how much power will be required for
this process?

568
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 16

17. If the sand-water mixture in problem 16 is diluted to 20% sand in water mixture, and the
flowrate increased to 2 m3/s (which will keep the amount of sand delivered the same), will
this reduce or increase the pressure gradient required and the power required? Again note
that you will need to assume that Q < 0 for the power law model.

18. If latex paint behaves as a Bingham plastic, How thick can a paint layer be before it starts to
run down a vertical wall? How thick would a layer of peanut butter (using its Bingham plastic
properties) have to be before it would flow down the wall? Note the density of paint is 1200
kg/m3, and the density of peanut butter is 1080 kg/m3.

19. Using basic principles, like was done in Section 16.4.1,

y
derive an equation for u(y) of a power-law fluid flowing
 p 
down a vertical wall, where   0  and gx = g (where g x
 x  g
acts in the x-direction, toward the ground, and y is oriented
outward from the wall, as shown). Equation 16.7 simplifies
n
 yx  u 
to g   0 , where yx  K   , with boundary
y  y 
u
conditions u = 0 at y = 0, and  0 at y = t, where t is the
y
fluid layer thickness.

20. Using basic principles, like was done in Section 16.5.1,

derive an equation for u(y) for a Bingham plastic material flowing down a vertical wall,
 p 
where   0  and gx = g (where g acts in the positive x-direction, toward the ground, and
 x 
y is oriented outward from the wall. See problem 19 schematic). Equation 16.108b
 yx  u  u
simplifies to g   0 , where  yx   B    0 when  yx   0 (and  0 when
y  y  y
u
yx  0 ). The boundary conditions are u = 0 at y = 0, and  0 at y = t, where t is the
y

fluid layer thickness. Here, you will have to use a variation on Eq. 16.55 as y 0  t  0 ,
g
since gravity is the driving force, not the pressure gradient.

569
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

Chapter 17

Turbulence: An Introduction

Contents
17.1 What is Turbulence? ................................................................................................. 571

17.2 Initiation of Turbulence: Flow Instabilities .................................................................. 574

17.2.1 Stability of Bounded Flows .............................................................................. 580
17.2.2 Stability of Unbounded Flows ......................................................................... 582
17.2.3 Effect of a Pressure Gradient on Flow Stability ................................................ 583

17.3 The Transition to Turbulence ..................................................................................... 584

17.3.1 Transition of Unbounded Flows ...................................................................... 585
17.3.2 Transition of Bounded Flows: Natural ............................................................. 589
17.3.3 Transition of Bounded Flows: Forced ............................................................... 592

17.4 The Character of Turbulence ..................................................................................... 592

17.5 Mathematical Modeling of Turbulent Flow ................................................................ 597

17.5.1 Reynolds Averaged Equations ......................................................................... 599
17.5.2 Turbulent Reynolds "Stresses" ........................................................................ 603
17.5.3 Phenomenological Equations .......................................................................... 606
17.5.3.1 Eddy Viscosity .................................................................................... 607
17.5.3.2 Mixing Length Concepts (near a flat plate) ......................................... 608

17.6 "Universal" Velocity Distribution Laws (for boundary layers) ..................................... 608
17.6.1 The Inner Region ............................................................................................. 611
17.6.1.1 The Viscous Sublayer .......................................................................... 613
17.6.1.2 Logarithmic Layer .............................................................................. 613
17.6.1.3 Buffer Layer ...................................................................................... 616
17.6.1.4 Models of the Entire Inner Region ..................................................... 616
van Driest Damping Factor ................................................................. 616
Spalding's Single Inner Region Equation ............................................. 618
17.6.2 The Outer Region ........................................................................................... 620
17.6.3 The Combined Universal Velocity Profile ......................................................... 624
17.6.4 A Shape Factor Comparison of Turbulent vs. Laminar Boundary Layers ............ 624

17.7 A Momentum Integral Analysis of a Flat Plate Turbulent Boundary Layer ................... 629
17.7.1 Transition to a Turbulent Boundary Layer on a Flat Plate ................................. 629
17.7.2 The Momentum Integral Equation for Turbulent Flow ..................................... 630
17.7.3 The Mean Velocity Profile: Nikuradse Power Law ............................................ 631
17.7.4 Assumed Shear Stress Relationship ................................................................. 632
17.7.5 Solution of the Momentum Integral Equation ................................................. 634
17.7.5.1 Turbulent Flow from Leading Edge ..................................................... 635
17.7.5.2 Laminar-Turbulent Transition at Critical Reynolds Number ................. 637
17.7.5.3 Comparison: Laminar, LE Turbulent, and Transitional Boundary Layers 639

17.8 Conclusion ................................................................................................................ 643

570
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

17.1 What is Turbulence?

Anyone who has flown in a commercial aircraft has probably heard the word "turbulence" used
to describe the bumpy, unsteady behavior often encountered during flight. This unsteady flight
behavior is caused by unsteady lift and drag forces, which are the result of the aircraft
encountering pockets of large velocity variations due to proximity to terrain variations (such as
mountains) or strong thermally-induced density variations, such as those associated with
thunderstorm activity. The result is generally large, somewhat random fluctuations in velocity,
which in turn interact with the surfaces of the aircraft, resulting in comparable fluctuations in lift
and drag. The end result is a bumpy, disconcerting ride, caused by what is termed clear-air
turbulence.

However, turbulence occurs in most practical flows over, around, and through engineering
systems. While the large regions of clear-air turbulence create a sometimes harrowing ride,
turbulence is also always present in the boundary layer flow over the aircraft's surfaces. In fact,
all vehicles, whether they are cars, trucks, trains, or ships, experience and are impacted by
turbulence, as are all fluid transport systems such as pipelines, air handling ducts, or water ways.

So, what is turbulence? The short answer is a self-sustaining state of seemingly random
fluctuations of a flowing fluid, which result in intense mixing of mass and momentum, and a
significant increase in surface friction. Now for the hard question: where and why does it occur,
and how does one characterize it and its effects on the fluid and the fluid interactions with
bounding surfaces? To answer this question, we will take the next multiple pages, and some
admitted arm waving, to even cursorily address the process. And we will again only be able to
give the short answer, which, although it may seem expansive, will be somewhat incomplete.
The reason for this vagueness is that next to the origin of the universe, or the establishment of a
unified field theory of physics, turbulence is probably the most perplexing problem in physics.
In fact, numerous giants of the fields of physics and fluid mechanics have often, in their later
years, commented that turbulence remains, and will probably remain, the most difficult physical
process to truly understand. Moreover, this continues to remain true, despite a continual
improvement in the tools to study turbulence (e.g. 3-D velocity measurement systems, and
continually more powerful computers).

So, where to begin? First, let us review the general process by which a flow becomes turbulent.
In a naturally occurring process, a flow will generally begin as a laminar flow. As the flow
develops, either along a surface or moving within a quiescent surrounding fluid, small
instabilities (think small, wave-like behavior) will develop within the flow. Under the
appropriate conditions, these instabilities will begin to amplify (i.e. grow in amplitude). As the
instabilities amplify and grow, they will eventually undergo a transition to a complex, three-

571
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

dimensional behavior, manifested by the development of localized concentrations of time-

dependent vorticity. Finally, this transitional behavior will grow into a complicated, three-
dimensional distribution of vorticity of extended scales, which both (1) entrains additional mass
and momentum from adjacent non-turbulent fluid, and ( 2) dissipates local fluid kinetic energy
by intense local viscous action. This process of entrainment, mixing, and dissipation becomes
self-sustaining, continuing to entrain surrounding fluid, and dissipate local fluid energy at a
much more elevated rate than a comparable laminar flow. A crude representation of this
evolution from laminar to turbulent flow for a flat plate boundary layer is shown schematically in
figure 17.1.

Instability
Laminar Transition Turbulence
amplification

Figure 17.1 A generic schematic of the evolution of turbulence from a laminar flat plate
flow (not to scale).

While turbulence is characteristic of almost all flows of practical engineering interest, a concrete
description is elusive—we generally discuss generic characteristics, since it is difficult to
describe exact characteristics that will apply to all types of turbulent flows. I once paid a visit to
A.D. Young, a famous Oxford University fluid dynamicist, and noted on a corner of his office
blackboard he had written the following statement: "When we measure a fluid behavior, and we
don't know what is happening, we call it turbulence."

While there are many descriptions of turbulence, many short and many long, we will use a set of
characteristics summarized by R.W. Stewart in his narration of the educational film,
"Turbulence". This film (converted to a YouTube video), although old, provides a practical
display of the visual characteristics of turbulence and a good, generic description of the
turbulence processes. I recommend you view the video.

In the film, Stewart cites the following basic characteristics: "disorder (randomness); enhanced
mixing; and three-dimensional motions involving angular momentum (vorticity)". Most
researchers would agree with these, and probably add to or modify the list. However, we will

572
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

use these as our basis for our discussion of turbulence. Turbulence is the epitome of a process
that is described by the statement: "I don't know what it is, but I know it when I see it (or
measure it)."

As an illustration of what is and what isn’t turbulence, consider the instantaneous measurement
of the local velocity behavior in three different types of flow, within otherwise steady flow
fields, as shown in figure 17.2. Such measurements can be made with an instrument known as a
hot-wire anemometer, which employs a very small heated wire to measure changes in local
velocity behavior using the relationship between electrical resistance and the temperature of the
wire. The benefit of employing a hot-wire anemometer to measure turbulence is that it has a
very fast response rate, so it can accurately measure flows that vary very rapidly with time, such
as turbulence.

(a) Laminar B.L. u

time

(b) Shedding Cylinder u

time

(c) Turbulent B.L. u

time

Figure 17.2 Typical hot-wire anemometer velocity measurements of the temporal velocity
behavior in: (a) a laminar boundary layer, (b) the wake of a shedding cylinder,
and (c) a fully-turbulent boundary layer.

The first of the measurements shown, figure 17.2a, is within a laminar boundary layer, which
shows essentially no temporal changes in the velocity at a point, since laminar flows move in
layers (lamina), and from an Eulerian view are steady.

The second measurement shown, figure 17.2b, is in the wake of a cylinder, in which the
separating boundary layer forms an unsteady, but periodic, set of shedding Strouhal vortices (as
we discussed in Section 15.4). As these vortices pass the velocity-sensing probe, they cause a
periodic variation in the velocity. While such flows can be turbulent, such periodic variations
can also occur in otherwise laminar flows, and can maintain two-dimensionality. So, a lack of
randomness and three-dimensionality would disqualify this flow as a turbulent flow.

573
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

The third measurement shown, figure 17.2c, represents the local velocity behavior within a
turbulent boundary layer. Here, the velocity will vary in an apparently erratic, temporal manner,
displaying the apparently random behavior that is characteristic of turbulence. Note that we
cannot discern whether the measured velocity is three-dimensional, or is the result of vorticity
variations, but a researcher encountering such a velocity behavior would presume that this is a
measurement of turbulence (note the supposition that anything not reflecting laminar or pure
periodicity is turbulence, as per A.D. Young's blackboard quote).

But what conditions are necessary for a laminar flow to transition to a turbulent flow? Again,
this is a debatable area, but most turbulence researchers would agree to three primary criteria:

1) High Reynolds number (inertia/viscous imbalance)

2) Velocity gradients
3) The presence of flow disturbances

The necessary levels or magnitudes of these criteria that are sufficient to produce a transition
from a laminar to a fully-turbulent flow differ, depending on the type of initial flow (i.e.
boundary layer, jet, pipe flow, etc.). For a given type of flow, there is generally a critical
Reynolds number that must be reached (based on a characteristic length appropriate to the flow,
such as the boundary layer thickness) before a flow will begin the process of transitioning to
turbulence. Velocity gradients are also particularly important; the stronger or sharper the
gradient is, the sooner turbulence will develop. Also, to initiate the process of transition to
turbulence requires the presence of disturbances, either preexisting within the flow or created by
irregularities of a bounding surface.

17.2 Initiation of Turbulence: Flow Instabilities

As was indicated in figure 17.1, the initiation of turbulence begins with the development of an
instability within the flow. Now, many types of flow can undergo flow instability, and in many
cases this instability is simply a transition from one stable laminar state (i.e. behavior) to another
stable laminar state. In such a transition process, some effect causes a flow to change from one
type of flow behavior to another. A good example is flow over a cylinder at low Reynolds
numbers, as shown in figure 17.3.

As pointed out in Chapter 15, at very low Reynolds numbers (ReD < 1), the flow around a
circular cylinder will behave very much like a potential flow (see Section 9.8.1), with the
streamlines passing evenly and symmetrically around the cylinder (figure 17.3a). However, as
the Reynolds number increases to ReD > 4, the flow takes on a different character, where two
attached, counter-rotating vortices appear adjacent to the downstream surface of the cylinder

574
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

(figure 17.3b). Then, as ReD > 40, the flow transitions to an unsteady flow, with laminar vortices
shed periodically from the rear of the cylinder (figure 17.3c). Thus, the flow over a circular
cylinder undergoes instabilities with increasing Reynolds number, which cause a sudden, and
marked change in the character of the flow field. See this link for images of the flow states for a
circular cylinder.

(a)

(b)

(c)

Figure 17.3 Laminar states for flow over a cylinder. (a) 0 < ReD < 1, symmetrical;
(b) 4 < ReD < 40, attached vortices; (c) 40 < ReD < 60-100, laminar vortex
shedding.

Note that there are many other types of flows for which a small change in the Reynolds number
leads to a marked change in behavior from one stable state to another. Another example is
changes in flow behavior with increased rotation. Consider a fluid contained within the annulus
formed by two concentric cylinders, as shown in figure 17.4. If the inner cylinder is rotated
steadily and the outer cylinder held fixed, at low rates of rotation the flow in the annulus will
form a steady, Couette flow, with a two-dimensional velocity profile, as characterized by
equation 6.55a:

v 
2
i ri ro  r 2 2


2
r ro  ri
2


575
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

  r r  r 
However, as the Reynolds number given by Re i  i i 0 i  of the flow increases beyond a
  
critical level, the flow will spontaneously undergo transition from a two-dimensional Couette
flow to a Taylor flow, which is a three-dimensional, relatively steady flow comprised of a series
of steady, laminar vortices engirdling the annulus, with their axes of rotation in the azimuthal
direction, as shown schematically in figure 17.4b . See also a visualization of Taylor-type flows
here.

Taylor
Ω State transition Ω vortices
to Taylor flow
with increased
rotation

(a) Laminar Couette flow (b) Laminar Taylor flow

Figure 17.4 Transition from stable laminar Couette flow to stable laminar Taylor flow in
the annulus of concentric cylinders due to increased rotation.

Both for flow over a cylinder and flow in a concentric cylinder annulus, a small change in the
appropriate Reynolds number results in a marked change of the flow state. Recall that the
Reynolds number is a measure of the ratio of inertia forces to viscous forces. So, as inertia
effects increase, the less stable the original flow becomes, and the greater the probability of a
change in the flow state.

To understand this loss of stability of one state, and transition to an alternative state, consider the
stacking of children’s blocks. If we have a set of children’s wooden blocks scattered on the
floor, the blocks would be in a stable state, since if they are disturbed (e.g. by giving them a push
with your hand), they will not change their state (all the blocks will stay on the floor level), other
than to move across the floor. However, if we stack several of the blocks, one upon the other, as
shown in figure 17.5, we know from our childhood experience that a small disturbance can cause
the blocks to tumble back onto the floor. Of course, if the stack of blocks is rather low (say two

576
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

to three), a larger disturbance will be required to knock the stack of blocks over than if the stack
of blocks is higher (say 5 to 7). Moreover, if one continues to stack the blocks to even higher
heights, they eventually reach a height where the addition of another block , no matter how
carefully done, will cause the stack to spontaneously collapse due to the inherent irregularities of
the individual blocks.

Decreasing Stability

Figure 17.5 The stability of stacked blocks. The higher the stack of blocks, the more
unstable the process becomes.

What is happening as we stack the blocks higher is that we are creating an unstable state, which
is subject to small disturbances or perturbations from its surroundings. When the disturbance is
of sufficient magnitude, the blocks will suddenly transition from one state (the stacked state) to
another state (with all blocks on the floor). We might consider that the stacked state is
marginally stable, which can sustain some disturbance when the stack is low, but becomes
unstable as the stack grows in height, until it is unable to tolerate any level of disturbance.

The same process of stability is true for a laminar flow. In the case of a fluid, it is generally the
Reynolds number that dictates the level of stability of a flow, and when a transition to another
state (i.e. turbulence) can be expected. Since the Navier-Stokes equations govern all Newtonian
fluids, both laminar and turbulent, an examination of the stability of a laminar flow using the
Navier-Stokes equations would seem to be a reasonable approach to determine when and how
transition from a laminar to a turbulent flow occurs, and the sensitivity of the process to both
Reynolds number and the type of disturbance. However, the non-linear nature of the Navier-
Stokes equation makes the process of analysis not only complicated, but also necessitates some
sweeping approximations. In particular, it is assumed that a flow is locally parallel, and for the
simplest situation, two-dimensional.

Thus, in a two-dimensional x-y Cartesian coordinate system, we assume that we can write the
equations for u and v as:

u = U+u' and v = V+v' (17.1)

Here, U and V are the mean values of the velocity, and u' and v' represent the fluctuation or
perturbation velocities (i.e. velocity components that fluctuate about the mean value). Note that

577
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

the assumption of parallel flow (i.e. parallel in the x-direction) also assumes that the mean flow
is fully developed, such that U(y), which means (via continuity) that V = 0. The fluctuating
velocities are, however, considered functions of x, y, and t (time).

Now, substituting Eq. 17.1 (with V = 0) into the two-dimensional continuity equation, Eq. 5.46a,
gives:


U  u   v  0
x y
U
or by the parallel flow assumption,  0 , and thus:
x
u  v
 0 (17.2)
x y
If we now assume a stream function  such that Eq. 17.2 is identically satisfied, we have:

 
u   and v 
y x

This is very much the process we used for potential flow theory. However, since u' and v' are
functions of x, y, and t, this means that   x, y, t  . Since we want to investigate how a
disturbance develops within a flow, we assume the perturbation can be expressed as a traveling
wave with amplitude varying with y and the wavelength varying with x and t. The general form
of an appropriate function for the stream function is:

  Fy  expix  ct  (17.3)

In Eq. 17.3, α is termed the wave number (proportional to 1/wave length) and c is termed the
wave or phase velocity. This expression assumes the behavior of the amplitude F(y) will either
grow or shrink, depending on the respective values of α and c. Since α and c are arbitrary
constants, the idea is to explore a range of values of α and c, and determine which will give rise
to a growth of F(y), which will damp F(y), and particularly where F(y) will remain constant.
What we surmise is that: a growth in F(y) will lead to a breakdown of the flow, leading to
turbulence; a damping of F(y) will result in the flow remaining in a laminar state; and the
combination where F(y) remains constant will be where the flow is marginally stable (i.e. the
boundary of stability).

Now to complicate the matter further, we note that for a traveling wave, the wave velocity is
assumed complex, such that:

c  cr  ici , where i  1 , (17.4)

578
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

If we substitute Eq. 17.4 into 17.3, we have:

  Fy  expix  c r t  c i t   Fy  expix  c r t expc i t 

Clearly, the term expc i t  controls whether the stream function (and thus the velocities) grows
or is damped. If c i > 0, the stream function grows exponentially; if c i < 0 the stream function
is damped exponentially. Now, since   0 , this means that the sign of c i controls the stability
of the flow, and that c i can be considered as a coefficient of amplification .

So, if we can develop and solve an appropriate governing equation for the stream function, we
can determine which values of α and c yield stable behavior, and which yield unstable behavior,
for a particular velocity profile. The appropriate governing equation is the Navier-Stokes
equation, into which we will substitute the derivative functions of  , which represent the
appropriate velocity relationships. To do this properly we use non-dimensional properties to
write F(y) as:

F( y)  U 0  * f () where   y /  *

Here, U 0 is a representative velocity of the mean flow (such as the velocity at the edge of a
boundary layer, or the mean velocity in a pipe flow),  * is the displacement thickness for a
given mean velocity, and time is non-dimensionalized as t*  tU 0  * . We also linearize the
process by retaining only first-order perturbation terms (i.e. products of u', v', and u x ). After
substituting into the x- and y-direction Navier-Stokes equations, differentiating, and
consolidating to eliminate pressure terms, we end up with (note that it is not an easy process):

U  cf ' ' f   U' ' f  f ' ' ' '2 f ' ' f 
2 4
2
(17.5)
i Re *

U0 *
In equation 17.5, Re *  , is the Reynolds number based on the boundary layer

displacement thickness. Equation 17.5 is known as the Orr-Sommerfield equation for the two
scientists, W. Orr (1907) and A. Sommerfeld (1908) who independently derived it. See White
(1990), pp.342-345 for details of the derivation.

Equation 17.5 looks pretty complicated, and it is. Some early solutions of this equation were
developed analytically by Tollmien (1929) and Schlicting (1930). By the 1950’s, solutions using
numerical computational approaches were possible, with recent solutions improving markedly as

579
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

computational power has increased. However, the basic strategy for solving the equation has
remained the same. While the mathematical details of the solution are too complicated to discuss
here, and are better described elsewhere (Wazzan, 1975), the key elements are as follows.

Equation 17.5 is a homogeneous equation (all terms in the equation are functions of f and its
derivatives). Additionally, the boundary conditions are that the disturbances, u' and v', and
appropriate derivatives (all of which are reflected in the function f) must vanish at infinity and
bounding surfaces. This makes all the boundary conditions for f() also homogeneous. Thus,
we have a homogeneous differential equation with homogeneous boundary conditions, which
means this is an eigenvalue problem. What this means is that for any given mean velocity profile
U   there are select combinations of the parameters , c, and Re * that will be a solution for
Eq. 17.5 and its boundary conditions. The sequence of , c, and Re * combinations which yield
valid solutions are then the eigenvlaues of the eigenfunctions f   .

Still with me? After a lot of computations, we can determine a functional relationship for the
eigenvalues of the form:

c  c, Re * 

Or, since c is complex, and we are really concerned with which combination of parameters
(eigenvalues) cause the disturbances to grow, damp, or remain constant, what we are really
interested in is the value of ci. Thus, we can develop a functional relationship for ci of the form:

ci  ci , Re *  (17.6)

So, referring back to our discussion above about ci, if ci > 0 the disturbance will grow (become
unstable), and if ci < 0 the disturbance will diminish (remain stable). So, the key value we are
interested in is ci = 0, which is the demarcation between unstable and stable behavior. As Eq.
17.6 suggests, the easiest way to display this stability behavior is using a graph of  vs . Re * , on
which we plot lines of constant ci.

17.2.1 Stability of Bounded Flows

Flows that are bounded by solid surfaces are the more stable of laminar flows, maintaining their
laminar state to higher values of Reynolds number. Such flows are Poiseuille flows between
plates and in pipes, and boundary layer flows over flat or curved surfaces.

Figure 17.6 is a stability graph for a Blasius flat-plate boundary layer, which displays lines of
constant ci for a laminar flat plate boundary layer flow, as determined by solution of Eq. 17.5.

580
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

Of particular interest on this graph is the ci = 0 line, shown as a heavier line. What this line
represents is a demarcation between disturbances that will grow, and those that will be damped.
Recall that α is the wave number, or the inverse of the wave length of the disturbance. So, as α
U *
increases, the wave length decreases. Note that Re *  0 is fixed by the parameters of the

flow.

0.45
0.40
0.35
0.30
α -0.5
0.25
0
0.20 1.0

0.15 1.96
1.8
1.5
0.10

0.05

102 103 104 105

Figure 17.6 A stability graph, showing curves of constant temporal amplification rate for a
flat plate boundary layer. The values associated with the curves are ci x 102.
The heavy black line is the line of neutral stability. Values of ci < 0 are in the
stable regime, and ci > 0 are in the unstable regime (shaded).[after Wazzen
(1975)]

On figure 17.6, note that for low Re * the flow is stable to disturbances at all wave numbers, and
thus the flow will damp out any disturbance and remain laminar. However, above a certain
critical Re * (for a Blasius-type boundary layer profile Re crit  520) there is a certain range of
disturbance wave numbers that the flow will amplify, thus producing an instability (i.e. causing
the flow to transition to another state). As figure 17.6 shows, this range of amplified wave
numbers is limited, and this range shifts from higher to lower α (i.e. longer wave lengths) as
Re * increases.

581
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

Note that in Eq. 17.5 the right hand side of the equation reflects the viscous terms of the Navier-
Stokes equation, and is inversely proportional to Re * . Thus, when Re * is small, viscosity has a
strong damping effect on the flow. However, as Re * increases, the impact of viscosity is
reduced, and a given flow will be more susceptible to unstable behavior. This effect of Re * is
clearly shown in figure 17.6, which illustrates that disturbances are damped at low Re * , and
amplified as Re * increases beyond Re crit .

One might think that by judicious control the region of unstable growth (the shaded region in
figure 17.6) might be avoided, and the flow could remain laminar to quite large Re * . However,
while this might be possible in theory, disturbances in real flows will generally contain a broad
set of wavelengths (i.e. wave numbers), and thus one or more of the wave numbers will likely be
amplified as Re * increases. Additionally, although not shown in figure 17.6, as Re *   , the
range of unstable wave numbers tends to zero. Again, while this suggests that we might be able
to finesse the flow to high Reynolds numbers by judicious control of the disturbance type, in
practice a real disturbance generally will pass through a Reynolds number where at least one of
its wave numbers will become unstable.

When one considers internal duct-type Poiseuille flows, similar stability graphs can be
established. For a two-dimensional Poiseuille flow between parallel plates of spacing t,
U t
calculations by Nachtsheim (1964) showed that Re t , crit = 5767, where Re t  0 . This Re t , crit

based on linear stability theory is quite large, since experimental studies indicate transition
generally occurs around Re t  1000 . Additionally, for a Poiseuille flow in a pipe, where
U0D
Re D  , numerous studies using linear stability theory indicate the flow should remain

stable for all Re D , and that there is no Re D , crit value where one should anticipate an instability.
In contrast, experiments indicate transition occurring at Re D  2000 in practice. Although
practical transition Reynolds numbers can be pushed higher (on the order of Re D  105 ) by
careful control of inlet disturbances and surface roughness, it is generally assumed that the
stability of such ducted flows is subject to a stability process not properly modeled by linear
stability theory. For an nice review of boundary layer stability and transition see (Zaric, 1990).

17.2.2 Stability of Unbounded Flows

Using Eq. 17.5, it can be shown that if a laminar velocity profile U   contains an inflection
d2U
point, such that U    0 , this significantly reduces the stability of the flow. Consequently,
dy 2

582
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

flows with natural inflections in their mean velocity profile, such as jets, wakes, and boundary
layers with an adverse pressure gradient, will become unstable almost immediately, whereas
flows without an inflection, such as boundary layers with a favorable pressure gradient will
remain stable to higher Re * .

Unbounded flows such as jets, wakes, and shear flows (two adjacent flows moving at different
mean velocities), as shown in figure 17.7, all have inflections in their respective velocity profiles,
and thus are subject to an almost immediate development of instabilities. A wake, such as
encountered in the flow behind a bluff body, and a jet impinging into a quiescent or lower
velocity flow have Re crit  4, so these are inherently unstable for all practical flows. Moreover,
linear stability theory predicts that shear flows have a Re crit  0, and are unstable under all
conditions. Thus, all unbounded flows will begin the process of transition to turbulence
essentially immediately.

inflection
point grid
inflection elements
point
inflection
point

(a) (b) (c) (d)

Figure 17.7 Typical unbounded laminar flow velocity profiles:

(a) jet flow, (b) wake flow, (c) shear flow, (d) grid flow

A subset of unbounded flows is the stability of flows in the wake of a grid placed normal to an
impinging uniform flow. These types of flows consist of a series of transverse jets and wakes, as
shown in figure 17.7d. These flows again have a Re crit  4, and again are almost immediately
unstable. These types of flows, and thus cross-stream grids, are commonly employed at the
entrances of wind tunnels to actually reduce the scale of entering flow disturbances and promote
lower flow disturbances in the downstream flow. However, more on this in Section 17.3.

17.2.3 Effect of a Pressure Gradient on Flow Stability

As mentioned in section 17.2.2, a velocity profile containing an inflection is very susceptible to

instability. We noted in section 14.5.2 that a Blasius flat-plate boundary layer has an inflection
point located at the boundary surface. As was illustrated in figure 17.6, a flat plate flow may
become unstable starting at Re crit  520. However, as we illustrated in Section 14.6, if a

583
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

boundary layer is subject to an external pressure gradient, this can strongly influence the
behavior of the mean boundary layer velocity profile. For a boundary layer flow subject to an
adverse pressure gradient (increasing stream-wise pressure), the inflection point will move away
from the surface, which will make the flow strongly unstable. In contrast, for a boundary layer
flow subject to a favorable pressure gradient (decreasing stream-wise pressure), there will be no
inflection point in the velocity profile, and the flow will be more stable.

For example, consider the Falkner-Skan boundary layer flows we considered in section 13.6,
which examined boundary layers developing under the effect of pressure gradients ranging from
strongly adverse to strongly favorable. A linear stability evaluation of Re crit for the two most
extreme cases [Wazzan (1975)] indicates that for m = - 0.0905 (strongly adverse) the Re crit  67,
whereas for m = 1 (strongly favorable) the Re crit  12,500. Thus, an adverse pressure gradient
boundary layer is very unstable, and will usually break down to turbulence almost immediately;
in contrast, a strong positive pressure gradient boundary layer is very stable, and is unlikely to
transition to turbulence, even in the presence of strong disturbances. Thus, the stability of a
boundary layer depends quite strongly on the level and the direction of the external pressure
gradient.

So what do these stability concerns mean in terms of a flow degenerating to turbulence? Well, it
should be clear that a laminar flow, unless it is a bounded flow either at low Reynolds number or
under the influence of a favorable pressure gradient, will encounter disturbances (e.g. due to
surface roughness or extraneous disturbances, such as acoustic noise, thermal gradients, etc.) that
will naturally lead to an unstable change of the flow.

So, what happens when a laminar flow becomes unstable? As we discussed earlier in section
17.2, the flow will transition to another state, which eventually leads to what is known as a
turbulent flow. However, how that happens is again (of course) quite complicated, and subject to
significant study. In the following section we will discuss this generic process of transition from
a laminar to a turbulent flow.

17.3 The Transition to Turbulence

As was illustrated in Section 17.2, as Reynolds number increases a flow is more susceptible to
amplification of disturbances, and subsequent degeneration to turbulence. However, this process
of transition may be a relatively sudden process, when there is an inflection point in the velocity
profile; or it may be a less well-defined and more projected process, as with bounded flows. Part
of the problem is that once disturbances start to grow, they give rise to additional inflections in
the mean velocity, which can modify the way the original wave number amplifies, which may
allow other wave numbers to amplify. As you can imagine, the process of disturbance growth

584
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

becomes very complicated very quickly. The process of transition is dependent on the types of
initial disturbances to the flow; even carefully designed flow systems are subject to variances in
the initial disturbances due to small temporal variations in sound, vibration, density, and thermal
gradients, among other external influences. Thus, the amplification of disturbances, and the
transition to turbulence can occur in markedly different ways on the way to a turbulent state.

17.3.1 Transition of Unbounded Flows

The transition in unbounded flows is complicated, but is generally consistent in the way in which
it occurs. In the vicinity of an inflection point in the mean velocity profile, a process known as a
Kelvin-Helmholtz instability develops almost immediately. This instability is initially an
inviscid process, modeled as a pressure-driven instability developing at the sharp velocity
interface between two dissimilar flow streams, such as encountered at the initiation of a jet, a
shear layer, or a separating wake. These types of inflectional flows are unstable to all
disturbances, and initiate as transverse waves within the shear layer, which rapidly roll up into a
sequence of transverse (or azimuthal, in the case of a circular jet) vortices. Once the vortices
form, viscosity begins to play a role as well, which we will discuss.
Interface Between (B)
Outer Flow and Large Entrainment of
Scale Structures Fluid/Energy by
Large Scales

Flow 
(A)
Amalgamation of
Small-Scale Stretching of Small Scales by
Flow Structures Large Scale Large Scales, Leading to
into Larger Flow Structure Energy Dissipation
(C)
Structures
Small-Scale
Structure (3-D)

Figure 17.8 Schematic illustration of the simultaneous processes within a turbulent flow of
(A) small-scale to large-scale amalgamation, (B) entrainment of outer region
fluid/energy into the turbulent region by the large-scales, and (C) stretching of
small-scales by large-scales leading to energy dissipation.

Once transverse Kelvin-Helmholtz vortices form, they will translate at roughly the average
velocity between the adjacent stream velocities. So, for a jet of velocity U exiting into a
quiescent environment, these vortices will initially travel at roughly 0.5 U. As figure 17.8
schematically illustrates, once formed, a complicated process of both amalgamation and

585
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

distortion of the vortices takes place, with combinations of sequential vortices rolling up
(amalgamating) into larger vortical structures, while the vortices (and their associated vorticity)
simultaneously undergo three-dimensional distortion due to vortex stretching by the rotation of
the larger-scale vortical structures. This leads to both (a) a spatial growth in the size and
complexity of the vortex structures, and (b) a stretching and concentration of the vorticity
comprising the amalgamated vortices. The growth to larger scales is important, since this growth
process causes the entrainment of additional fluid from bordering non-turbulent flow, which
results in both the rapid mixing of the adjacent flows, and lateral spreading of the initial
disturbances. The stretching and concentration of the vorticity within the amalgamated
structures accomplishes two things: (1) a transfer of kinetic energy from the motion of the larger
rotating structures to smaller three-dimensional vortices contained within or adjacent to these
larger structures, and (2) reduction in scale and rotational intensification of the smaller vortices.
This process is again important since the smaller scale vortices will contain very large shear
gradients, which will cause their kinetic energy to dissipate as heat through viscous effects.

To illustrate how the stretching of smaller-scale vortices passes energy from larger-scale flow
structures to smaller scales, consider a simple illustration. Assume that the smaller scale vortices
are modeled as a vortex tube with solid body rotation, as shown in Figure 17.9.



initial
stretching of
vortex tube  stretched

L2
stretched > initial
L
1

(a) initial vortex tube (b) stretched vortex tube

Figure 17.9 The effect of stretching on a vortex tube of constant cross-section

For simplicity, consider the angular momentum of the vortex tube to be modeled as a rotating
rod with a constant angular velocity (i.e. vorticity) throughout the tube, such that:

Angular momentum = I
and
1 2
Kinetic energy = I
2
1 2
Here I  mr is the moment of inertia a rod rotating along its axis, and m is the rod mass, and r
2
is the rod radius. Now, let an initial section of the rotating vortex tube of radius r1, and length L1

586
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

be stretched to a length L2, where L2 > L1. Assume (ideally) that angular momentum is
conserved during the stretching process, such that:

1 1
I11  I 2  2  m1 r12 1  m 2 r22  2
2 2

Since the mass of the tube will remain constant, m1 = m2, we have:

2 2 r12
r 1  r  2
1 2   2  1 2
r2

Note that we would also get this result assuming that circulation is constant during the process.

Since the volume of the stretched tube will remain constant, we can also write that:

L1
r12 L1  r22 L 2  r2  r1
L2

Substituting into the previous equation, we have:

r12 L
 2  1  1 2
r22
L1

So, as the ideal vortex tube is stretched, the vorticity is increased (we showed this previously in
section 10.3.1).

Now the angular kinetic energy of the initial vortex tube and the stretched vortex tube will be:

1 2 2
KE 1  mr1 1
2
2
1 1 L  2  L2  1 L  L 
KE 2  m 2 r22  22  m r12  1 1    m r12 12  2   KE 1  2 
2 2  L2   L1  2  L1   L1 

Thus, when the vortex tube is stretched such that L2 > L1, the kinetic energy will increase
proportional to the increase in length.

Since kinetic energy is fed from the larger rotational structures to these smaller vortices, the
dissipation process within the small vortices provides a mechanism for the transfer and
dissipation of the kinetic energy of the overall flow. Since this process is like a waterfall

587
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

dissipating its energy as it plunges into a quiescent pool, the process of dissipation of the initial
kinetic energy of a laminar flow by turbulence is generally known as the energy cascade.

Interestingly, this process contains two divergent, but symbiotic processes: (1) the growth of
larger, complex scales which extract energy from the main flow, coupled with (2) the generation
of continually smaller scales, which subsequently dissipate the extracted energy. As any
turbulence expert would argue, this is an oversimplification of the process, but it does capture the
essence of what nature is physically trying to achieve.

An excellent example of these processes of scale change and energy dissipation in action is the
dissipation of the kinetic energy of a wall jet in a hot tub or swimming pool. If you place your
hand in front of one of the jets, you will notice that when your hand is very close to the jet's
origin, you will feel a pressure on your hand that is relatively uniform, with some small scale
fluctuations. However, as you move your hand away from the origin, your hand will experience
larger and more intermittent pulsations in the pressure (due to the growth in the larger vortical
structures), but a reduction in the average pressure, as the jet kinetic energy is dissipated by the
smaller vortices. As you move your hand even farther from the origin, the pulsations will
become of even larger extent, but also weaker, as the large scales grow, but the total energy of
the jet is dissipated. Eventually, at a distance well removed from the origin, this vortex
interaction succeeds in both distributing the concentrated energy of the initial jet over a very
broad area, and dissipating it via the small scales through this complicated energy cascade. At
that point, it is unlikely that your hand can detect much of the previous momentum of the initial
jet.

A similar type of process applies for wakes, which are generally created by flows over bluff
bodies. However, wake flows are often dominated by large scale inviscid instabilities which
may initiate periodic vortex shedding, prior to a change of state to a randomly turbulent flow.
For example, the vortex shedding from circular cylinder, as shown in figure 17.3c and discussed
in Section 15.4, will develop and remain laminar well before turbulent behavior develops. And
when turbulent behavior does develop (generally around Re > 200, where Re is based on the
width of the bluff body), it will first develop within the shed vortices, rendering them subject to
small-scale three-dimensional vortex stretching, leading to kinetic energy dissipation within the
shed vortex. At much higher Reynolds numbers (roughly Re > 3x105), the shear layer exiting the
bluff body will break down to turbulence prior to formation of the shedding vortices, and thus
mediate, and eventually eliminate the periodic shedding process.

Recall that wakes, like jets, were predicted by linear stability theory to become unstable at
Re  4. As figure 17.3 illustrates, the flow over a cylinder does just that—but it transitions to a
different stable state, not to turbulence. Thus, a transition from an initial laminar flow of one

588
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

state can result in a different, but still laminar, state. Turbulence is not always the result of a
transition of a laminar flow, but it is the most likely state.

Flow across a grid of wires presents an interesting process. One can think of this as producing a
spatially periodic distribution of jets and wakes, like that shown in figure 17.7d, and thus a
condition very sensitive to instability. However, unlike a jet, wake, or shear flow, the scales of
this type of flow start basically at the spacing of the grid and slowly decay with distance, and do
not extract additional kinetic energy from the mean flow (once passing through the grid). The
reason for this is that all initial inflections and grid-created shear layers are off setting, and there
are no significant velocity differences beyond the scale of the grid. Thus, the vortex structures
within the flow are generated initially at the scale of the grid, and while they may expand and
grow with distance, there can be no entrainment of additional kinetic energy from the mean flow.
Additionally, the vorticity within this flow will quickly reorient in three-dimensions such that
energy dissipation takes place in an almost isotropic manner (i.e. kinetic energy will be
dissipated uniformly by vortices/vorticity oriented roughly uniformly in all three-dimensions).
Since the substantive vortex stretching occurring in jets, wakes, and shear layers does not occur
with grid turbulence, the initial kinetic energy imparted to the instabilities by the flow interaction
with the grid will dissipate slowly with downstream distance.

A nice aspect of a grid flow is that if a flow approaching a grid contains disturbances/instabilities
(e.g. large vortices) that are greater than the grid scale size, after passing through the grid these
disturbances will be reduced to the scale of the grid instability, and will dissipate out of the flow
at a rate corresponding to that of the grid disturbances. This process of disturbance reduction is
put to good use in reducing the level of disturbances and turbulence entering wind tunnels, or
any other ducted flow, where minimal flow fluctuations are important. That is why all good
wind tunnel facilities are built with both flow straighteners (closely-packed collections of tubes
oriented in the streamwise direction), followed by a set of several screens, often of sequentially
smaller grid size. This NASA technical brief discusses the use to screens in wind tunnels to
control inlet turbulence levels.

17.3.2 Transition of Bounded Flows: Natural

Although the initiation of transition for unbounded flows begins essentially at their point of
formation (due to their very unstable nature), the initiation point and extent of transition for
bounded flows is much less defined. Referring back to figure 17.1, the precise location of the
point where transition begins and where it ends is very fuzzy, and dependent on the initial
conditions of the flow and disturbance types present. However, a remarkable observation, both
experimentally, and more recently computationally, is that the end state of turbulence for most
flows is quite consistent, and somewhat predictable for a wide variety of initially laminar
conditions. Once a full-turbulent flow has developed, the statistics and the behavior of the flow

589
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

are remarkably consistent, and in most cases ignorant of the initial conditions which precipitated
the development of the turbulent flow. In other words, once a bounded flow achieves a
developed state of turbulence, the method by which the flow evolved to a turbulent state is
generally irrelevant.

So, what does happen during the transition process, and how does a bounded flow go from being
fully laminar to fully turbulent? To understand the beginning of transition, it would help to
know what actually happens as a disturbance amplifies within a laminar bounded flow. In
section 17.2 we modeled the disturbances as a traveling wave, with a specific wave number and
wave velocity. This allowed the development of a stability graph, such as figure 17.6, showing
regions of damping and amplification of such waves for a specific type of flow. To test the
validity of the traveling wave model, and the resultant theoretical predictions of wave growth,
Schubauer and Skramstad (1947) performed some very meticulous experiments on laminar
boundary layers in a large wind tunnel facility. Using a thin, magnetically-excited metal ribbon
located close to the boundary surface, they generated periodic waves of varying wave number
and amplitude, and demonstrated the validity of the stability graph shown in figure 17.6. The
single wave number disturbances Schubauer and Skramstad generated, and tracked, are known as
Tollmien-Schlicting waves, after the two researchers who developed the first stability solutions
of Eq. 17.5.

An important, if not more important experimental study, was that done by Klebanoff et al.
(1962), where they showed that following the wave generation, initially two-dimensional
Tollmien-Schlicting waves very rapidly become three-dimensional, developing a transverse
peak-and-valley velocity profile very near the bounding surface, as shown schematically in
region 2-3 in Figure 17.10. It was subsequently shown that this peak-valley behavior is the result

Figure 17.10 Transition mechanism on a flat plate. (Source: H. Schlichting et al, "Boundary
–Layer Theory", McGraw-Hill, 2006)

590
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

of the longitudinal (stream-wise) tilting and stretching of initially transverse boundary-layer

vorticity into vortices in the stream-wise direction. This three-dimensionality is accentuated with
stream-wise distance, until the stretched vortices begin to interact into larger three-dimensional
agglomerations of vorticity, while simultaneously breaking down into smaller units of vorticity,
as shown in region 4 of figure 17.10. This process is similar to that noted for the breakdown of
unbounded flows, with (1) the larger structures interacting with the outer flow at the edge of the
boundary layer to entrain kinetic energy, while (2) the smaller vorticity units are stretched to a
small enough scale where effective viscous dissipation of the kinetic energy can take place.

The experiments by Schubauer and Skramstad (1947) and Klebanoff et al. (1962) were done in
extremely controlled environments, starting with two-dimensional, single wave-number
disturbances. In most real systems, as we discussed, naturally occurring disturbances within a
laminar flow are neither two-dimensional, nor of a single wave number. Consequently,
transition does not start uniformly at a specific Re * . Thus, the breakdown to turbulence in a real
laminar flow will begin somewhat sporadically, developing in what are termed turbulent "spots",
shown schematically as region 5 of figure 17.10. Turbulent spots were first observed by
Emmons (1951) as localized regions of turbulence, which develop naturally and apparently
randomly as localized regions of turbulence. Once formed, these turbulent spots travel
downstream, expanding in both the stream-wise and span-wise directions. Interestingly, the
leading edge of a spot travels more rapidly (0.9 U∞, where U∞ is the velocity of the outer edge of
the boundary layer) than the trailing edge (0.5 U∞), thus expanding the stream-wise extent of the
spot. A spot also acts as a moving disturbance, which entrains, disturbs, and otherwise interacts
strongly with fluid adjacent to it, resulting in a transverse expansion of the spot, spreading at half
angles of 8° to 12º relative to its steam-wise axis.

As turbulent spots develop and expand, both in the stream-wise and transverse directions, the
manifold spots that initially develop will eventually join, or coalesce. When all regions of the
flow are encompassed by turbulent behavior, the flow is considered fully turbulent, shown
schematically as region 6 of figure 17.10, and transition to turbulence is considered to have been
completed. At the end of transition (a rather fuzzy demarcation), boundary layer turbulence is
remarkable in that it becomes what is known as self-sustaining. What this means is that once a
flow becomes fully turbulent, the flow will remain turbulent, and perpetuate the complex three-
dimensional generation, agglomeration, and breakdown of vorticity which is turbulence, as long
as the flow remains bounded.

Now figure 17.10 may seem a bit busy, and possibly confusing. However, it is a simplification
of the complicated processes of transition to turbulence. To see a simulation of the process of
initial transition to turbulent spots to early turbulence, view this video by Lee and Zaki (2015). It
is a remarkable numerical simulation that shows the manifold complexity of the transition of a
flat plate boundary layer from laminar to early turbulence. Although a simulation, I am guessing

591
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

that this is close to what actually happens in reality. I’m sure you will find this simulation
impressive, fascinating, and very intimidating with regard to the complexity of the behavior.

17.3.3 Transition of Bounded Flows: Forced

While naturally transitioning bounded flows depend on the initial presence of disturbances to
initiate the process, the point of initiation will be somewhat sporadic. However, the process of
transition can be forced to occur by the use of surface roughness, either natural or artificially
applied. One can generally categorize roughness into two types (White (1991)): (1) two-
dimensional roughness, usually comprised of a wire or rod placed on the bounding surface
transverse to the flow direction, and (2) three-dimensional roughness, which can be a sphere,
hemisphere, sand grains, or similar elements, located either singly or multiply distributed on the
bounding surface. Interestingly, the effect of two- and three-dimensional roughness on transition
is quite different.

Generally, small amounts of roughness of any type will not affect the transition process or
location unless it is of significant height relative to the local displacement thickness, *. Letting
the height of the roughness be given by k, for a two-dimensional roughness, if k/* is less than
0.3, the roughness will have essentially no effect on the location of transition. However, for k/*
> 0.3, the point of transition will move upstream closer to the roughness element. It has been
shown that the minimal Reynolds number based on the roughness height for "tripping" a laminar
flat plate boundary layer to initiate transition is (roughly) [White (1991)]:

Uk
Re k   850


However, as Re k increases, the point of transition will move back toward the trip. To have
transition occur immediately following the roughness element requires Re k  2800 . These are
only rough rules of thumb, since there is no consistent agreement in the literature.

For three-dimensional roughness, little effect on transition is expected before k/*  0.6, about
twice that for two-dimensional elements.

17.4 The Character of Turbulence

As you can guess, once a flow transitions to turbulence, the behavior is, as we pointed out in
section 17.1 (after R. W. Stewart), a process of "disorder, enhanced mixing, and three-
dimensional motions involving angular momentum (vorticity)". The character of turbulence,
often called the "structure,” is such that it does satisfy these characteristics. However, as we
discussed in section 17.3, there are identifiable ways that turbulence evolves which have a
particular rhyme and reason. In this section, we will review a bit of this structure, which will
help motivate our approach to modeling the turbulence process. Be aware, however, that the

592
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

study of turbulent structure is a topic that is still widely studied and debated, and that the small
amount that we discuss here is just the tip of a very complicated iceberg.

First, through hypothesis and experiment turbulence has been shown to have a flow structure that
extends over a range of scales of motion. After the breakdown of a laminar flow, turbulence
results in the generation of both large- and small-scale flow structures, both of which have a
purpose. Following either the inviscid-type (Kelvin-Helmholtz) breakdown of unbounded shear
layers, or the amplification of Tollmien-Schlicting waves in a boundary layer, the initial flow
structures (usually vortical) proceed in two somewhat divergent directions: (1) amalgamation
with other structures, to yield larger scales, and (2) stretching, tilting, and intensification, to yield
progressively smaller scale vortical structures. The larger scales usually develop adjacent to the
interface between the turbulence region and the main outer flow (which is generally non-
turbulent). These larger scales interact with the outer flow to entrain more fluid into the
turbulent region, which increases the extent of, and adds kinetic energy to, the turbulent region.

While the large scales are of the order of the thickness of the turbulent region (e.g. the boundary
layer thickness, ), the small dissipation scales, known as Kolmogorov length scales, are of size
1 1 3
  3  4 L  3  4    4
L   3  or   3 3      Re 0.75 , where  is the thickness of the turbulent
U   U    U
region (e.g. the boundary layer thickness for a turbulent boundary layer). To give this some
perspective, a boundary layer of thickness  = 3 cm in water flowing at 1 m/s and kinematic
viscosity of 10-6 m2/s has a Re   30,000 . For this flow the corresponding scale of the
L
dissipating flow structures is  0.00044 , or L=0.013 mm. Small scales indeed!


In section 17.3.2, we mentioned that a turbulent boundary layer flow is a self-sustaining flow.
What this means is once a boundary layer flow achieves a turbulent state, it not only remains
turbulent, but sustains the turbulent state, as opposed to dissipating away like an unbounded
flow. We mentioned that the larger scales of boundary layer turbulence facilitate the
introduction of new kinetic energy into the boundary layer from the outer flow; but while this is
necessary, it is not sufficient to sustain the continuing turbulence. The sustaining mechanism of
boundary layer turbulence is a process of interaction of the fluid adjacent to the solid boundary
known as "bursting". Bursting is an intermittent process of fluid ejection from very near the
boundary that introduces additional vorticity into the boundary layer. Note that since the total
vorticity (i.e. the circulation) within the boundary layer must remain essentially constant, this
bursting process introduces, on average, a balanced quantity of vorticity of both signs into the
flow. This offsetting collection of discrete vorticity (1) is eventually stretched and intensified,
subsequently dissipating the energy at these smaller scales, or (2) undergoes a complicated
process of amalgamation with other structures of both like and opposite signs of rotation, which
supports the development and growth of the larger outer-region structures. Despite the fact that

593
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

these bursting processes initiate at relatively small scales, the process is so prolific near the
boundary surface that collectively it becomes the source of continuing new vorticity to sustain
boundary layer turbulence.

Outer Region Structures

Turbulent Bursting

(a) Side-view

(b) Plan-view

Figure 17.11 Hydrogen bubble flow visualization of a low Reynolds number turbulent
boundary layer. Water flow: Rex = 2.2x105 and Re* = 746. (a) Side view
illustrating turbulent "bursting" near the bounding surface, and development
of large outer structures. (b) Plan-view illustrating alternating "streak" like
behavior (indicated by arrows ) near the surface (visualization wire located
at y+ = 5 —see section 17.6.1 for discussion of y+ scaling).

594
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

As mentioned, this bursting process occurs intermittently, in a somewhat random, but repetitive
process, with fluid appearing to erupt from very near the boundary into the boundary layer fluid.
Of course, if fluid is ejected from the boundary, by conservation of mass there must be a
comparable amount of fluid that moves toward the surface. The large outer region flow
structures entrain high-energy fluid from the edge of the boundary layer, which penetrates down
to the bounding surface, creating what is termed a high-speed "sweep". Moreover, although
Stewart speaks of disorder in turbulence, the process of fluid moving toward the surface, and the
bursting of fluid away from the surface, are somewhat organized processes. The turbulence
production process near the surface develops as almost parallel flows toward and away from the
surface, which gives rise to what appears to be a "streak" like pattern very near the surface, with
higher-velocity regions flanked by low-velocity regions. This is not dissimilar from the three-
dimensional distortion of Tollmien-Schlicting waves during the transition process, discussed in
section 17.3.2.

Now, since the low-velocity streak regions are adjacent to higher-velocity flows immediately
above the streaks, this creates an inflection in the local mean velocity profile, which is of course
unstable. The breakdown of this velocity inflection results in the "burst" of fluid and vorticity
into the boundary layer. This breakdown is very sporadic, but repetitive, providing the new
vorticity to sustain the boundary layer. Additionally, although the process is still widely debated,
the resulting interaction adjacent to the boundary during the streak destabilization process is such
that it sustains the continuing re-formation of the high- and low-velocity streak regions.

Figure 17.11 shows two hydrogen bubble visualization pictures (in a water flow) which reveal
some of the flow structure of a turbulent boundary layer discussed in this section (click here to
see an in depth presentation on hydrogen bubble flow visualization--a video from an older film,
so a bit grainy, but informative). Very fine time-lines of hydrogen bubbles are generated at high
frequency by electrolysis from a very fine wire (on the far left of each picture). Illumination of
the subsequent lines of bubbles reveals the deformation of the local flow within the boundary
layer. Figure 17.11(a) is a side-view, with a wire oriented vertically from the bounding wall and
spanning the boundary layer. Here, the bubble time-line bubble lines reveal patterns
characteristic of both bursting behavior (combined with vortex stretching) near the surface, and
larger outer-region vortical structures. Figure 17.11(b) is a plan-view, with the wire stretched
transverse across the plate and very near the surface. This bubble time-line image reveals the
stream-wise streak-like behavior, indicative of the vertical inflow and outflow of fluid adjacent
to the plate boundary.

595
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

Figure 17.12 A general model of the turbulence processes sustaining a developed turbulent
boundary. Not to scale.

Figure 17.12 is a simple model illustrating the basic behavior within a turbulent boundary layer.
This figure outlines the basic processes which sustain and maintain the turbulence process. The
key elements are: (1) the bursting process occurring near the boundary, which sustains the
turbulence, (2) the amalgamation of these initial burst-generated structures to form larger, outer
region structures, (3) the three-dimensional stretching of small-scale turbulence by these larger
outer region structures, which both reduces the scales and dissipates energy, and (4) the inward
entrainment of fluid which is manifested as a sweep of higher-speed fluid toward the boundary to
replace the ejected bursting fluid.

So, boundary layer turbulence sustains itself by creating both large- and small-scale flow
structures, which respectively entrain more fluid/kinetic energy and ultimately dissipate this
kinetic energy. A quasi-organized process of fluid bursting/streak formation adjacent to the
bounding surface introduces new vorticity to the boundary layer, which is stretched, yielding
small scales that dissipate kinetic energy, and/or amalgamate into larger flow structures that
maintain the entrainment process. This is a bit of a simplistic description of a very complex
process, but it captures the essence of the turbulence process in a bounded flow, and is not
dissimilar from the process for an unbounded turbulent flow. A more detailed discussion of the

596
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

hypothesized processes and structure of a turbulent boundary layer can be found in Smith and
Walker (1997).

17.5 Mathematical Modeling of Turbulent Flow

So how do we take this generic, and complicated, flow process discussed in section 17.4, and
model it such that we can predict properties of engineering interest, such as mean velocity
behavior and surface shear stress? Obviously, turbulence is not a steady environment, yet it is not
exactly unsteady. Reviewing the behavior of local velocity measurements, such as those
illustrated in figure 17.2 for a turbulent boundary layer, it would appear that the collective large
and small-scale vortical interactions yield a velocity behavior that appears as a random local
fluctuation about some average, or mean, value. Therefore, a very simplified modeling approach
that views the velocity behavior within a turbulent flow as a combination of a mean part and a
fluctuating part seems promising, and reflective of the "disorder" or randomness of turbulence.
Now this approach of course discounts all the flow structure we alluded to in section 17.4.
However, modeling such a semi-organized process of three-dimensional vortical flow structure is
quite complicated, and very computationally intensive. So, we opt to make our model of
turbulence as simple as possible. This approach, by the way, is the same approach suggested by
Osborne Reynolds (1883), who was the first to observe and document the breakdown of
turbulence using observations of the flow of water in a pipe (and for whom the Reynolds number
is named).

Here, we use a statistical approach, which assumes that the velocity at a point in a turbulent flow
can be modeled as a set of random velocity fluctuations about a local mean value. For a velocity
probe measuring only the x-direction component, u, we would characterize the velocity as shown
in figure 17.13.
 fluctuation about the mean
u t   U  u t 

average fluctuation u
 time

time

Figure 17.13 Representation of turbulent velocity as combination of a mean component, U,

plus a fluctuating component, u  .

Thus, the mean component of the velocity, U, is the time average over a period T, where T is a
sufficient period for the mean to reach a constant value, given as:

597
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

T
1
U   udt
T0

The fluctuating component of the velocity, u  , is the instantaneous velocity variation about the
mean value, which can be viewed as u t   u t   U . Since u  is some random function of
time, we often use the root mean square of u  to characterize the level of the unsteadiness, such
that:
1
1 
T 2 ___
u rms   u  2 dt   u2
T 0 

urms is generally termed the turbulence intensity, and is often used, particularly in experimental
measurements, to characterize the magnitude of the turbulence fluctuations.

Similar to the way we choose to characterize the u velocity, we use the same approach for the y
and z direction velocities, such that:

v  V  v , w  W  w 

Recall that we applied a similar approach to the velocity behavior in section 17.2, where we
considered the stability of laminar flows. For the case of stability, we made an assumption for
the behavior of the unsteady components that they are periodic and subject to either damping or
amplification. For turbulence, we cannot make assumptions of periodicity or
amplification/damping, but only that the fluctuating velocity components are time dependent.

Since we want to determine the average mean velocity behavior of a flow, and the resultant
average stresses associated with the flow, our approach is as follows:

1) Substitute the time-dependent velocities into the differential equations of motion (i.e.
continuity and Navier-Stokes)

2) Time-average the equations

3) Drop terms which average to zero

T T T
1 1 U
e.g. u    u dt  0 and U u    Uu dt   u dt  0
T0 T0 T0

4) Solve the resulting equation.

This process of averaging is known as Reynolds averaging (often termed RANS, for Reynolds-
Averaged Navier-Stokes equations), after the work by O. Reynolds (1895), and results in a set of

598
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

equations that can potentially be solved for the average turbulent velocity behavior for a flow of
interest (more on this later).

17.5.1 Reynolds Averaged Equations

In this section, we will develop the Reynolds averaged equations in two dimensions. This
approach is, of course, generically expandable to three dimensions (see Eqs. 17.18 below).
However, the results for two-dimensions will suffice for our purposes of introduction. Here, we
develop the x- and y-component equations, where U and V are the mean velocity components,
and u' and v' are the fluctuating components. We make the same assumption for the pressure,
such that we can write the total velocities and pressure for a two-dimensional turbulent flow as:

u  U  u' , v  V  v' , p  P  p' (17.7)

Substituting Eqs. 17.7 into the x-direction Navier-Stokes equation, Eq. 5.46b (reduced to two
dimensions), and expanding gives:

U  u' U  u'  V  v' U  u'   1 P  p'    U 2 u'    U 2 u'
2 2

x y  x x y
Or expanding,

U U u' u' U U u' u'

U  u' U  u' V  v' V  v'
x x x x y y y y
1 P 1 p' 2
 U  u'
2 2
 U  u'
2
   2  2  2  2
 x  x x x y y
If we time average this equation, the time averages of fluctuating terms, or the product of a mean
term and a fluctuating term, will average to zero, such that:
0 0 0 0

U U u ' u ' U U u ' u '

U  u' U  u' V  v' V  v'
x x x x y y y y
0 0 0

1 P 1 p' 2U  2 u' 2U  2 u'

   2  2  2  2
 x  x x x y y

Since U, V and their derivatives are time-mean terms, this means that we can drop the averaging
terminology indicated by the over bar on terms that contain only mean velocity values. So, we
are left with:

599
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

U u' U u' 1 P 2U 2U

U  u' V  v'   2  2
x x y y  x x y

Rearranging, this equation such that all the time-averaged fluctuation terms are moved to the
right side of the equation gives:

U U 1 P 2U 2U u ' u '

U V    2   2  u'  v' (17.8)
x y  x x y x y

Now, substituting Eqs. 17.7 into the two-dimensional continuity equation, Eq. 5.46a (reduced to
two dimensions), and expanding gives:

u v U  u ' V  v' U V u ' v'

 0      (17.9)
x y x y x y x y

Time averaging Eq.17.9 gives:

0 0
U V u ' v'
   0
x y x y

Since the time averages of the mean velocity derivatives must also satisfy continuity, we have:

U V U V
   0 (17.10)
x y x y

So, subtracting Eq. 17.10 From Eq. 17.9 gives:

u ' v'
 0 (17.11)
x y
u '
Now, we rewrite the last two terms of Eq. 17.8, by adding and subtracting u '
x

u ' u ' u ' u ' u ' u ' u ' u ' u '

 u'  v'  u '  v'  u'  u'   2u '  v'  u' (17.12)
x y x y x x x y x

u ' v'
Substituting  from Eq.17.11 into Eq. 17.12, and rearranging the terms gives:
x y

 u'
u '
 v'
u '
 2u '
u '
 v'
u '
 u'
v'

 
 u '2 u ' v'
 (17.13)
x y x y y x y

600
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

Substituting Eq. 17.13 back into Eq. 17.8 gives:

U
U
V
U

1 P 2U  2 U u ' v'  u ' 2
 2  2  
  (17.14)
x y  x x y y x

Similarly, if we apply the same assumptions of fluctuating properties and time averaging to the
y-direction Navier-Stokes equation, Eq. 5.46c (in two-dimensions), we get:

U
V
V
V

1 P  2V  2 V u' v'  v'2
 2  2  
  (17.15)
x y  y x y x y

Assuming that our fluctuation model is appropriate, Eqs. 17.14 and 17.15 are equations for the
mean velocity components U and V. Note that these equations differ from those for a laminar
flow by the last two terms in each equation. These last two terms account for the effects of the
velocity fluctuations on the mean flow due to cross-stream momentum exchange caused by the
fluctuations. We learned in Section 6.3.3 and in Section 11.4.1, that the introduction of a flow
normal to the direction of the mean flow, by suction through a porous surface, results in an
increase in the local shear stress at the suction surface due to additional momentum changes.
Recall that the shear stress developed in section 6.3.3 for a Poiseuille-type flow with transverse
injection-suction, was a rather complicated term, which included both the effects of conventional
shear and the effects of cross-stream momentum exchange due to the fluid suction. The shear
stress developed in section 11.4.1 for a fully-developed boundary layer with wall suction was
shown to reduce to the cross-stream momentum changes across the transpiration boundary layer.
The last two terms in Eqs. 17.14 and 17.15 are of a similar origin—cross-stream momentum
exchanges resulting from time-dependent variations of u and v from the local mean velocities, U
and V.

Note that the first of these two terms is due to cross-stream momentum exchange (since v' is
normal to u'), whereas the last tem is due to inline fluctuations e.g. u  2  u   u , where the
momentum changes are due to the fluctuations in the direction of the mean flow. Normally,
cross-stream momentum changes dominate inline momentum changes, but not always, as in the
case of isotropic grid flows (i.e. flows in the wake of a crossed-wire grid). However, for most
boundary layer, jet, and simple wake flows, where the u-velocity changes most strongly in the y-

direction (cross stream) it is can be shown that

u ' v'

 
 u' 2
and
u ' v'

 
 v' 2
.
y x x y

To illustrate this, recall our order of magnitude analysis for a boundary layer (Section 13.2),
where we argued we could eliminate terms from the two-dimensional Navier-Stokes and
continuity equations to yield the two-dimensional boundary layer equations:

601
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

u u 1 p  2u
u v   2
x y  x y
(13.13a)
And
u v
 0
x y
(13.13b)

Now, consider a similar order of magnitude analysis of the latter two fluctuation velocity terms
of Eq. 17.14 for a boundary layer-type flow. We first assume that the velocity fluctuations in
both the x and y directions are of comparable order, since the fluctuating velocity components in
turbulence are somewhat isotropic. Similar to what was done in section 13.2, we now make the
assumption that:

x  0(L), y  0(), u'  0(L), v'  0(L)

Here L is a characteristic length of development along the boundary layer surface, and  is the
corresponding boundary layer thickness at that location, such that 0(L) >> 0(),
Thus, we can assess that:


u' v'  u'2

 

0L 0( L) 0L 0( L)

 0L  
 0L   1
y x 0() 0( L)  0() 

0L  u ' v' u ' 2 

We note that  1 , and by comparison  . This assessment has been
0() y x
proven to be valid experimentally, with the exception of grid turbulence flows and boundary
layer flows with a large adverse pressure gradient. So, to a good approximation we can neglect

the
 
 u'2
term, and write Eq.17.14 as:
x
U U 1 P 2U  2 U  u ' v'
U V   2  2  (17.16)
x y  x x y y

Likewise, it can be shown that

 
 v'2
in Eq. 17.15 can be neglected, since
u ' v'

 
 v' 2
.
y x y
This reduces Eq. 17.15 to:

V V 1 P  2V  2 V  u ' v'
U V   2  2  (17.17)
x y  y x y x

602
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

Note that when we consider all three dimensions of a turbulent flow, we must consider an
additional velocity component in the z (transverse) direction, given by w  W  w' . For such a
three-dimensional flow we have three component equations given by:

U
U
V
U
W
U

1 P 2U 2U
 2  2  2  
 
 2 U  u ' 2 u ' v' u ' w '
 (17.18a)
x y z  x x y z x y z

U
V
V
V
W
V

1 P 2V 2V
 2  2  2   
 
 2 V u ' v'  v' 2  v' w ' (17.18b)
x y z  y x y z x y z

U
W
V
W
W
W

1 P

2W
 
2W
    
 
 2 W u ' w ' v' w '  w ' 2 (17.18c)
x y z  z x 2 y 2 z 2 x y z

17.5.2 Turbulent Reynolds "Stresses"

Equations 17.18 are the Navier-Stokes equations with nine additional terms due to the time-
averaged unsteady velocity components. Recall that these nine terms originate from the
momentum terms in the Navier-Stokes equation. As we alluded to above, these fluctuation
momentum terms result in additional forces on the fluid, which significantly modify the mean
velocity profile. However, we do not know what these fluctuation terms are, so we have a
quandary, with many more unknowns than we have equations. Since we only have four
equations (continuity and the three Navier-Stokes components), we must find some way to solve
for, or model, the nine (actually six, as we will discuss below) fluctuation terms. Trying to add
higher-order equations, such as the energy equation, just makes the process more complicated,
and doesn't lead to closure.

So, how do we proceed with modeling the fluctuation terms? Well, moving the fluctuation terms
to the right side of the Navier-Stokes equation reveals a strong similarity to stress components.
To examine this, let's consider Eq. 17.16, the x-direction N-S equation applied to a boundary
layer-type flow.

We note that the last two terms in Eq. 17.16 can be written as:

 2 U u ' v' 1   U 

     u ' v' (17.19)
y 2
y  y  y 
U
Since for a laminar flow  yx  , we could infer that the other component of Eq. 17.19
lam y
could be construed as a "turbulent" component of stress due to the fluctuating behavior, or:

603
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

 yx  u' v' (17.20)

turb

This approach was taken by early turbulence researchers, initiated by O. Reynolds, and thus
these apparent stresses due to the fluctuating velocity components are typically known as
"Reynolds" stresses. Now for three-dimensional flows, the x-direction Eq. 17.18a contains three
Reynolds stresses:

 xx turb
 
  u ' 2 ,  yx
turb
 u' v' , and  zx turb
 u ' w '

Collectively, in all three Eqs.17.18, there are a total of six different Reynolds stresses:

 xx turb
 
  u ' 2 ,  yy
turb
 
  v' 2 ,  zz turb
 
  w' 2

 yx  u' v'   xy ,  yz  v' w'   zy ,  zx turb

 u ' w '   xz turb
turb turb turb turb

Note that using the equilibrium argument of section 5.6 allows the equating of the shear in
adjacent planes, which reduces the Reynolds stresses from nine to six. However, this still leaves
us far from a tractable set of equations.

Now, let's examine how fluctuations in velocity give rise to these added stresses. To do so, let's
just consider a two-dimensional boundary layer-type flow, such that the x-direction stresses can
be modeled as the sum of the laminar and turbulent stresses, as shown in Eq. 17.21:

 yx   yx   yx
lam turb

U ____
   u v y (17.21)
y

Reynolds stress

u'
2
v'
1

Figure 17.14 Motion of a particle: (a) upward from a region of low u-velocity (1) to a
region of high u-velocity (2), or (b) downward from a region of high u-
velocity (2) to a region of low u-velocity(1).

604
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

Using the simplified model of turbulent fluid motions shown in figure 17.14, we can infer how the
motion of a fluid particle through a velocity gradient results in the generation of an added “stress”.

First, consider a fluid particle moving upward from (1) to (2). Since v > 0, and u1 < u2 , then the
fluctuating velocities for this particle would be:

v  0
u  0 (since the arriving particle has a velocity deficit i.e. u   u 1  u 2 )
 uv  0   turb   uv  0

Thus, a particle arriving from (1) will be accelerated to the u-velocity at (2). This requires
momentum being transferred from the flow to the particle, which is the result of an applied stress
equivalent to  uv on the particle by the flow at (2). Conversely, the flow at (2) will be
decelerated by the arrival of the velocity-deficit particle from (1), which is the result of an equal
and opposite stress applied by the particle on the flow at (2).

Since continuity requires that every outflow must be balanced by an inflow, the upward
movement of a particle from (1) to (2) must be balanced by a corresponding downward
movement of a separate fluid particle from (2) to (1). For this downward motion, (b) in figure
17.14, we have v < 0, and u2 > u1, such that the fluctuating velocities for this particle would be:

v  0
u  0 (the arriving particle now has a velocity excess i.e. u   u 2  u 1 ) )
 uv  0   turb  u v  0

Thus, while the process is reversed, a positive stress again results from the arrival of the particle.
In this case, the particle from (2) arrives at (1) with an excess of velocity relative to the
surrounding flow, and thus will be decelerated to the velocity at (1), which is the result of an
applied stress on the particle equivalent to  uv by the flow at (1). And the local flow at (1)
will be accelerated by the arrival of the velocity-excess particle from (2), which is again the
result of an equal and opposite stress applied by the particle on the flow.

So, what is to be inferred here? That whenever there is fluid movement normal to a shear flow
(i.e. across a velocity gradient), this results in momentum exchange and resultant additional
stresses within the fluid.

To exemplify this process further, consider two trucks carrying rocks, running at different
speeds, and passing each other on a highway, as shown in figure 17.15. If passengers in the back
of the trucks throw rocks from one truck to the other, the residual momentum in the rocks will

605
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

change the overall momentum of the opposing truck. Thus, the rocks from the slow truck will
cause the fast truck to decelerate, and the rocks from the fast truck will cause the slow truck to
accelerate.

Slow Truck Fast rocks

accelerate truck

Slow rocks
decelerate truck Fast Truck

Figure 17.15 Model of momentum exchange due to the tossing rocks between a slow truck
and a fast truck, resulting in the acceleration of the slow truck, and
deceleration of the fast truck--an emulation of the process by which Reynolds
stress is created.

17.5.3 Phenomenological Equations

As pointed out in in section 17.5.2, Reynolds stresses are pseudo-shear stress variables that arise
from the Reynolds averaged Navier-Stokes equation, and are in addition to the conventional
variables of mean velocity and pressure. Thus, when we include the Reynolds stresses as
unknowns in the time-averaged equations, we have a mathematical dilemma:

Too few equations: 4 Need to determine a way to

Too many unknowns: 10 approximate the Reynolds stresses

This is of course for three-dimensional flow. For a simplified, and idealized, two-dimensional
flow we would of course have three equations and seven unknowns--still too many unknowns.
So, to develop solutions for the mean velocity and pressure for a turbulent flow requires that we
develop other equations that provide relationships between the Reynolds stresses and the mean
motion of the fluid.

Since we don't have appropriate equations for our Reynold's stress terms based on first
principles, such as continuity or Newton's first law, we most commonly rely on what are termed
phenomenological equations to allow closure (i.e. provision of enough additional equations
relating mean flow variables to the required Reynolds stresses such that a solution is possible).
Because of the immense complexity of turbulence, these equations are generally very simplified
interpretations of the observed, or measured, characteristics of turbulent behavior. Frequently,
these are little more than curve fits of particular data that relates measured stresses to mean flow
behavior. In other cases, they are extrapolations of known relationships for a laminar fluid, such
as the proportionality of shear stress to the velocity gradient for a laminar flow. The most

606
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

sophisticated approaches utilize the Reynolds averaged differential equation of energy to provide
a further degree of flexibility in calculating mean flow properties. However, despite the added
complexity of this latter approach, other phenomenological relationships must still be assumed to
solve the increased set of equations. Note that while there have been many attempts at
developing such phenomenological equations, with varying effectiveness, we will only examine
a couple of these approaches, and only for boundary layer-type flows. For further discussion of
the basics of more sophisticated techniques, see F. A.White (1991), W.C. Reynolds (1976), and
G.C. Speziale (1991) as starting points.

17.5.3.1 Eddy Viscosity

The use of an eddy viscosity is a simplistic attempt to create an artificial property, or function,
that approximates the Reynolds stresses by modeling them like a laminar flow. This process
assumes a fictitious “turbulent” or "eddy" viscosity. We know that laminar shear stresses
(Newtonian) are phenomenologically proportional to the local gradient of velocity, i.e. for a
boundary layer type flow:

u u
 laminar    
y y

Accordingly, it seems reasonable to hypothesize a “turbulent” viscosity that will relate the
turbulent Reynolds stresses to the gradient of the mean velocity, such that:
____
U ____
 u v
 turbulent     u v
or   (17.22)
y U
y
Here,  is known as the “eddy” viscosity, or sometimes "eddy" diffusivity, where the term eddy
refers to the eddying interlayer mixing characteristic of turbulence. We note in general that the
eddy viscosity for a turbulent flow is much larger than the comparable kinematic viscosity, with
 >> .

Our basic problem is how do we model  ? For some cases, experiments reveal that   y for
roughly y  0.2 (except very near the boundary), and   constant for y  0.2 (for a flat plate
flow). For other types of flows, models of  vary in myriad ways. In general, many attempts
have been made to develop generic models for eddy viscosities. Several of the better models are
outlined in White (1991). However, the models tend to be restricted to specific types of flows
and geometries, and there has yet to be a generic model that works effectively over a range of
flow geometries and conditions.

607
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

17.5.3.2 Mixing Length Concepts (near a flat plate)

The mixing length concept, and the subsequent phenomenological equations, was first proposed
by Ludwig Prandtl [Prandtl (1925)], and like eddy viscosity was one of the early attempts at
modeling turbulent behavior near a boundary. Drawing on flow visualization studies, like our
previous figure 17.11, Prandtl made an analogy between the large-scale motions (that carry
energy within the flow, not the dissipating scales) and molecular motion in gases, which is
modeled by mean-free paths (i.e. the distance a molecule will move, before interacting with
another molecule). Prandtl's idea was similar to our previous analogy of throwing rocks between
trucks; i.e. a lump of fluid mixes after traveling a certain length (i.e. a normal distance from the
boundary), with the distance the fluid travels dependent on its initial starting distance from a
surface. The model assumes that as the fluid moves a distance, l, it conserves its momentum
over that distance, and then mixes out. So, if a fluid moves from some level y to y  l, the
u
change in the velocity, u, over l will be u  l . Thus, to a crude approximation we can say that
y
U
the fluctuation velocity, similar to our model in figure 17.14 will be u   l . We additionally
y
need a model for v' to complete the analysis. Prandtl, on a leap of faith, hypothesized that v'
U
should be roughly proportional to u', or v   l as well. So, Prandtl's model of the Reynolds
y
stress for a two-dimensional boundary layer-like flow was:

2
____
 dU 
-  u v  const.l  2
 (17.23)
 dy 

Now, equation 17.23, while conceptually appealing, still leaves the modeling process in limbo,
since we do not know how l varies with the mean velocity, nor the constant of proportionality.
Prandtl, through an evaluation of experimental data, made some conjectures on the behavior of l,
which would allow Eq. 17.23 to model the behavior of the mean velocity profile U(y). However,
as we will discuss in the following section, Prandtl could not determine a universal model for l,
but realized that the modeling of l depended on the specific regions of behavior within the
boundary layer.

17.6 “Universal” Velocity Distribution Laws (for boundary layers)

Similar to laminar boundary layers, when one examines experimental data for turbulent flat plate
boundary layers, clear characteristics of the mean velocity profiles are quite evident, as illustrated
in figure 17.16. When plotted in linear form, a turbulent boundary layer velocity profile appears

608
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

Turbulent Boundary Layer Velocity Behavior

somewhat flat over the span of the boundary layer, with most of the change between 0 < U < U∞
u+ vs. y+ (linear scale)
(with U∞ the velocity at y = ) occurring very close to the bounding surface. This is much different
30

25

20

Outer
u+

15
Region
10 Inner
Region
5

0
0 1000 2000 3000 4000 5000
Turbulent Boundary Layer Velocity Behavior
y+
u+ vs. y+ (semi-log scale)
a) Linear Plot
30
Outer or
25 Wake
Region
Logarithmic
20 layer

Buffer
u+

15 layer
Viscous
10 sublayer Inner Region
5

0
1 10 100 1000 10000
y+
b) Semi-Logarithmic Plot

Figure 17.16 Typical averaged turbulent velocity profiles, plotted in non-dimensional

u u y
coordinates, of u   and y    . a) plotted on linear scales. b)
u 
+
plotted with a logarithmic y scale.

609
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

than a laminar boundary layer, where the change is more distributed (see figure 13.4). The reason
for this more uniform velocity profile is a result of the production and sustaining mechanisms of
turbulence, as we discussed generically in Section 17.4.

Very near the surface a process of turbulent "bursting" or breakdown of the local flow occurs,
which generates turbulent eddies of small scale. These initially small eddies, grow quickly by
amalgamation to larger scales, which subsequently entrain energy from the outer flow. What this
process suggests is that flow at and very near the boundary will be dominated by viscosity and
very high shear rates. However, the region just beyond this viscous region will be where the
initiation of the bursting behavior and rapid growth in scales will take place, and where Reynolds
stresses will start to dominate laminar stresses. Further out, the growth of scales will slow, and
the turbulent mixing process will become more uniform, and energy dissipation will ensue at the
smallest scales. The consequent effect of these varied processes will be the rapid transition
within the boundary layer from essentially laminar-like behavior very near the boundary to a
Reynolds-stress dominated behavior over the bulk of the outer region of the boundary layer.

The modeling of the actual processes of the boundary layer can be quite complicated. However,
by using an assumption of "regions" of behavior, and some rather simple mathematical
approximations, relatively simple equations can be derived that are very effective in modeling
the mean velocity behavior.

The modeling of the actual processes of the boundary layer can be quite complicated. However,
by using an assumption of "regions" of behavior, and some rather simple mathematical
approximations, relatively simple equations can be derived that are very effective in modeling
the mean velocity behavior.

The delineation of the regions and layers of a turbulent boundary layer was first proposed by
Ludwig Prandtl (1933) and Theodore von Karmen (1930), and the characterization of these
regions has not significantly changed to the present day. Basically, they deduced that the
velocity profile within a turbulent boundary layer consists of a thin, viscosity-dominated near-
wall region, a Reynolds-stress dominated outer region, and an overlap region, where both types
of stresses are important. In this way, Prandtl and von Karmen sought simple ways to model
these regions, and develop "universal" mean velocity profiles, which would apply to a broad
variety of boundary layer-type flows.

To simplify the description of the regions of behavior, Prandtl and von Karmen divided the
velocity profile into several “layers” of behavior, depending on the relative dependence (based
on experimental results) on viscous stresses and Reynolds stresses. They determined that a
y
region near to the bounding surface (or wall), within  0.2 , could be described by two distinct

layers, with a connecting layer, all of which are dependent on the wall shear stress. The

610
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

remainder of the boundary layer could be described as a single region or layer, whose behavior
was modified by external conditions, such as the external pressure gradient. These two regions
are generally termed the "Inner Region" and the "Outer Region" of the boundary layer, and are
comprised of:

a) Viscous sublayer y
b) Buffer layer Inner Region,  0 .2

c) Logarithmic layer

y
d) Outer (Wake) layer Outer Region, 0.2  1


Figure 17.16 shows the generic shape of the mean velocity profile for a turbulent boundary layer,
illustrating the respective regions and layers of behavior, plotted in both linear (17.16a) and
logarithmic (17.16b) distance scales. Sections 17.6.1 through 17.6.3 discusses of each of these
regions and how they are modeled.

17.6.1 The Inner Region

y
Within a relative small region adjacent to the bounding wall, roughly  0.2 , a turbulent

boundary layer transitions from being dominated by purely viscous stresses very near the
y
bounding surface to Reynolds stress dominated behavior as  0.2 . This region within

y
 0.2 , is known as the "Inner Region", and is actually modeled as two distinct layers, with a

transitional layer, or buffer layer, linking the two. Since this region is influenced most strongly
by the wall shear stress,  w , and not the outer region free stream velocity, U  , it is argued that
the local mean velocity, U, has a functional relationship of the form: U  f inner  w , , , y .
Dimensional analysis indicates that the functional equation for this set of parameters is:

 
  w y 
U   
 f inner   (17.24)
w   
  
 

w
Note that has the dimensions of velocity. What this combination of shear stress and

density represents is a characteristic velocity based on the wall shear stress. So, if  w is the wall

611
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

or solid boundary shear stress, we hypothesize a velocity, u  , which represents a momentum

flux equivalent to the local shear stress.

w
i.e.  w   u 2  u  (17.25)


So u is a “velocity” representing the change in momentum per unit area equivalent to the shear
stress  w . Utilizing this shear velocity, we can rewrite Eq. 17.24 as:

U  u y 
 f inner   
u   
or

 
u   f inner y  (17.26)

U u  y yu 
where, u   and y    . Here, the plus superscript simply indicates that this is
u  
a non-dimensional property. Also note that the non-dimensional displacement, y+, is a positional
Reynolds number based on u .

The inner region has been found to scale well on the “wall” variables, u+ and y+, as described by
Eq. 17.26. However, to properly reflect the variations in viscous and Reynolds stress across this
region, this inner region is broken into three sub-regions described as the “viscous sublayer,” the
“buffer layer”, and the “logarithmic layer.” As we will discuss in Section 17.6.2, the outer
region scales on the actual boundary layer thickness, , and is comprised of both the outer
portion of the logarithmic layer (an overlap region) and a layer described by a “velocity defect
law,” which also displays logarithmic behavior based on the actual boundary layer thickness, .
This probably sounds a bit confusing, and to some degree it is.

However, the use of inner variables (u+ and y+) for the inner region, and outer variables (U∞
and ) for the outer region, has proven to be a sound method to model the particular mean
velocity characteristics of the regions of the boundary layer, which are shown in figure 17.16 in
both linear and semi-logarithmic scales. The latter semi-logarithmic presentation (figure 17.16b)
is the preferred manner for representation of the velocity across a turbulent boundary layer, since
it expands and emphasizes the inner region, within which the major effects of profile
deformation and turbulence interaction take place. In the following, we discuss and demonstrate
the modeling processes employed for each of the layers comprising the inner region.

612
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

17.6.1.1 The Viscous Sublayer

Very near the wall, viscous effects dominate such that the flow will behave in a highly-stressed,
but laminar-like, manner. In this very thin region, the shear stress across this layer is essentially
constant at the boundary (y = 0) value, such that we can approximate that    w . Thus, we can
write the equation for the shear stress as:

dU
   w  shear stress @ boundary
dy
Rearranging, and making use of Eq. 17.25:

dU  w   w  w 1 u 2
   
dy      

Integrating gives:

u 2
U yC (C = 0, since U  0 at y = 0)


Or finally (by dividing through by u):

U uy
u    y (17.27)
u 
Equation 17.27 has been shown to be quite accurate for this very thin viscous sublayer, which
experiments have shown to exist for roughly y   5  7 . Note that this region of viscous stress is
very thin. On a light aircraft traveling at about 200 miles per hour, this layer would only be
about 0.025 mm (0.001" ) thick.

17.6.1.2 Logarithmic Layer

Farther from the solid surface, a turbulent boundary layer transitions from the viscosity-
dominated, laminar-like viscous sublayer, to a Reynolds stress dominated layer within the inner
region. The mean velocity within this layer is observed and modeled as a layer in which the
velocity scales logarithmically with distance from the boundary. This layer can extend from y+ 
30 to 300-500, with the outer extent of this layer dependent on the maturity of the boundary
layer. The maturity of a boundary layer is quantified by the Reynolds number based on either
plate length or boundary layer thickness. Generally, as the Reynolds number increases, the
extent of the logarithmic region expands (although non-linearly). The characterization of this
logarithmic region is roughly modeled by the Prandtl mixing length concept as described by Eq.

613
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

17.23. Since this logarithmic layer is well removed from the wall, the Reynolds stresses due to
turbulent mixing are found to dominate viscous stresses, such that the stream wise shear can be
approximated as:
U ____ ____
   lam   turb     u v   u v
y
To determine a functional form for the Reynolds stress, we employ the mixing length concept of
Prandtl from Eq. 17.23:
2
____
 dU 
-  u v  const.l  2

 dy 
As we discussed in section 17.5.3.2, the modeling of the mixing length, l, is problematic, since it
depends on the type of turbulent behavior present. However, for a turbulent flow adjacent to a
surface, Prandtl and Von Karmen hypothesized that a lump of fluid near the wall would mix after
traveling a distance proportional to its original distance from the wall, l  y , or l  ky , where k
is an undefined constant. This was to acknowledge that Reynolds stresses appeared to increase
with distance from the surface within the inner region. Thus,
2 2 2
____
 dU  2  dU   dU 
    u v  const.l    ky  
2
  k 2 y 2   (17.28)
 dy   dy   dy 
Note that the constant in Eq. 17.23 is assumed to be incorporated into the constant k in Eq.
17.28.

We now assume, as we did for the sublayer region, that    w , i.e. that the shear stress across
the inner region is equivalent to the shear stress at the wall. This has been shown to be the case
experimentally, such that the inner region is often referred to as the "constant stress" region.
Employing this additional assumption, we write:
2
 dU 
2 2
 w  k y  
 dy 
Rearranging, to solve for the mean velocity (and substituting from Eq. 17.25):
2 2
 dU  w u 2 u 
   2 2  2 2    
 dy  k y k y k y

dU u 
 (17.29)
dy ky
We assume that Eq. 17.29 holds from a region starting somewhere above the viscous sublayer, at
yo, where U = U(yo). Thus, we integrate from yo to y:

614
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

U y
u dy
 dU  
U  y0 
k 
y0
y
   
U  U y 0  
u
lny  lny0   u  ln y   u  ln yu   
k k  y0  k   y 0 u  

u   yu   u    
U ln  ln   Uy 0 
k    k  y 0 u   (17.30)
Dividing Eq. 17.30 by u, we get:

U 1  yu   1    U y 0 
 ln   ln   (17.31)
u  k    k  y 0 u   u

B=constant

We note that the indicated terms in Eq. 17.31 are an unidentified constant, B, such that Eq. 17.31
becomes:

U 1 yu 
 ln B
u k 
or
1
u  ln y   B (17.32)
k
Fitting Eq. 17.32 to experimental measurements was done originally by Prandtl based on
measurements in a turbulent pipe flow by J. Nikuradse, and k and B were found to be
approximately k  0.40 and B  5.0 . Using more modern measurements, Coles and Hirst
(1968) suggest values of k  0.41 and B  5.5 . However, more recent work at high Reynolds
numbers by a number of researchers, Marusic et al. (2010), suggests that the Karman constant (k)
may range from 0.37 to 0.42, depending on the type of flow (with corresponding variations in the
constant B). Based on the recommendations of Marusic et al., based on data which is quite
comprehensive, we will assume values of k = 0.385 and B = 4.2 in this text. For more detailed
assessment of the arguments for one value over another, the reader is referred to Marusic et al.

The thickness of the turbulent boundary layer that constitutes the logarithmic layer has also
undergone much debate — on both its lower and upper limits. The lower limit is general
accepted as roughly y   30 , although limits of y   50 have been suggested. The outer limit
is dependent on the Reynolds number of the flow, and increases with increasing local Reynolds
number (i.e. as the flow progresses downstream). However, a range of 30  y   300  600 for
the logarithmic layer seems to be reasonable, although arguments are also made for a narrower

615
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

range. So, in general we will assume the equation for the mean velocity within the logarithmic
layer, and its range of applicability, is:

U 1  yu 
 ln    4.2  u   2.6 ln y   4.2 for 30  y   300  600 (17.33)
u  0.385   

17.6.1.3 Buffer Layer

This layer lies roughly between about 5  y   30 , and is essentially the transition zone between
the viscous sublayer and the logarithmic layer. While this is a relatively narrow region, it is
particularly important because of the high-velocity fluctuations and Reynolds stresses that
develop in this region. Basically, it is within this layer that the boundary layer generates new
vorticity concentrations by means of the bursting process discussed in section 17.4. It is these
vorticity concentrations that subsequently propagate away from the buffer layer, create the
constant Reynolds stress region of the logarithmic layer, and amalgamate to form the large, outer
region structures. Modeling this layer has not been particularly successful from a physical
perspective. There are models that can predict the velocity transition from the viscous sublayer
to the logarithmic layer, but these are essentially hindsight relationships, which essentially use
complicated curve fits to match experimental results. These approaches generally use mixing
length or eddy viscosity models, or a combination of both. We will discuss two successful curve-
fit models using these approaches.

17.6.1.4 Models of the Entire Inner Region

van Driest Damping Factor

First, for the inner region, we use the idea of the "viscosity" being a combination of the fluid
viscosity, , and the turbulent, or "eddy" viscosity, , as defined by Eq. 17.22. Thus, for the
inner region we can write:

dU dU  ____
 dU
     turbulent 
          u v  (17.34)
dy dy   dy
In Eq. 17.34, the eddy viscosity is intended to "model" the Reynolds stress, which can be related
to a Prandtl-type mixing length, or other approach. To use the mixing length approach, let's
reconsider Eq. 17.28, where we assumed (incorporating the constant into the mixing length l ):

2
dU
____
 dU   2 dU  dU
τ  ρ u v  ρε
 ρ l 2     ρ l  (17.35)
dy  dy   dy  dy
From Eq.17.28, we can relate the eddy viscosity to the mixing length as:

616
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

dU
 l 2 (17.36)
dy
If we now substitute Eq. 17.36 into Eq. 17.34, noting that within the inner region the shear stress
is essentially constant,    w  u 2 ,we get:

dU  dU  dU
  u 2     
    l 2 
dy  dy  dy
Which we can expand and rearrange as an quadratic equation for dU/dy as:

2
 dU 
2 dU
l      u 2  0
 dy  dy
Using the quadratic equation, we can solve for dU/dy as:

2 2 2
dU      4l u 

dy 2l 2
Retaining only the positive value of the root, and using some judicious algebraic manipulation, it
can be shown that:

dU 2u 2

dy 4l 2 u 2
  1
2
and with some more work that:

du  2 2 lu
  where l   (17.36)
dy 
1 1
4l 2 u 2 1  1  4 l  2   
2
Now, within the viscous sublayer viscosity dominates, such that there is no mixing length and
l   0 , and Eq.17.36 reduces to:

du 
1  u   y which is Eq.17.27 for the viscous sublayer.
dy 
However, within the logarithmic layer the mixing length is large such that l   1 , Eq. 17.36
then reduces to:

du  1 1
   u  ln y   B (Eq.17.32, for a Prandtl mixing length l  ky )
dy 
l k

617
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

Clearly, one could create a model of the entire inner region if one could provide a model of the
mixing length, which would make a smooth transition from the viscous sublayer to the
logarithmic layer. One model that effectively did this was published by Van Driest (1956), who
suggested an asymptotic, decaying exponential model for the transition from pure viscosity to the
Prandtl mixing length, l  ky . His model was:

  y 
l   ky  1  exp   (17.37)
  A 

He termed this l  as a "damping factor", since it gradually increased the mixing length across
viscous - buffer - logarithmic layers. Substituting Eq. 17.37 into Eq.17.36 and integrating for u+
gives:
y 2dy 
u   (17.38)
0 2
2  y 

 
1  1  4 ky  1  exp  
  A 

Since there was no boundary condition to apply to determine van Driest's constant A, he
basically fit Eq. 17.38 by varying the constant A until the integrated Eq.17.38 matched the
logarithmic layer equation for large values of u+. van Driest suggested a value of A = 26 as the
appropriate value, but this was using the then accepted mixing length constants of k = 0.41 and B
= 5.5 in Eq. 17.32. However, using our most recent values of k = 0.385and B = 4.2, a value of A
= 21 is a better fit. A comparison of the numerically integrated van Driest equation (there is no
closed form solution) with the sublayer and logarithmic layer equations is shown in figure
17.17a, and is shown to match both equations quite well.

Spalding's Single Inner Region Equation

Another example of a model covering the entire inner region was suggested by Spalding (1961).
His model is based indirectly on the mixing length concept, using a modification of Eq. 17.32
which will satisfy both the logarithmic behavior within the logarithmic layer, and the linear
behavior within the viscous sublayer. To do this Spalding rewrote Eq.17.32 as follows:

1
u  ln y   B
k
ku   ln y   lnexpkB
y
 
exp ku  
expkB 
Expanding the exponential for kB as a series:

  
y   exp kB exp ku   exp kB  1  ku   12 ku   
2
 
 16 ku 
3

 ... 17.39

618
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

(a)
25
van Driest
u+= y+
20 u+ = 2.6 ln(y+) + 4.2

15
+
u

10

0 -1 0 1 2 3
10 10 10 10 10
y+

(b)
25
Spalding
u+= y+
20 u+ = 2.6 ln(y+) + 4.2

15
+
u

10

0 -1 0 1 2 3
10 10 10 10 10
y+

Figure 17.17 Comparison of: (a) van Driest and (b) Spalding inner region equations with
the viscous sublayer and logarithmic layer equations.

619
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

This reverses the dependency from Eq.17.32, and making y+ the dependent variable, and u+ the
independent variable in Eq.17.39. As shown on the right bracketed portion of Eq.17.39, Spalding
 
also wrote the second exponential as a series expansion of exp ku  in four terms, which is
applicable when ku+ is small. However, he noted that when ku+ became large, greater than
roughly 10, the higher order terms in the expansion (e.g. 1
24
ku   4
 ... ) , which were neglected
in the expansion, would become dominant. So, he reasoned that by adding back the exp(ku+)
inside the { } brackets and subtracting the four-term series expansion would create a function
that was negligible when ku+ was small, and returned to essentially the logarithmic relationship
of Eq.17.32 when ku+ exceeded roughly 20. To have the equation yield the viscous sublayer
equation for small ku+, Spalding also added a u+ to the equation, which would yield the viscous
sublayer equation, Eq.17.27, when ku+< 7 or so. Spalding's resulting inner region equation is:

    
y   u   exp kB  exp ku   1  ku   12 ku 
2
 
 16 ku 
3
(17.40)

Figure 17.17b shows Eq.17.40, Spalding's inner region equation, in comparison with the viscous
sublayer and logarithmic layer equations. As shown, Spalding's inner region equation matches
both equations quite well.

So, like Van Driest, Spalding used some judicious reasoning and the empirical equations for the
viscous sublayer and the logarithmic layer to create a function that matches both equations in
their regions of applicability. Note that Spalding did not rely on the on eddy viscosity
arguments, although one can derive an eddy viscosity by taking the derivative of Eq. 17.40.
One important issue with Equation 17.40 is that u+ is implicit in this equation, as well as being
quite complicated. Consequently, use of such equations is of limited utility, other than for
comparison with experimental data.

To see other ways of generalizing the inner region equation, in order to account for such effects
as pressure gradients and roughness, I recommend you read the NASA technical memorandum
by Shih et al.(1999).

17.6.2 The Outer Region

The outer region, extends over the bulk of the boundary layer, from 0.2 < y/ < 1, overlapping on
the lower end with the logarithmic layer, and at the upper end with a “wake” region. This outer
region is dominated by Reynolds stresses, with significant turbulent mixing. However, this
vigorous mixing creates a mean velocity profile that does not change rapidly, and is somewhat
uniform in appearance. The argument is that within this region viscosity is a secondary agent in
establishing the mean profile. Although viscosity is the mechanism for the final energy
dissipation (at very small scales), the mixing of the larger scales is what gives rise to the

620
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

redistribution of momentum within this region. Within the outer layer, we view the effect of the
Reynolds stresses as creating a "defect" of the mean flow from the driving velocity at the edge of
the boundary layer, U∞. However, the level of the Reynolds stresses is still characterized by the
wall shear stress,  w , and fluid density, , which characterize the shear velocity, u  , as defined
in Eq. 17.25. Thus, we can reason that the functional relationship for the mean velocity in the
outer region should be:

 dp 
U  f  U  , ,  w , y, , 
 dx 
dp
Here,  is the boundary layer thickness, and is the pressure gradient, which is of course
dx
related to U∞. Von Karman in 1930 showed that the most effective functional relationship for
the outer region was a modeling of the "defect" of the mean velocity, U, from the external
velocity, U. Employing this defect concept (i.e. U   U ), dimensional analysis yields a
relationship for the velocity defect of the form:

U  U y   dp
 f  , where   , a dimensionless pressure gradient (17.41)
u    w dx
Equation 17.41 turns out to be a good functional relationship for the outer region. However,
determining a universal relationship for this equation has proven quite challenging, since the
function changes with flow geometry and history. In reality, the process has been more of an
exercise in curve fitting of empirical data, rather than development by some a priori
mathematical model. The simplest way to approach a model is to assume a self-preserving or
equilibrium boundary layer (i.e. one where the velocity defect function of Eq. 17.41 is the same
function of y/ for all x locations). This assumption, while convenient, is not exact, and only
becomes a good approximation when the Reynolds number becomes large.

y 
The problem is to determine the functional relationship f  ,   in Eq. 17.41. The approach is a
 
combination of empirical curve fitting, and the understanding that the inner and outer region
must overlap and give identical results within that overlap region. In order for the logarithmic
layer and outer layer to overlap, Eq. 17.41 must asymptote to Eq. 17.32 within this overlap
region. Thus, Von Karmen suggested that the equation for the mean outer region velocity should
be of the form:

U 1  yu    y
 ln   B  A   (17.42)
u k    

621
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

y
Here we assume that A   is a function of the external pressure gradient, and   is a function

that varies with the location within the boundary layer. Now, if we set y = , where U = U∞, Eq.
17.42 gives:

U  1  u  
 ln   B  A 1 (17.43)
u k   
Subtracting Eq.17.42 from Eq. 17.43, gives us a functional form for the velocity defect equation,
Eq.17.41, as:

U   U 1  u   1  yu   y
 ln   A1  ln    A 
u k    k    
1  y   y 
  ln   A 1    (17.44)
k     
Now, within the overlap of the inner and outer regions, if Eq. 17.42 is to match Eq. 17.32, then
 y y y
   0 as  0 . Thus, in the overlap region, where  0 , Eq. 17.44 becomes:
  

U  U 1  y
  ln   A1 (17.45)
u k 

U
Solving Eq. 17.45 for , and equating it to Eq. 17.32 and rearranging, we have:
u
U U 1  y  1 yu
  ln   A 1  ln   B
u u k    k 
U  1  yu   1  u   1  yu 
  ln   ln   A 1  ln    B
u k    k    k   

And solving for A 1 , we have:

U   1  u   
A 1    ln   B (17.46)
u k    
What we have in Eq. 17.46 is the difference between the velocity at the edge of the boundary
y
layer, and the value predicted by the logarithmic law, Eq. 17.32. If we assume that   is a

function such that 1  1 at the edge of the boundary layer, then we have:

622
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

U   1  u   
A     ln   B (17.47)
u k    

So, the value of A   is a measure of the difference between the actual external velocity, and
that predicted by the law of the wall (the logarithmic layer). Of course, recall that the value of
 dp
A   is a function of the pressure gradient imposed on the boundary layer, where   .
 w dx
The value of A   varies with the applied pressure gradient, and often the particular geometry,
but in general is a fairly constant value for a given pressure gradient. Some selected values of
A   , determined from experiment (White, 2011) are:

Pressure Gradient  A   value

Strong favorable -4.8 1.0
Flat plate 0 2.5
Mild adverse 6.3 5.6
Strong Adverse 29 13

Except for the flat plate value, these are representative of values obtained for a limited selection
of experimental data. Note that a linear curve fit of this data gives an approximate relationship
of:
A  0.36  2.8 (17.48a)
Thus, if  is relatively constant, Eq. 17.48a allows the approximation of the respective value of
A().

y
The function   has been debated for decades, and still is undergoing scrutiny. However,

Coles (1956) probably came as close as anyone to defining its functionality, in proposing what
he termed the Law of the wake. Here, the wake is the outer region of the boundary layer where
the flow diverges from the logarithmic law of the wall. Examining a broad set of data, Coles
  y 
proposed that the wake portion of Eq. 17.43   could be approximated as an S-shaped
   
function, approximated by:

2 3
 y  y  y  y
   sin2   (sometimes approximated as  3   2  ) (17.48b)
 2   
Thus, to a good approximation we can now rewrite the outer region mean velocity profile of Eq.
17.42 as:

623
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

U 1  yu    y
 ln   B  A sin2   (17.49)
u k    2 
And the velocity defect form of Eq. 17.41 as:

U  U 1  y   y 
  ln   A 1  sin2   (17.50)
u k   2  

17.6.3 The Combined Universal Velocity Profile

Figure 17.18, shown in both linear and semi-logarithmic coordinates, plots the mean turbulent
velocity profile predicted by Eqs. 17.27 (sublayer), 17.33 (logarithmic layer), and 17.49 (outer
layer). The outer region is shown for a range of pressure gradients, from a strong favorable to a
strong adverse; additionally, the outer region without a wake correction (A=0) is shown for
comparison. Note the significant variations in the outer region created by the application of an
adverse pressure gradient.

To gain further perspective on the effect of pressure gradients on turbulent boundary layers, the
velocity profiles of figure 17.18a are replotted in figure 17.19 relative to the outer flow velocity,
U  , and the boundary layer thickness, . As figure 17.19 shows, a favorable pressure gradient
will result in a flatter velocity profile, while an adverse pressure gradient will cause a retardation
of the profile. As the velocity profile for the strong adverse pressure gradient illustrates, while
the outer region is strongly retarded, the near-wall region does not display as much obvious
retardation (although the wall shear stress,  w , is reduced). For very strong adverse pressure
gradients, a turbulent flow will undergo separation, although it is more robust than a laminar
boundary layer, as we discuss below.

17.6.4 A Shape Factor Comparison of Turbulent vs. Laminar Boundary Layers

To illustrate how significantly the boundary layer mean velocity profile is modified by transition
to turbulence, figure 17.20 shows a comparison between a turbulent boundary layer profile
modeled by Eqs. 17.27, 17.33, and 17.49 for a turbulent flat plate flow, and a laminar Blasius
boundary layer, as calculated in Table 13.1. Both are plotted on linear axes, and referenced to
the mean outer flow, U  , and assume the same boundary layer thickness, . Note the striking
flatness of the turbulent profile versus the laminar profile. This variation is characterized by the
*
shape factor we discussed in section 13.6, defined by H  , the ratio of the displacement to

the momentum thickness. The shape factor, H, for the laminar Blasius boundary layer is 2.6,
whereas H for the turbulent boundary layer shown is 1.3, illustrating how much "flatter" a
turbulent boundary layer is, which makes it much less susceptible to flow separation.

624
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

40

35

30

25

20
Experiment--Flat Plate
u+

15
Viscous Sublayer
Logarithmic Layer
10 Outer Region (no wake)
Outer Region (flat plate)
5 Strong Favorable (A=1)
Mild Adverse (A=5.6)
Strong Adverse (A=13)
0
0 500 1000 1500 2000 2500 3000
y+
(a) Linear scale
40

35

30

25

20

15
u+

Experiment--Flat Plate
Viscous Sublayer
Logarithmic Layer
10 Outer Region (no wake)
Outer Region (flat plate)
5 Strong Favorable (A=1)
Mild Adverse (A=5.6)
Strong Adverse (A=13)
0
1 10 100 1000 10000
y+
(b) Logarithmic scale

U
Figure 17.18 Mean turbulent velocity profiles plotted using wall variables, u   and
u
u  y yu 
y   , for sublayer (Eq.17.27), logarithmic (Eq.17.33), and outer
 
u
regions (Eq. 17.49). Shown for      3000 .


625
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

0.8

0.6

U/U
0.4 Experiment--Flat Plate
Flat Plate (A=2.5)
Strong Favorable (A=1)
0.2
Mild Adverse (A=5.6)
Strong Adverse (A=13)
0
0 0.2 0.4 0.6 0.8 1
y/
Figure 17.19 Mean turbulent velocity profiles of figure 17.18a replotted relative to U  and  .
u
The profiles shown assume      3000 .


0.8

0.6
U/U
Laminar
0.4
Turbulent

0.2

0
0 0.2 0.4 0.6 0.8 1
y/
Figure 17.20 A comparison of flat plate boundary layer velocity profiles plotted relative to
U  and  . The laminar boundary layer is a Blasius boundary layer, plotted
from table 13.1. The turbulent boundary layer profile is plotted using a
u
combination of Eqs.17.27, 17.33, and 17.49 assuming      3000 .


626
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

Generally, a decrease in H from the flat plate value indicates a more stable flow, whereas an
increase in H indicates reduced stability. For a laminar flow, a higher H value increases the
likelihood of: (a) transition to turbulence, or (b) a flow separation of the boundary layer. For a
turbulent flow, since the flow is already in a turbulent state, an increase in H from the flat plate
value just implies a greater likelihood of flow separation. Note that for a turbulent flow, a lower
H value due to a favorable pressure gradient in some cases can result in a return of the boundary
layer to a laminar state, termed "relaminarization"; for strong pressure gradients, relaminarization
can occur quite quickly, in roughly 20-100 initial boundary layer thicknesses (depending on the
physical situation). Relaminarization can also occur due to rotation, suction, flow curvature, and
heating [Narasimha and Sreenavasan (1979)].

Figure 17.21 shows a range of equilibrium velocity profiles for both laminar and turbulent
boundary layers, due to a range of applied pressure gradients. These pressure gradients range
from favorable to strong adverse pressure gradients for the laminar profiles, and from a flat plate
flow to a strong adverse pressure gradient for the turbulent profiles, and illustrate the range of
possible shape factors. It is interesting to note that the shape factor for a laminar flat plate flow
(H = 2.6) is twice that of a turbulent flat plate flow (H = 1.3). This quantizes what is clearly
shown in the comparative figure 17.20 --- that the displacement of the boundary layer fluid (*)
is much greater than the loss of momentum () for a laminar flow than for a turbulent flow; and
that shape factors are markedly higher for laminar vs. turbulent behavior.

(a) Laminar (b) Turbulent

Figure 17.21 A comparison of mean flat plate boundary layer velocity profiles plotted
relative to U  and  for a range of shape factors, H:
(a) Laminar boundary layers: 2.2 < H < 3.5 (laminar separation);
(b) Turbulent boundary layers 1.3 < H < 2.4 (turbulent separation).
[from White, 1991]

627
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

In Section 13.3, it was pointed out that for a Blasius flat plate flow an inflection in the velocity
profile develops at the surface, and that such an inflection point is an indication of a marginally
stable flow. As figure 17.21(a) illustrates, as the pressure gradient becomes more adverse, and H
correspondingly increases, the inflection in the laminar velocity profile develops away from the
surface, which results in an increasingly less stable laminar boundary layer, quickly leading to
rapid transition to turbulence and/or boundary layer separation. Interestingly, figure 17.21(b)
shows no apparent inflection points as the adverse pressure gradient increases. This is because
the resultant inflections will occur very near the boundary, within the viscous sublayer and buffer
region, which are not plotted in the figure. It is worth noting that, for a laminar boundary layer,
an increase in H by only 35% above the flat plate value will result in an immediate boundary
layer separation. In contrast, for a turbulent boundary layer, H must increase by 85% before
immediately separation will occur. This is another testament to the fragility of a laminar
boundary layer, and the robustness of a turbulent one.

While the velocity profiles shown in figures 17.18, 17.19, and 17.21(b) are reasonably close to
measured turbulent boundary layer characteristics, these reflect basically equilibrium boundary
layer flows, for which the conditions, such as the pressure gradient, remain relatively constant.
In most real applications, such as airfoils, a flow will generally undergo significant changes in
pressure gradient over the flow surface, and thus will not be an equilibrium flow. Additionally,
considerations of other effects such as surface roughness or surface transpiration require
additional assumptions, and generally more sophisticated modeling [see White (1991)].

However, for smooth, solid boundaries with gradual streamwise changes, the collective
equations reflected by figure 17.18 do a reasonable job of modeling mean turbulent velocity
behavior. However, for more accurate predictions of turbulent boundary layer behavior, there is
a great body of research and literature that attempts to provide more universal characteristics for
turbulent boundary layers under varying conditions. In addition, many sources exist that provide
ever-improving capabilities for computational prediction of turbulence in general and turbulent
boundary layer characteristics in particular. Most of these prediction approaches require
significant computational resources and sophisticated numerical techniques to determine
significant engineering properties such as surface shear, drag, and separation behavior.

For our purposes, we will only consider a simple, approximate method, which employs the same
momentum integral equation we applied to laminar boundary layers in Chapter 14. Here, we
will adapt the momentum integral equation to predict averaged turbulent boundary layer
characteristics. By utilizing a simple model of a mean turbulent velocity profile, we can
approximate the basic boundary layer characteristics relevant to engineering applications, using
some fairly simple calculations.

628
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

17.7 A Momentum Integral Analysis of a Flat Plate Turbulent Boundary Layer

17.7.1 Transition to a Turbulent Boundary Layer on a Flat Plate

As discussed in section 17.3.2, a laminar flow does not suddenly transition to turbulence, but
does so by means of a complicated process that is still not clearly understood. What is
understood is that there is a Reynolds number range over which transition may occur.
The Reynolds number at which transition to turbulence initiates is known as the critical Reynolds
number, or Recrit. Here, Recrit can be based on the x-distance along a flat pate, or on the
boundary, displacement, or momentum thicknesses. This can make a discussion of the location
of Recrit somewhat confusing, so for our purposes we will use Re x crit , which is based on the
location from the leading edge of a flat plate, as illustrated in figure 17.22.

Laminar Transition Turbulence

x
Figure 17.22 Schematic representation of transition to turbulent flow on a flat plate.

Note that the transition process will not occur immediately, but requires some distance (i.e.
Reynolds number range) over which a full transition from a laminar to a fully-turbulent flow will
occur. In addition, a number of things can affect the critical Reynolds number and the transition
process, including surface roughness, external flow fluctuations, and pressure gradients.

In general, increasing surface roughness, or an increase in the level of external flow fluctuations
(measured as the rms about the mean flow velocity) will cause a flow to begin transition to
turbulence at a lower critical Reynolds number. An adverse (increasing) pressure gradient will
also stimulate an early transition to turbulence; recall our discussion of how sensitive laminar
flow stability is to an adverse pressure gradient in Section 17.2.3. However, a favorable
(decreasing) pressure gradient will delay the transition to turbulence to a higher critical Reynolds
numbers. And, as pointed out in Section 17.6.4, very strongly favorable pressure gradients can
actually cause turbulence to “relaminarize”, or revert back to laminar behavior.

For an initially uniform flow over a flat plate, the accepted Reynolds number range for the
initiation of transition to turbulence is roughly Re x crit  3  105 to 3  106 , where Re xcrit  5 105

629
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

is generally accepted as the critical Reynolds number for transition to turbulence under normal
surface roughness and external flow conditions.

In the following analysis of a flat plate turbulent boundary layer, we will examine two situations.
The first is that assumes that the boundary layer is turbulent from the leading edge of a flat plate.
The second, that the flow is initially laminar, and then suddenly transitions to a fully-developed
and self-sustaining turbulent boundary layer. While both of these are an idealized models, they
provide reasonable predictions of boundary layer growth and wall shear stress, and illustrate the
differences in behavior between turbulent and laminar boundary layers.

17.7.2 The Momentum Integral Equation for Turbulent Flow

As discussed in Section 17.6.4, the detailed analysis of turbulence is a complicated process

requiring very complicated mathematical modeling and substantial computational resources.
However, we can develop some relatively simple relationships for the characteristics of a
turbulent boundary layer using a momentum integral analysis similar to that done in section
14.2.1 for a laminar flow. We will do this using an assumed velocity profile, and shear stress
information obtained from the outer region equation, Eq. 17.49.

Consider the von Karman momentum integral equation, Eq. 14.11 developed in Chapter 14.

w 1 dU  d
U 2

 *  2
U  dx
 
dx
For a flat plate boundary layer, with no pressure gradient and constant U  , this reduces to:

d
 w  U 2 (17.51)
dx
Here,  is the momentum thickness, given by Eq. 13.27 as:

U U 
 1  dy (17.52)
0
U   U  
Eqs. 17.51 and 17.52 are equally applicable to turbulent boundary layers as well as laminar
boundary layers, since they only require expressions for U, representing the approximation of the
mean velocity profile, and  w , the wall shear stress. As we show in the following section, a
reasonable approximation for U exists. However, determining the shear stress is problematic,
since in a turbulent flow we cannot relate the shear stress to the gradient of the mean velocity,
since the "stresses" in a turbulent flow also take into account the "Reynolds" stresses. Therefore,
we have to seek an alternative relationship for the shear stress, which we do by employing the
outer region velocity relationship from Eq. 17.49.

630
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

17.7.3 The Mean Velocity Profile: Nikuradse Power Law

Based on an extensive set of experimental studies in pipe flows, Nikuradse (1932) established
that to good approximation the mean velocity profile of a turbulent flow in a fully-developed
pipe flow (of radius R) can be approximated by:
1
U  R  r n
  where n > 6 (17.53)
U  R 
Prandtl (1921) had previously suggested the use of a power-law mean velocity profile to
approximate a flat-plate turbulent boundary layer, with Eq. 17.53 being recast for a flat plate as:
1
U  y n
  (17.54)
U   
Eq.17.54 presumes that since the pipe boundary layer would extend to the center of the pipe, that
to a good approximation  = R, and y  R  r can be substituted in Eq. 17.53 to obtain Eq.17.54.
It has subsequently been shown that Eq.17.54 is a good approximate representation of a turbulent
boundary layer velocity profile for use in the integral analyses of turbulent boundary layers, and
we will use this relationship for our present integral analysis.

30 30

25 25
TBL TBL
20 20
u+ n=7 u+ n=7
15 15
u+

u+

10 10

5 5

0 0
0 1000 2000 3000 1 100 10000
y+ y+
(a) Linear (b) Semi-logarithmic

Figure 17.23 Comparison of mean turbulent boundary layer velocity and a Nikuradse n = 7
1 1
U y n y
n
power law model,    U   U     . Shown assuming
U     
u 
   3000 .


631
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

Note that while Eq.17.54 is a good approximation of the mean velocity shape, it t does not meet
the boundary conditions exactly, as shown by figure 17.23. Although Eq.17.54 does satisfy the
U y U y
first order boundary conditions (  0 at  0 , and  1 at  1 ), it does not satisfy the
U  U 
appropriate second and higher-order derivative boundary conditions :
1 n
dU U   y  n y y
e.g.    which is infinite at  0 , and non-zero at  1
dy n     

However, because of its simplicity, Eq.17.54 has been widely employed to effectively model the
approximate shape of the mean velocity behavior.

From empirical data, the value of the constant n was found to be a function of Reynolds number,
varying from roughly 7  n  9 for a flat plate flow. Although there is no consensus, generally
accepted values are:

n  7 5x105  Re x  107
n9 106  Re x  108

We shall examine the difference these respective n values have on predicted boundary layer
characteristics later in section 17.7.5.

17.7.4 Assumed Shear Stress Relationship

Determining an appropriate shear stress relationship to employ with Eq. 17.51 is a bit
problematic. Recall that for a laminar flow, we would simply apply the Newtonian relationship
relating the viscosity to the rate of change of velocity normal to the flow direction at the wall
dU
(y=0), such that  w   . However, if we apply the Nikuradse-type velocity profile, we
dy y 0

have:
1
dU  y n
w   , with U  U   
dy y 0 

1 n
dU 1 1  U  1
w    U  1 ( y) n
     as y  0
dy n  
n 1
 y n
y 0
n
 


632
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

As we will see, to appropriately integrate Eq. 17.51 we will require an expression for the wall
shear stress which is a function of the boundary layer thickness, . With this in mind, we make
use of the empirical equation for the mean outer region velocity profile, Eq. 17.49, given as:

U 1  yu    y
 ln   B  A sin 2  
u k    2 

Letting y = , and U = U∞, we obtain:

U  1  u    
 ln   B  A sin 2  
u k     2 
or,
U  1  u  
 ln   B  A 
u k   
Rearranging, we get:

U  1  U  u   1  U  1  u 
 ln   B  A   ln    ln    B  A 
u k   U  k    k  U 

U   u 0 cf
Noting that  Re  , u   0 , and    ,we can write:
  U U 2
2

2 1 1  c 
 lnRe    ln f   B  A  (17.55)
cf k k  2 

Equation 17.55 relates cf directly to Re, but the relationship is implicit. To develop an explicit
relationship, we assume a series of cf values from 0.0015 to 0.005, and then solve Eq.17.55 for
for the respective Re values using a root finding program. We then use the cf vs. Re values and
develop a regression power law fit for cf as a function of Re, given by (using flat plate values of
k = 0.385, B = 4.2, and A  2.5 ) :
0.173
w  U 
 0.0195Re  
0.173
cf   0.0195   (17.56)
1   
U 2
2

633
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

17.7.5 Solution of the Momentum Integral Equation

Solution of Eq.17.51 requires that we first determine the momentum thickness, , as a function
of the boundary layer thickness, , via Eq. 17.52. To facilitate the integration of Eq.17.52, we
rewrite the Nikaradse relationship, Eq.17.54, as:
1
1
U  y n y
    n where 
U    

Thus, Eq.17.52 becomes:


U U 
 1
U U 
1 1
 1

 1  dy    1  d    1   d
n

n
 (17.57)
0
U  U   0
U  U  0  
After integration of Eq.17.57, we obtain:

 n
  (17.58)
 n  1n  2) 
Here,  is a constant, depending on the value of n.

Now, we substitute Eqs.17.58 and 17.56 into Eq.17.51:

d
 w  U 2
dx
0.173
1 U  d
0.0195 U 2     U 2 
2    dx

0.173
U  d
0.00975   
   dx
0.173
0.00975  U  
  dx   0.173 d (17.59)
   

To integrate Eq. 17.59 requires that we assume the state of the flow from the leading edge of the
flat plate. The simplest approach is to assume that the boundary layer is turbulent from the
leading edge of the plate. This is a reasonable approach if a boundary layer tripping mechanism
(such as a rod or large sand roughness) is applied very near the leading edge. However, for the
more general case, the boundary layer will be laminar from the leading edge of the plate, and
then transition to a turbulent boundary layer. This can be addressed by assuming laminar
behavior up to a transition location, and performing the integration in two parts: integrating the
laminar equation from the leading edge to the transition point, and then integrating the equation

634
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

from that point onward assuming turbulent flow. Here, we will examine both of these situations,
and determine the difference in the predicted behavior.

17.7.5.1 Turbulent Flow from Leading Edge

First, we will assume that the flow is turbulent from the leading edge, and thus start our
integration of Eq.17.59 from  = 0 at x = 0. Thus:

0.173 x 
0.00975  U  
 dx   
0.173
  d
    x 0  0
0.173
0.00975  U   1.173
  x  0.853 1.173
    1.173
Solving for  gives:

1 0.173

 0.00975  1.173  U  x  1.173
      x  0.0216 0.853 Re x 0.147 x (17.60)
 0.853    
or
 0.0216 0.853

x Re 0x.147
The friction coefficient is defined as:

w
cf 
1
U 2
2
The wall shear stress,  w , is determined from the simplified integral equation, Eq. 17.51, giving:

w d
cf  2 (17.61)
1 dx
U 2
2
d d d
We obtain in Eq.17.61 from Eq. 17.58 as  ,so:
dx dx dx

w d d
cf  2  2 (17.62)
1 2 dx dx
U 
2
d
Finally, we determine by differentiating Eq. 17.60 and substituting into Eq. 17.62, giving the
dx
result:

635
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

w d d 0.0368 0.147
cf  2  2  (17.63)
1 dx dx Re 0x.147
U 2
2

For 7  n  9 , the values of , and corresponding equations for and cf are:
x

  0.158 0.0261
n=7   0.0972  cf  (17.64a)
 x Re 0x.147 Re 0x.147

  0.170 0.0258
n=8   0.0889  cf  (17.64b)
 x Re 0x.147 Re 0x.147

  0.183 0.0255
n=9   0.0818  cf  (17.64c)
 x Re 0x.147 Re 0x.147

Note that the difference in the friction coefficient between n = 7 and n = 9 is only 2.4%, which is
less than the empirical accuracy used to establish the curve fits for the velocity profile and shear
stress relations ships used to develop Eq. 17.63. This suggests that the selection of n for our
modeling does not have a critical impact.

The total drag on a plate is obtained by integrating the shear stress over the plate surface, such
that:

L
Drag    w Wdx where W is the plate width, and L is the plate length (17.65)
0

Note that we can do this integration the hard way, by integrating using Eq. 17.63 for cf, or the
easier way by noting that shear stress is directly related to the change in the momentum thickness
through Eq. 17.51. Therefore, substituting the simplified integral equation, Eq. 17.51, into Eq.
17.65, we have:

L L 
d L

Drag    w Wdx   U W dx  U 2 W  d  U 2 W  x L
2
 (17.66)
0 0
dx 0

Equation 17.66 indicates that we only need to determine the value of the momentum thickness at
x = L to establish the overall drag on the plate. This is because  is a direct reflection of the
momentum "lost" from the boundary layer (for a dp/dx = 0 flow).

From Eq. 17.58 for the momentum thickness,    , which combined with Eq. 17.60 gives, for
x = L:

636
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

0.0216  0.147 L
 x  L   x  L  (17.67)
Re 0L.147

To calculate the non-dimensional drag coefficient, CD, we divide Eq.17.66 divide through by
1
U 2 WL  , to get a generic expression for CD:
2
L

Drag  Wdx
U 2 W  x L 2 x L
w

CD   0
  (17.68)
1 1 1 L
U  WL 
2 2
U  WL 2
U  WL
2 2 2
Now, substituting Eq. 17.67 into 17.68 for  at x = L, we get:

0.147
2  x L 0.0432 0.147   
CD   0.147
 0.0432   (17.69)
L ReL  ReL 
Equation 17.69 gives the predicted flat plate drag coefficient for a boundary layer that is
turbulent from the leading edge of a flat plate.

17.7.5.2 Laminar-Turbulent Transition at Critical Reynolds Number

If we now consider a boundary layer that is initially laminar, which then transitions to turbulence
at some critical Reynolds number, we can derive an approximate relationship for the boundary
layer growth, cf, and CD for the combined processes. However, we have two problems. The first
is that the critical Reynolds number, Recrit, at which transition from laminar to turbulent flow
occurs is not a well-defined value. It may change markedly, due to variations in the impinging
flow conditions, such as turbulence intensity levels, mean profile non-uniformities, and surface
irregularities and roughness. The second is that transition, as we pointed out earlier, occurs over
a finite length of the plate, yielding a sustaining turbulent boundary layer only after some
development length. However, to facilitate a solution, we will assume that the combined
laminar-turbulent flow makes an immediate transition from laminar to turbulent behavior at a
specified critical Reynolds number, and that our Eq. 17.59 we derived from the integral equation,
applies immediately from that critical Reynolds number onward.

We will assume that the flow over the first portion of the flat plate behaves according to the
Blasius solution we developed in Chapter 13, and that the boundary layer for the laminar region
develops according to Eq.13.29, given by:

 5 5x
 or 
x Re x Re x

637
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

Equation 13.29 assumes that the boundary initiates at the leading edge of the plate, such that
 = 0 at x = 0. If we assume that  = crit at x = xcrit or Rex = Recrit, we can integrate Eq. 17.59 to
establish the behavior of the turbulent portion of the boundary layer, starting the integration at
 = crit at x = xcrit as follows:

0.173 x 
0.00975  U  
 dx   
0.173
  d
    x crit  crit

0.173 1.173
0.00975  U    1crit
.173
  x  x crit   
    1.173
Solving for  gives:

0.853
 0.01144  U   0.173 
   x  x crit   1crit.173 
     
With a bit of work, this can be shown to be:

0.853
  0.01144
  Re x  Re crit   Re1.crit173  (17.70)
U   
or
0.853
 0.01144
  Re x  Re crit   Re1.crit173 
  

x Re x ,
U x U x U 
where Re x  , Re crit   crit , and Re crit   crit . Equation 17.70 will give the value
  
of  for the turbulent portion of the boundary layer based on the respective Reynolds number,
where Rex > Recrit. However, we can simplify the variables required for Eq. 17.70 by using Eq.
13.29 to show:

0.5
Re crit  5 Re crit (17.71)
So, Substituting Eq.17.71 into Eq.17.70 gives:

0.853
 0.01144 0.587 
  Re x  Recrit   6.61Recrit 
   
 (17.72)
x Re x

638
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

Now, to obtain the friction coefficient, cf, differentiate Eq.17.72 and substitute into Eq.17.62,
which after some manipulation gives:

0.147
 0.01144 0.587 
c f  0.0195 Re x  Re crit   6.61Re crit  (17.73)
  

To determine the drag coefficient, CD, for a combined laminar-turbulent flow over a plate, we
combine Eq.17.58 with Eq.17.68 and Eq. 17.72, applied at x = L to obtain:

0.853
 0.01144 0.587 
2 x  L 2 x  L
2  Re L  Re crit   6.61Re crit 
Drag 
CD       (17.74)
1 L L Re L
U 2 WL
2

17.7.5.3 Comparison: Laminar, LE Turbulent, and Transitional Boundary Layers

Figures 17.21 and 17.22 illustrate the growth of a flat plate boundary layer with distance, x, and
the consequent behavior of cf(x) and CD(L), considering the flow is either (a) turbulent from the
leading edge (TLE), or (b) a combination of laminar and turbulent (L-T), with a transition at a
critical Reynolds number. For the turbulent portions, a value of n = 7 was assumed for the
power-law velocity profile ( = 0.0972).

To properly illustrate the non-dimensional change of boundary layer thickness relative to the
plate length, we divide  in Eqs.17.60 and 17.72 by the plate length, L, which gives respectively:

Turbulent from leading edge:

  x  0.0216 0.853  x 
0.853 0.853
0.158  x 
    0.147   for n=7, =0.0972 (17.75)
L Re0.147
L L ReL  L 

Laminar-Turbulent Transition:

x
0.5
5 x
   (laminar) for Re x  Re crit (17.76a)
L Re L  L 

639
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

0.853
 0.01144  x  0.587 
  Re L    Re crit   6.61Re crit 
x     L  

L Re L
(turbulent) for Re x  Re crit (17.76b)
0.853
  x  0.587

0.1177 Re L    Re crit   6.61Re crit  for n = 7,  = 0.0972
 L 
 
Re L

Figure 17.24 shows plots of (x)/L vs. x/L for (1) Eq.17.75 (TLE), and(2) Eq.17.76 (L-T). We
assume Recrit = 5x105 for the latter equation. Results are shown for plate Reynolds numbers of
both ReL = 106 and ReL = 5x106 to illustrate the variance of the boundary layer growth over the
plate as plate length is increased. Note that the laminar boundary layer grows much more slowly
than the turbulent, so for ReL = 106 the difference in the assumed growth modes is quite
pronounced, with the L-T boundary layer being 32% thinner at the end of the plate than the TLE
boundary layer.

As the plate Reynolds number increases, this discrepancy in boundary layer thickness decreases,
since the contribution of the laminar portion to the L-T boundary layer is reduced accordingly.
So, for ReL = 5x106, the difference in the end of plate boundary layer is less than 5%. For
Reynolds numbers exceeding ReL = 107, which reflect most practical external flows, the
difference drops to roughly 1%, which is essentially negligible.

Figure 17.25 Illustrates the behavior of the friction coefficient, cf, as predicted by Eqs. 13.33
(laminar flow), 17.63 and 17.72. Note that cf for a turbulent boundary is significantly greater
than the cf for a comparable laminar boundary layer at the same Reynolds number. This is a
result of the more intense momentum exchange within a turbulent boundary layer, and the higher
shear gradient at the surface. In fact, for Rex = 5x105 (the Recrit value for the combined L-T), the
cf for the TLE turbulent boundary is roughly four times greater than the comparable cf for a
laminar boundary.

Note that the cf predicted by Eq. 17.73 (after transition) for assumed laminar-turbulent transition
(L-T) is initially in excess of that predicted by Eq.17.64a for TLE. At first, this may seem
unusual, but consider the process of boundary layer growth. At the point where the laminar
portion of the L-T boundary layer transitions to turbulence, the boundary layer will be thinner
than the comparable TLE boundary layer. And as figure 17.24 illustrates, the thinner the
boundary layer, the faster it grows. Thus, when the boundary layer transitions from laminar to
turbulent, the subsequent turbulent boundary layer will be growing faster, with a consequent
higher cf, than that of the TLE boundary layer at the same Rex. Of course, as Rex increases, and

640
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

the growth rates of the L-T and TLE become comparable, the difference in cf between the TLE
and L-T predictions become minimal.

(a) ReL = 106

(b) ReL = 5x106

Figure 17.24 Comparison of predicted flat plate boundary layer growth (for n = 7) assuming:
(1) turbulent from the plate leading edge (TLE) [Eq. 17.75] and
(2) transition from laminar to turbulent (L-T) [Eq.17.76] at Recrit = 5x105.
(a) ReL= 106 (b) ReL= 5x106.

641
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

Figure 17.25 Predicted friction coefficient, cf(x) (for n = 7), assuming:

(1) turbulent from the plate leading edge (TLE) [Eq. 17.64a] and
(2) transition from laminar to turbulent (L-T) [Eq.17.73] at Recrit = 5x105.

Figure 17.26 Predicted total drag coefficient, CD(L), for n = 7, assuming

(1) turbulent from the plate leading edge (TLE) [Eq. 17.69] and
(2) transition from laminar to turbulent (L-T) [Eq.17.74] at Recrit = 5x105.

642
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

Figure 17.26 is a plot of CD(L) comparing the predictions for laminar, turbulent from the leading
edge (TLE), and combined laminar-turbulent (L-T) boundary layers. Note that for ReL < 107, a
L-T boundary layer will have lower drag than the TLE boundary layer, since the initial laminar
boundary layer will have significantly lower shear stress than a turbulent boundary layer. For
example, if Recrit = 5x105, and ReL = 106, the L-T drag will be roughly 30% less than for a TLE
flow. However, as ReL increases, the difference between the L-T and the TLE drag will decrease
until they are essentially equivalent at roughly REL = 1.5x107.

As an illustration of when one should be concerned with an initial laminar region, consider that
the Reynold number for the wing width (the chord) of a small private plane traveling at 150
miles per hour (220 ft/s or 67 m/s) is roughly 4 to 5x106. So, from figure 17.26 there would be
little reason to account for any initial laminar portion of such practical aerodynamic flows, since
the CD values for the TLE and L-T predictions are essentially identical. However, the flow over
the rudder of a small boat moving at 15 ft/s or 4.5 m/s will be roughly 2x105, for which the
consideration of an initial portion of laminar flow could have as much as a 15% reduction from a
flow turbulent from the rudder leading edge. The bottom line is that unless a practical flow is
fairly slow, or small, adaptation for an initial laminar portion of the boundary layer is probably
no warranted.

17.8 Conclusion

While this chapter is rather long, it is still a simplistic discussion of turbulence, and how it is
modeled and predicted. The literature on turbulence is quite extensive, and quite contentious.
Because of the complexity of turbulence, the many approaches to the prediction of turbulence
properties, that have been suggested and pursued, have met with varying degrees of success. It
was the intent of this chapter to introduce you to the generic characteristics of turbulence, and
some simple, and reasonably effective, techniques for assessing turbulence properties for simple
geometries. My hope is that having completed this chapter you have a reasonable grasp of
turbulence, the basic methods for approaching solutions, and the difficulties that can develop in
doing so. For a more thorough discussion of semi-empirical approaches to prediction of
turbulence properties, I suggest reading Chapters 5 and 6 of an excellent book, "Viscous Fluid
Flow," by Frank White (1991). But recognize that even White's book just scratches the surface of
our understanding, and does not cover the wealth of material subsequently available. A Google
search will undoubtedly produce many, many recent sources with which to familiarize yourself
with present-day capabilities and understanding.

However, although there have been many marvelous computational and experimental studies,
which have produced some incredible: (1) computational simulations of turbulence behavior

643
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

through massive numerical solutions of the time-dependent, three-dimensional Navier-Stokes

equations, and (2) three-dimensional experimental measurement and visualization of real
turbulent flows, we are still struggling to understand the turbulence process better than it is
described in this chapter. Even though studies can demonstrate a variety of vortical structures
comprising turbulence, researchers are still trying to sort out the cause and effect of these
structures, and to put that understanding to the development of prediction methods that do not
require the use of hours or days on a super computer. Hopefully, there is a von Karman or
Prandtl out there who can synthesize the vast amount of information that has been developed,
and can subsequently develop effective universal prediction methods that can be run on a
personal computer. Time will tell.

References

Coles, D. E. (1956) "The Law of the Wake in the Turbulent Boundary Layer," J. Fluid Mech.,
vol. 1, pp. 191-226.

Coles, D. E. and Hirst, E. A. (1969) Proceedings of Computation of Turbulent Boundary Layers,

1968, AFOSR-IFP-Stanford Conference, vol. II. Thermosciences Division, Stanford University,
California.

Emmons, H. W. (1951) "The Laminar-Turbulent Transition in a Boundary Layer: Part I," J.

Fluid Mech., vol. 2, pp. 15-36.

Klebanoff, P. S., Tidstrom, K. D., and Sargent, L. M. (1962) "The Three-Dimensional Nature of
Boundary Layer Instability," J. Fluid Mech., vol. 12, pp. 1-24.

Marusic, I., McKeon, B. J., Mondewitz, P. A., Nagib, H. M., Smits, A. J., and Sreenivasan, K. R.
(2010) "Wall-bounded turbulent flows at high Reynolds numbers: Recent advances and key
issues," Physics of Fluids, vol. 22, 065103, pp. 1-24.

Nachtsheim, P. R. (1964) "An Initial Value Method for the Numerical Treatment of the Orr-
Sommerfeld Equation for the Case of Plane Poiseuille Flow," NASA Technical Note D-2414.

Orr, W. M'F. (1907). "The stability or instability of the steady motions of a liquid. Part II".
Proceedings of the Royal Irish Academy. A 27: 69–138{{inconsistent citations}}

Nikuradse, J. (1932) "Gesetzmassigkeit der turbulenten Stromung in glatten Rohren," Ing.-Arch.

Vol. 1, pp. 306-332.

644
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

Prandtl, L. (1921) "Uber den Reibungswiderstand stromender Luft," Ergebnisse AVA Gottingen,
1st Series, pp. 136-142.

Prandtl, L. (1925) "Uber die ausgebildete Turbulenz," Z. Angew. Math. Mech., vol. 5, pp. 136-
139.

Prandtl, L. [with Tietjens, O. G.] (1933) In Applied Hydro- and Aero-Mechanics, McGraw-Hill,
New York.

Reynolds O. (1883) "An experimental investigation of the circumstances which determine

whether the motion of water in parallel channels shall be direct or sinuous and of the law of
resistance in parallel channels," Philos. Trans. R. Soc., vol. 174, pp. 935–982.

Reynolds O. (1895) "On the dynamical theory of incompressible viscous fluids and the
determination of the criterion," Philos. Trans. R. Soc., vol. 186, pp. 123–164.

Reynolds, W. C. (1976) "Computation of Turbulent Flows," Ann. Rev. Fluid Mech., vol. 8, pp.
183-208.

Schlicting, H. (1933) "Laminare Strahlenausbreitung," Z.Angew. Math. Mech, vol. 13, pp. 260-
263.

Schubauer, G. B. and Skramstad, H. (1947) "Laminar Boundary Layer Oscillation and Transition
on a Flat Plate," J. Res. Nat. Bur. Standards, vol. 38, pp. 251-292.

Shih, T.S., Povinelli, L.A., Liu, N-S, Potapczuk, M.G., and Lumley, J.L. (1999), "A generalized
wall function," NASA/TM--1999-209398.

Smith, C.R. and Walker, J.D.A (1997) “Sustaining Mechanisms of Turbulent Boundary Layers:
The Role of Vortex Development and Interactions,” in Self-Sustaining Mechanisms of Wall
Turbulence, R.L. Panton, ed., Computational Mechanics Publications, England England, pp. 13-
47, 1997.

Spalding, D. B. (1961) "A Single Formula for the Law of the Wall," J. Appl. Mech., vol. 28,
issue 3, pp. 455-457.

Speziale, C. G. (1991) "Analytical Methods for the Development of Reynolds-Stress Closures in

Turbulence," Ann. Rev. Fluid Mech., vol. 23, pp. 107-157.

645
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

Narashima, R. and Sreenivasan, K.R. (1979) "Relaminarization of Fluid Flows," in Advances in

applied mechanics. Volume 19. (A80-40840 17-31) New York, Academic Press, Inc., 1979, p.
221-309.

Sommerfeld, A. (1908). "Ein Beitrag zur hydrodynamische Erklärung der turbulenten

Flüssigkeitsbewegungen". Proceedings of the 4th International Congress of Mathematicians III.
Rome. pp. 116–124

Tollmien, W. (1929) "Uber die Entstehung der Turbulenz," Nachr. Ges. Wiss. Göttingen, Math.
Phys. Klasse II, pp. 21-44 [translated in NACA Technical Memo. 609].

Van Driest, E. R. (1956) "On turbulent flow near a wall," J. Aero. Sci., vol. 23, pp.1007-1011.

Von Karman, T. (1930) "Mechanische Ahnlichkeit und Turbulenz," In Proceedings of the 3rd
International Congress on Applied Mechanics.

Wazzan, A. R. (1975) "Spatial Stability of Tollmein-Schlicting Waves," Prog. Aerosp. Sci., vol.
16, no. 2, pp. 99-127.

White, F. M. (1991), Viscous Fluid Flow, McGraw-Hill Inc., New York.

White, F. M.(2011), Fluid Mechanics, 7th Ed., McGraw-Hill, inc., New York

Problems

1. If Re crit  Re *  520 for a flat plate boundary layer, determine the corresponding value for
Re crit  Re x .You will need to make use of Eq. 13.30.
cm cm 2
2. Consider a flat plate boundary layer in water with U   10 and   102 . If
s s
Re crit  Re *  520 , for a flat plate boundary layer, what will be the displacement thickness
when Re crit is reached? What will be the boundary layer thickness? At what x location will
this occur along the plate? You will need to use equations for a laminar boundary layer from
Chapter 13.

646
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

3. Using Eq. 17.32 for the logarithmic layer of the inner region, determine an expression for the
eddy viscosity, , in terms of k, y, and u  . Next, determine an expression for , in terms of
k, y+, and the kinematic viscosity, . What is the value of / for y+ =300?

4. Using Eqs. 17.36 and 17.37 for the van Driest model of the entire inner region, determine an
expression for the eddy viscosity, , in terms k, A, y+, and the kinematic viscosity, . What
is the value of / for y+ =5, 21, and 300?

5. Consider Eq. 17.55 where cf = f(Re), but the relationship is implicit. Eq. 17.56 was
determined by assuming flat plate values for k, B, and A(), determining Re values
implicitly by assuming several values of cf from 0.0015 to 0.005 and using a root finding
program to determine the corresponding Re values. The resultant cf vs. Re were then fit to
a power law curve to determine cf = f(Re). Repeat the curve fit process to verify Eq. 17.56
for a flat plate (A()=2.5), and then determine the power law curve fit of cf = f(Re), for a
strong favorable pressure gradient, with A()=1.0. Assume flat plate values for k and B for
both flows. It is suggested you use Excel and its function "Goal Seek" to determine cf vs.
Re, and then "Trendline" on an Excel plot of cf vs. Re to establish the power law fit, giving
cf = f(Re). Compare your results for A=1.0 to the flat plate Eq. 17.55.

m m2 kg
6. Consider a flat plate boundary layer in water with U   0.1 ,   106 ,   1000 3 , a
s s m
plate length of 5 meters, and width of 2 meters.
a. If Re crit  Re *  520 , determine the corresponding value for Re crit  Re x .
b. Then determine the drag force on the plate assuming:
(1) the flow is turbulent from the leading edge, or
(2) that the flow transitions at the Re crit  Re x you determined.
Assume n = 7 in Eq. 17.54. Is the difference between (1) and (2) substantial?

m m2 kg
7. Consider a flat plate boundary layer in air with U   10 ,   1.52x105 ,   1.2 3 , a
s s m
plate length of 5 meters, and width of 2 meters.
a. If Re crit  Re *  520 , determine the corresponding value for Re crit  Re x .
b. Then determine the drag force on the plate assuming:
(1) the flow is turbulent from the leading edge, or
(2) that the flow transitions at the Re crit  Re x you determined.
Assume n = 8 in Eq. 17.54. Is the difference between (1) and (2) substantial?

647
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

cm  2 cm
2
8. Consider a flat plate boundary layer in water with U   10 ,   10 , and a plate
s s
length of 5 meters. Assume n = 8 in Eq.17.54, and determine the displacement thickness (*)
and momentum thickness () at the trailing edge if:

(a) the flow is turbulent from the leading edge, or

(b)the flow transitions to turbulence at Re crit  Re x  3x105 .

Is the difference between (a) and (b) substantial?

m m2
9. Consider a flat plate boundary layer in air with U   10 ,   1.2x10 5 , and a plate
s s
length of 5 meters. Assume n = 8 in Eq.17.54, and determine the displacement thickness (*)
and momentum thickness () at the trailing edge if:

(a) the flow is turbulent from the leading edge, or

(b) the flow transitions to turbulence at Re crit  Re x  6x105 .

Is the difference between (a) and (b) substantial?

10. Consider two flat plate boundary layer flows over a plate of length 5 meters. One flow is
cm cm 2 m
water with U   10 ,   102 , and the other is air with U   10 ,
s s s
m2
  1.2x10 5 . Assume n = 9 in Eq.17.54, and determine the boundary layer thicknesses
s
() at the trailing edge for both flows if they both transition to turbulence at
Re crit  Re x  3x105 . Also, determine the Kolmogorov length scales at the trailing edge for
both flows.

11. Consider the flow of air over a flat plate, which has roughly the area of one wing of a Boeing
777. The plate is of span 25 meters and chord 5 meters. The air flows at roughly a Mach
m m2 kg
number of 0.3, or U   100 with   1.2x10 5 and   1.2 2 . Assuming n = 9, and
s s m
a flow that is turbulent from the leading edge, what is total shear drag (consider both sides)
on the plate? (note the flow is across the chord of the plate, so L = 5 meters).

648
Introduction to Graduate Fluid Mechanics: C.R. Smith Chapter 17

12. Assuming an approximate velocity profile for a flat plate turbulent boundary layer modeled
U   y 
as  1  0.1ln  . First, determine * and  for this profile, and subsequently the
U    
shape factor H to show this is an acceptable approximation of the velocity profile. Note that
y y
for your calculation of * and , you will need to determine the limits of   ln  and
 
2
 y    y   y
 ln  for    0 , which you must do using L'Hospital's rule. Then, following a
       
procedure similar to Section 17.7.5, integrate the Momentum Integral Equation to determine
expressions for /x = f(Rex) and cf = f(Rex). Since dU/dy =  at y = , again assume that cf
= f(Re) is given by Eq. 17.56, and the flow is turbulent from the leading edge. Compare
your results to the corresponding equations for /x = f(Rex) and cf = f(Rex) in Section
17.7.5.1.

649

View publication stats