2.

20 - Marine Hydrodynamics, Spring 2005

Lecture 1

2.20 – Marine Hydrodynamics

Lecture 1

Introduction
Marine Hydrodynamics is the branch of Fluid Mechanics that studies the motion of incom­
pressible fluids (liquids) and the forces acting on solid bodies immersed in them.

Marine hydrodynamics is a large and diverse subject and only a limited number of topics
can be covered in an introductory course. The topics that will be covered throughout the
semester include:

• Model testing and similitude

• Interaction between bodies and ideal fluids

• Viscosity and boundary layers

• Effect of waves on resistance and ship motion

Why study Marine Hydrodynamics?

Studying Marine Hydrodynamics provides a greater understanding of a wide range of phe­
nomena of considerable complexity involving fluids. Another benefit is that it allows pre­
dictions to be made in many areas of practical importance.

Fluids vs. Solids

In brief, Fluid Mechanics studies the kinematics and dynamics of a group of particles with­
out having to study each particle separately.

Most of us have taken some courses on solids or related to solids. Even those who haven’t
can get an intuitive feeling about some physical properties of a solid. Thus a comparison
of solids and fluids will give some guidelines as to which properties can be translated to
fluids and on what terms.

Similarities

1. Fundamental laws of mechanics apply to both fluids and solids

- Conservation of mass
- Conservation of momentum (Newton’s law of motion)
- Conservation of energy (First law of thermodynamics)

2. Continuum hypothesis is used for both fluids and solids

measured property

“particle” size
local value

length scale

variations due to variations due to

molecular fluctuations varying flow
O(10-10 – 10-8)m

e.g. The smallest measurement scales are in the order of M ∼ 10−5 m → VM ∼ 10−15 m3 .
This corresponds to ∼ 3 × 1010 molecules of air in STP or ∼ 1013 molecules of water.

 Differences

1. Shape

- Solids have definite shape

- Fluids have no preferred shape

2. Constitutive laws

Constitutive laws are empirical formulas that relate certain unknown variables.

The constitutive laws used in 13.021 relate:

dynamics (force, stress ...) to kinematics (position, displacement, velocity ...)

Different constitutive laws are used for solids and different for fluids.

- For solids Hooke’s law is used to relate stress and strain

�stress
�� � = f (strain)
� �� �
force/area relative displacement/length
� �
N/m2 [m/m]
•
•• For solid mechanics ‘statics’ is a dominant aspect

- For fluids stress is related to rate of strain

�stress
�� � = f (rate of strain)
� �� �
force/area velocity gradient
� �
N/m2 [1/s]
•
•• Fluids at rest cannot sustain shear force. Fluids have to be moving to be non-trivial.

The branch of Fluid Mechanics that studies fluids at rest is referred to as ‘Hydrostatics’.

(Archimedes, c 200 BC)

Hydrostatics study the trivial case where no stresses due to fluid motion exist.

Sometimes distinction between liquids and solids is not a sharp one(honey, jelly, paint, . . . ).
Fortunately most common fluids, such as air and water are very close to ”ideal” fluids.

Liquids vs. Gasses

Liquids and gasses are two categories of fluids.

A fluid is ‘a body whose particles move easily among themselves. Fluid is a generic term,
including liquids and gasses as species. Water, air, and steam are fluids.’ [1] .

A liquid is ‘Being in such a state that the component parts move freely among themselves,
but do not tend to separate from each other as the particles of gases and vapors do; neither
solid nor aeriform.’[1]

A gas is ‘The state of matter distinguished from the solid and liquid states by relatively low
density and viscosity, relatively great expansion and contraction with changes in pressure
and temperature, the ability to diffuse readily, and the spontaneous tendency to become dis­
tributed uniformly throughout any container.’[2]

In brief, a liquid is generally incompressible and does not fill a volume by expanding into
it while on the other hand, a gas is compressible and expands to fill any volume containing it.

The science that studies the dynamics of liquids is referred to as ‘Hydrodynamics’, while
the science that studies the dynamics of gasses is referred to as ‘Aerodynamics’.

The main difference between the study of ‘Hydrodynamics’ and the study of ‘Aerodynam­
ics’ is the property of incompressibility. In general hydrodynamic flows are treated as
incompressible while aerodynamic flows are treated as compressible.

[1] Webster Dictionary

[2] American Heritage Dictionary

4
Why is a liquid flow incompressible?

It can be shown that the ratio of the characteristic fluid velocities U in a flow to the
speed of sound C in the medium gives a measure of compressibility of the medium for that
particular flow. This ratio is called the Mach number M . Although the speed of sound
in water is of comparable magnitude to the speed of sound in air, the characteristic fluid
velocities in water are significantly smaller. Thus in the case of water, the Mach number
is very small, indicating that water is virtually incompressible.

U : Characteristic fluid flow velocity

C : Speed of sound in the medium
U
M≡ : Mach number
C
The average speed of sound in air and water is:

Cair ∼ 300m/s = 984f t/sec = 583knots

Cwater ∼ 1200m/s = 3, 937f t/sec = 2, 333knots
Cwater
Therefore the average ratio of the speed of sound in water to air is Cair
∼ 4. Further on,
ρwater 1kg/m3 3
because the average water to air density ratio is ρair
∼ 103 kg/m3
= 10 , it is ‘harder’ to
move in water and therefore, typically, it is:

Uwater <<Uair
giving thus typical values of Mach numbers in the order of:

Mair ∼ O(1) ⇒ COMPRESSIBLE flow

Mwater <<1 ⇒ INCOMPRESSIBLE flow

Note: An incompressible flow does not mean constant density.

 2.20 - Marine Hydrodynamics, Spring 2005

Lecture 2

2.20 – Marine Hydrodynamics

Lecture 2

Chapter 1 - Basic Equations

1.1 Description of a Flow
To define a flow we use either the ‘Lagrangian’ description or the ‘Eulerian’ description.
• Lagrangian description: Picture a fluid flow where each fluid particle caries its own
properties such as density, momentum, etc. As the particle advances its properties
may change in time. The procedure of describing the entire flow by recording the
detailed histories of each fluid particle is the Lagrangian description. A neutrally
buoyant probe is an example of a Lagrangian measuring device.

The particle properties density, velocity, pressure, . . . can be mathematically repre­

sented as follows: ρp (t), vp (t), pp (t), . . .

The Lagrangian description is simple to understand: conservation of mass and New­

ton’s laws apply directly to each fluid particle . However, it is computationally
expensive to keep track of the trajectories of all the fluid particles in a flow and
therefore the Lagrangian description is used only in some numerical simulations.

r
υ p (t )
p

Lagrangian description; snapshot

• Eulerian description: Rather than following each fluid particle we can record the
evolution of the flow properties at every point in space as time varies. This is the
Eulerian description. It is a field description. A probe fixed in space is an example
of an Eulerian measuring device.

This means that the flow properties at a specified location depend on the location
and on time. For example, the density, velocity, pressure, . . . can be mathematically
represented as follows: v (x, t), p(x, t), ρ(x, t), . . .

The aforementioned locations are described in coordinate systems. In 13.021 we use

the cartesian, cylindrical and spherical coordinate systems.

The Eulerian description is harder to understand: how do we apply the conservation

laws? However, it turns out that it is mathematically simpler to apply. For this
reason, in Fluid Mechanics we use mainly the Eulerian description.

y
r r
υ ( x, t )
r
( x, t )
x

Eulerian description; Cartesian grid

1.2 Flow visualization - Flow lines

• Streamline: A line everywhere tangent to the fluid velocity v at a given instant (flow
snapshot). It is a strictly Eulerian concept.

• Streakline: Instantaneous locus of all fluid particles that have passed a given point
(snapshot of certain fluid particles).

• Pathline: The trajectory of a given particle P in time. The photograph analogy would
be a long time exposure of a marked particle. It is a strictly Lagrangian concept.

Can you tell whether any of the following figures ( [1] Van Dyke, An Album of Fluid Motion
1982 (p.52, 100)) show streamlines/streaklines/pathlines?

1.3 Some Quantities of Interest

• Einstein Notation
– Range convention: Whenever a subscript appears only once in a term, the sub­
script takes all possible values. E.g. in 3D space:
xi (i = 1, 2, 3) → x1 , x2 , x3
– Summation convention: Whenever a subscript appears twice in the same term
the repeated index is summed over the index parameter space. E.g. in 3D space:
ai bi = a1 b1 + a2 b2 + a3 b3 (i = 1, 2, 3)
Non repeated subscripts remain fixed during the summation. E.g. in 3D space
ai = xij n̂j denotes three equations, one for each i = 1, 2, 3 and j is the dummy
index.

Note 1: To avoid confusion between fixed and repeated indices or different re­
peated indices, etc, no index can be repeated more than twice.

Note 2: Number of free indices shows how many quantities are represented by
a single term.

Note 3: If the equation looks like this: (ui ) (x̂i ) , the indices are not summed.

– Comma convention: A subscript comma followed by an index indicates partial

differentiation with respect to each coordinate. Summation and range conven­
tions apply to indices following a comma as well. E.g. in 3D space:
∂ui ∂u1 ∂u2 ∂u3
ui,i = = + +
∂xi ∂x1 ∂x2 ∂x3
• Scalars, Vectors and Tensors

Scalars Vectors (ai xi ) Tensors (aij )

magnitude magnitude magnitude
direction direction
orientation
density ρ (x, t) velocity v (x, t) /momentum momentum flux
pressure p (x, t) mass flux stress τij (x, t)

1.4 Concept and Consequences of Continuous Flow

For a fluid flow to be continuous, we require that the velocity v (x, t) be a finite and con­
tinuous function of x and t.
i.e. ∇ · v and ∂
v
∂t
are finite but not necessarily continuous.
∂v
Since ∇ · v and ∂t < ∞, there is no infinite acceleration i.e. no infinite forces , which is
physically consistent.

1.4.1 Consequences of Continuous Flow

• Material volume remains material. No segment of fluid can be joined or broken apart.

• Material surface remains material. The interface between two material volumes al­
ways exists.

• Material line remains material. The interface of two material surfaces always exists.

Material
surface

fluid a

fluid b

• Material neighbors remain neighbors. To prove this mathematically, we must prove

that, given two particles, the distance between them at time t is small, and the
distance between them at time t + δt is still small.

 Assumptions At time t, assume a continuous flow (∇ · v , ∂ v

∂t
<< ∞) with fluid velocity
v (x, t). Two arbitrary particles are located at x and x + δx(t), respectively.

Result If δx(t) ≡ δx → 0 then δx(t + T ) → 0, for all subsequent times t + T .

v r v
Proof v x + {υ (x )δt}
x

v
v
δ x δx(t + δt)

v v
v v x +δ x +
x +δ x
{(υr (xv ) + δ xv ⋅∇υr (xv ) )δt}
After a small time δt:

• The particle initially located at x will have travelled a distance v (x)δt and at time
t + δt will be located at x + {v (x)δt}.

• The particle initially located at x+δx will have travelled a distance (v (x) + δx · ∇v (x)) δt.
(Show this using Taylor Series Expansion about(x, t)). Therefore after a small time
δt this particle will be located at x + δx + {(v (x) + δx · ∇v (x))δt}.

• The difference in position δx(t + δt) between the two particles after a small time δt
will be:

δx(t + δt) = x + δx + {(v (x) + δx · ∇v (x))δt} − (x + {v (x)δt})
⇒ δx(t + δt) = δx + (δx · ∇v (x))δt ∝ δx

Therefore δx(t + δt) ∝ δx because ∇v is finite (from continuous flow assumption).
Thus, if δx → 0 , then δx(t + δt) → 0. In fact, for any subsequent time t + T :
� t+T
δx(t + T ) ∝ δx + δx · ∇v dt ∝ δx,
t

and δx(t + T ) → 0 as δx → 0. In other words the particles will never be an infinite distance
apart. Thus, if the flow is continuous two particles that are neighbors will always remain
neighbors.

6
1.5 Material/Substantial/Total Time Derivative: D/Dt
A material derivative is the time derivative – rate of change – of a property following a
fluid particle ‘p’. The material derivative is a Lagrangian concept.

By expressing the material derivative in terms of Eulerian quantities we will be able to

apply the conservation laws in the Eulerian reference frame.

Consider an Eulerian quantity f (x, t). The time rate of change of f as experienced by a
particle ‘p’ travelling with velocity −
→
vp is the substantial derivative of f and is given by:

r r
f (x p + υ pδt , t + δt )
r r
Df (−
→
xp (t), t) f (−
→
xp + −
→
vp δt, t + δt) − f (−
→
xp , t) r x p (t ) + υ pδt
= lim (1) f ( x p ,t )
Dt δt→0 δt particle r
x p (t )

Performing a Taylor Series Expansion about (− →

xp , t) and taking into account that → −
vp δt = δx,
we obtain:
→+−
− → ∂f (x, t)
f (x p vp δt, t + δt) = f (x, t) + δt + δx · ∇f (x, t) + O(δ 2 )(higher order terms)(2)
∂t
From Eq.(1, 2) we see that the substantial derivative of f as experienced by a particle
travelling with →
−
vp is given by:
Df ∂f −
= +→
vp · ∇f
Dt ∂t
The generalized notation:
D ∂
≡ +−→vp · ∇
Dt
���� �∂t �� �
Lagrangian Eulerian

  (x, t) as experienced by a fluid par­

Example 1: Material derivative of a fluid property G
ticle.

Let ‘p’ denote a fluid particle. A fluid particle is always travelling with the local fluid
 (x, t) as experienced
velocity vp (t) = v (xp , t). The material derivative of a fluid property G
by this fluid particle is given by:

DG 
∂G
= + �v ·��
∇G
Dt ∂t �
���� ����
Convective
Lagragian Eulerian
rate of change
rate of change rate of change

Example 2: Material derivative of the fluid velocity v (x, t) as experienced by a fluid par­
ticle. This is the Lagrangian acceleration of a particle and is the acceleration that appears
in Newton’s laws. It is therefore evident that its Eulerian representation will be used in
the Eulerian reference frame.

Let ‘p’ denote a fluid particle. A fluid particle is always travelling with the local fluid
velocity vp (t) = v (xp , t). The Lagrangian acceleration DvDt
(
x,t)
as experienced by this fluid
particle is given by:

Dv ∂v
= + �v ·��
∇v�
Dt
���� ∂t
����
Convective
Lagragian Eulerian
acceleration
acceleration acceleration

1.6 Difference Between Lagrangian Time Derivative and

Eulerian Time Derivative

Example 1: Consider an Eulerian quantity, temperature, in a room at points A and B

where the temperature is different at each point.

∂T Point B: 1o
Point A: 10o Point C:
∂t

∂T
At a fixed in space point C, the temperature rate of change is ∂t
which is an Eulerian time
derivative.

Example 2: Consider the same example as above: an Eulerian quantity, temperature, in

a room at points A and B where the temperature varies with time.

DT ∂T v
= + υfly⋅ ∇ T
Point A: 10o Dt ∂t Point B: 1 o

Following a fly from point A to B, the Lagrangian time derivative would need to include
the temperature gradient as both time and position changes: DT
Dt
= ∂T
∂t
+ vf ly · ∇T

 ∂
1.6.1 Concept of a Steady Flow ( ∂t ≡ 0)
A steady flow is a strictly Eulerian concept.

Assume a steady flow where the flow is observed from a fixed position. This is like watching

∂ D
from a river bank, i.e. ∂t = 0 . Be careful not to confuse this with Dt which is more like
D
following a twig in the water. Note that Dt = 0 does not mean steady since the flow could
speed up at some points and slow down at others.

∂
=0
∂t

1.6.2 Concept of an Incompressible Flow ( Dρ Dt

≡ 0)
An incompressible flow is a strictly Lagrangian concept.

Assume a flow where the density of each fluid particle is constant in time. Be careful not
to confuse this with ∂ρ
∂t
= 0, which means that the density at a particular point in the flow
is constant and would allow particles to change density as they flow from point to point.
Also, do not confuse this with ρ = const , which for example does not allow a flow of two
incompressible fluids.

10

 2.20 - Marine Hydrodynamics, Spring 2005

Lecture 3

2.20 – Marine Hydrodynamics

Lecture 3

1.7 Stress Tensor

1.7.1 Stress Tensor τij
The stress (force per unit area) at a point in a fluid needs nine components to be completely
specified, since each component of the stress must be defined not only by the direction in
which it acts but also the orientation of the surface upon which it is acting.

The first index i specifies the direction in which the stress component acts, and the second
index j identifies the orientation of the surface upon which it is acting. Therefore, the ith
component of the force acting on a surface whose outward normal points in the j th direction
is τij .

X2

22

12
32
21
23
11
13 31
33 X1

X3

Figure 1: Shear stresses on an infinitesimal cube whose surfaces are parallel to the coordinate
system.

 X2
2

P n̂
A1

1
A3
X1
Q
3
R
area A0
A2
X3

Figure 2: Infinitesimal body with surface PQR that is not perpendicular to any of the Cartesian
axis.

Consider an infinitesimal body at rest with a surface PQR that is not perpendicular to any
of the Cartesian axis. The unit normal vector to the surface PQR is n̂ = n1 x̂1 +n2 x̂2 +n3 x̂3 .
The area of the surface = A0 , and the area of each surface perpendicular to Xi is Ai = A0 ni ,
for i = 1, 2, 3.

Newton’s law: Fi = (volume force)i for i = 1, 2, 3
on all 4 faces

Note: If δ is the typical dimension of the body : surface forces ∼ δ 2

: volume forces ∼ δ 3

An example of surface forces is the shear force and an example of volumetric forces is the
gravity force. At equilibrium, the surface forces and volumetric forces are in balance. As
the body gets smaller, the mass of the body goes to zero, which makes the volumetric
 equal to zero and leaving the sum of the surface forces equal zero. So, as δ →
forces
0, all4f aces Fi = 0 for i = 1, 2, 3 and ∴ τi A0 = τi1 A1 + τi2 A2 + τi3 A3 = τij Aj . But the area
 surface ⊥ to Xi is Ai = A0 ni . Therefore τi A0 = τij Aj = τij (A0 nj ), where τij Aj is
of each
the notation (represents the sum of all components). Thus τi = τij nj for i = 1, 2, 3,
where τi is the component of stress in the ith direction on a surface with a normal n . We
call τ i the stress vector and we call τij the stress matrix or tensor.

2
1.7.2 Example: Pascal’s Law for Hydrostatics
In a static fluid, the stress vector cannot be different for different directions of the surface
normal since there is no preferred direction in the fluid. Therefore, at any point in the
fluid, the stress vector must have the same direction as the normal vector n and the same
magnitude for all directions of n .

no summation
  
Pascal’s Law for hydrostatics: τij = − (pi ) (δij )
⎡ ⎤
−p1 0 0
τ = ⎣ 0 −p2 0 ⎦
0 0 −p3

where pi is the pressure acting perpendicular to the ith surface. If p0 is the pressure acting
perpendicular to the surface PQR, then τi = −ni p0 , but:

τi = τij nj = −(pi )δij nj = −(pi )(ni )

Therefore po = pi , i = 1, 2, 3 and n is arbitrary.

1.7.3 Symmetry of the Stress Tensor

To prove the symmetry of the stress tensor we follow the steps:

j
ij

ji ji

o i
ij

Figure 3: Material element under tangential stress.


1. The of surface forces = body forces + mass× acceleration. Assume no symmetry.
Balance of the forces in the ith direction gives:

(δ)(τij )T OP − (δ)(τij )BOT T OM = O(δ 2 ),

since surface forces are ∼ δ 2 , where the O(δ 2 ) terms include the body forces per unit
depth. Then, as δ → 0, (τij )T OP = (τij )BOT T OM .

2. The of surface torque = body moment + angular acceleration. Assume no sym­
metry. Balance of moments about o gives:

(τji δ)δ − (τij δ)δ = O(δ 3 ),

since the body moment is proportional to δ 3 . As δ → 0 , τij = τji .

1.8 Mass and Momentum Conservation

Consider a material volume ϑm and recall that a material volume is a fixed mass of mate­
rial. A material volume always encloses the same fluid particles despite a change in size,
position, volume or surface area over time.

1.8.1 Mass Conservation

The mass inside the material volume is:


M (ϑm ) = ρdϑ
ϑm(t)

Sm(t)

ϑm ( t )

Figure 4: Material volume ϑm (t) with surface Sm (t).

Therefore the time rate of increase of mass inside the material volume is:

d d
M (ϑm ) = ρdϑ = 0,
dt dt
ϑm (t)

which is the integral form of mass conservation for the material volume ϑm .

1.8.2 Momentum Conservation

The fluid velocity inside the material volume in the ith direction is denoted as ui . Linear
momentum of the material volume in the ith direction is

ρui dϑ
ϑm(t)

Newton’s law of motion: The time rate of change of momentum of the fluid in the material
control volume must equal the sum of all the forces acting on the fluid in that volume.
Thus:

d
(momentum)i =(body force)i + (surface force)i
dt     
d
ρui dϑ = Fi dϑ + τij nj dS
dt 
ϑm (t) ϑm (t) Sm (t) τi

 
Divergence Theorems For vectors: ∇· vdϑ = ⊂⊃ 
v .n̂ dS
ϑ
∂vj
S vj nj
∂xj

 
∂τij
For tensors: dϑ = ⊂⊃ τij nj dS
∂xj
ϑ S

Using the divergence theorems we obtain

  
d ∂τij
ρui dϑ = Fi + dϑ
dt ∂xj
ϑm(t) ϑm(t)

which is the integral form of momentum conservation for the material volume ϑm .

1.8.3 Kinematic Transport Theorems

Consider a flow through some moving control volume ϑ(t) during a small time interval Δt.

Let f (x, t) be any (Eulerian) fluid property per unit volume of fluid (e.g. mass, momentum,

etc.). Consider the integral I(t):


I(t) = f (x, t) dϑ
ϑ(t)

According to the definition of the derivative, we can write

d I(t + Δt) − I(t)

I(t) = lim
dt Δt→0
⎧ Δt ⎫
⎪       ⎪
1 ⎨ ⎬
= lim f (x, t + Δt)dϑ − f (x, t)dϑ
Δt→0 Δt ⎪⎩ ⎪
⎭
ϑ(t+Δt) ϑ(t)

S(t+Δt)

ϑ ( t + Δt )

ϑ( t ) S(t)

Figure 5: Control volume ϑ and its bounding surface S at instants t and t + Δt.

 Next, we consider the steps

1. Taylor series expansion of f (x, t + Δt) about (x, t).

∂f
f (x, t + Δt) = f (x, t) + Δt (x, t) + O((Δt)2 )
∂t
  
2. dϑ = dϑ + dϑ
ϑ(t+Δt) ϑ(t) Δϑ
 
where, dϑ = [Un (x, t)Δt] dS and Un (x, t) is the normal velocity of S(t).
Δϑ S(t)

S(t+Δt)

S(t)

v
U n ( x, t )Δt + O( Δt ) 2 dS

Figure 6: Element of the surface S at instants t and t + Δt.

Putting everything together:

⎧ ⎫
⎪  
⎨    ⎪
⎬
d 1 ∂f 2
I(t) = lim dϑf + Δt dϑ + Δt dSUn f − dϑf + O(Δt) (1)
dt ⎪
Δt→0 Δt ⎩ ∂t ⎪
⎭
ϑ(t) ϑ(t) S(t) ϑ(t)

From Equation (1) we obtain the Kinematic Transport Theorem (KTT), which is equivalent
to Leibnitz rule in 3D.

  
d ∂f (x, t)
f (x, t)dϑ = dϑ + f (x, t)Un (x, t)dS
dt ∂t
ϑ(t) ϑ(t) S(t)

For the special case that the control volume is a material volume it is ϑ(t) = ϑm (t) and Un
= v · n
ˆ, where v is the fluid particle velocity. The Kinematic Transport Theorem (KTT),
then takes the form
  
d ∂f (x, t)
f (x, t)dϑ = dϑ + f (x, t)(v · n
ˆ)dS
dt ∂t   
ϑm (t) ϑm (t) Sm (t) f (vi ni )
(Einstein Notation)

Using the divergence theorem,

 
∇· α
dϑ = ⊂⊃ 
 ·n
α ˆdS

ϑ
∂
α S αi ni
∂xi i

we obtain the 1st Kinematic Transport Theorem (KTT)

   
d ∂f (x, t)
f (x, t) dϑ = + ∇ · (fv ) dϑ,
dt ∂t   
ϑm (t) ϑm (t) ∂
(f vi )
∂xi

where f is some fluid property per unit volume.

1.8.4 Continuity Equation for Incompressible Flow

• Differential form of conservation of mass for all fluids Let the fluid property
per unit volume that appears in the 1st KTT be mass per unit volume ( f = ρ):
   
d ∂ρ
0 = ρdϑ = + ∇ · (ρv ) dϑ
↑ dt ↑ ∂t
conservation ϑm (t) 1st KTT ϑm (t)
of mass
But since ϑm is arbitrary the integrand must be ≡ 0 everywhere.

Therefore:
∂ρ
+ ∇ · (ρv ) = 0
∂t
∂ρ
+ [v · ∇ρ + ρ∇ · v ] = 0
∂t  
Dρ
Dt

Leading to the differential form of

Dρ
Conservation of Mass: + ρ∇ · v = 0
Dt

10

• Continuity equation ≡ Conservation of mass for incompressible flow In

general it is ρ = ρ(p, T, . . .), but we consider the special case of an incompressible
flow, i.e. Dρ
Dt
= 0 (Lecture 2).

Note: For a flow to be incompressible, the density of the entire flow need not be
const). As an example consider a flow of more than one incom­
constant (ρ(x, t) =
pressible fluids, like water and oil, as illustrated in the picture below.

Constant ρ

fluid particle
ρ2

fluid particle oil water

ρ1
Figure 7: Interface of two fluids (oil-water)

Since for incompressible flows Dρ

Dt
= 0, substituting into the differential form of the
conservation of mass we obtain the

∂vi
Continuity Equation: ∇ · v ≡ = 0
∂xi
  
rate of volume dilatation

11

1.8.5 Euler’s Equation (Differential Form of Conservation of Momentum)

• 2nd Kinematic Transport Theorem ≡ 1st KTT + differential form of conservation

of mass for all fluids. If G = fluid property per unit mass, then ρG = fluid property
per unit volume

     
d ∂
ρGdϑ = (ρG) + ∇ · (ρGv ) dϑ
dt ↑ ∂t
ϑm (t) 1st KT T ϑm (t)
⎡ ⎤
 ⎢  ⎥
⎢ ∂ρ ∂G ⎥
= ⎢ G + ∇ · ρv +ρ + v · ∇G ⎥
⎢ ∂t ∂t ⎥ dϑ
⎣      ⎦
ϑm (t)
=0 from mass conservation = DG
Dt

The 2nd Kinematic Transport Theorem (KTT) follows:

 
d DG
ρGdϑ = ρ dϑ
dt Dt
ϑm ϑm

Note: The 2nd KTT is obtained from the 1st KTT (mathematical identity) and the
only assumption used is that mass is conserved.

12

• Euler’s Equation
We consider G as the ith momentum per unit mass (vi ). Then,
   
∂τij d Dvi
Fi + dϑ = ρvi dϑ = ρ dϑ
∂xj ↑ dt ↑ Dt
ϑm (t) conservation ϑm (t) 2nd KTT ϑm (t)
of momentum
But ϑm (t) is an arbitrary material volume, therefore the integral identity gives
Euler’s equation:

⎛ ⎞
Dvi ⎜ ∂vi ⎟ ∂τij
ρ ≡ ρ⎜
⎝ + v · ∇vi ⎟
⎠ = Fi +
Dt ∂t    ∂xj
∂v
vj ∂xi
j

And in vector tensor form:


Dv ∂v
ρ ≡ρ + v · ∇v = F + ∇ · τ
Dt ∂t

13

 2.20 - Marine Hydrodynamics, Spring 2005

Lecture 4

2.20 - Marine Hydrodynamics

Lecture 4

Introduction
Governing Equations so far:

Knowns Equations # Unknowns #

density ρ(x, t) Continuity 1 velocities vi (x, t) 3

(conservation of mass)


body force Fi Euler 3 stresses τij (x, t) 6

(conservation of momentum)

4 9


3 of the 9 unknowns of the stress tensor are eliminated by symmetry

The number of unknowns (9) is > than the number of equations (4), i.e. we don’t have
closure. We need constitutive laws to relate the kinematics vi to the dynamics τij .

1.9 Newtonian Fluids

1. Consider a fluid at rest (vi ≡ 0). Then according to Pascal’s Law:

τij = −ps δij (Pascal’s Law)

⎡ ⎤
−ps 0 0
τ = ⎣ 0 −ps 0 ⎦
0 0 −ps

where ps is the hydrostatic pressure and δij is the Kroenecker delta function, equal
 j.
to 1 if i = j and 0 if i =

2. Consider a fluid in motion. The fluid stress is defined as:

τij ≡ −pδij + τ̂ij

  
isotropic components all non-isotropic components
on diagonal both on/off diagonal
where p is the thermodynamic pressure and τ̂ij are the dynamic stresses. It should
be emphasized that −pδij includes all the isotropic components of the stress tensor
on the diagonal, while τ̂ij represents all the non-isotropic components, which may or
may not be on the diagonal (shear and normal stresses). The dynamic stresses τ̂ij is
related to the velocity gradients by empirical relations.
Experiments with a wide class of ‘Newtonian’ fluids showed that the dynamic stresses
are proportional to the rate of strain.

τ ij
Newtonian Fluid

Fluid

∂u k
∂x m

 τ̂ij ≈ linear function of the ( rate of strain ≡ velocity gradient)


 
  
∂ ∂X ∂ ∂X ∂uk
=
∂t ∂x ∂x ∂t ∂xm
 
u

∂uk
i.e. τ̂ij ≈ αijkm i, j, k, m = 1, 2, 3
  ∂xm
34 =81
empirical coefficients
(constants for Newtonian fluids)

For isotropic fluids, this reduces to:



∂ui ∂uj ∂ul
τîj =μ + +λ ,
∂xj ∂xi ∂xl
 
∇·v

The fluid properties in the previous relation are:

• μ - (coefficient of) dynamic viscosity.

• λ - bulk elasticity, ‘second’ coefficient of viscosity

∂ui

For incompressible flow, we have shown that ∇ · v = = 0.

∂xi
Therefore, for an incompressible, isotropic, Newtonian fluid the dynamic or viscous
stresses τ̂ij are expressed as:


∂ui ∂uj
τ̂ij = μ +
∂xj ∂xi

1.9.1 Discussion on viscous stresses τ̂ij

1. Verify that for v = 0 we recover Pascal’s law.

∂ui

Proof: v = 0 ⇒ = 0 ⇒ τ̂ij = 0 ⇒ τij = −pδij + 0 ⇒ hydrostatic conditions

∂xj

2. Verify that τ̂ij = τ̂ji .


∂ui ∂uj
Proof: τ̂ij = μ + = τ̂ji ⇒ symmetry of stress tensor
∂xj ∂xi

3. When τ̂ij , i = j the viscous stress is a normal stress and is given by:

∂ui
τ̂ii = 2μ
∂xi

The normal viscous stresses τ̂ii are the diagonal terms of the viscous stress tensor.
The τ̂ii in general are not isotropic.

 j the viscous stress is a shear stress and is given by:

4. When τˆij , i =


∂ui ∂uj
τ̂ij = μ +
∂xj ∂xi

The shear viscous stresses τˆij , i = j are the off diagonal terms of the viscous stress
tensor.

⎡ ⎤
∂u ∂u ∂v
⎢ 2 ∂x +
∂y ∂x ⎥
5. A 2D viscous tensor has the form: μ
⎢
⎣
∂u ∂v
⎥
⎦

∂v
+ 2
∂y ∂x ∂y

6. Notation 1: The viscous stresses τ̂ij are often referred to (somewhat confusingly) as
shear stresses, despite the fact that when i = j the viscous stress is a normal stress.

 
∂uj
7. Notation 2: Often τij ↔ τ̂ij and τij is used to denote μ ∂ui
∂xj
+ ∂xi
.

4
1.10 Navier-Stokes equations(for Incompressible, Newtonian Fluid)

Equations # Unknowns #
Continuity 1 velocities vi (x, t) 3
Euler 3 stresses τ̂ij (x, t) 6
Newtonian 6 pressure p(x, t) 1
fluid symmetry
10 10
 
closure
To form the Navier-Stokes equations for incompressible, Newtonian fluids, we first substi­
tute the equation for the stress tensor for a Newtonian fluid, i.e.


∂ui ∂uj
τij = −pδij + τ̂ij = −pδij + μ +
↑ ∂xj ∂xi
Newtonian Fluid
into Euler’s equation:

Dui ∂τij
ρ = Fi +
Dt ∂xj


∂p ∂ ∂ui ∂uj
= Fi − +μ +
∂xi ∂xj ∂xj ∂xi

2


∂p ∂ ui ∂ ∂uj
= Fi − +μ +
∂xi ∂x2j ∂xi ∂xj
 
Continuity = 0

Therefore the Navier-Stokes equations for an incompressible, Newtonian fluid in cartesian

coordinates are given as:

Dui ∂ui ∂ui 1 ∂p ∂ 2 ui 1

= + uj =− + ν 2 + Fi Tensor form
Dt ∂t ∂xj ρ ∂xi ∂xj ρ

Dv ∂v 1 1

= + v · ∇v = − ∇p + ν∇2v + F Vector form

Dt ∂t ρ ρ
where ν ≡ μ
ρ
denoted as the kinematic viscosity [ L2 /T ].

5
• Unknowns and governing equations for incompressible, Newtonian fluids

Equations # Unknowns #

Continuity 1 pressure p(x, t) 1

Navier-Stokes 3 velocities vi (x, t) 3

4 4

• Values of constants (density, dynamic and kinematic viscosity) used in 13.021

water air units

density ρ 103 1 [kg/m3 ]

dynamic viscosity μ 10−3 10−2 [kg/ms]

kinematic viscosity ν 10−6 10−5 [m2 /s]

• Notation 1: The Continuity and the Navier-Stokes equations form the Governing
Equations for incompressible, Newtonian fluids.

Continuity + Navier-Stokes = Governing Equations

Notation 2: Alternatively, we refer to each equation Dv

Dt
i
= − ρ1 ∂x
∂p
i
+ ν∇2 vi + ρ1 Fi
as the ith Momentum Equation. In this case, the Continuity and the Momentum
equations form the Navier-Stokes System of Equations.

Continuity + Momentum Equations = Navier-Stokes System of Equations

Both notations are equivalent, and in this text it will be made clear from the context
when the term Navier-Stokes refers to the Momentum Equations or to the System of
Governing Equations.

6
1.11 Boundary Conditions
In the previous paragraphs we formulated the governing equations that describe the flow
of an incompressible, Newtonian fluid. The governing equations (N-S) are a system of
partial differential equations (PDE’s). This 4 × 4 system of equations describes all the
incompressible flows, from rain droplets to surface waves.

One of the reasons this system of equations provides such different solutions lies on the
variety of the imposed boundary conditions. To complete the description of this problem
it is imperative that we specify appropriate boundary conditions. For the N-S equations
we need to specify ‘Kinematic Boundary Conditions’ and ‘Dynamic Boundary Conditions’.

1. Kinematic Boundary Conditions specify the boundary kinematics (position, ve­

locity, . . . ). On an impermeable solid boundary, velocity of the fluid = velocity of the
body. i.e. velocity continuity.

v = u ‘no-slip’ + ‘no-flux’ boundary condition

where v is the fluid velocity at the body and u is the body surface velocity

• v · n
ˆ = u · n ˆ ‘no flux’ ← continuous flow
• v · tˆ = u · t̂ ‘no slip’ ← finite shear stress

v
v

v
u

 2. Dynamic Boundary Conditions specify the boundary dynamics ( pressure, sheer

stress, . . . ).

Stress continuity: p = p + p interface

τij = τij + τij , interface

The most common example of interfacial stress is surface tension.

p interface, τ ij interface
τij '
p'
τ ij
p

1.12 Surface Tension

• Notation: Σ [Tension force / Length] ≡ [Surface energy / Area].

• Surface tension is due to the intermolecular attraction forces in the fluid.

• At the interface of two fluids, surface tension implies in a pressure jump across the
interface. Σ gives rise to Δp across an interface.

• For a water/air interface: Σ = 0.07 N/m. This is a function of temperature, impurities

etc. . .

• 2D Example:
dθ dθ
cos · Δp · Rdθ = 2Σsin ≈ 2Σ dθ
2
  2  2

≈1 ≈ dθ
2
Σ
∴ Δp = R
Higher curvature implies in higher pressure jump at the interface.

8
 p

p’=p+Δp dθ/2

Σ
R Σ

dθ

• 3D Example: Compound curvature


1 1
Δp = + Σ
R1 R2

where R1 and R2 are the principle radii of curvature.

1.13 Body Forces – Gravity

• Conservative forces

F = −∇ϕ for some ϕ,

where ϕ is the force potential.

  2  2
F · dx = 0 or F · dx = − ∇ϕ · dx = ϕ(x1 ) − ϕ(x2 )
1 1

• A special case of a conservative force is gravity F = −ρgk̂.

– In this case the gravitational potential is given by ϕg = ρgz. Therefore:

F = ∇(−ϕg ) = ∇(−ρgz) = ∇ps

 
≡ hydrostatic pressure ps

– Substitute in Navier-Stokes equation

Dv
ρ = −∇p + F
 +ρν∇2v
Dt
body force

= −∇p + ∇(−ρgz) + ρν∇2v

– Define: total pressure ≡ hydrostatic pressure + hydrodynamic pressure

p = ps + pd
p = −ρgz + pd =⇒

pd = p + ρgz

– Re-write Navier-Stokes:

Dv  
ρ = −∇ p + ρgz + ρν∇2v
Dt  
p d = p + ρgz
Dv
ρ = −∇pd + ρν∇2v
Dt
Therefore:
– Presence of gravity body force is equivalent to replacing the total pressure by a
dynamic pressure (pd = p − ps = p + ρgz) in the Navier-Stokes(N-S) equation.
– Solve the N-S equation with pd . To calculate the total pressure p simply add
back the hydrostatic component p = pd + ps = pd − ρgz.

10

 2.20 - Marine Hydrodynamics, Spring 2005

Lecture 5

2.20 - Marine Hydrodynamics

Lecture 5

Chapter 2 - Similitude (Keyword: EQUAL RATIOS)

Similitude: Similarity of behavior for different systems with equal similarity parameters.

Prototype ↔ Model

(real world) (physical/ analytical/ numerical . . . experiments)

Similitude Similarity Parameters (SP’s)

Geometric Similitude Length ratios, angles
Kinematic Similitude Displacement ratios, velocity ratios
Dynamic Similitude Force ratios, stress ratios, pressure ratios
..
.

Internal Constitution Similitude ρ, ν

Boundary Condition Similitude

..
.
For similitude we require that the similarity parameters SP’s (eg. angles, length ratios,
velocity ratios, etc) are equal for the model and the real world.

Example 1 Two similar triangles have equal angles or equal length ratios. In this case
the two triangles have geometric similitude.
Example 2 For the flow around a model ship to be similar to the flow around the prototype
ship, both model and prototype need to have equal angles and equal length
and force ratios. In this case the model and the prototype have geometric
and dynamic similitude.

1
2.1 Dimensional Analysis (DA) to Obtain SP’s
2.1.1 Buckingham’s π theory

Reduce number of variables → derive dimensionally homogeneous relationships.

1. Specify (all) the (say N) relevant variables (dependent or independent): x1 , x2 , . . . xN

e.g. time, force, fluid density, distance. . .

We want to relate the xi ’s to each other I( x1 , x2 , . . . xN ) = 0

2. Identify (all) the (say P) relevant basic physical units (“dimensions”)

e.g. M,L,T (P = 3) [temperature, charge, . . . ].
3. Let π = xα1 1 xα2 2 . . . xαNN be a dimensionless quantity formed from the xi ’s. Suppose

xi = Ci M mi Lli T ti , i = 1, 2, . . . , N
where the Ci are dimensionless constants. For example, if x1 = KE = 12 M V 2 =
1
2
M 1 L2 T −2 (kinetic energy), we have that C1 = 12 , m1 = 1, l1 = 2, t1 = −2. Then

π = (C1α1 C2α2 . . . CNαN )M α1 m1 +α2 m2 +...+αN mN Lα1 l1 +α2 l2 +...+αN lN T α1 t1 +α2 t2 +...+αN tN
For π to be dimensionless, we require
⎧ ⎫
⎪
⎪ 
 ⎪
N
⎪
⎪
⎪ ⎪
⎪

⎪
⎨
αi mi = 0 ⎪
⎬

P αi li = 0 aP × N system of Linear Equations (1)

⎪
⎪ ⎪
⎪
⎪
⎪ αi ti = 0 ⎪
⎪

⎪
⎩

 ⎪ ⎭

Σ notation
Since (1) is homogeneous, it always has a trivial solution,

αi ≡ 0, i = 1, 2, . . . , N (i.e. π is constant)
There are 2 possibilities:

(a) (1) has no nontrivial solution (only solution is π = constant, i.e. independent of
xi ’s), which implies that the N variable xi , i = 1, 2, . . . , N are Dimensionally Independent
(DI), i.e. they are ‘unrelated’ and ‘irrelevant’ to the problem.
(b) (1) has J(J > 0) nontrivial solutions, π1 , π2, . . . , πJ . In general, J < N , in fact,

J = N − K where K is the rank or ‘dimension’ of the system of equations (1).

2
2.1.2 Model Law
Instead of relating the N xi ’s by I(x1 , x2 , . . . xN ) = 0, relate the J π’s by

F (π1 , π2 , . . . πJ ) = 0, where J = N − K < N

For similitude, we require

(πmodel )j = (πprototype )j where j = 1, 2, . . . , J.

If 2 problems have all the same πj ’s, they have similitude (in the πj senses), so π’s serve
as similarity parameters.

Note:

• If π is dimensionless, so is π × const, π const , 1/π , etc. . .

• If π1 , π2 are dimensionless, so is π1 × π2 , π1
π2
, π1const1 × π2const2 , etc. . .

In general, we want the set (not unique) of independent πj ’s, for e.g., π 1 , π 2 , π 3 or π 1 , π 1
× π 2 , π 3 , but not π 1 , π 2 , π 1 × π 2 .

Example: Force on a smooth circular cylinder in steady, incompressible flow

Application of Buckingham’s π Theory.

ρ,ν
D

Figure 1: Force on a smooth circular cylinder in steady incompressible fluid (no gravity)

A Fluid Mechanician found that the relevant dimensional quantities required to evaluate
the force F on the cylinder from the fluid are: the diameter of the cylinder D, the fluid
velocity U , the fluid density ρ and the kinematic viscosity of the fluid ν. Evaluate the
non-dimensional independent parameters that describe this problem.

3
 xi : F, U, D, ρ, ν → N = 5
xi = ci M mi Lli T ti → P = 3

N =5
 



F U D ρ ν
P = 3 mi 1 0 0 1 0
li 1 1 1 -3 2
ti -2 -1 0 0 -1

π = F α1 U α2 Dα3 ρα4 ν α5
For π to be non-dimensional, the set of equations

αi mi =0
αi li =0
αi ti =0

has to be satisfied. The system of equations above after we substitute the values for the
mi ’s, li ’s and ti ’s assume the form:
⎛
⎞

⎛ ⎞ α 1 ⎛ ⎞
1 0 0 1 0
⎜ α2 ⎟ 0

⎜ ⎟
⎝
1 1 1 −3 2
⎠
⎜ ⎟ ⎝
⎠

⎜
α3 ⎟
=
0

−2 −1 0 0 −1
⎝
α4 ⎠
0

α5
The rank of this system is K = 3, so we have j = 2 nontrivial solutions. Two families of
solutions for αi for each fixed pair of (α4 , α5 ), exists a unique solution for (α1 , α2 , α3 ).
We consider the pairs (α4 = 1, α5 = 0) and (α4 = 0, α5 = 1), all other cases are linear
combinations of these two.

 1. Pair α4 = 1 and α5 = 0.
⎛ ⎞⎛ ⎞ ⎛ ⎞
1 0 0 α1 −1

⎝ 0 1 0 ⎠ ⎝ α2 ⎠ = ⎝ 4 ⎠

0 0 1 α3 2

which has solution

⎛ ⎞ ⎛ ⎞

α1 −1

⎝ α2 ⎠ = ⎝ 2 ⎠

α3 2

ρU 2 D2
∴ π1 = F α1 U α2 Dα3 ρα4 ν α5 =
F
Conventionally, π1 → 2π1−1 and ∴ π1 = 1
F
ρU 2 D2
≡ Cd , which is the Drag coefficient.
2

2. Pair α4 = 0 and α5 = 1.
⎛ ⎞⎛ ⎞ ⎛ ⎞
1 0 0 α1 0

⎝ 0 1 0 ⎠ ⎝ α2 ⎠ = ⎝ −2 ⎠

0 0 1 α3 −1

which has solution

⎛ ⎞ ⎛ ⎞

α1 0

⎝ α2 ⎠ = ⎝ −1 ⎠

α3 −1
ν
∴ π2 = F α1 U α2 Dα3 ρα4 υ α5 =
UD
Conventionally, π2 → π2−1 , ∴ π2 = UD
ν
≡ Re , which is the Reynolds number.
Therefore, we can write the following equivalent expressions for the non-dimensional inde­
pendent parameters that describe this problem:

F (π1 , π2 ) = 0 or π1 = f (π2 )
F (Cd , Re ) = 0 or Cd = f (Re )
F UD F UD
F( , )=0 or = f( )
1/2 ρU D
2 2 ν 1/2 ρU D
2 2 ν

5
Appendix A
Dimensions of some fluid properties

Quantities Dimensions
(M LT )
Angle θ none (M 0 L0 T 0 )
Length L L
Area A L2
Volume ∀ L3
Time t T
Velocity V LT −1
Acceleration V̇ LT −2
Angular velocity ω T −1
Density ρ M L−3
Momentum L M LT −1
Volume flow rate Q L3 T −1
Mass flow rate Q M T −1
Pressure p M L−1 T −2
Stress τ M L−1 T −2
Surface tension Σ M T −2
Force F M LT −2
Moment M M L2 T −2
Energy E M L2 T −2
Power P M L2 T −3
Dynamic viscosity μ M L−1 T −1
Kinematic viscosity ν L2 T −1

 2.20 - Marine Hydrodynamics, Spring 2005

Lecture 6

2.20 - Marine Hydrodynamics

Lecture 6

2.2 Similarity Parameters from Governing Equations and

Boundary Conditions
In this paragraph we will see how we can specify the SP’s for a problem that is governed by
the Navier-Stokes equations. The SP’s are obtained by scaling, non-dimensionalizing
and normalizing the governing equations and boundary conditions.
1. Scaling First step is to identify the characteristic scales of the problem.
For example: Assume a flow where the velocity magnitude at any point in space or
time |v(x, t)| is about equal to a velocity U , i.e. |v(
 x, t)| = α U , where α is such that
0 ≤ α ∼ O(1). Then U can be chosen to be the characteristic velocity of the flow
and any velocity v can be written as:

v = Uv 
where it is evident that v  is:
(a) dimensionless (no units), and
(b) normalized (|v  | ∼ O(1)).
Similarly we can specify characteristic length, time, pressure etc scales:
Characteristic scale Dimensionless and Dimensional quantity
normalized quantity in terms of characteristic scale
Velocity U v  v = Uv 
Length L x x = Lx
Time T t t = T t
Pressure po − pv p p = (po − pv )p

2. Non-dimensionalizing and normalizing the governing equations and bound­

ary conditions
Substitute the dimensional quantities with their non-dimensional expressions (eg.
substitute v with Uv  , x with Lx , etc) into the governing equations, and boundary
conditions. The linearly independent, non-dimensional ratios between the character­
istic quantities (eg. U , L, T , po − pv ) are the SP’s.

(a) Substitute into the Continuity equation (incompressible flow)

∇ · v = 0 ⇒
U  
∇ · v = 0 ⇒
L
∇ · v  = 0

Where all the () quantities are dimensionless and normalized (i.e., O(1)),
∂v 
for example, ∂x  = O(1).

(b) Substitute into the Navier-Stokes (momentum) equations

∂v 1
+ (v · ∇) v = − ∇p + ν∇2v − g ĵ ⇒
∂t ρ
U ∂v  U 2  po − pv   νU  2 
+ (v · ∇ ) v  = − ∇ p + 2 (∇ ) v − g ĵ
T ∂t  L ρU 2 L
U2
divide through by L , i.e., order of magnitude of the convective inertia term ⇒

� � � � � �
L ∂v p�
o − pv  �ν � 2 � gL
+ (v · ∇)v = − (∇p) + ∇ 
v − ĵ
U T ∂t ρU 2 UL U2

The coefficients ( � ) are SP’s.

2
Since all the dimensionless and normalized terms () are of O(1), the SP’s
( � ) measure the relative importance of each term compared to the con­
vective inertia. Namely,
L Eulerian inertia ∂v
• ≡ S = Strouhal number ∼ ∼ ∂t
UT convective inertia (v · ∇)v

The Strouhal number S is a measure of transient behavior.

For example assume a ship of length L that has been travelling with velocity
U for time T . If the T is much larger than the time required to travel a ship
length, then we can assume that the ship has reached a steady-state.

L
<< T ⇒
U
L
= S << 1 ⇒
UT
∂v
ignore → assume steady-state
∂t

po − p v
• 1 ≡ σ = cavitation number.
2
ρU 2
The cavitation number σ is a measure of the likelihood of cavitation.

If σ >> 1, no cavitation. If cavitation is not a concern we can choose po

as a characteristic pressure scale, and non-dimensionalize the pressure p as
p = p o p
po pressure force
• 1 ≡ Eu = Euler number ∼
2
ρU 2 inertia force

UL inertia force
• ≡ Re = Reynold’s number ∼
ν viscous force
If Re >> 1, ignore viscosity.
�
U2 U � � 12
• = √ ≡ Fr = Froude number ∼ inertia force
gL gL gravity force

3
(c) Substitute into the kinematic boundary conditions

 boundary ⇒
u = U
u = U boundary


(d) Substitute into the dynamic boundary conditions

� ��
1 1 R=LR
p = pa + Δp = pa + + =⇒
R R p=(po −pv )p
� 1 �� 2 �
Δp
� � � �� �
1 1 2 � /ρ
p 
= pa + +  = pa +
(po − pv ) L R1 R2

σ U 2L

U 2L inertial forces
• � ≡ We = Weber number ∼
/ρ surface tension forces

• Some SP’s used in hydrodynamics (the table is not exhaustive):

SP Definition

Reynold’s number Re UL
∼ inertia
ν viscous
�
Froude number Fr U2
∼ inertia
gL gravity
pressure
Euler number Eu 1
po
ρU 2
∼ inertia
2

po −pv pressure
Cavitation number σ 1
ρU 2
∼ inertia
2

Strouhal number S L
∼ Eulerian inertia
UT convective inertia
Weber number We U 2L
∼ inertia
Σ/ρ surface tension

2.3 Similarity Parameters from Physical Arguments

Alternatively, we can obtain the same SP’s by taking the dimensionless ratios of significant
flow quantities. Physical arguments are used to identify the significant flow quantities.
Here we obtain SP’s from force ratios. We first identify the types of dominant forces acting
on the fluid particles. The SP’s are merely the ratios of those forces.

1. Identify the type of forces that act on a fluid particle:

� 2�
1.1 Inertial forces ∼ mass × acceleration ∼ (ρL ) UL = ρU 2 L2
3

∂u � U� 2
1.2 Viscous forces ∼ μ × area ∼ μ L (L ) = μU L
∂y
����
shear stress

1.3 Gravitational forces ∼ mass × gravity ∼ (ρL3 )g

1.4 Pressure forces ∼ (po − pv )L2

2. For similar streamlines, particles must be acted on forces whose resultants are in the
same direction at geosimilar points. Therefore, the following force ratios must be
equal:

inertia ρU 2 L2 UL
• ∼ = ≡ Re
viscous μU L ν

� �1/2 � �1/2
inertia ρU 2 L2 U
• ∼ =√ ≡ Fr
gravity ρgL3 gL

�1 �−1
2
inertia (po − pv )L2 p o − pv
• ∼ 1 2 2
= 1 2 ≡σ
pressure 2
ρU L 2
ρU

2.4 Importance of SP’s

• The SP’s indicate whether different systems have similar flow properties.

• The SP’s provide guidance in approximating complex physical problems.

Example A hydrofoil of length L is submerged in a known fluid (density ρ, kinematic

viscosity ν). Given that the hydrofoil is travelling with velocity U and the gravitational
acceleration is g, determine the hydrodynamic force F on the hydrofoil.

F ρ,ν
g U
L

SP’s for this problem:

L p o − pv U 2L U UL
S= , σ= 1 , We = � , Fr = √ , Re =
UT 2
ρU 2 /ρ gL ν

We define the dimensionless force coefficient:

F
CF ≡ 1
2
ρU 2 L2
The force coefficient must depend on the other SP’s:
� � �
CF = CF (S, σ, We , Fr , Re ) or CF = CF S, σ −1 , We−1 , Fr , Re−1

Procedure We will first study under what conditions each SP → 0.

We will estimate CF for the case that all of the SP’s → 0.

1. Significance of the Strouhal number S = L/U T .

Change S keeping all other SP’s (σ −1 , We−1 , Fr , Re−1 ) fixed.

CF transient

Steady-State

S-1 = UT / L
S~O(1)
S~O( 1)

Exact position of the cut

depends on the problem and
the quantities of interest.

For S << 1, assume steady-state: ∂t

∂
=0
For S >> 1, unsteady effect is dominant.

For example, for the case L = 10m and U = 10m/s we can neglect the unsteady
effects when:

L L
S << 1 ⇒ << 1 ⇒ T >> ⇒ T >> 1s
UT U

Therefore for T >> 1s we can approximate S 1 and we can assume steady state.
In the case of a steady flow:

� �
CF = CF S 0, σ −1 , We−1 , Fr , Re−1 ⇒

� �
CF ∼
= CF σ −1 , We−1 , Fr , Re−1

7
 po − pv
2. Significance of the cavitation number σ = 1 .
2
ρU 2

Change σ −1 keeping all other SP’s (S 0, We−1 , Fr , Re−1 ) fixed.

Some comments on cavitation

pv : Vapor pressure, the pressure at which water boils
p o ≤ pv : State of fluid changes from liquid to gas ⇒ CAVITATION
Consequences : Unsteady → Vibration of structures, which may lead to fatigue, etc
Unstable → Sudden cavity collapses
→ Large force acting on the structure surface
→ Surface erosion

CF

Strong
cavitation No cavitation

σinception σ

For σ << 1, cavitation occurs.

For σ >> 1 ⇒ σ −1 << 1, cavitation will not occur.

In general cavitation occurs when we have large velocities, or when po ∼ pv

8
For example, assume a hydrofoil travelling in water of density ρ = 103 kg/m3 .

The characteristic pressure is po = 105 N/m2 and the vapor pressure is pv = 103 N/m2 .
Cavitation will not occur when:
�
1 2
2
ρU p o − pv
σ −1 << 1 ⇒ << 1 ⇒ U << 1 ⇒ U << 14m/s
p o − pv 2
ρ

Therefore for U << 14m/s it is σ >> 1 ⇒ σ −1 0 and cavitation will not occur.

In the case of a steady, non-cavitating flow:

� �
CF = CF 0, σ −1 0, We−1 , Fr , Re−1 ⇒

� �
CF ∼
= CF We−1 , Fr , Re−1

 U 2L
3. Significance of the Weber number We = .
Σ/ρ

Change We−1 keeping the other SP’s (S 0, σ −1 0, Fr , Re−1 ) fixed.

For We << 1, surface tension is significant.

For We >> 1 ⇒ We−1 << 1, surface tension is not significant.

For example, assume a hydrofoil travelling with velocity U = 1m/s� near an air/water
interface (water density ρ = 103 kg/m3 , surface tension coefficient = 0.07N/m).

Surface tension can be neglected when:

Σ Σ
ρ ρ
We−1 << 1 ⇒ << 1 ⇒ L >> ⇒ L >> 7 · 10−5 m
U 2L U2

Therefore for L >> 7 · 10−5 m it is We >> 1We−1 1 and surface tension effects can
be neglected.

So in the case of a steady, non-cavitating, non-surface tension flow:

� �
CF = CF 0, 0, We−1 0, Fr , Re−1 ⇒

� �
CF ∼
= CF Fr , Re−1

10

 U
4. Significance of the Froude number Fr = √ , which measures the ‘gravity effects’.
gh

Change Fr keeping the other SP’s (S 0, σ −1 0, We−1 0, Re−1 ) fixed.

‘Gravity effects’, hydrostatic pressure do not create any flow (isotropic) nor do they
change the flow dynamics unless Dynamic Boundary Conditions apply.

√ √
‘Gravity effects’ are not significant when U << gh ⇒ Fr 0, or U >> gh ⇒
Fr−1 0. Physically, this is the case when the free surface is

• absent or
• far away or
• not disturbed, i.e., no wave generation.

The following figures (i - iv) illustrate cases where gravity effects are not significant.

11

In any of those cases the gravity effects are insignificant and equivalently Fr is not
important (i.e. Fr 0 or Fr−1 0).

So in the case of a steady, non-cavitating, non-surface tension, with no gravity

effects flow:

� �
CF = CF 0, 0, 0, Fr 0 or Fr−1 0, Re−1 ⇒

� �
CF ∼
= CF Re−1

A look ahead: Froude’s Hypothesis

Froude’s Hypothesis states that

CF = CF (Fr , Re ) = C1 (Fr ) + C2 (Re )

Therefore dynamic similarity requires

(Re )1 = (Re )2 , and

(Fr )1 = (Fr )2

Example: Show that if ν and g are kept constant, two systems

(1, 2) can be both geometrically and dynamically similar only
if:

L1 = L2 , and
U1 = U 2

12

 UL
5. Significance of the Reynolds number Re = .
ν

Change Re keeping the other SP’s (S 0, σ −1 0, We−1 0, Fr 0 or Fr−1 0)

fixed.

Recall that for a steady, non-cavitating, non-surface tension, with no gravity

effects flow: � �
CF = CF Re−1

CF Sphere

Plate

Laminar (Re)cr Turbulent Re

Transition
Re << 1, Stokes flow (creeping flow)
Re < (Re )cr , Laminar flow
Re > (Re )cr , Turbulent flow
Re → ∞, Ideal fluid

For example, a hydrofoil of cord length L = 1m travelling in water (kinematic vis­

cosity ν = 10−1 m2 /s) with velocity U = 10m/s has a Reynolds number with respect
to L:

ν
Re = = 107 → ideal fluid, and Re−1 0
UL

13
Therefore for a steady, non-cavitating, non-surface tension, with no-gravity
effects flow in an ideal fluid:

CF = CF (0, 0, 0, 0, 0) = constant = 0

→D’Alembert’s Paradox
No drag force on moving body in ideal fluid.

14

 2.20 - Marine Hydrodynamics, Spring 2005

Lecture 7

2.20 - Marine Hydrodynamics

Lecture 7

Chapter 3 – Ideal Fluid Flow

The structure of Lecture 7 has as follows: In paragraph 3.0 we introduce the concept of
inviscid fluid and formulate the governing equations and boundary conditions for an ideal
fluid flow. In paragraph 3.1 we introduce the concept of circulation and state Kelvin’s
theorem (a conservation law for angular momentum). In paragraph 3.2 we introduce the
concept of vorticity.

⎧
⎪
⎨
Inviscid Fluid ν=0
Ideal Fluid Flow ≡ +
⎪
⎩ Incompressible Flow (§ 1.1) Dρ
= 0 or ∇ · v = 0
Dt

3.0 Governing Equations and Boundary Conditions for Ideal Flow

• Inviscid Fluid, Ideal Flow
Recall Reynolds number is a qualitative measure of the importance of viscous forces
compared to inertia forces,

UL inertia forces
Re = =
ν viscous forces
For many marine hydrodynamics problems studied in 13.021 the characteristic lengths
and velocities are L ≥ 1m and U ≥ 1m/s respectively. The kinematic viscosity in
water is νwater = 10−6 m2 /s leading thus to typical Reynolds numbers with respect to
U and L in the order of

UL
Re = ≥ 106 >>> 1 ⇒
ν
1 viscous forces
∼ 0
Re inertia forces

This means that viscous effects are << compared to inertial effects - or confined
within very small regions. In other words, for many marine hydrodynamics prob­
lems, viscous effects can be neglected for the bulk of the flow.

Neglecting viscous effects is equivalent to setting the kinematic viscosity ν = 0, but

ν = 0 ⇔ inviscid fluid

Therefore, for the typical marine hydrodynamics problems we assume

incompressible flow + inviscid fluid ≡ ideal fluid flow

which turns out to be a good approximation for many problems.

• Governing Equations for Ideal Fluid Flow

– Continuity Equation:

∇ · v = 0

– Momentum (Navier-Stokes ⇒ Euler) equations:

∂v 1
+ v · ∇v = − ∇p − g ĵ
∂t ρ

By neglecting the viscous stress term (ν∇2v ) the Navier-Stokes equations reduce

to the Euler equations. (Careful not to confuse this with the Euler equation in

§1.6).

The N-S equations are second order PDE’s with respect to space

(2nd order in ∇2 ), thus: (a) require 2 kinematic boundary conditions, and (b)

produce smooth solutions in the velocity field.

The Euler equations are first order PDE’s, thus: (a) require 1 kinematic bound­

ary condition, and (b) may allow discontinuities in the velocity field.

• Boundary Conditions for Euler equations (Ideal Flow):

– KBC:
� �
v · n̂ = u · n
ˆ=U ← ‘no flux’ + free (to) slip
  n
given

 · t̂ does not apply.

Note: ‘No slip’ condition v · tˆ = U

The ‘no slip’ condition is required to ensure that the velocity gradients are finite
and therefore the viscous stresses τ̂ij are finite.

But since ν = 0 the viscous stresses are identically zero (τ̂ij ∝ μ = ρν = 0)

and the velocity gradients can be infinite. Or else the velocity field need not be
continuous.

 ∂u
Viscousflow τ w ∝ μ
y u(dy) ∂y
<∞ <∞

u(0) ∂u
U Inviscidflow τ w ∝0
∂y

– DBC:

p = . . . Pressure given on the boundary

Similarly to the argument for the KBC, viscous stresses τ̂ij cannot be specified
on any boundary since ν = 0.

• Summary of consequences neglecting viscous effects this far:

– Neglecting viscous effects is equivalent to setting the kinematic viscosity equal

to zero:

ν=0

Setting ν = 0 ⇐⇒ inviscid fluid

– Setting ν = 0 the viscous term in the Navier-Stokes equations drops out and we
obtain the Euler equations.

The Euler equations are 1st order PDE’s in space, thus (a) require only one
boundary condition for the velocity and (b) may allow for velocity jumps.

– Setting ν = 0 all the viscous stresses τ̂ij ∝ μ = ρν are identically 0. This may
allow for infinite velocity gradients.

This affects (a) the KBC, allowing free slip, and (b) the DBC, where no viscous
stresses can be specified on any boundary.

4
3.1 Circulation – Kelvin’s Theorem
3.1.1 Γ ≡ Instantaneous circulation around any arbitrary closed contour C.

C v
v v
dx

Γ= · dx
v  
C tangential
velocity

The circulation Γ is an Eulerian idea and is instantaneous, a ‘snapshot’.

3.1.2 Kelvin’s Theorem (KT) :

For ideal fluid under conservative body forces,

dΓ
= 0 following any material contour C,
dt
i.e., Γ remains constant under for Ideal Fluid under Conservative Forces (IFCF).

This is a statement of conservation of angular momentum.

(Mathematical Proof: cf JNN pp 103)

Kinematics of a small deformable body:

(a) Uniform translation → Linear momentum

(b) Rigid body rotation → Angular momentum
(c) Pure strain→ No linear or angular momentum involved (no change in volume
(d) Volume dilatation

For Ideal Fluid under Conservative body Forces:

(a) Linear momentum → Can change

(b) Angular momentum → By K.T., cannot change
(c) Pure strain→ Can change
(d) Volume dilatation → Not allowed (incompressible fluid)

Kelvin’s Theorem is a statement of conservation of angular momentum

under IFCF.

Example 1: Angular momentum of point mass.

v1
v2
r1 m1
θ r2 m2

Angular momentum of point mass:

 
 


L
= |r × (mv )| = mvr = mr
2 θ̇

Conservation of angular momentum:

   
 
  

L


=

L

1 2
1 2 m =m
m1 v1 r1 = m2 v2 r2 =⇒
v1 r1 = v2 r2 or
r12 θ̇1 = r22 θ̇2

Conservation of angular momentum does not imply constant angular

velocity:

Angular Momentum � angular velocity ω



Example 2: Conservation of circulation around a shrinking circular material volume

Vm .

υ2
υ1 r1
r2

Vm
Vm

2π
2π
Γ1 = dθr1 v1 = dθr2 v2 = Γ2
0 0

Example 3: Conservation of circulation around a shrinking arbitrary material vol­

ume Vm , Cm .

Γ1 = v1 · dx = v2 · dx = Γ2

C1 C2

3.2 Vorticity
3.2.1 Definition of Vorticity
  
∂w ∂v ∂w ∂u ∂v ∂u
 = ∇ × v =
ω − î − − ĵ + − k̂
∂y ∂z ∂x ∂z ∂x ∂y

Relationship of vorticity to circulation - Apply Stokes’ Theorem:

Γ= v · dx = (∇ × v) · ndS

ˆ = ω · ndS
ˆ ≡ Flux of vorticity out of S
C any S ‘covering’ C any S ‘covering’ C

3.2.2 What is Vorticity?

∂
For example, special case: 2D flow - w = 0; ∂z
= 0; ωy = ωx = 0 and

∂v ∂u
ωz = −
∂x ∂y

(a) Translation: u = constant, v = constant

time t + Δt

uiˆ + υ ˆj uiˆ + υ ˆj

uiˆ + υ ˆj uiˆ + υ ˆj
time t

∂v ∂u
= 0, = 0 ⇒ ωz = 0 → no vorticity
∂x ∂y

9
(b) Pure Strain (no volume change):

Areat + Δt

Areat

No volume change Areat = Areat + Δt

∂u ∂v ∂u ∂v
= − ; u = -v; = 0; = 0 ⇒ ωz = 0
∂x ∂y ∂y ∂x

10

(c) Angular deformation

δ x = ⎛⎜ ∂u ∂y dy ⎞⎟Δt
⎝ ⎠

r
υ = ∂u ∂y dy iˆ + ∂υ ∂y dy ˆj time
t + Δt

dy

time t
( )
δ y = ∂υ ∂x dx Δt
dx
r r
υ =0 υ = ∂u ∂x dx iˆ + ∂υ ∂x dx ĵ

∂u ∂v
 = 0 only if
ω = → δx = δy( for dx = dy)
∂y ∂x

11

 (d) Pure rotation with angular velocity Ω

time t + Δt r
υ = − Ωdy iˆ
Ω Δt time t

dy

Ω Δt

Ω r
υ =0
dx r
υ = Ωdx ĵ

∂v ∂u
= Ω; = −Ω; ωz = 2Ω
∂x ∂y
i.e. vorticity ∝ 2(angular velocity).

3.2.3 Irrotational Flow

A flow is irrotational if the vorticity is zero everywhere or if the circulation is zero
along any arbitrary closed contour:

�
ω ≡ 0 everywhere ⇔ Γ ≡ 0 for any C
Further on, if at t = to , the flow is irrotational, i.e., Γ ≡ 0 for all C, then Kelvin’s
theorem states that under IFCF, Γ ≡ 0 for all C for all time t:

once irrotational, always irrotational

(Special case of Kelvin’s theorem)

12
 2.20 - Marine Hydrodynamics, Spring 2005
Lecture 8

2.20 - Marine Hydrodynamics

Lecture 8
In Lecture 8, paragraph 3.3 we discuss some properties of vortex structures. In paragraph
3.4 we deduce the Bernoulli equation for ideal, steady flow.

3.3 Properties of Vortex Structures

3.3.1 Vortex Structures

• A vortex line is a line everywhere tangent to ω

.
vortex line
r
ω2
Ω2 r
u2

r
ω1
Ω1 r
u1

• A vortex tube (filament) is a bundle of vortex lines.

vortex tube

r
u
r
ω

vortex lines

1
• A vortex ring is a closed vortex tube.
A sketch and two pictures of the production of vortex rings from orifices are

shown in Figures 1, 2, and 3 below.

(Figures 2,3: Van Dyke, An Album of Fluid Motion 1982 p.66, 71)

side view
v
u v
u
U v
ω
v
u Γ

cross section
v v
u
ω
U
v v
u ω

Figure 1: Sketch of vortex ring production

3.3.2 No Net Flux of Vorticity Through a Closed Surface

Calculus identity, for any vector v :

∇ · (∇ × v) = 0 ⇒
 
ω

∇ · ω =0⇒
 
∇ · ω =   · n̂
ω
 dS = 0
⇑
V Divergence S vorticity flux

Theorem

i.e. The net vorticity flux through a closed surface is zero.

(a) No net vorticity flux through a vortex tube:

(Vorticity Flux)in = (Vorticity Flux)out ⇒ (ω · n̂)in δAin = (ω

 · n̂)out δAout

(ωv ⋅ n̂ )out

v
ω ⋅ nˆ = 0
(ωv ⋅ n̂ )in

(b) Vorticity cannot stop anywhere in the fluid. It either traverses the fluid begin­
ning or ending on a boundary or closes on itself (vortex ring).

r r
ω ω

3.3.3 Conservation of Vorticity Flux

 
0 = Γ3 = v · dx =  · ndS
ω ˆ =0
C3 S3

n̂1 C3 n̂2
C1 C2

  
Γ1 = v · dx =  · n̂1 dS =
ω  · n̂2 dS = Γ2
ω
C1 S1 S2

Therefore, circulation is the same in all circuits embracing the same vortex tube. For
the special case of a vortex tube with ‘small’ area:

Γ = ω1 A1 = ω2 A2

ω1 ω2

A1 A2

An application of the equation above is displayed in the figure below:

ω2
ω1
A2 = 2A1
A1
ω2 = ω1/2

3.3.4 Vortex Structures are Material Structures

Consider a material patch Am on a vortex tube at time t.

Am

∂Am

By definition,

ω · n̂ = 0 on An

Then,
 
Γ∂Am = v · dx =  · nds
ω ˆ =0
∂Am Am

At time t + Δt, Am moves, and for an ideal fluid under the influence of conservative
body forces, Kelvin’s theorem states that

Γ∂Am = 0
So, ω · n̂ = 0 on Am still, i.e., Am still on the vortex tube. Therefore, the vortex tube
is a material tube for an ideal fluid under the influence of conservative forces. In the
same manner it can be shown that a vortex line is a material line, i.e., it moves with
the fluid.

3.3.5 Vortex stretching

Consider a small vortex filament of length L and radius R, where by definition ω is
tangent to the tube.

R A

Γ = ωA = constant (in time)

↑ ↑
Stokes Kelvins’
Theorem Theorem

But tube is material with volume = AL = πR2 L = constant in time (continuity)

Γ ωA ω
∴ = = = constant
Volume LA L
As a vortex stretches, L increases, and since the volume is constant (from continuity),
A and R decrease, and due to the conservation of the angular momentum, ω increases.
In other words,

Vortex stretching ⇔ L ↑ ⇒ ω ↑ (conservation of angular momentum)

⇒ A and R ↑ (continuity)

3.3.6 Summary on Vortex Structures

r
ω
Γ Γ

R
r
r
ω A

Vortex ring length L = 2πR [L]

Cross sectional area A = πr2 [L2 ]

Vortex ring volume ∀ = AL = const [L3 ]

⇑
continuity

Vorticity ω = ∇ × v [T−1 ]

Circulation Γ = const [L2 T−1 ]

⇑
Kelvin’s theorem

Γ = ωA = const [L2 T−1 ]

⇑
vorticity flux through A

Γ ∝ U r = const [L2 T−1 ]

Continuity relates length ratios

⎧
⎪
⎪ ∀
⎪
⎪ A ∝ ∴ as L ↑ A↓
⎪
⎪ L
⎪
⎨
∀ = LA = const

⎪
⎪
⎪
⎪ 
⎪
⎪ ∀
⎪
⎩ r ∝ ∴ as L ↑ r↓
L

Kelvin’s theorem + Continuity relate length ratios to Γ, ω, U

Γ r∝ ∀/L L
U r ∝ Γ = const → U∝ → U ∝Γ ∴ as L ↑ U↑
r ∀

Γ A∝∀/L L
ωA ∝ Γ = const → ω∝ → ω∝Γ ∴ as L ↑ ω↑
A ∀

Example 1: Example 2:

A2 ω2
r A2

A1
Γ Γ2
r
L2 r
ω1 ω2
L1
A1 r Γ1
ω1

A1 < A2 L1 < L 2 Given

ω1 > ω 2 A1 > A2 From continuity only
Γ1 = Γ2 From Kelvin’s theorem
ω1 < ω 2 From Kelvin’s theorem + continuity
U1 < U 2 From Kelvin’s theorem + continuity

8
 ∂
3.4 Bernoulli Equation for Steady ( ∂t = 0), Ideal(ν = 0),
Rotational flow


p=f v Viscous flow: Navier-Stokes’ Equations (Vector Equations)
p = f (|v |) Ideal flow: Bernoulli Equation (Scalar equation)

Steady, inviscid Euler equation (momentum equation):

 
p
v · ∇v = −∇ + gy (1)
ρ
From Vector Calculus we have

∇ (u · v ) = (u · ∇) v + (v · ∇) u + u × (∇ × v ) + v × (∇ × u) ⇒

∇( 12 |v |2 ) = v · ∇v + v × (∇ × v ) ⇒
 2
v
v · ∇v = ∇ − v × (∇ × v ) where v 2 ≡ v · v = |v |2
2

From the previous identity and Equation (1) we obtain

 2  
v p
v · (1) → v · ∇ − v · v × (∇ × v ) = −v · ∇ + gy
2    ρ
 ⊥v 
0
v · momentum (1) → energy

Therefore,    
v2 p D v2 p
v · ∇ 2
+ ρ
+ gy = 0 = Dt 2
+ ρ
+ gy
streamline pathline
v2 p
i.e., 2
+ ρ + gy = constant on a streamline
v2 p
In general, 2
+ ρ
+ gy = F (Ψ) where Ψ is a tag for a particular streamline.

Assumptions: Ideal fluid, Steady flow, Rotational in general.

3.4.1 Example: Contraction in Water or Wind Tunnel

Contraction Ratio: γ = R1 /R2 >> 1 ( γ = O(10) for wind tunnel ; γ = O(5) for water
tunnel)

Let U¯1 and U¯2 denote the average velocities at sections 1 and 2 respectively.
 2
2 2 U¯2 R1
1. From continuity: Ū1 πR1 = Ū2 πR2 → ¯ = = γ 2 >> 1
U1 R2
2.

∂u
Since 0,ω
=  = 0 → vortex ring.
∂r

10
 ⎧
⎪
⎪ ω1 ω2 ω2 R2 1
⎪
⎪ = → = ∼ << 1
⎨
2πR1 2πR2 ω1 R1 γ
ω/L = constant ⇒
   
⎪
⎪
⎪
⎪ since ω ∼
∂u
→
∂u ∂u
⎩
<<
∂r ∂r 2 ∂r 1

i.e.,

Section 2
Section 1

3. Near the center, let U1 = U¯1 (1 + ε1 ) and U2 = U¯2 (1 + ε2 ) where ε1 and ε2 measure
the relative velocity fluctuations. Apply the Bernoulli equation along a reference
average streamline

P1 + 12 ρŪ12 = P2 + 12 ρŪ22 (2)

Apply Bernoulli Equation to a particular streamline

 2 
2
P1 + 12 ρ U¯1 (1 + ε1 ) = P2 + 12 ρ
Ū2 (1 + ε2 )

From (2) and (3) we obtain

ε2 Ū 2 1
ε1 Ū12 = ε2 Ū22 + O(ε2 ) → ∼ 12 ∼ 4 << 1
ε1 Ū2 γ

11

 2.20 - Marine Hydrodynamics, Spring 2005

Lecture 9

2.20 - Marine Hydrodynamics

Lecture 9
Lecture 9 is structured as follows: In paragraph 3.5 we return to the full Navier-Stokes
equations (unsteady, viscous momentum equations) to deduce the vorticity equation and
study some additional properties of vorticity. In paragraph 3.6 we introduce the concept of
potential flow and velocity potential. We formulate the governing equations and boundary
conditions for potential flow and finally introduce the stream function.

3.5 Vorticity Equation

Return to viscous incompressible flow. The Navier-Stokes equations in vector form
 
∂v p
+ v · ∇v = −∇ + gy + ν∇2v
∂t ρ
By taking the curl of the Navier-Stokes equations we obtain the vorticity equation. In
detail and taking into account ∇ × u ≡ ω  we have

 
∂v p  
∇ × (Navier-Stokes) → ∇ × + ∇ × (v · ∇v ) = −∇ × ∇ + gy + ∇ × ν∇2v
∂t ρ
The first term on the left side, for fixed reference frames, becomes
∂v ∂ ∂ω
∇× = (∇ × v ) =
∂t ∂t ∂t
In the same manner the last term on the right side becomes
 
∇ × ν∇2v = ν∇2 ω 
Applying the identity ∇ × ∇ · scalar = 0 the pressure term vanishes, provided that the
density is uniform  
p
∇ × ∇( + gy) = 0
ρ

1
The inertia term v · ∇v , as shown in Lecture 8, §3.4, can be rewritten as
 2
1 v
v · ∇v = ∇ (v · v ) − v × (∇ × v ) = ∇  where v 2 ≡ |v |2 = v · v
− v × ω
2 2

and then the second term on the left side can be rewritten as
 2
v
∇ × (v · ∇) v = ∇ × ∇ − ∇ × (v × ω ) = ∇ × (ω × v )
2
= (v · ∇) ω − (ω · ∇) v + ω (∇ · v ) + v (∇ · ω )
   
=0 =0 since
incompressible ∇·(∇×v )=0
fluid

Putting everything together, we obtain the vorticity equation

Dω
 · ∇) v + ν∇2 ω
= (ω 
Dt

Comments-results obtained from the vorticity equation

• Kelvin’s Theorem revisited - from vorticity equation:

D
ω
If ν ≡ 0, then Dt
= (ω · ∇) v , so if ω
 ≡ 0 everywhere at one time, ω
 ≡ 0 always.

• ν can be thought of as diffusivity of vorticity (and momentum), i.e., ω once generated

(on boundaries only) will spread/diffuse in space if ν is present.

v
v ω
ω
v v
Dv v Dω
D ω v
= υ∇ 2v + ... = υ∇ 2ω + ...
Dt Dt

 ∂T
• Diffusion of vorticity is analogous to the heat equation: = K∇2 T , where K is the
∂t
heat diffusivity.

2
Numerical example  ν ∼ 1 mm /s. For diffusion time t = 1 second, diffusion
√ for
distance L ∼ O νt ∼ O (mm). For diffusion distance L = 1cm, the necessary
diffusion time is t ∼ O (L2 /ν) ∼ O(10)sec.

• In 2D space (x, y),

∂
v = (u, v, 0) and ≡0
∂z

 = ∇ × v is ⊥ to v (ω
So, ω  is parallel to the z-axis). Then,

⎛ ⎞
⎜ ∂ ∂ ∂ ⎟
(ω · ∇) v = ⎝ ωx + ωy + ωz ⎠ v ≡ 0,
 ∂x  ∂y ∂z
0 0

0

so in 2D we have

Dω
= ν∇2 ω

Dt
If ν = 0, D
ω
Dt
= 0, i.e., in 2D following a particle the angular velocity is conserved.
Reason: In 2D space the length of a vortex tube cannot change due to continuity.

• In 3D space,

Dωi ∂vi ∂ 2 ωi
= ωj + ν
Dt ∂x ∂xj ∂xj
  j  
vortex turning and stretching diffusion

for example,

Dω2 ∂u2 ∂u2 ∂u2

= ω1 + ω2 + ω3 + diffusion
Dt ∂x1 ∂x ∂x3
    2  
vortex turning vortex stretching vortex turning

z ≡ x3 z ≡ x3
∂u2
dz > 0
dz ∂x3
dy
u2 = 0 y ≡ x2 u2 = 0 y ≡ x2
∂u2
dy > 0
∂x2
x ≡ x1 x ≡ x1

∂u2 Dω2 ∂u2 ω3 >0 Dω2

>0 ⇒ >0 > 0 ⎯⎯ ⎯→ >0
∂x2 1Dt
424 3 ∂x3 1Dt
424 3
vortex stretching rate vortex turning rate

3.5.1 Example: Pile on a River

Scouring

What really happens as length of the vortex tube L increases?

IFCF is no longer a valid assumption.

Why?
Ideal flow assumption implies that the inertia forces are much larger than the viscous
effects. The Reynolds number, with respect to the vortex tube diameter D is given by
UD
Re ∼
ν
As the vortex tube length increases ⇒ the diameter D becomes really small ⇒ Re is not
that big after all.
Therefore IFCF is no longer valid.

3.6 Potential Flow

Potential Flow (P-Flow) is an ideal and irrotational fluid flow

⎧ ⎫
⎪
⎪ Inviscid Fluid ν=0 ⎬
⎪
⎪
⎨ + Ideal Flow
⎭
P-Flow ≡ Incompressible Flow ∇ · v = 0
⎪
⎪
⎪
⎪ +
⎩
Irrotational Flow ω = 0 or Γ = 0

3.6.1 Velocity potential

For ideal flow under conservative body forces by Kelvin’s theorem if ω ≡ 0 at some
time t, then ω ≡ 0 ≡ irrotational flow always. In this case the flow is P-Flow.

Given a vector field v for which ω  = ∇ × v ≡ 0, there exists a potential function

(scalar) - the velocity potential - denoted as φ, for which

v = ∇φ

Note that
 = ∇ × v = ∇ × ∇φ ≡ 0
ω
for any φ, so irrotational flow guaranteed automatically. At a point x and time t,
the velocity vector v (x, t) in cartesian coordinates in terms of the potential function
φ(x, t) is given by
 
∂φ ∂φ ∂φ
v (x, t) = ∇φ (x, t) = , ,
∂x ∂y ∂z

 φ (x)

u u

∂φ u=0 ∂φ
>0 <0
∂x ∂x
u >0 u <0

from low φ ⎯
⎯→ to high φ

The velocity vector v is the gradient of the potential function φ, so it always points
towards higher values of the potential function.

3.6.2 Governing Equations and Boundary Conditions for Potential Flow

(a) Continuity

∇ · v = 0 = ∇ · ∇φ ⇒ ∇2 φ = 0

Number of unknowns → φ

Number of equations → ∇2 φ = 0

Therefore we have closure. In addition, the velocity potential φ and the pressure p
are decoupled. The velocity potential φ can be solved independently first, and after
φ is obtained we can evaluate the pressure p.

p = f (v ) = f (∇φ) → Solve for φ, then find pressure.

7
(b) Bernoulli equation for P-Flow
This is a scalar equation for the pressure under the assumption of P-Flow for
steady or unsteady flow.

Euler equation:
   
∂v v2 p
+∇ − v × ω = −∇ + gy
∂t 2 ρ

Substituting v = ∇φ and ω
 = 0 into Euler’s equation above, we obtain
     
∂φ 1 p
∇ +∇ |∇φ|2 = −∇ + gy
∂t 2 ρ
or
 
∂φ 1 p
∇ + |∇φ|2 + + gy = 0,
∂t 2 ρ
which implies that

∂φ 1 p
+ |∇φ|2 + + gy = f (t)
∂t 2 ρ
everywhere in the fluid for unsteady, potential flow. The equation above can be
written as
 
∂φ 1 2
p = −ρ + |∇φ| + gy + F (t)
∂t 2
which is the Bernoulli equation for unsteady or steady potential flow.

DO NOT CONFUSE WITH

BERNOULLI EQUATION FROM § 3.4,

USED FOR STEADY, ROTATIONAL FLOW

Summary: Bernoulli equationS for ideal flow.

(a) For steady rotational or irrotational flow along streamline:
 
1 2
p = −ρ v + gy + C(ψ)
2
(b) For unsteady or steady irrotational flow everywhere in the fluid:
 
∂φ 1 2
p = −ρ + |∇φ| + gy + F (t)
∂t 2
∂
(c) For hydrostatics, v ≡ 0, ∂t = 0:

p = −ρgy + c ← hydrostatic pressure (Archimedes’ principle)

∂
(d) Steady and no gravity effect ( ∂t = 0, g ≡ 0):

ρv 2 ρ
p=− + c = − |∇φ|2 + c ← Venturi pressure (created by velocity)
2 2
(e) Inertial, acceleration effect:

Eulerian inertia
 
∂φ
p ∼ − ρ +···
∂t
∂
∇p ∼ − ρ v + ···
∂t

∂p
p p+ δx
∂x
δx

9
(c) Boundary Conditions

• KBC on an impervious boundary

∂φ
v·n

ˆ= u·n
ˆ no flux across boundary ⇒ = Un given
∂n
n̂·∇φ Un given

• DBC: specify pressure at the boundary, i.e.,

 
∂φ 1
−ρ + |∇φ|2 + gy = given
∂t 2

Note: On a free-surface p = patm .

10

3.6.3 Stream function

• Continuity: ∇ · v = 0; Irrotationality: ∇ × v = ω = 0
• Velocity potential: v = ∇φ, then ∇ × v = ∇ × (∇φ) ≡ 0 for any φ, i.e.,
irrotationality is satisfied automatically. Required for continuity:

∇ · v = ∇2 φ = 0
 defined by

• Stream function ψ

v = ∇ × ψ 




 ≡ 0 for any ψ
Then ∇ · v = ∇ · ∇ × ψ  , i.e., satisfies continuity automatically.
Required for irrotationality:
   
 =∇ ∇·ψ
∇ × v = 0 ⇒ ∇ × ∇ × ψ  − ∇2 ψ=0 (1)



still 3 unknown
 =(ψx ,ψy ,ψz )
ψ

• For 2D and axisymmetric flows, ψ is a scalar ψ (stream functions are more ‘use­
ful’ for 2D and axisymmetric flows).

∂
For 2D flow: v = (u, v, 0) and ∂z
≡ 0.

 
 î       
 ∂ ĵ∂ k̂∂  ∂ ∂ ∂ ∂


 
v = ∇ × ψ =
 ∂x ∂y ∂z  =
ψz î +
−
ψz ĵ +
ψy −
ψx k̂


ψ ψ ψ 
∂y
∂x
∂x
∂y

x y z

∂ψ
Set ψx = ψy ≡ 0 and ψz = ψ, then u = ∂y
; v = − ∂ψ
∂x

So, for 2D:

 = ∂ ψx + ∂ ψy + ∂ ψz ≡ 0
∇·ψ
∂x ∂y ∂z
Then, from the irrotationality (see (1)) ⇒ ∇2 ψ = 0 and ψ satisfies Laplace’s
equation.

11
 ∂
• 2D polar coordinates: v = (vr , vθ ) and ∂z
≡ 0.

y
ê êr
r

  vr vθ
 
v
z
 êr rêθ êz       

1
 ∂ ∂  1 ∂ψ ∂ψ 1 ∂ ∂

=

v = ∇ × ψ ∂ 
=

z
êr −

z
êθ + rψθ − ψr êz
r

∂r ∂θ ∂z 
r ∂θ
∂r
r ∂r
∂θ

ψr ψθ ψz

Again let
ψr = ψθ ≡ 0 and ψz = ψ , then
1 ∂ψ ∂ψ
vr = and vθ = −
r ∂θ ∂r

 also reduces to ψ (read JNN 4.6 for details).

• For 3D but axisymmetric flows, ψ

12

• Physical Meaning of ψ.
In 2D

∂ψ ∂ψ
u= and v = −
∂y ∂x
We define
 
x  
x
ψ(x, t) = ψ(x0 , t) + v · nd

ˆ = ψ(x0 , t) + (udy − vdx)


x0 
x0
 
total volume flux
from left to right
accross a curve C
between x and x0

v
x

C’
v
t

v
xo C n̂

For ψ to be single-valued, must be path independent.

    
= or − = 0 −→ v · n
ˆ d
= ∇· v ds = 0
C C� C C� C−C � S =0, continuity

Therefore, ψ is unique because of continuity.

13

Let x1 , x2 be two points on a given streamline (v · n̂ = 0 on streamline)

streamline

x2
ψ (x2 ) = ψ (x1 ) + v · n̂ d


   
ψ2 ψ1 
x1 =0
along a
streamline

Therefore, ψ1 = ψ2 , i.e., ψ is a constant along any streamline. For example, on

an impervious stationary body v · n̂ = 0, so ψ = constant on the body is the
appropriate boundary condition. If the body is moving v · n̂ = Un

ψ = ψ0 + Un d
on the boddy

given

∂φ
ψ = constant ≡ =0
u=0 ∂n

ψ = given

ψo

14

 Flux Δψ = −vΔx = uΔy.

∂ψ ∂ψ

Therefore, u = and v = −
∂y ∂x

(x, y + Δy)

u
streamline
-v
streamline (x +Δx, y)
(x,y)

ψ + Δψ
ψ

15

Summary of velocity potential formulation vs. stream-function formulation for ideal flows

⎧ ⎫
⎨
For irrotational flow use φ ⎬

For incompressible flow use ψ

⎩
⎭

For P-Flow use φ or ψ

velocity potential stream-function

definition v = ∇φ v = ∇ × ψ
continuity ∇ · v = 0 ∇2 φ = 0 automatically
  satisfied
 
irrotationality ∇ × v = 0 automatically satisfied ∇× ∇×ψ  =∇ ∇·ψ  − ∇2 ψ
=0

∂
2D: w = 0, ∂z =0
continuity ∇2 φ = 0 automatically satisfied
irrotationality automatically satisfied ψ ≡ ψz : ∇2 ψ = 0

Cauchy-Riemann equations for (φ, ψ) = (real, imaginary) part of an analytic complex

function of z = x + iy

∂φ ∂ψ
u= ∂x
u= ∂y
Cartesian (x, y)
∂φ
v= ∂y v = − ∂ψ
∂x

∂φ 1 ∂ψ
vr = ∂r
vr = r ∂θ
Polar (r,θ)
1 ∂φ
vθ = r ∂θ
vθ = − ∂ψ
∂r

Given φ or ψ for 2D flow, use Cauchy-Riemann equations to find the other:

e.g. If φ = xy, then ψ = ?

⎫
∂φ ∂ψ 1 ⎪
u= =y= → ψ = y 2 + f1 (x) ⎪
⎪
∂x ∂y 2 ⎬ 1
⇒ ψ = (y 2 − x2 ) + const
⎪
⎪ 2
→ ψ = − x2 + f2 (y) ⎪
∂φ ∂ψ 1
v= =x=− ⎭

∂y ∂x 2

16
 2.20 - Marine Hydrodynamics, Spring 2005
Lecture 10

2.20 - Marine Hydrodynamics

Lecture 10

3.7 Governing Equations and Boundary Conditions for P-Flow

3.7.1 Governing Equations for P-Flow

(a) Continuity �2 φ = 0
� �
1 2
(b) Bernoulli for P-Flow (steady or unsteady) p = −ρ φt + |�φ| + gy + C(t)
2

3.7.2 Boundary Conditions for P-Flow

Types of Boundary Conditions:

∂φ
(c) Kinematic Boundary Conditions - specify the flow velocity �v at boundaries. = Un
∂n
(d) Dynamic Boundary Conditions - specify force F� or pressure p at flow boundary.
� �
1 2
p = −ρ φt + (�φ) + gy + C (t) (prescribed)
2

The boundary conditions in more detail:

• Kinematic Boundary Condition on an impermeable boundary (no flux condition)

�
�v ·n̂ = ����
U ·n̂ = Un = Given
���� ����
fluid velocity boundary velocity nornal boundary velocity
�v =�φ

�φ · n̂ = Un ⇒

∂ ∂ ∂
(n1 + n2 + n3 )φ = Un ⇒
∂x1 ∂x2 ∂x3

∂φ
= Un
∂n

v
U

n = (n1 , n 2 , n 3 )
v
v
( )

• Dynamic Boundary Condition: In general, pressure is prescribed

� �
1 2
p = −ρ φt + (�φ) + gy + C (t) = Given
2

3.7.3 Summary: Boundary Value Problem for P-Flow

The aforementioned governing equations with the boundary conditions formulate the

Boundary Value Problem (BVP) for P-Flow.

The general BVP for P-Flow is sketched in the following figure.

KBC : (Lecture 19)


Free surface DBC : − ρ (φ + 1 (∇φ ) 2 + gy ) + C (t ) = GIVEN
 t
21 23
 non − linear

∇ 2φ = 0

p = − ρ (φt + (∇φ ) + gy ) + C (t )
1 2

∂φ
Solid boundary KBC : = U n = GIVEN
∂n

It must be pointed out that this BVP is satisfied instantaneously.

3.8 Linear Superposition for Potential Flow

In the absence of dynamic boundary conditions, the potential flow boundary value
problem is linear.

• Potential function φ.

∇ 2 φ = 0 in V

∂φ
= U n =f on B
∂n

• Stream function ψ.

∇2 ψ = 0 in V

ψ=g on B

Linear
� Superposition: if φ1 , φ2 , . . . are harmonic functions, i.e., �2 φi = 0, then φ =
αi φi , where αi are constants, are also harmonic, and is the solution for the boundary
value problem provided the kinematic boundary conditions are satisfied, i.e.,
∂φ ∂
= (α1 φ1 + α2 φ2 + . . .) = Un on B.
∂n ∂n
The key is to combine known solution of the Laplace equation in such a way as to satisfy

the kinematic boundary conditions (KBC).

The same is true for the stream function ψ. The K.B.C specify the value of ψ on the

boundaries.

3.8.1 Example

��� �
Let φi x denote a unit-source flow with source at xi , i.e.,

��� �� � � 1
�� � �

φi x ≡ φsource x,
xi
=
ln �x −
xi
� (in 2D)

2π

� �� � ��−1
= − 4π �x − xi � (in 3D),

then find mi such that

� �
φ= mi φi (x) satisfies KBC on B
i

Caution: φ must be regular for x ∈ V , so it is required that �x ∈

/ V.

v
•x 2

v
•x 1
∇ 2 φ = 0 in V

v
•x 3 •x 4
v

∂Φ
=f
∂n

Figure 1: Note: �xj , j = 1, . . . , 4 are not in the fluid domain V .

3.9 - Laplace equation in different coordinate systems (cf Hildebrand §6.18)

3.9.1 Cartesian (x,y,z)
� � � �
î ˆ
j ˆ
k ∂φ ∂φ ∂φ
�v = u, v, w = �φ = , ,
∂x ∂y ∂z

∂ 2φ ∂ 2φ ∂ 2φ
�2 φ = + + 2
∂x2 ∂y 2 ∂z

êz
z

P ( x, y , z )

O y ê y

x
êx

3.9.2 Cylindrical (r,θ,z)

r 2 = x2 + y 2 ,
θ = tan−1 (y/x)

� ê ê ê � � ∂φ 1 ∂φ ∂φ �
� r θ z
v = vr , vθ , vz = , ,
∂r r ∂θ ∂z

∂ 2 φ 1 ∂φ 1 ∂ 2 φ ∂ 2 φ
�2 φ = 2
+ + 2 2 + 2 ⇔
�∂r �� r ∂r� r ∂θ ∂z
r ∂r ( ∂r )
1 ∂φ
r ∂φ
� �
2 1 ∂φ ∂φ 1 ∂2φ ∂ 2φ
�φ = r + 2 2 + 2
r ∂r ∂r r ∂θ ∂z

êz
z

P (r , θ , z )

O y ê y
θ r
x
êx

3.9.3 Spherical (r,θ,ϕ)

r 2 = x2 + y 2 + z 2 ,
θ = cos−1 (z/r) ⇔ z = r (cos θ)
ϕ = tan−1 (y/x)

� � � �
� êr êθ êϕ ∂φ 1 ∂φ 1 ∂φ
v = �φ = vr , vθ , vϕ = , ,
∂r r ∂θ r(sin θ) ∂ϕ

� �
2 ∂ 2 φ 2 ∂φ 1 ∂ ∂φ 1 ∂ 2φ
�φ = + + sin θ + ⇔
2 2
�∂r �� r ∂r� r sin θ ∂θ ∂θ r2 sin2 θ ∂ϕ2
1 ∂
r 2 ∂r
(r2 ∂φ
∂r )
� � � �
2 1 ∂ 2 ∂φ 1 ∂ ∂φ 1 ∂ 2φ
�φ = 2 r + 2 sin θ + 2 2
r ∂r ∂r r sin θ ∂θ ∂θ r sin θ ∂ϕ2

êz
z

P (r , θ , φ )

θ
r

O y ê y
φ
x
êx

3.10 Simple Potential flows

1. Uniform Stream �2 (ax + by + cz + d) = 0

�
1D: φ = U x + constant ψ = U y +
constant;
v = (U, 0, 0)
�
2D: φ = U x + V y + constant ψ = U y − V x +
constant;
v = (U, V, 0)
�
3D: φ = U x + V y + W z +
constant
v = (U, V, W )

2. Source (sink) flow

2D, Polar coordinates

� � �
2 1 ∂ ∂ 1 ∂2
� = r + 2 2 , with r = x2 + y 2
r ∂r ∂r r ∂θ
An axisymmetric
� solution: φ = a ln r + b. Verify that it satisfies �2 φ = 0, except at
r = x2 + y 2 = 0. Therefor, r = 0 must be excluded from the flow.

Define 2D source of strength m at r= 0:

m
φ= ln r
2π

∂φ m m
�φ = êr = êr ⇐⇒ vr = , vθ = 0
∂r 2πr 2πr

source
y
(strength m)

Net outward volume flux is

� �
�� �
�� �
v · n̂ds = � · vds = � · vds
C S Sε
� �
�2π
v · n̂ds = vr rε dθ = m
����
Cε 0
����
m
2πrε source
strength
y
n̂ C
Sε
x
ε

If m < 0 ⇒ sink. Source m at (x0 , y0 ):

�
m
φ = ln (x − x0 )2 + (y − y0 )2
2π
m m
φ = ln r (Potential function) ←→ ψ = θ (Stream function)
2π 2π

m
Ψ= θ
y 2π
θ
m
Vr =
2π
x
1
ψ=0

10

3D: Spherical coordinates

� � � � �
2 1 ∂ 2 ∂ ∂ ∂
� = 2 r + , , · · · , where r = x2 + y 2 + z 2
r ∂r ∂r ∂θ ∂ϕ
a
A spherically symmetric solution: φ = + b. Verify �2 φ = 0 except at r = 0.
r

Define a 3D source of strength m at r = 0. Then

m ∂φ m
φ=− ⇐⇒ vr = = , vθ = 0, vϕ = 0
4πr ∂r 4πr2
Net outward volume flux is
��
m
� vr dS = 4πrε2 · = m (m < 0 for a sink )
4πrε2

11

3. 2D point vortex
� �
2 1 ∂ ∂ 1 ∂2
� = r + 2 2
r ∂r ∂r r ∂θ
Another particular solution: φ = aθ + b. Verify that �2 φ = 0 except at r = 0.

Define the potential for a point vortex of circulation Γ at r = 0. Then

Γ ∂φ 1 ∂φ Γ
φ= θ ⇐⇒ vr = = 0, vθ = = and,
2π ∂r r ∂θ 2πr
1 ∂
ωz = (rvθ ) = 0 except at r = 0
r ∂r
Stream function:

Γ
ψ=− ln r
2π
Circulation:

� � � �2π
� � � � � � Γ
v · dx = v · dx + v · dx = rdθ = Γ
����
2πr
C1 C2 C1 −C2
� R R �� �
0 vortex
ωz dS=0 strength
S

12

4. Dipole (doublet flow)

A dipole is a superposition of a sink and a source with the same strength.

2D dipole:

� � � �
m 2 2 2 2
φ= ln (x − a) + y − ln (x + a) + y
2π
� �
µ ∂ �
lim φ = ln (x − ξ) + y ��
2 2
a→0 2π
���� ∂ξ ξ=0
µ = 2ma
constant
µ x µ x
=− = −
2π x2 + y 2 2π r2

2D dipole (doublet) of moment µ at the origin oriented in the +x direction.

∂
NOTE: dipole = µ ∂ξ (unit source)

13

unit
source α
x

−µ x cos α + y sin α −µ cos θ cos α + sin θ sin α

φ= 2 2
=
2π x +y 2π r
3D dipole:

⎛ ⎞
m
⎝
1 1
⎠
where µ = 2ma fixed

φ = lim
−
� −
�
a→0 4π
2 2 2 2 2 2
(x − a)
+ y + z (x + a)
+ y + z
�

�
µ ∂ 1
� µ x µ x

= −
� � = − = −
4π ∂ξ
� 4π (x2 + y 2 + z 2 )3/2 4π r3
(x
− ξ)2 + y 2 + z 2 �
ξ=0

3D dipole (doublet) of moment µ at the origin oriented in the +x direction.

14

5. Stream and source: Rankine half-body

It is the superposition of a uniform stream of constant speed U and a source of

strength m.

U
m

m � 2
2D: φ = U x + ln x + y 2
2π

U
m
x D

v
stagnation point v = 0

Dividing
Streamline

∂φ m x
u= =U+
∂x 2π x + y 2
2
m
u|y=0 = U + , v |y=0 = 0 ⇒
2πx
m
V� = (u, v) = 0 at x = xs = − , y=0
2πU

m
For large x, u → U , and U D = m by continuity ⇒ D = .
U

15

 m
3D: φ = U x − �
4π x
2 + y 2 + z 2

stagnation point

div. streamlines

∂φ m x
u= =U+
∂x 4π (x + y + z 2 )3/2
2 2
m x
u|y=z=0 = U + , v |y=z=0 = 0, w|y=z=0 = 0 ⇒
4π |x|3
�
� m
V = (u, v, w) = 0 at x = xs = − , y=z=0
4πU

m
For large x, u → U and U A = m by continuity ⇒ A = .
U

16

6. Stream + source/sink pair: Rankine closed bodies

U S +m -m S x
a

dividing streamline
(see this with PFLOW)

�
To have a closed body, a necessary condition is to have min body = 0

2D Rankine ovoid:

� � � � � �
m 2 2 2 2
m (x + a)2 + y 2
φ = U x+ ln (x + a) + y − ln (x − a) + y = U x+ ln
2π 4π (x − a)2 + y 2

3D Rankine ovoid:
⎡ ⎤
m⎣ 1 1
⎦

φ = Ux − � −�
4π 2 2
(x + a) + y2 + z2 (x − a) + y2 + z2

17

For Rankine Ovoid,

� �
∂φ m x+a x−a
u= =U+ −
∂x 4π �(x + a)2 + y 2 + z 2 �3/2 �(x − a)2 + y 2 + z 2 �3/2
� �
m 1 1
u|y=z=0 =U + −
4π (x + a)2 (x − a)
2
m (−4ax)

=U +
4π (x2 − a2 )2

� �2 � m
�
u|y=z=0 =0 at x2 − a2 = 4ax
4πU

At x = 0,

m 2a
u=U+ where R = y 2 + z 2
4π (a2 + R2 )3/2

Determine radius of body R0 :

�R0
2π uRdR = m
0

18

7. Stream + Dipole: circles and spheres

r
U µ θ

µx � µ �
2D: φ = U x + = cos θ U r +
2πr2 ↑ 2πr
x=r cos θ
The radial velocity is then
∂φ � µ �
ur = = cos θ U − .
∂r 2πr2
� µ
Setting the radial velocity vr = 0 on r = a we obtain a = 2πU
. This is the K.B.C.
for a stationary circle of radius a. Therefore, for

µ = 2πU a2

the potential � µ �
φ = cos θ U r +
2πr
is the solution to ideal flow past a circle of radius a.
• Flow past a circle (U, a).

19

 � �
a2
φ = U cos θ r + r
� 2
�
1 ∂φ
Vθ = r ∂θ
= −U sin θ 1 + ar2
�
= 0 at θ = 0, π − stagnation points
Vθ |r=a = −2U sin θ π 3π
= �2U at θ = 2 , 2 − maximum tangential velocity

2U

2U

Illustration of the points where the flow reaches maximum speed around the circle.

µ cos θ � µ �
3D: φ = U x + = U r cos θ 1 +
4π r2 4πr3

y
µ
r
U θ
x
z

The radial velocity is then

∂φ � µ �
vr = = cos θ U −
∂r 2πr3

20
 � µ
Setting the radial velocity vr = 0 on r = a we obtain a = 3
2πU
. This is the K.B.C.
for a stationary sphere of radius a. Therefore, choosing

µ = 2πU a3

the potential � µ �
φ = cos θ U r +
2πr
is the solution to ideal flow past a sphere of radius a.
• Flow past a sphere (U, a).

� �
a3
φ = Ur cos θ 1 + 3
2r
� �
1 ∂φ a3
vθ = = −U sin θ 1 + 3
r ∂θ 2r
�
3U = 0 at θ = 0, π
vθ |r=a = − sin θ
2 = − 3U
2
at θ = π2

3/ U
2

θ
x

3/ U
2

21

8. 2D corner flow Velocity potential φ = rα cos αθ; Stream function ψ = rα sin αθ

� 2 �
∂ 1 ∂ 1 ∂2
(a) �2 φ = ∂r 2 + r ∂r
+ r2 ∂θ2
φ=0
(b)

∂φ
ur = = αrα−1 cos αθ
∂r

1 ∂φ

uθ = = −αrα−1 sin αθ
r ∂θ
∴ uθ = 0 { or ψ = 0} on αθ = nπ, n = 0, ±1, ±2, . . .

i.e., on θ = θ0 = 0, απ , 2π
α
, . . . (θ0 ≤ 2π)

i. Interior corner flow – stagnation point origin: α > 1. For example,

α = 1, θ0 = 0, π, 2π, u = 1, v = 0

ψ=0

22

 π 3π
α = 2, θ0 = 0, , π, ,2π u = 2x, v = −2 y
o 2 2
(90 corner)

ψ=0

ψ=0

θ=2π/3, ψ = 0

2π 4π θ=0, ψ = 0
α = 3 2 , θ0 = 0, , ,2π 120o 120o
3 3 θ=2π, ψ = 0
o
120
(120o corner)

θ=4π/3, ψ = 0

23

ii. Exterior corner flow, |v| → ∞ at origin:

α<1
π
θ0 = 0, only
α
π
Since we need θ0 ≤ 2π, we therefore require α
≤ 2π, i.e., α ≥ 1/2 only.

1/2 ≤ α < 1
π
θ0 = 0,
α
For example,

α = 1/2, θ0 = 0, 2π (1/2 infinite plate, flow around a tip)

θ=0, ψ = 0

θ=2π, ψ = 0

α = 2/3, θ0 = 0, 3π
2
(90o exterior corner)

θ=0, ψ = 0

θ=3π/2, ψ = 0

24

Appendix A1: Summary of Simple Potential Flows

Cartesian Coordinate System

Flow Streamlines Potential Stream function
φ(x, y, z) ψ(x, y)

Uniform flow U∞ x + V∞ y + W∞ z U∞ y − V∞ x

y−yo
2D Source/Sink (m) at (xo , yo ) m
2π
ln((x − xo )2 + (y − yo )2 ) m
2π
arctan( x−x )
o

3D Source/Sink (m) at (xo , yo , zo ) m q

− 4π 1 NA
(x−xo )2 +(y−yo )2 +(z−zo )2

Γ y−yo Γ ln((x − x )2 + (y − y )2 )
Vortex (Γ) at (xo , yo ) 2π
arctan( x−x ) − 2π o o
o

µ (x−xo ) cos α+(y−yo ) sin α µ (y−yo ) cos α+(x−xo ) sin α

2D Dipole (µ) at (xo , yo ) at an angle α − 2π
(x−xo )2 +(y−yo )2 2π (x−xo )2 +(y−yo )2
α

µ (x−xo )
3D Dipole (+x) (µ) at (xo , y0 , zo ) − 4π NA
((x−xo )2 +(y−yo )2 +(z−zo )2 )3/2

25

Appendix A2: Summary of Simple Potential Flows

Cylindrical Coordinate System

Flow Streamlines Potential Stream function
φ(r, θ, z) ψ(r, θ)

Uniform flow U∞ r cos θ + V∞ r sin θ + W∞ z U∞ r sin θ − V∞ r cos θ

2D Source/Sink (m) at (xo , yo ) m ln r m θ

2π 2π

3D Source/Sink (m) at (xo , yo , zo ) m

− 4πr NA

Vortex (Γ) at (xo , yo ) Γ θ Γ ln r

− 2π
2π

µ cos θ cos α+sin θ sin α µ sin θ cos α+cos θ sin α

2D Dipole (µ) at (xo , yo ) at an angle α − 2π r 2π r
α

µ cos θ
3D Dipole (+x) (µ) at (xo , yo , zo ) − 4π 2 NA
r

26

Appendix A3: Combination of Simple Potential Flows

m m m
Stream + Source (2D) φ = U∞ x + 2π ln r xs = − 2πU ∞
D= U∞

=
�
m √ 1 m m
Rankine Half Body (3D) φ = U∞ x − 4π xs = − 4πU∞ A= U∞
x2 +y 2 +z 2

m
� �
Stream + Source + Sink (2D) φ = U∞ x + 2π ln((x + a)2 + y 2 ) − ln((x − a)2 + y 2 )

=
m √ 1 1
Rankine Closed Body (3D) φ = U∞ x + 4π ( (x+a)2 +y 2 +z 2 −√ )
(x−a)2 +y 2 +z 2

µx a2
Stream + Dipole (2D) φ = U∞ x + 2πr 2 if µ = 2πa2 U∞ φ = U∞ cos θ(r + r )

=
µ cos θ a3
Circle (Sphere) R = a (3D) φ = U∞ x + 4πr 2 if µ = 2πa3 U∞ φ = U∞ cos θ(r + 2r 2 )

2D Corner Flow (2D) φ = Crα cos(αθ) ψ = Crα sin(αθ) θ0 = 0, nπ

α

27

Appendix B: Far Field Behavior of Simple Potential Flows

Far field behavior

φ �v = �φ
r >> 1

1
(2D) ∼ ln r ∼
r
Source
1 1
(3D) ∼ ∼
r r2

1 1
(2D) ∼ ∼
r r2
Dipole
1 1
(3D) ∼ ∼
r2 r3

1
Vortex (2D) ∼1 ∼
r

28

 2.20 - Marine Hydrodynamics, Spring 2005

Lecture 11

2.20 - Marine Hydrodynamics

Lecture 11

3.11 - Method of Images

m  2
• Potential for single source: φ = ln x + y 2
2π

m
m

   
m 2 2
• Potential for source near a wall: φ = ln x2 + (y − b) + ln x2 + (y + b)
2π

m
m
y

dφ b
=0 x
dy

b
m

Added source for

symmetry

Note: Be sure to verify that the boundary conditions are satisfied by symmetry or
by calculus for φ (y) = φ (−y).

  
Γ −1 y − b −1 y + b
• Vortex near a wall (ground effect): φ = Ux + tan ( ) − tan ( )
2π x x

U b y

x
b

-Γ

Added vortex for

symmetry

dφ
Verify that = 0 on the wall y = 0.
dy
 
a2 a2
• Circle of radius a near a wall: φ ∼
= Ux 1 + +
x2 + (y − b)2 x2 + (y + b)2

y
U b
y
x

∂φ
This solution satisfies the boundary condition on the wall ( ∂n = 0), and the degree it
satisfies the boundary condition of no flux through the circle boundary increases as
the ratio b/a >> 1, i.e., the velocity due to the image dipole small on the real circle
for b >> a. For a 2D dipole, φ ∼ d1 , ∇φ ∼ d12 .

2
 • More than one wall:

Example 1:
b
b
U
U
b'
b'

b'

Example 2: Example 3:
b b
b b
-Γ Γ
b' b'
b' b'

b' b'
b' b'
Γ -Γ
b b b b

3.12 Forces on a body undergoing steady translation

“D’Alembert’s paradox”
3.12.1 Fixed bodies & translating bodies - Galilean transformation.
y y’

x’ U
x o’
o z’
z

Fixed in space Fixed in translating body

x = x` + Ut

Reference system O: v , φ, p Reference system O’: v  , φ , p

X S X’ S U
O O’

∇2 φ = 0 ∇2 φ = 0

v · n̂ = ∂φ  · n̂ = (U, 0, 0) · (nx , ny , nz )
=U
∂n  ∂φ
v · n̂ = ∂n
=0
= U nx on Body
 
v → 0 as |x| → ∞ v → (−U, 0, 0) as |x | → ∞
φ → 0 as |x| → ∞ φ → −U x as |x | → ∞
Galilean transform:
 
v(x, y, z, t) = v (x = x − U t, y, z, t) + (U, 0, 0)
φ(x, y, z, t) = φ (x = x − U t, y, z, t) + U x ⇒
−U x + φ(x = x + U t, y, z, t) = φ (x , y, z, t)
Pressure (no gravity)
p∞ = − 12 ρv 2 + Co = Co = − 12 ρv 2 + Co = Co − 12 ρU 2

∴ Co = Co − 12 ρU 2
In O: unsteady flow In O’: steady flow
∂φ 1
ps = −ρ ∂φ
∂t
v 2 +Co
− 12 ρ  ps = −ρ v 2 +Co = Co
− 2 ρ 
∂t

U2 0
0
∂ ∂x ∂
∂φ
∂t
=( + ∂x
) (φ + U x ) = −U 2
∂t ∂t
 
0 −U

∴ ps = ρU 2
− 12 ρU 2 + Co = 12 ρU 2 + Co

ps − p∞ = 12 ρU 2 stagnation pressure ps − p∞ = 12 ρU 2 stagnation pressure

3.12.2 Forces

n̂


Total fluid force for ideal flow (i.e., no shear stresses): F =  pˆ
ndS
B
For potential flow, substitute for p from Bernoulli:
⎛ ⎞

⎜ ∂φ 1 ⎟
 ⎜ ⎟
F =  −ρ ⎜ + |∇φ|2 + gy +c(t)⎟ ndS
ˆ
⎝∂t 
2  ⎠
B
hydrodynamic hydrostatic
force force
 
For the hydrostatic case v ≡ φ ≡ 0 :


F s =  (−ρgyn̂) dS = (−) ∇ (−ρgy) dυ = ρg∀ĵ where ∀ = dυ
↑ ↑  
Gauss outward Archimedes
B theorem normal
υB υB
principle

We evaluate only the hydrodynamic force:

 
 ∂φ 1 2
F d = −ρ + |∇φ| n̂dS
∂t 2
B
 ∂φ 
For steady motion ∂t
≡0 :

 1
Fd = − ρ v 2 ndS
ˆ
2
B

5
3.12.3 Example Hydrodynamic force on 2D cylinder in a steady uniform stream.

n̂
U S x
a

   −ρ  2π 
−ρ 

F d =
2
|∇φ|2 nd
ˆ = 2
2
|∇φ|
r=a n̂adθ
B 0
 2π
Fx =
F ·
î = −ρa
2
dθ |∇φ|
2r=a 
n̂
·
î

0
− cos θ
2π 2
=
ρa
2
|∇φ|
r=a cos θdθ

Velocity potential for flow past a 2D cylinder:




a2
φ = U r cos θ 1 + 2
r
Velocity vector on the 2D cylinder surface:


  

∂φ
 1 ∂φ

∇φ|r=a = ( vr |r=a , vθ |r=a ) =  ,

∂r


r ∂θ





r=a

0 r=a −2U sin θ

Square of the velocity vector on the 2D cylinder surface:


|∇φ|2 
r=a = 4U 2 sin2 θ

6
Finally, the hydrodynamic force on the 2D cylinder is given by

2π
2π
ρa   1 
Fx = dθ 4U 2 sin2 θ cos θ = ρU 2 (2a) 2 dθ sin 2
  θ cos
 θ = 0
2
0
2   0 even odd
π 3π
diameter
ps −p∞ w.r.t ,
or
projection  
2 2

≡0

Therefore, Fx = 0 ⇒ no horizontal force ( symmetry fore-aft of the streamlines). Similarly,

2π
1 
Fy = ρU 2 (2a)2 dθ sin2 θ sin θ = 0
2
0

In fact, in general we find that F ≡ 0, on any 2D or 3D body.

D’Alembert’s “paradox”:

No hydrodynamic force∗ acts on a body moving with steady translational (no circulation)
velocity in an infinite, inviscid, irrotational fluid.

∗
The moment as measured in a local frame is not necessarily zero.

3.13 Lift due to Circulation

3.13.1 Example Hydrodynamic force on a vortex in a uniform stream.
Γ Γ
φ = Ux + θ = U r cos θ + θ
2π 2π

U Γ

Consider a control surface in the form of a circle of radius r centered at the point vortex.
Then according to Newton’s law:
d  steady flow
ΣF = LCV −→
dt
(FV + FCS ) + M
 N ET = 0 ⇔ F ≡ −FV = FCS + M
 N ET

Where,
F = Hydrodynamic force exerted on the vortex from the fluid.
FV = −F = Hydrodynamic force exerted on the fluid in the control volume from the vortex.
FCS = Surface force (i.e., pressure) on the fluid control surface.
M N ET = Net linear momentum flux in the control volume through the control surface.
d 
L
dt CV
= Rate of change of the total linear momentum in the control volume.

Γ Fy θ
U
Fx
x

Control
volume

The hydrodynamic force on the vortex is F = FCS + M

 IN

8
a. Net linear momentum flux in the control volume through the control surfaces, M  N ET .
Recall that the control surface has the form of a circle of radius r centered at the point
vortex.

a.1 The velocity components on the control surface are

u = U− sin θ
2πr
Γ
v = cos θ
2πr

The radial velocity on the control surface is therefore, given by

∂x 
ur = U = U cos θ = V · n̂
∂r

Γ
vθ =
2πr
U
θ

a.2 The net horizontal and vertical momentum fluxes through the control surface are
given by

2π
2π  
Γ
(MN ET )x = − ρ dθruvr = − ρ dθr U − sin θ U cos θ = 0
2πr
0 0

2π
2π  
Γ
(M N ET )y = − ρ dθrvvr = − ρ dθr cos θ U cos θ
2πr
0 0

2π
ρU Γ ρU Γ
=− cos2 θdθ = −
2π 2
0

9
b. Pressure force on the control surface, FCS .

b.1 From Bernoulli, the pressure on the control surface is

1
p = − ρ |v |2 + C
2
b.2 The velocity |v |2 on the control surface is given by
 2  2
2 2 2 Γ Γ
|v | =u + v = U − sin θ + cos θ
2πr 2πr
 2
2 Γ Γ
=U − U sin θ +
πr 2πr

b.3 Integrate the pressure along the control surface to obtain FCS

2π

(FCS )x = dθrp(− cos θ) = 0

0

2π
2π   
(FCS )y = dθrp(− sin θ) = − 2ρ ΓU
− πr (−r) dθ sin2 θ = − 12 ρU Γ
0
0 
π

c. Finally, the force on the vortex F is given by

Fx = (FCS )x + (Mx )IN = 0

Fy = (FCS )y + (My )IN = −ρU Γ
i.e., the fluid exerts a downward force F = −ρU Γ on the vortex.
Kutta-Joukowski Law

2D : F = −ρU Γ
3D : F = ρU
 × Γ
Generalized Kutta-Joukowski Law:
 

n
F = ρU
 × Γi
i=1

where F is the total force on a system of n vortices in a free stream with speed U
.

10
 2.20 - Marine Hydrodynamics, Spring 2005
Lecture 12

2.20 - Marine Hydrodynamics

Lecture 12

3.14 Lifting Surfaces

3.14.1 2D Symmetric Streamlined Body

No separation, even for large Reynolds numbers.

stream line

• Viscous effects only in a thin boundary layer.

• Small Drag (only skin friction).
• No Lift.

3.14.2 Asymmetric Body

(a) Angle of attack α,

chord line

U α

(b) or camber η(x),

chord line mean camber line

(c) or both

amount of camber
chord line
mean camber line

U α

angle of attack

 and Drag  to U
Lift ⊥ to U 

3.15 Potential Flow and Kutta Condition

From the P-Flow solution for flow past a body we obtain

P-Flow solution, infinite velocity at trailing edge.

Note that (a) the solution is not unique - we can always superimpose a circulatory flow
without violating the boundary conditions, and (b) the velocity at the trailing edge →
∞. We must therefore, impose the Kutta condition, which states that the ‘flow leaves
tangentially the trailing edge, i.e., the velocity at the trailing edge is finite’.
To satisfy the Kutta condition we need to add circulation.

Circulatory flow only.

Superimposing the P-Flow solution plus circulatory flow, we obtain

Figure 1: P-Flow solution plus circulatory flow.

3.15.1 Why Kutta condition?

Consider a control volume as illustrated below. At t = 0, the foil is at rest (top control
volume). It starts moving impulsively with speed U (middle control volume). At t = 0+ ,
a starting vortex is created due to flow separation at the trailing edge. As the foil moves,
viscous effects streamline the flow at the trailing edge (no separation for later t), and the
starting vortex is left in the wake (bottom control volume).

t=0 Γ=0

ΓS
+ ΓS
t=0 U starting vortex
due to separation
(a real fluid effect,
no infinite vel of
potetial flow)

ΓS
for later
ΓS
U
t

no Γ

starting vortex left in wake

Kelvin’s theorem:
dΓ
= 0 → Γ = 0 for t ≥ 0 if Γ(t = 0) = 0
dt
After a while the ΓS in the wake is far behind and we recover Figure 1.

4
3.15.2 How much ΓS ?

Just enough so that the Kutta condition is satisfied, so that no separation occurs. For

example, consider a flat plate of chord  and angle of attack α, as shown in the figure

below.

chord length

Simple P-Flow solution

Γ = πlU sin α
L = ρU Γ = ρU 2 πl sin α
|
|L
CL = 1 2 = 2π sin α ≈ 2πα for small α
ρU l � �� �
2 only for
small α

However, notice that as α increases, separation occurs close to the leading edge.

Excessive angle of attack leads to separation at the leading edge.

When the angle of attack exceeds a certain value (depends on the wing geometry) stall
occurs. The effects of stalling on the lift coefficient (CL = 1 ρU 2 Lspan ) are shown in the
2
following figure.

 C
L

This region independent of R,

ν used only to get Kutta

condition

stall location f(R)

stall
2π

α
O(5 o )

• In experiments, CL < 2πα for 3D foil - finite aspect ratio (finite span).

• With sharp leading edge, separation/stall to early.

sharp trailing edge

round leading edge to forstall to develop circulation
stalling

3.16 Thin Wing, Small Angle of Attack

• Assumptions

– Flow: Steady, P-Flow.

– Wing: Let yU (x), yL (x) denote the upper and lower vertical camber coordinates,
respectively. Also, let x = /2, x = −/2 denote the horizontal coordinates of
the leading and trailing edge, respectively, as shown in the figure below.

y=yU(x)

For thin wing, at a small angle of attack it is

yU yL
, << 1
 
dyU dyL
, << 1
dx dx
The problem is then linear and superposition applies.
Let η(x) denote the camber line
1 t(x)
η(x) = (yU (x) + yL (x)),
2
and t(x) denote the half-thickness
t(x)
Camber line η(x)
1
t(x) = (yU (x) − yL (x)).
2

For linearized theory, i.e. thin wing at small AoA, the lift on the wing depends
only on the camber line but not on the wing thickness. Therefore, for the
following analysis we approximate the wing by the camber line only and ignore
the wing thickness.

7
• Definitions

In general, the lift on the wing is due to the total circulation Γ around the wing.
This total circulation can be given in terms due to a distribution of circulation γ(x)
(Units: [LT −1 ]) inside the wing, i.e.,
� /2
Γ= γ(x)dx
−/2

γ (x) Γ
U

Noting that superposition applies, let the total potential Φ for this flow be expressed
as the sum of two potentials

Φ = −U
� �� x� + φ
����
Free stream Disturbunce
potential potential

The flow velocity can by expressed as

v = ∇Φ = (−U + u, v)

where (u, v) are given by ∇φ = (u, v) and denote the velocity disturbance, due to the
presence of the wing. For linearized wing we can assume
u v
u, v << U ⇒ , << 1
U U

Consider a flow property q, such as velocity, pressure etc. Then let qU = q(x, 0+ ) and
qL = q(x, 0− ) denote the values of q at the upper and lower wing surfaces, respectively.

• Lift due to circulation

Applying Bernoulli equation for steady, inviscid, rotational flow, along a streamline
from ∞ to a point on the wing, we obtain
1 � �
p − p∞ = − ρ |v |2 − U 2 ⇒
2
1 �� � � 1
p − p∞ = − ρ (u − U )2 + v 2 − U 2 = − ρ(u2 + v 2 − 2uU ) ⇒
2 2
1 u v v
p − p∞ = − ρuU ( + −2)
2 U
���� U ����
���� u
<<1 <<1 ∼1

Dropping terms of order Uu , v

U
<< 1 we obtained the linearized Bernoulli equation
for thin wing at small AoA
p − p∞ = ρuU

Integrating the pressure along the wing surface, we obtain an expression for the total
lift L on the wing

� �l/2
�� � � ��
L = (p − p∞ )ny dS = p(x, 0− ) − p∞ − p(x, 0+ ) − p∞ dx
−l/2

�l/2 �l/2
� � � �
L = p(x, 0− ) − p(x, 0+ ) dx = ρU u(x, 0− ) − u(x, 0+ ) dx (1)
−l/2 −l/2

To obtain the total lift on the wing we will seek an expression for u(x, 0± ).

Consider a closed contour on the wing, of negligible thickness, as shown in the figure

below. γ (x)

u ( x,0 + )
x
t→0
u ( x,0 − )

δx

In this case we have

γ(x)δx = |u(x, 0+ )|δx + u(x, 0− )δx ⇒ γ(x) = |u(x, 0+ )| + u(x, 0− )

For small u/U we can argue that u(x, 0+ ) ∼

= −u(x, 0− ), and obtain
γ(x)
u(x, 0± ) = ∓ (2)
2
From Equations (1), and (2) the total lift can be expressed as
� l/2
L = ρU γ(x)dx = ρU Γ
−l/2
� �� �
=Γ

The same result can be obtained from the Kutta-Joukowski law (for nonlinear foil)
� /2
δL = ρU δΓ = ρU γ(x)δx ⇒ L = ρU γ(x)δx = ρU Γ
−/2

δ L = ρU δ Γ = ρUγ (x)δ x

x U
t→0
δ Γ = γ (x)δ x

δx

10

• Moment, with respect to mid-chord, due to circulation

y
L

l xcp l x
2 2
M

δL(x) = ρU γ(x)δx
δM = xδL(x) = ρU xγ(x)δx ⇒
� /2
M = ρU xγ(x)dx ⇒
−/2
M
CM = 1
2
ρU 2 2

The center of pressure xcp , can be obtained by

M = Lxcp ⇒
� /2
M −/2
xγ(x)dx
xcp = = � /2
L γ(x)dx
−/2

11

 � /2
3.17 Simple Closed-Form Solutions for −/2 γ(x)dx from Linear
Theory
1. Flat plate at angle of attack α, i.e., η = αx.

Linear lifting theory gives γ(x), which can be integrated to give the lift coefficient

CL ,
� /2
L/span = ρU γ(x)dx = · · · = ρU 2 πα ⇒
−/2
L/span
CL = 1 ⇒
2
ρU 2 
CL = 2πα ( exact nonlinear hydrofoil CL = 2π sin α)

the moment coefficient CM ,

� /2
M/span = ρU xγ(x)dx = · · · = 14 ρU 2 2 πα ⇒
−/2
M/span
CM = 1 ⇒
2
ρU 2 2
1
CM = 2
πα

and the center of pressure xcp

xcp = 14  i.e., at quarter chord

12

2. Parabolic camber η = η0 {1 − ( 2xl )2 }, at zero AoA α = 0.

Linear lifting theory gives γ(x), which can be integrated to give the lift coefficient

CL ,
� /2
L/span = ρU γ(x)dx = · · · = 2ρU 2 πη0 ⇒
−/2
η0 η0
CL = 4π , where ≡ ‘camber ratio’
 

the moment coefficient CM ,

M/span = 0 (from symmetry) ⇒

CM = 0

and the center of pressure xcp

xcp = 0

13

 � � �2 �
2x
3. Linear superposition: Both AoA and camber η = αx + η0 1 − .


η0
CL = CLα + CLη = 2πα + 4π

We can also write the previous relation in a more general form

CL (α) = 2πα + CL (α = 0)
� �� �
≡ 4π ηl0

η0
Lift coefficient CL as a function of the angle of attack α and l
.

In practice even if the camber is not parabolic, we still make use of the
previous relations, i.e., CL (α = 0) ∼= 4πη0 /.
Also note that the angle of attack for any camber is defined as
η(/2) − η(−/2) yU − yL
α≡ =
 
and η0 is determined from η ∗ , where

η ∗ = η − αx.

14
 2.20 - Marine Hydrodynamics, Spring 2005
Lecture 13

2.20 - Marine Hydrodynamics

Lecture 13

3.18 Unsteady Motion - Added Mass

D’Alembert: ideal, irrotational, unbounded, steady.

Example Force on a sphere accelerating (U = U (t), unsteady) in an unbounded fluid that

is at at rest at infinity.

ϕ θ
r U(t)

x
3D Dipole
a
n̂

U(t)


∂φ

K.B.C on sphere:
= U (t) cos θ

∂r

r=a
Solution: Simply a 3D dipole (no stream)

a3
φ = −U (t) cos θ
2r2

∂φ 
Check: = U (t) cos θ

∂r

r=a

Hydrodynamic force:
  
∂φ 1 2
Fx = −ρ + |∇φ| nx dS
∂t 2
B

On r = a,


∂φ  3
˙ a cos θ| 1˙
 = − U 2 r=a = − U a cos θ
∂t r=a 2r 2
   
∂φ 1 ∂φ 1 ∂φ 1
∇φ|r=a = , , = U cos θ, U sin θ, 0
∂r r ∂θ r sin θ ∂ϕ 2
 1
|∇φ|2 r=a = U 2 cos2 θ + U 2 sin2 θ; n̂ = −êr , nx = − cos θ
4
 π
dS = (adθ) (2πa sin θ)
B 0

adθ
a sin θ
a
θ
x

Finally, ⎡ ⎞⎤ ⎛
⎛ ⎞

π ⎢ 1 ⎜ 1 2 2 ⎟ ⎥
2 ⎝ ⎠ ⎢ ˙ 1 ⎜ 2 2 ⎟ ⎥
Fx = (−ρ) 2πa dθ (sin θ) − cos
 θ ⎢− U a cos θ + 2 ⎝U cos θ + U sin θ⎠⎥
0 ⎣ 2    4  ⎦
nx
∂φ
∂t
|∇φ|2
π π 
3 2 2 2 2 1 2
Fx = −U̇ (ρa )π dθ sin θ cos θ + (ρU )πa dθ sin θ cos θ cos θ + sin θ
4
0   0
  
⎡
2/3 ⎤ = 0, D’alembert revisited

⎢ 2 3 ⎥

Fx ⎢
= − U̇ (t) ⎣ ρ πa ⎥

   ⎦

Hydrodynamic Force Acceleration Fluid Density

3  
Volume =1/2∀sphere

Thus the Hydrodynamic Force on a sphere of diameter a moving with velocity U (t) in
an unbounded fluid of density ρ is given by
 
2 3
Fx = −U̇ (t) ρ πa
3

Comments:

• If U̇ = 0 → Fx = 0, i.e., steady translation → no force (D’Alembert’s Condition ok).

• Fx ∝ U̇ with a (−) sign, i.e., the fluid tends to ‘resist’ the acceleration.

• [· · · ] has the units of (fluid) mass ≡ ma

• Equation of Motion for a body of mass M that moves with velocity U :

 
M
 U̇ = ΣF = F + F = − U̇ m + FB ⇔

H

B

a
Body mass Hydrodynamic force All other forces on body Fluid mass
S pn̂dS

(M + ma ) U̇ = FB

i.e., the presence of fluid around the body acts as an added or virtual mass to the
body.

3
3.19 General 6 Degrees of Freedom Motions
3.19.1 Notation Review
(3D) U1 , U2 , U3 : Translational velocities
U4 ≡ Ω1 , U5 ≡ Ω2 , U6 ≡ Ω3 : Rotational velocities

2
5
1
6 4

(2D) U1 , U2 : Translational velocities

U6 ≡ Ω3 : Rotational velocity
U3 = U 4 = U 5 = 0

2
6
1

3.19.2 Added Mass Tensor (matrix)

mij ; i, j = 1, 2, 3, 4, 5, 6

mij : associated with force on body in i direction due to unit acceleration in j

direction. For example, for a sphere:

m11 = m22 = m33 = 1/2ρ∀ = (mA ) all other mij = 0

3.19.3 Added Masses of Simple 2D Geometries

• Circle
2

a
1

m11 = m22 = ρ∀ = ρπa2

• Ellipse

a
1
b

m11 = ρπa2 , m22 = ρπb2

• Plate

1
2a

m11 = ρπa2 , m22 = 0

• Square

2a
1

2a

m11 = m22 ≈ 4.754ρa2

A reasonable approximation to estimate the added mass of a 2D body is to

use the displaced mass (ρ∀) of an ‘equivalent cylinder’ of the same lateral
dimension or one that ‘rounds off’ the body. For example, consider a square
and approximate with an
(a) inscribed circle: mA = ρπa2 = 3.14ρa2 .

2a

√ 2
(b) circumscribed circle: mA = ρπ 2a = 6.28ρa2 .

(√2)a

Arithmetic mean of (a) + (b) ≈ 4.71ρa2 .

3.19.4 Generalized Forces and Moments

In this paragraph we are looking at the most general case where forces and moments
are induced on rigid body moving with 6 DoF motions, in an unbounded fluid that
is at rest at infinity.

Body fixed reference frame, i.e., OX1 X2 X3 is fixed on the body.

x2
v
U( t )

o x1

x3
v
Ω( t )

 (t) = (U1 , U2 , U3 ) , translational velocity

U

= (Ω1 , Ω2 , Ω3 ) ≡ (U4 , U5 , U6 ) , rotational velocity with respect to O
Ω(t)

 , Ω),
Consider a body with a 6 DoF motion (U  and a fixed reference frame OX1 X2 X3 .
Then the hydrodynamic forces and moments with respect to O are given by the
following relations (JNN §4.13)

• Forces

Fj = −U̇i mji − Ejkl Ui Ωk mli with i = 1, 2, 3, 4, 5, 6

1. 2.
and j, k, l = 1, 2, 3

• Moments

Mj = −U̇i mj+3,i − Ejkl Ui Ωk ml+3,i − Ejkl Uk Ui mli with i = 1, 2, 3, 4, 5, 6

3. 2. 3.
and j, k, l = 1, 2, 3

7
Einstein’s Σ notation applies.

⎧
⎪
⎪ 0 if any j, k, l are equal
⎪
⎪
⎨ 1 ifj, k, l are in cyclic order, i.e.,
Ejkl = ‘alternating tensor’ = (1, 2, 3), (2, 3, 1), or (3, 1, 2)
⎪
⎪
⎪
⎪ −1 ifj, k, l are not in cyclic order i.e.,
⎩

(1, 3, 2), (2, 1, 3), (3, 2, 1)

Note:

(a) if Ωk ≡ 0 , Fj = −U̇i mji (as expected by definition of mij ).

Also if U̇i ≡ 0, then Fj = 0 for any Ui , no force in steady translation.

(b) Bl ∼ Ui mli ‘added momentum’ due to rotation of axes.

  
Then all the terms marked as 2. are proportional to ∼ Ω × B where B is linear
momentum (momentum from i coordinate into new xj direction).

(c) If Ωk ≡ 0 : Mj = −U̇i mj+3,i mij − Ejkl Uk Ui mli .

 
 

even with U˙ =0, Mj =0 due to this term

Moment on a body due to pure steady translation – ‘Munk’ moment.

3.19.5 Example Generalized motions, forces and moments.

A certain body has non-zero added mass coefficients only on the diagonal, i.e. mij =
δij . For a body motion given by U1 = t, U2 = −t, and all other Ui , Ωi = 0, the
forces and moments on the body in terms of mi are:
F1 = , F2 = , F3 = , M1 = , M2 = , M3 =

Solution:

mij = δij

U1 = t U2 = −t Ui = 0 i = 3, 4, 5, 6 Ωk = 0 k = 1, 2, 3

U̇1 = 1 U̇2 = −1 U̇i = 0 i = 3, 4, 5, 6

Use the relations from (JNN §4.13):

k Ω =0
Fj = −U̇i mij − Ejkl Ui Ωk mil −→
Fj = −U̇i mij

k Ω =0
Mj = −U̇i mi(j+3) − Ejkl Ui Ωk mi(l+3) − Ejkl Uk Ui mli −→
Mj = −U̇i mi(j+3) − Ejkl Uk Ui mli

where i = 1, 2, 3, 4, 5, 6 and j, k, l = 1, 2, 3

For F1 , F2 , F3 use the previous relationship for Fj with j = 1, 2, 3 respectively:

F1 = − U̇1 m11 − U̇2 m21 −U̇3 m31 −U̇4 m41 −U̇5 m51 −U̇6 m61 → F1 = −m11
     
=1 =0 =0 =0 =0 =0

− U̇2 m22 → F2 = m22

Check
F2 =

=−1

− U˙ 3 m33 → F3 = 0
Check
F3 =

=0

9
For M 1, M2 , M3 use the previous relationship for Mj with j = 1, 2, 3 respectively:

M1 = −U̇i mi(1+3) − E1kl Uk Ui mli

= −U̇i mi4 − E1kl Uk Ui mli

= −U̇1 m14 −U̇2 m24 −U̇3 m34 − U̇4 m44 − U̇5 m54 −U̇6 m64

     

=0 =0 =0 =0 =0 =0
 
−E123 U2 U1 m13 +U2 m23 + U3 m33 + U4 m43 +U5 m53 +U6 m63
     
=0 =0 =0 =0 =0 =0
 
−E132 U3 U1 m12 + U2 m22 + U3 m32 +U4 m42 +U5 m52 +U6 m62 → M1 = 0
      
=0 =0 =−1 =0 =0 =0 =0

M2 = −U̇i mi5 − E2kl Uk Ui mli

= U̇5 m55 − E231 U3 Ui m1i − E213 U1 Ui m3i
= −E213 U1 U3 m33 → M2 = 0

M3 = −U̇i mi6 − E3kl Uk Ui mli

= U̇6 m66 − E312 U1 Ui m2i − E321 U2 Ui m1i
 
+1 −1

= − U1 U2 m22 + U2 U1 m11 → M3 = t2 (m22 − m11 )

   
t −t −t t

10

3.19.6 Example Munk Moment on a 2D submarine in steady translation

2
1

θ U

3 (out of page)

U1 =U cos θ
U2 = − U sin θ

Consider steady translation motion: U̇ = 0; Ωk = 0. Then

M3 = −E3kl Uk Ui mli

For a 2D body, m3i = mi3 = 0, also U3 = 0, i, k, l = 1, 2. This implies that:

M3 = − E312 U1 (U1 m21 + U2 m22 ) − E321 U2 (U1 m11 + U2 m12 )

 
=1 =−1

= −U1 U2 (m22 − m11 )

⎛ ⎞

= U 2 sin θ cos θ ⎝m22 − m11 ⎠

  
>0

Therefore, M3 > 0 for 0 < θ < π/2 (‘Bow up’). Therefore, a submarine under
forward motion is unstable in pitch (yaw). For example, a small bow-up tends to
grow with time, and control surfaces are needed as shown in the following figure.

11

 B
H

• Restoring moment ≈ (ρg∀H)sinθ.

• critical speed Ucr given by:

2
(ρg∀) H sin θ ≥ Ucr sin θ cos θ (m22 − m11 )

θ
Ucr
H

2
Usually m22 >> m11 , m22 ≈ ρ∀. For small θ, cos θ ≈ 1. So, Ucr ≤ gH or Fcr ≡
U
√ cr
gH
≤ 1. Otherwise, control fins are required.

12

 2.20 - Marine Hydrodynamics, Spring 2005

Lecture 14

2.20 - Marine Hydrodynamics

Lecture 14

3.20 Some Properties of Added-Mass Coefficients

1. mij = ρ·[function of geometry only]

F, M = [linear function of mij ] × [function of instantaneous U, U˙ , Ω]

not of motion history

2. Relationship to fluid momentum.

F(t)

where we define Φ to denote the velocity potential that corresponds to unit velocity
U = 1. In this case the velocity potential φ for an arbitrary velocity U is φ = U Φ.
The linear momentum L  in the fluid is given by
   
 =
L ρvdV = ρ∇φdV = + ρφˆ
ndS
↑
V V Green’s
theorem
B ∞

φ→0 at ∞

 
Lx (t = T ) = ρU Φnx dS = U ρΦnx dS
B B

The force exerted on the fluid from the body is −F (t) = −(−mA U̇ ) = mA U̇ .

1
 T T Newton’s Law 
↓
dt [−F (t)] = mA U˙ dt = mA U ]T0 = Lx (t = T )−Lx (t = 0) = U ρΦnx dS
  
0 0 mA U B

Therefore, mA = total fluid momentum for a body moving at U = 1 (regardless of

how we get there from rest) = fluid momentum per unit velocity of body.

∂φ ∂φ ∂U Φ ∂Φ
K.B.C. ∂n
= ∇φ · n̂ = (U, 0, 0) · n̂ = U nx , ∂n
= U nx ⇒ ∂n
= U nx ⇒ = nx
∂n

∂Φ
∴ mA = ρ Φ dS
∂n
B

For general 6 DOF:

 
∂Φj
mji =ρ Φi nj dS = ρ Φi dS = j fluid momentum due to
  ∂n
j−force/moment B potential due to body B i body motion
i−direction of motion moving with Ui =1

3. Symmetry of added mass matrix mij = mji .

    
∂Φj
mji = ρ Φi dS = ρ Φi (∇Φj · n̂)dS = ρ ∇ · (Φi ∇Φj ) dV
∂n ↑
Divergence
B B V
⎛ ⎞ Theorem


=ρ ⎝∇Φi · ∇Φj + Φi ∇2 Φj ⎠dV
  
V =0

Therefore,

mji = ρ ∇Φi · ∇Φj dV = mij
V

2
4. Relationship to the kinetic energy of the fluid. For a general 6 DoF body motion
Ui = (U1 , U2 , . . . , U6 ),

φ = Ui Φi ; Φi = potential for Ui = 1

notation


1
K.E. = ρ ∇φ · ∇φdV = 12 ρ Ui ∇Φi · Uj ∇Φj dV
2 V
V

1
= ρUi Uj ∇Φi · ∇Φj dV = 12 mij Ui Uj
2
V

K.E. depends only on mij and instantaneous Ui .

5. Symmetry simplifies mij . From 36 → 21 → ‘?’. Choose such coordinate system

symmetry
that some mij = 0 by symmetry.

Example 1 Port-starboard symmetry.

⎡ ⎤
m11 m12 0 0 0 m16 Fx
⎢ m22 0 0 0
m26 ⎥ Fy
⎢ ⎥
⎢ m33 m34 m35 0 ⎥ Fz
mij = ⎢
⎢
⎥
⎥
⎢
m44 m45 0
⎥
Mx 12 independent coefficients
⎣
m55 0
⎦
My
m66 Mz
U1 U2 U3 Ω1 Ω2 Ω3

3
Example 2 Rotational or axi-symmetry with respect to x1 axis.

⎡ ⎤
m11 0 0 0 0 0
⎢ m22 0 0 0 m35 ⎥
⎢ ⎥
⎢ m22 0 m35 0 ⎥ where m22 = m33 , m55 = m66
mij = ⎢
⎢
⎥
⎥ and m26 = m35 , so 4 different coefficients
⎢
0 0 0
⎥

⎣
m55 0 ⎦

m55

Exercise How about 3 planes of symmetry (e.g. a cuboid); a cube; a sphere?? Work
out the details.

3.21 Slender Body Approximation

Definitions
(a) Slender Body = a body whose characteristic length in the longitudinal direction is
considerably larger than the body’s characteristic length in the other two directions.
(b) mij = the 3D added mass coefficient in the ith direction due to a unit acceleration in
the j th direction. The subscripts i, j run from 1 to 6.
(c) Mkl = the 2D added mass coefficient in the k th direction due to a unit acceleration in
the lth direction. The subscripts k, l take the values 2,3 and 4.

x2
x5
x2
L

x4
x1
O
x 4 x3
x
x6
x3

Goal To estimate the added mass coefficients mij for a 3D slender body.

Idea Estimate mij of a slender 3D body using the 2D sectional added mass coefficients
(strip-wise Mkl ). In particular, for simple shapes like long cylinders, we will use known 2D
coefficients to find unknown 3D coefficients.

mij = [Mkl (x) contributions]
3D 2D

Discussion If the 1-axis is the longitudinal axis of the slender body, then the 3D added
mass coefficients mij are calculated by summing the added mass coefficients of all the thin
slices which are perpendicular to the 1-axis, Mkl . This means that forces in 1-direction
cannot be obtained by slender body theory.

5
Procedure In order to calculate the 3D added mass coefficients mij we need to:

1. Determine the 2D acceleration of each crossection for a unit acceleration in the ith
direction,
2. Multiply the 2D acceleration by the appropriate 2D added mass coefficient to get the
force on that section in the j th direction, and
3. Integrate these forces over the length of the body.

Examples

• Sway force due to sway acceleration

Assume a unit sway acceleration u̇3 = 1 and all other uj , u̇j = 0, with j = 1, 2, 4, 5, 6.
It then follows from the expressions for the generalized forces and moments (Lecture
12, JNN §4.13) that the sway force on the body is given by

f3 = −m33 u̇3 = −m33 ⇔ m33 = −f3 = − F3 (x)dx
L

A unit 3 acceleration in 3D results to a unit acceleration in the 3 direction of each

2D ‘slice’ (U̇3 = u̇3 = 1). The hydrodynamic force on each slice is then given by
F3 (x) = −M33 (x)U̇3 = −M33 (x)

Putting everything together, we obtain

 
m33 = − −M33 (x)dx = M33 (x)dx
L L

• Sway force due to yaw acceleration

Assume a unit yaw acceleration u̇5 = 1 and all other uj , u̇j = 0, with j = 1, 2, 3, 4, 6.
It then follows from the expressions for the generalized forces and moments that the
sway force on the body is given by

f3 = −m35 u̇5 = −m35 ⇔ m35 = −f3 = − F3 (x)dx
L

For each 2D ‘slice’, a distance x from the origin, a unit 5 acceleration in 3D, results
to a unit acceleration in the -3 direction times the moment arm x (U̇3 = −xu̇5 = −x).
The hydrodynamic force on each slice is then given by
F3 (x) = −M33 (x)U̇3 = xM33 (x)

6
 Putting everything together, we obtain

m35 = − xM33 (x)dx
L

• Yaw moment due to yaw acceleration

Assume a unit yaw acceleration u̇5 = 1 and all other uj , u̇j = 0, with j = 1, 2, 3, 4, 6.
It then follows from the expressions for the generalized forces and moments that the
yaw force on the body is given by

f5 = −m55 u̇5 = −m55 ⇔ m55 = −f5 = − F5 (x)dx
L

For each 2D ‘slice’, a distance x from the origin, a unit 5 acceleration in 3D, results
to a unit acceleration in the -3 direction times the moment arm x (U̇3 = −xu̇5 = −x).
The hydrodynamic force on each slice is then given by

F3 (x) = −M33 (x)U̇3 = xM33 (x)

However, each force F3 (x) produces a negative moment at the origin about the 5 axis

F5 (x) = −xF3 (x)

Putting everything together, we obtain


m55 = x2 M33 (x)dx
L

In the same manner we can estimate the remaining added mass coefficients mij - noting
that added mass coefficients related to the 1-axis cannot be obtained by slender
body theory.

In summary, the 3D added mass coefficients are shown in the following table. The empty
boxes may be filled in by symmetry.

⎡ ⎤

⎢ m22 = M22 dx m23 = M23 dx m24 = M24 dx m25 = −xM23 dx m26 = xM22 dx ⎥
⎢ ⎥
⎢ L L L L L ⎥
⎢ ⎥
⎢ ⎥
⎢ ⎥
⎢ ⎥
⎢ ⎥
⎢ m33 = M33 dx m34 = M34 dx m35 = −xM33 dx m36 = xM32 dx ⎥
⎢ L L L L ⎥
⎢ ⎥
⎢ ⎥
⎢ ⎥
⎢ ⎥
⎢ ⎥
⎢ m44 = M44 dx m45 = −xM34 dx m46 = xM24 dx ⎥
⎢ ⎥
⎢ L L L ⎥
⎢ ⎥
⎢ ⎥
⎢ ⎥
⎢ ⎥
⎢ m55 = x2 M33 dx m56 = −x2 M32 dx ⎥
⎢ ⎥
⎢ L L ⎥
⎢ ⎥
⎢ ⎥
⎢ ⎥
⎢ ⎥
⎢ ⎥
⎢ m66 = x2 M22 dx ⎥
⎣ L
⎦

3.22 Buoyancy Effects Due to Accelerating Flow

Example Force on a stationary sphere in a fluid that is accelerated against it.

⎛ ⎞
⎜ a3 ⎟
φ (r, θ, t) = U (t) ⎜
⎝r +
⎟
⎠ cos θ
2r2

dipole for sphere

∂φ  3a
 = U˙ cos θ
∂t r=a 2
 
3
∇φ|r=a = 0, − U sin θ, 0
2
1  9
|∇φ|2 r=a = U 2 sin2 θ
2 8
Then,

π  
  3a 9
Fx = (−ρ) 2πr2 dθ sin θ (− cos θ) U̇ cos θ + U 2 sin2 θ
2 8
0
π π
3 2 2 9π 2
= U̇ 3πρa dθ sin θ cos θ +ρU a dθ cos θ sin3 θ
4
0   0  
=2/3 =0
 
Fx = U˙ ρ 2πa3
  
4
3
πa3 ρ + 23 πa3 ρ
⇓

=ρ∀

9
Part of Fx is due to the pressure gradient which must be present to cause the fluid to
accelerate:

∂U ∂U ∂U ∂U 1 ∂p
x-momentum, noting U = U (t) : +U +v +w =− (ignore gravity)
∂t ∂x
 ∂y   ∂z ρ ∂x
  
0 0 0
dp
= −ρU̇ for uniform (1D) accelerated flow
dx
Force on the body due to the pressure field
  
 ∂p
F = ˆ =−
pndS ∇pdV ; Fx = − dV = ρ∀U̇
∂x
B VB VB

‘Buoyancy’ force due to pressure gradient = ρ∀U̇

10

Analogue: Buoyancy force due to hydrostatic pressure gradient. Gravitational accelera­

tion g ↔ U̇ = fluid acceleration.

ps = −ρgy

∇ps = −ρg ĵ → F s = −ρg∀ĵ Archimedes principle
Summary: Total force on a fixed sphere in an accelerated flow
 
Fx = U̇ ρ∀ + m(1) = U˙ 32 ρ∀ = 3U˙ m(1)
 
Buoyancy added mass
1
2
ρ∀

In general, for any body in an accelerated flow:

Fx = Fbuoyany + U˙ m(1) ,
where m(1) is the added mass in still water (from now on, m)

Fx = −U˙ m for body acceleration with U̇

Added mass coefficient

m
cm =
ρ∀
in the presence of accelerated flow Cm = 1 + cm

11

Appendix A: More examples on symmetry of added mass tensor

•Symmetry with respect to Y (= “X-Z” plane symmetry) 12 non-zero, independent coefficients

•Symmetry with respect to X and Y (= “Y-Z” and “X-Z” plane symmetry) 7 non-zero, independent coefficients

12
 •Axisymmetric with respect to X-axis 4 non-zero, independent coefficients

•Axisymmetric with respect to X axis and X (=“Y-Z” plane symmetry) 3 non-zero, independent coefficients

13
 2.20 - Marine Hydrodynamics, Spring 2005
Lecture 15

2.20 - Marine Hydrodynamics

Lecture 15

Chapter 4 - Real Fluid Effects (ν =

 0)
Potential Flow Theory → Drag = 0.

Observed experiment (real fluid ν << 1 but = 0) → Drag = 0.

In particular the total drag measured on a body is regarded as the sum of two components:
the pressure or form drag, and the skin friction or viscous drag.

Total Drag = Pressure Drag + Skin Friction Drag

Profile Drag or Form Drag or Viscous Drag
Drag Force due to Pressure Drag Force due to Viscous Stresses
� �� �� � � �� �� �
pn̂ds τ t̂ds
S S

where n̂ and t̂ are the normal and tangential unit vectors on the body surface respectively.
The pressure and the viscous stresses on the body surface are p and τ respectively.

The form drag is evaluated by integrating the pressure along the surface of the body. For
bluff bodies that create large wakes the form drag is ∼ total drag.

The skin friction drag is evaluated by integrating the viscous stresses on and along the
body boundary. For streamlined bodies that do not create appreciable wakes, friction drag
is dominant.

4.1 Form Drag

4.1.1 Form Drag on a Bluff Body
Consider a sphere of diameter d:
D (Drag)
ν
U
d
ρ

If no DBC apply then we have seen from Dimensional Analysis that the drag coefficient
is a function of the Reynolds number only:

CD = CD (Re )
The drag coefficient CD is defined with respect to the body’s projected area S:
D D
CD = 1 =
ρU 2 S 1
ρU 2 πd2 /4
2 2 � �� �
Projected area

The Reynolds number Re is defined with respect to the body’s diameter d:

Ud
Re =
ν
The following graph shows the dependence of CD on Re as measured from numerous ex­
periments on spheres.

CD

0.5

0.25

Re
3x105

Figure 1: Drag coefficient (CD ) for a sphere for Re > 102 .

i) Bluff Body → Form Drag

For a bluff body (examples: sphere, cylinder, flat plate, etc.) there is appreciable flow
separation and a wake is formed downstream of the body. The pressure within the
wake is significantly smaller than that upstream of the body. Therefore the integral
of the pressure along the body boundary (= form drag) does not evaluate to zero as
predicted by P-Flow.

In general, for bluff bodies form drag >> friction drag

Flow separation

Real Flow Potential Flow

Assume: pressure in the wake ∼ p∞ (pressure at infinity)

pressure on the upstream boundary of the body ∼ ps (stagnation pressure)
Then:
� � � 1 2�
D= CD (Projected/frontal area)(ps − p∞ ) = CD S ρU
���� � �� � ↑ 2
Friction Coefficient Bernoulli
S
to be determined

ii) CD = CD (Re ) → Regime Dependence

In general, for typical bluff bodies such as spheres, it is found that CD = CD (Re )
has a form similar to that shown in Figure (1). This means that CD has a ‘regime
dependence’ on Re .

There are two main regimes of interest:

• Laminar regime: Re no separation < Re < Re critical ∼

= ( 3 × 105 )
for a smooth sphere with smooth inflow

Laminar boundary layer

Separation pt

• Wide wake
No Stagnation pt Wake • Early separation
Drag
Stagnation pt
Width ~ Diameter • ‘Large’ CD =O(1)

Separation pt
Laminar boundary layer

• Turbulent regime: Re > Re critical

Turbulent boundary layer

Separation pt

No Stagnation pt Wake • Narrow wake

Width ~ Diameter/2 • Delayed separation
Stagnation pt
Separation pt
• Smaller CD

Turbulent boundary layer

 D/L
iii) Cylinder The drag coefficient for a cylinder is defined as:
CD = 1
2
ρU 2 d

D
U

L d

CD

1.2

0.6

Re
3x105

Figure 2: Drag coefficient (CD ) for a cylinder for Re > 102

iv) Bodies with Fixed Separation Points

For bodies with fixed separation points, the drag coefficient is roughly constant, i.e.,
does not depend on Re . For example, for a flat plate or disc CD ≈ 1.2

Separation pt

Separation pt

4.1.2 Boundary Layers

� �∗
L

∂v � �
 ∗ ν � 2 �∗
+ v · ∇v = ... + ∇ v
UT ∂t UL
����
1
Re
L

For most flows of interest to us ReL >> 1, i.e., viscosity can be ignored if U, L govern
the problem, thus potential flow can be assumed. In the context of potential flow theory,
drag = 0! Potential flow (no τij ) allows slip at boundary, but in reality, the no-slip condi­
tion applies on the boundaries. Otherwise, if ν =  0 and a free-slip KBC is imposed then
τ ∼ ν ∂y → ∞ at the boundary.
∂u

Prandtl: There is a length scale δ (boundary layer thickness δ << L) over which
velocity goes from zero on the wall to the potential flow velocity U outside the boundary
layer.

U
u=U

U δ<<L
y
L x

u=0

Estimate δ: Inside the boundary layer, viscous effects are of the same order as the
inertial effects.

�
∂ 2U ∂U U U2 ν δ2 δ ν 1
ν 2 ∼U →ν 2 ∼ → ∼ 2 → ∼ =� << 1 As ReL ↑, δ ↓
∂y ∂x δ L UL L L UL ReL

Generally: ReL >> 1, Lδ << 1, thus potential flow is good outside a very thin bound­
ary layer (i.e., provided no separation - a real fluid effect). For Reynolds number not
>> 1(Re ∼ O(1)), then thick boundary layer ( δ ∼ O(L)) and Prandtl’s boundary layer
idea not useful. If separation occurs, then boundary layer idea is not valid.

6
4.1.3 Boundary Layers and Flow Separation

Boundary layers help understand flow separation.

Example for flow past a circle.

Outside the boundary layer P-Flow is valid. Let capital U denote the potential flow

tangential velocity on the circle and let x denote the distance along the circle surface (i.e.,

x = body coordinate).

From the steady inviscid x-momentum equation (steady Euler) along the body boundary
(y = 0, V = 0), we obtain :
dU 1 dp
U =− (1)
dx ρ dx

Note 1: Equation (1) is used frequently in boundary layer theory.

Note 2: From Equation (1) → for flow past a flat plate dxdp
= 0 along the plate.

dU 1 dp
P − Flow solution on body y=0 :U =−
dx dx

y U = U max boundary layer : u, υ

x δ
U0
ν

U =0 U =0

dU dU
>0 <0
dx dx
dp dp
<0 >0
dx dx

dU dU
dx
>0 Acceleration dx
<0 Deceleration

dp dp
dx
<0 ‘Favorable’ pressure gradient dx
>0 ‘Adverse’ pressure gradient

 X2 X3
X4
X1
X5

X2 > X1
X=X1 y
y

P
P
p
p

u
u v
v U1 ω U2 > U1
ω
P>p Flow is being
pushed to attach

X3 > X2 X4 > X3
y y

P
P
p

u u
U3 U2 ∂u v U4 U3
= 0, ω = 0
τ3 > 0 ∂y τ4 = 0
X4 is defined as the point of
separation

X5 > X4
y

u
U5 U4

Separated Flow τ5 < 0

Flow reversal

A better way to think about separation is in terms of diffusion of vorticity.

ω=0 outside B.L.

y y

P P

ω2 ω(y) ω4 = 0 ω(y)
ω1 ω3
ω removed from fluid
ω added to fluid
by diffusion

Think of vorticity as heat; ω(y) is equivalent to a temperature distribution. Note:



DV  Dω
= ... + ν∇2 V and = ... + ν∇2 ω


Dt Dt

 2.20 - Marine Hydrodynamics, Spring 2005

Lecture 16

2.20 - Marine Hydrodynamics

Lecture 16

4.1.4 Vortex Shedding and Vortex Induced Vibrations

Consider a steady flow Uo over a bluff body with diameter D.

Fy (lift)

U0
D Fx (drag)

We would expect the average forces to be: However, the measured oscillatory forces are:

F
F
Fx
Average

Fx
Average

Fy t
Fy
t

• The measured drag Fx is found to oscillate about a non-zero mean value with
frequency 2f .

• The measured lift Fy is found to oscillate about a zero mean value with frequency
f.

• f = ω/2π is the frequency of vortex shedding or Strouhal frequency.

1
Reason: Flow separation leads to vortex shedding. The vortices are shed in a staggered
array, within an unsteady non-symmetric wake called von Karman Street. The frequency
of vortex shedding is the Strouhal frequency and is a function of Uo , D, and ν.

Fy Von
Karman
Street
Uo Fx
D

Strouhal frequency
����
f D
i) Strouhal Number We define the (dimensionless) Strouhal number S ≡ .
U0

The Strouhal number S has a regime dependence on the Re number S = S(Re ).

For a cylinder:

•
Laminar flow S ∼ 0.22

S(Re)
•
Turbulent flow S ∼ 0.3

0.3
0.22

105 106 107 Re

ii) Drag and Lift The drag and lift coefficients CD and CL are functions of the correlation
length.
For ‘∞’ correlation length:

• If the cylinder is fixed, CL ∼ O(1) comparable to CD .

• If the cylinder is free to move, as the Strouhal frequency fS approaches one of
the cylinder’s natural frequencies fn , ‘lock-in’ occurs. Therefore, if one natural
frequency is close to the Strouhal Frequency fn ∼ fS , we have large amplitude
motions ⇒Vortex Induced Vibration (VIV).

2
4.2 Drag on a Very Streamlined Body

UL
R eL ≡
ν Uo D
D b
Cf ≡ 1
ρU 2 (Lb)
2 ���� L
S=wetted area
one side of plate

Cf = Cf (ReL , L/b)
� �
Unlike a bluff body, Cf is a strong function of ReL since D is proportional to ν τ = ν ∂u
∂y
.
See an example of Cf versus ReL for a flat plate in the figure below.

Cf

Laminar
0.01
Turbulent

0.001

Re
105 106

Skin friction coefficient as a function of the Re for a flat plate

• ReL depends on plate smoothness, ambient turbulence, . . .

• In general, Cf ’s are much smaller than CD ’s (Cf /CD ∼ O(0.1) to O(0.01)). Therefore,
designing streamlined bodies allows minimal separation and smaller form drag at the
expense of friction drag.

• In general, for streamlined bodies CTotal Drag is a combination of CD (Re ) and Cf (ReL ),
� �
1 2
and the total drag is D = 2 ρU CD S + C f Aw , where CD has a regime
frontal area wetted area
dependence on Re and Cf is a continuous function ReL .

3
4.3 Known Solutions of the Navier-Stokes Equations
4.3.1 Boundary Value Problem

• Navier-Stokes’:

∂�v 1 1
+ (�v · �) �v = − �p + ν�2�v + f�
∂t ρ ρ
•
Conservation of mass:

� · �v = 0

• Boundary conditions on solid boundaries “no-slip”:

�
�v = U

Equations very difficult to solve, analytic solution only for a few very special cases (usually
when �v · ��v = 0. . . )

 4.3.2 Steady Laminar Flow Between 2 Long Parallel Plates: Plane Couette Flow

Steady, viscous, incompressible flow between two infinite plates. The flow is driven by a
pressure gradient in x and/or motion of the upper plate with velocity U parallel to the
x-axis. Neglect gravity.

Assumptions Governing Equations Boundary Conditions

∂ ∂u ∂v ∂w
i. Steady Flow: ∂t =0 Continuity: ∂x + ∂y + ∂z =0 �v = (0, 0, 0) on y = 0
∂�v ∂�v ∂�v
ii. (x, z) >> h: ∂x = ∂z =0 NS: ∂t + �v · ��v = − ρ1 �p + ν�2�v �v = (0, 0, 0) on y = h

iii. Pressure: independent of z

Continuity
∂u ∂v ∂w ∂v
+ + =0⇒ = 0 ⇒ v = v(x, z) ⇒ v = 0 (1)
∂x ∂y ����
���� ∂z ∂y ↑
BC: v(x,0,z)=0
=0, from assumption ii

Momentum x
� 2 �
∂u ∂u ∂u ∂u 1 ∂p ∂ u ∂ 2u ∂ 2u
+u + ����
v +w =− +ν + + 2 ⇒
∂t
���� ∂x
���� ∂y ∂z
���� ρ ∂x ∂x2 ∂y 2 ����
���� ∂z
=0, (1)
=0, i =0, ii =0, ii =0, ii =0, ii
2
∂ u 1 ∂p
ν 2
= (2)
∂y ρ ∂x

Momentum y
∂v 1 ∂p
+�v · � ����
v =− + ν�2 ����
v
⇒

∂t
���� ρ ∂y
=0, (1) =0, (1)
=0, i

∂p ∂p dp
=0 ⇒ p = p(x) and = (3)
∂y ↑ ∂x dx
assumption iii

Momentum z
� 2 �
∂w ∂w ∂u ∂w 1 ∂p ∂ w ∂ 2w ∂ 2w
+u + ����
v +w =− +ν + + ⇒
∂t
���� ∂x
���� ∂y ∂z
���� ρ ����
∂z ∂x2 ∂y 2 �∂z
���� ��
2
�
=0, (1)
=0, i =0, ii =0, ii =0, iii =0, ii =0, ii
2
∂ w
= 0 ⇒ w = ay + b ⇒ w=0 (4)
∂y 2 ↑
w(x,0,z)=0
w(x,h,z)=0

From Equations (1), (4)

∂u du
�v = (u, 0, 0). Also ⇒ u = u(y) and = (5)
↑ ∂y dy
assumption ii

From Equations (2), (3, and (5)

� � � �
d2 u 1 d2 p 1 dp 2 1 dp y
= ⇒ u=− − y +C1 y+C2 ⇒ u= − (h − y)y + U
dy 2 ρν dx2 ↑ 2µ dx ↑ 2µ dx h
µ=ρν u(x,0,z)=0

w(x,h,z)=U

• Special cases for Couette flow

dp Px −Px+L
1
u(y) = 2µ (h − y)y(− dx ) + U hy , where (− dx
dp
)= L

� dp
� � dp
�
I. U = 0, − dx
>0 II. U �= 0, − dx
=0

y y

dp dp
(− )>0 (− )=0
dx u ( h) = U = 0 dx u ( h) = U U

h p p
h

u ( y) u ( y)
u ( 0) = 0 u ( 0) = 0

Parabolic profile Linear profile

•Velocity
dp
u(y) = 1
2µ
(h − y)y(− dx ) u(y) = U hy
•Max velocity
h2 dp
umax = u(h/2) = 8µ
(− dx ) umax = U
•Volume flow rate
�h h3 dp
Q = 0 u(y)dy = 8µ (− dx ) Q = h2 U
•Average velocity
Q h2 dp U
ū = h
= 6µ
(− dx ) ū = 2

•Viscous stress on bottom plate (skin friction)

� � dp � �
du � du �
τw = µ dy � = h2 − dx >0 τw = µ dy �
= µ Uh
y=0 y=0

 � dp
�
III. U �= 0, − dx
�= 0

u ( h) = U U u ( h) = U U
back flow
h
τw τw

dp dp
U > 0, (− ) > 0, G > 0 U > 0, (− ) < 0, G < 0
dx dx

� dp
� � dp
�
− dx
>0 − dx
<0
•Viscous stress on bottom plate (skin friction)
� dp �
τw = h2 − dx + µ Uh
�
< dp < 2µU attached
τw =0 when (− )= − 2 , in which case the flow is insipient
> dx > h separated

 � dp
�
For the general case of U �= 0 and − dx
�= 0,
h � dp � U
− +µ τw =
2 dx h
We define a Dimensionless Pressure Gradient G

h2 � dp �
G≡ −
2µU dx
such that

• G > 0 denotes a favorable pressure gradient

• G < 0 denotes an adverse pressure gradient
• G = −1 denotes an incipient flow
• G < −1 denotes a separated or back-flow

u ( h) = U U u ( h) = U U u ( h) = U U

h
back flow

dp dp dp
U > 0, (− ) > 0, G > 0 U > 0, (− ) < 0, G < 0 U > 0, (− ) < 0, G = −1 ⇒ τ w = 0
dx dx dx

Lessons learned in § 4.3.2:

1. Reviewed how to simplify the Navier-Stokes equations.

2. Obtained one solution to the Navier-Stokes equations.

3. Realized that once the Navier-Stokes are solved we know

everything.

In the next paragraph we are going to study one more solution to the Navier-Stokes equa­
tion, in polar coordinates.

9
 4.3.3 Steady Laminar Flow in a Pipe: Poiseuille Flow
∂x
y

a
x

z
r=a
L
Vx(r)
Steady, laminar pipe flow. KBC: vx (a) = 0 (no slip) and
dvx
(r2 = y 2 + z 2 , �v = (vx , vr , vθ )) dr (0) = 0 (symmetry).

Assumptions Governing Equations Boundary Conditions

∂ 1 ∂rvr 1 ∂vθ ∂vx

i. Steady Flow: ∂t =0 Continuity: r ∂r + r ∂θ + ∂x =0 vx (r = a) = 0 no-slip
∂�v ∂�v dvx
ii. (x, z) >> h: ∂x = ∂θ =0 NS: In polar coordinates (see SAH pp.74) dr |r=0 = 0 symmetry

⇒ �v = �v (r)

iii. Pressure: independent of θ

Following a procedure similar to that for plane Couette flow (left as an exercise) we can
show that
� � ��
1 dp 1 d dvx
vr = vθ = 0, vx = vx (r), p = p(x), and =ν r
ρ dx r dr dr
� �� �
r component of �2
in cylindrical coordinates

After applying the boundary conditions we find:

� �
1 dp � 2 �
vx (r) = − a − r2
4µ dx
Therefore the volume flow rate is given by

� 2π � a
π 4 � dp �
Q= dθ rdrvx (r) = a −
0 0 8µ dx
and the skin friction evaluates to
� �
∂vr ∂vx �� ∂vx �� a � dp �
τw = τx (−r) = −τxy = −µ ( + )� = −µ � ⇒ τw = −
∂x ∂r r=a ∂r r=a 2 dx

10

4.4 Boundary Layer Growth Over an Infinite Flat Plate for

Unsteady Flow
Boundary layer thickness is related to the area where the viscosity and vorticity effects are
diffused.

For a flow over an infinite flat plate, the boundary layer thickness increases unless it is
constrained in the y direction and/or by time (unsteady flow).

1. Steady flow, constrained in y

For a steady flow past a flat plate, the boundary layer thickness increases with x.
If the flow is constrained in y, eventually the viscous effects are diffused along the
entire cross section and the flow becomes invariant in the streamwise direction.
In paragraphs 4.3.2 and 4.3.3, we studied two cases of steady laminar viscous flows,
where the viscous effects had diffused along the entire cross section.

y
h
U0
x
boundary layer Couette flow for x >> h
thickness increases with x

� � � �
Couette h
Steady flow, we assumed that viscous effects diffused through entire .
Poiseuille a

11

2. Unsteady flow, unconstrained in y

Consider the simplest example of an infinite plate in unsteady motion.

U(t)
x

∂�v ∂�v
Assumptions �p = 0, ∂x
= ∂z
= 0 ⇒ �v = �v (y, t)

Can show v = w = 0 and u = u(y, t).

Finally, from u momentum (Navier-Stokes in x) we obtain

⎛ ⎞
2 2 2
∂u ∂u ∂u ∂u 1 ∂p ⎜∂ u ∂ u ∂ u⎟
+u + ����
v +w =− +ν ⎝ 2 + 2 + 2 ⎠ ⇒
∂t ∂x
���� ∂y � ��
∂z� ρ ����
∂x ∂x
���� ∂y ∂z
����
=0
=0 =0 =0 =0 =0

∂u ∂ 2u
=ν 2 ‘ �momentum
�� � ’ diffusion equation (6)
∂t ∂y velocity
(heat)

Equation (6) is:

� first order PDE in time → requires 1 Initial Condition

� second order PDE in y → requires 2 Boundary Conditions
- u(y, t) = U (t) at y = 0, for t > 0
- u(y, t) →
0 as y → ∞

From Equation (6), we observe that the flow over a moving flat plate is due to viscous
dissipation only.

12
4.5.1 Sinusoidally Oscillating Plate

i. Evaluation of the Velocity Profile for Stokes Boundary Layer

The flow over an oscillating flat plate is referred to as ‘Stokes Boundary Layer’.

Recall that eiα = cos α + i sin α where α is real.

Assume that the plate is oscillating with U (t) = Uo cos ωt = Real {Uo eiωt }. From linear
theory, it is known that the fluid velocity must have the form
� �
u (y, t) = Real f (y) eiωt , (7)

where f(y) is the unknown complex (magnitude & phase) amplitude of oscillation.

To obtain an expression for f (y), simply substitute (7) in (6). This leads to:

d2 f
iωf = ν 2 (8)
dy

Equation (8) is a 2nd order ODE for f (y). The general solution is
√  √ 
(1+i) ω/2ν y −(1+i) ω/2ν y
f (y) = C1 e + C2 e (9)

The velocity profile is obtained from Equations (7), (9) after we apply the Boundary
Conditions.
� √ω �
u(y, t) must be bounded as y → ∞ ⇔ C1 = 0 � ω �
u (y, t) = Uo (e−y 2ν ) cos − y + ωt
u(y = 0, t) = U (t) ⇔ f (y = 0) = Uo ⇔ C2 = Uo 2ν
Stokes Boundary Layer

13

ii. Some Calculations for the Stokes Boundary Layer

Once the velocity profile is evaluated, we know everything about the flow.

ω
−y
2ν
e
λ = 2πδ

δ1/ e ≡ δ
u( y)
−1 1/ e 1 Uo
u(y )
Stokes Boundary Layer. Velocity ratio Uo as a function of the distance from the plate y .

Observe: �
u(y, t) √ω �
ω �
−y
= (e 2ν )

cos − y + ωt (10)
Uo � �� � 2ν

Exponentially decaying � �� �
envelope Oscillating component

SBL thickness
The ratio Uuo is composed of an exponentially decaying part → thickness of SBL decays
exponentially with y. We define various parameters that can be used as measures of
the SBL thickness:
u(δ )
• We define δ1/e as the distance y from the plate where U1o/e = 1e . Substituting
�
into (10), we find that δ ≡ δ1/e = 2νω
�
• The oscillating component has wave length λ = 2π 2ν = 2πδ. At λ, u(λ) ∼ 0.002.
ω Uo =
u(δ1% )
• We define δ1% as the distance y from the plate, where = 1%. Substituting
� Uo
u(δ1% )
into (10), we find that δ1% = − ln( Uo ) 2ν ∼ 4.6δ.
ω =

14

 Numerical examples:
For oscillating plate in water (ν = 10−6 m2 /s= 1mm2 /s) we have
4.6 √ ∼ �
δ1% = √ T = 2.6 ����
T
���� π
in mm in sec

2π
T = ω
δ1%

1s 3mm

10s ≤1cm

Excursion length and SBL

The plate undergoes a motion of amplitude A.
Uo
X = A sin(ωt) ⇒ U = X˙ = ����
Aω cos(ωt) ⇒ ω =
A
Uo

Comparing the SBL thickness ∼ δ with A, we find

� � �
δ ν/ω νA/Uo ν 1
∼ = = ∼�
A A ↑ A Uo A ReA
Uo
ω= A

Skin friction
The skin friction on the plate is given by
� �
∂u �� ω � �
τw = µ � = . . . = µUo sin ωt − cos ωt
∂y y=0 2ν

The maximum skin friction on the wall is

�
ω
|τw |max = µUo
ν

3π 7π
and occurs at ωt = 4
, 4 ,···

15
4.5.2 Impulsively Started Plate

U(t)

Uo

Recall Equation (6) that describes the the flow u(y, t) over an infinite flat plate undergoing
unsteady motion.
∂u ∂ 2u
=ν 2
∂t ∂y
For an impulsively started plate, the Boundary Conditions are:
�
u(o, t) = Uo
for t > 0, i.e. u(y, 0) = 0
u(∞, t) = 0

Notice that the problem stated by Equation (6) with the above Boundary Conditions has no
explicit time scale. In this case it is standard procedure to (a) use Dimensional Analysis
to find the similarity parameters of the problem, and (b) look for solution in terms of the
similarity parameters:

u � y � u
u = f (Uo , y, t, ν) ⇒ =f √ ⇒ = f (η) Self similar solution
↑ Uo 2 νt Uo
DA � �� �
≡η
similarity parameter

The velocity profile is thus given by� :

�η
u 2 2
= erfc (η) = 1 − erf (η) = 1 − √ e−α dα
Uo � �� � π
Complementary
0
error function

16

 �
Hints on obtaining the solution:
⎫
η = 2√yνt ⎪
⎪
⎪
⎪
⎪
⎬ ∂u 2
=ν ∂ u d(u/Uo ) d2 (u/Uo )
∂ ∂η ∂ y ∂ ∂t ∂y 2
∂t
= ∂t ∂η
= − √
4t νt ∂η −→ −η = −→ . . .
⎪
⎪ dη dη 2
⎪
⎪ � �� �
∂2
� ∂η �2 ∂ 2 ⎪
1 ∂ 2 ⎭ 2nd order ODE
∂y 2
= ∂y ∂η2 = 4νt ∂η2

Boundary layer thickness

In the same manner as for the SBL, we define various parameters that can be used to
measure the boundary layer thickness:
√
• δ ≡ 2 νt. At y = δ −→ uU(δo) ∼

= 0.16.
• δ1% ∼

= 1.82δ.

Excursion length and boundary layer thickness

At time t, the plate has travelled a distance L = Uo t → t = ULo .
Comparing the boundary layer thickness ∼ δ with L, we find
√ � �

δ νt νL/Uo ν 1

∼
=
=
∼
�
L L L Uo L ReL

Skin friction
The skin friction on the plate is given by
�
∂u �
� Uo
τw = µ � = . . . = −µ √
∂y y=0 πνt

17

 2.20 - Marine Hydrodynamics, Spring 2005

Lecture 17

2.20 - Marine Hydrodynamics

Lecture 17

4.6 Laminar Boundary Layers

y
U potential flow
u, v viscous flow
δ
x

Uo

4.6.1 Assumptions
∂
• 2D flow: w, ≡ 0 and u (x, y) , v (x, y) , p (x, y) , U (x, y).
∂z

• Steady flow: ≡ 0.
∂t

• For δ << L, use local (body) coordinates x, y, with x tangential to the body
and y normal to the body.

• u ≡ tangential and v ≡ normal to the body, viscous flow velocities (used inside
the boundary layer).

• U, V ≡ potential flow velocities (used outside the boundary layer).

1
4.6.2 Governing Equations

• Continuity

∂u ∂v
+ = 0 (1)
∂x ∂y
• Navier-Stokes:

 2 
∂u ∂u 1 ∂p ∂ u ∂2u
u +v =− +ν + (2)
∂x ∂y ρ ∂x ∂x2 ∂y 2
 2 
∂v ∂v 1 ∂p ∂ v ∂ 2v
u +v =− +ν + (3)
∂x ∂y ρ ∂y ∂x2 ∂y 2
4.6.3 Boundary Conditions

• KBC
Inside the boundary layer:
No-slip u(x, y = 0) = 0
No-flux v(x, y = 0) = 0
Outside the boundary layer the velocity has to match the P-Flow solution.

Let y  ≡ y/δ, y ∗ ≡ y/L, and x∗ ≡ x/L. Outside the boundary layer y  → ∞

but y ∗ → 0. We can write for the tangential and normal velocities

u(x∗ , y  → ∞) = U (x∗ , y ∗ → 0) ⇒ u(x∗ , y  → ∞) = U (x∗ , 0),
and v(x∗ , y  → ∞) = V (x∗ , y ∗ → 0) ⇒ v(x∗ , y  → ∞) = V (x∗ , 0) = 0
No-flux
P-Flow

In short:

u(x, y  → ∞) = U (x, 0)

v(x, y  → ∞) = 0
• DBC
As y  → ∞, the pressure has to match the P-Flow solution. The x-momentum
equation at y ∗ = 0 gives
∂U ∂U 1 dp ∂ 2U dp ∂U
U +V =− + ν 2
⇒ = −ρU
∂x ↓ ∂y ρ dx ↓ ∂y dx ∂x
0
0

2
4.6.4 Boundary Layer Approximation
Assume that ReL >> 1, then (u, v) is confined to a thin layer of thickness δ (x) << L.
For flows within this boundary layer, the appropriate order-of-magnitude scaling /
normalization is:

Variable Scale Normalization

u U u = Uu∗

x L x = Lx∗

y δ y = δy ∗

v V =? v = Vv ∗

• Non-dimensionalize the continuity, Equation (1), to relate V to U

 ∗  ∗  
U ∂u V ∂v δ
+ = 0 =⇒ V = O U
L ∂x δ ∂y L

• Non-dimensionalize the x-momentum, Equation (2), to compare δ with L

⎡ ⎤
 ∗  ∗ ∗   2 ∗ ⎥
U2 ∂u UV ∂u 1 ∂p νU ⎢⎢ δ2 ∂ 2u ∂ u ⎥
u + v =− + 2 ⎢ 2 + ⎥
L ∂x δ

 ∂y ρ ∂x δ ⎣ L ∂x 2 ∂y 2 ⎦
2


O(U /L) ignore

The inertial effects are of comparable magnitude to the viscous effects when:

U2 νU δ ν 1
∼ 2 =⇒ ∼ = << 1
L δ L UL R eL
∂p
The pressure gradient ∂x
must be of comparable magnitude to the inertial effects
 2
∂p U
=O ρ
∂x L

3
• Non-dimensionalize the y-momentum, Equation (3), to compare ∂p
∂y
to ∂p
∂x
 ∗  ∗  2 ∗  2 ∗
UV ∂v V2 ∂v 1 ∂p νV ∂ v νV ∂ v
u + v =− + 2 2
+ 2
L

 ∂x δ

 ∂y ρ ∂y L

 ∂x δ

 ∂y 2
2 δ 2 δ 2 δ3 2 δ

O( UL ) O( UL ) O( UL ) O( UL )

L L L3 L

∂p
The pressure gradient ∂y
must be of comparable magnitude to the inertial effects
 2 
∂p U δ
=O ρ
∂y L L
∂p ∂p
Comparing the magnitude of ∂x
to ∂y
we observe
 2   2
∂p U δ ∂p U
= O ρ while =O ρ =⇒
∂y L L ∂x L
∂p ∂p ∂p
<< =⇒ ≈0 =⇒
∂y ∂x ∂y

p = p(x)

• Note:

- From continuity it was shown that V/U ∼ O(δ/L) ⇒ v << u, inside the
boundary layer.
∂p
- It was shown that ∂y = 0, p = p(x) inside the boundary layer. This means that
the pressure across the boundary layer is constant and equal to the pressure
outside the boundary layer imposed by the external P-Flow.

4.6.5 Summary of Dimensional BVP

Governing equations for 2D, steady, laminar boundary layer
∂u ∂v
Continuity : + =0
∂x ∂y
∂u ∂u 1 dp ∂ 2u
x-momentum : u +v = − +ν 2
∂x ∂y ρ dx ∂y


U dU/dx, y=0
∂p
y - momentum : =0
∂y

Boundary Conditions
KBC

At y=0 : u(x, 0) = 0
v(x, 0) = 0
At y/δ → ∞ : u(x, y/δ → ∞) = U (x, 0)
v(x, y/δ → ∞) = 0

DBC
dp ∂U IN the b.l.
1
= −ρU or p(x) = C − ρU 2 (x, 0)
dx ∂x 
2 
Bernoulli for the P-Flow at y =0

4.6.6 Definitions
∞ 
∗ u
Displacement thickness δ ≡ 1− dy
U
0

∞
u u
Momentum thickness θ≡ 1− dy
U U
0

Physical Meaning of δ ∗ and θ

Assume a 2D steady flow over a flat plate.

dp
Recall for steady flow over flat plate dx
= 0 and pressure p = const.

Choose a control volume ([0, x] × [0, y/δ → ∞]) as shown in the figure below.

y /δ → ∞ CV
Q
Uo 4
Uo
3
Uo P - Flow

u(y)
Boundary Layer
1
2
0 x
CV for steady flow over a flat plate.

Control Volume ‘book-keeping’

Surface n̂ v v · n̂ v (v · n̂) −pn̂


1 −î Uo î −Uo −Uo2 î pî

2 −ĵ 0 0 0 pĵ

3 î u(x, y)î + v(x, y)ĵ u(x, y) u2 (x, y)î + u(x, y)v(x, y)ĵ −pî

4 ĵ Uo î + v(x, y)ĵ v(x, y) v(x, y)Uo î + v 2 (x, y)ĵ −pĵ

Conservation of mass, for steady CV

  ∞  ∞  x

  
v · ndS
ˆ =0⇒− Uo dy + v(x
, y)dx = 0 ⇒ u(x, y )dy +
1234 0 0
0

Q
 ∞  ∞  ∞  ∞ 
   u
Q= Uo dy − udy = (Uo − u)dy = Uo 1− dy  ⇒ Q = U0 δ ∗
0 0 0 U
0
o 
∗
δ

where () are the dummy variables.

Conservation of momentum in x, for steady CV

 
u(v · n
ˆ)dS = Fx ⇒
1234
 ∞  ∞  x  ∞  ∞ 
−Uo2 dy  + 2 
u (x, y )dy + 
v(x , y)Uo dx =  
pdy − pdy  + Fx,f riction ⇒
0 0 0 0 0
 ∞  ∞  x 

−Uo2 dy  + 2  
u (x, y )dy + Uo v(x , y)dx = Fx,f riction ⇒

0 0
0

Q
 ∞  ∞  ∞ 
−Uo2 dy  + u2 (x, y  )dy  + Uo (Uo − u)dy  = Fx,f riction ⇒
0 0 0
 ∞   
− Uo2 +u +2
Uo2 − Uo u dy  = Fx,f riction ⇒
0
 ∞   

u2 u
Uo2 − dy 
= Fx,f riction ⇒

0 Uo2 Uo
  ∞   
2 u u
Fx,f riction = −Uo 1− dy  ⇒ Fx,f riction = −Uo2 θ
Uo Uo
0

θ

4.7 Steady Flow over a Flat Plate: Blasius’ Laminar Boundary

Layer

Uo

L x

Steady flow over a flat plate: BLBL

4.7.1 Derivation of BLBL

dp
• Assumptions Steady, 2D flow. Flow over flat plate → U = U0 , V = 0, =0
dx
• LBL governing equations
∂u ∂v
+ =0
∂x ∂y
∂u ∂u ∂2u
u +v =ν 2
∂x ∂y ∂y
• Boundary conditions
u = v = 0 on y = 0
y 
v → V = 0, u → Uo outside the BL, i.e., >> 1
δ
• Solution Mathematical solution in terms of similarity parameters.
 
u Uo y η νx
and η≡y ⇔ =√ ⇔y=η
U νx x Rx Uo
Similarity solution must have the form

u (x, y)
= F (η)
U
 o

self similar solution

8
 We can obtain a PDE for F by substituting into the governing equations. The
PDE has no-known analytical solution. However, Blasius provided a numerical
solution. Once again, once the velocity profile is evaluated we know everything
about the flow.

4.7.2 Summary of BLBL Properties: δ, δ0.99 , δ ∗ , θ, τo , D, Cf

 
u (x, y) Uo νx y η
=



F (η) ; η = y
;
y ≡ η
;
=


Uo νx
Uo x Rx
evaluated
numerically



local R#


⎫
δ≡
νx ⎪
⎪
⎪
⎪
Uo ⎪
⎪
⎪
⎪

⎪
⎪
⎪
⎪
νx ⎪ 
δ
.99 ∼
= 4.9 , i.e., η.99 = 4.9 ⎪ ⎪
⎪
√

Uo ⎪
⎪
δ ∝ x, δ ∝ 1 Uo
⎬
 

∗ ∼ νx
⎪ δ
δ = 1.72 , i.e.,
η ∗ = 1.72
⎪
⎪
⎪ ∝
ν 1

= √

Uo ⎪
⎪ x Uo x
⎪
⎪
Rx

⎪
⎪
⎪
⎪
∼ νx ⎪
⎪
θ = 0.664 ⎪
⎪
Uo ⎪
⎪


−1/2 ⎫

U x ⎪
⎪
τ o ≡ τw ∼ 2 ⎪
o
= 0.332ρUo ⎪
⎪
1

ν
⎬
τo ∝ √
x



−1/2 ⎪
⎪
= 0.332 ρUo2 Rx ⎪
⎪ 3/2


⎭
τo ∝ Uo
⎪
local R#

Total drag on plate L x B

L  −1/2
  Uo L √
D = 

B τo dx ∼
= 0.664 ρUo2 (BL) ⇒ D ∝ L, D ∝ U 3/2
ν
width 0 

−1/2
ReL

Friction (drag) coefficient:

D ∼ 1.328 1 1
Cf = 1 2 = ⇒ Cf ∝ √ , Cf ∝ √
2
(ρUo ) (BL) R eL L U

Cf
Blasius Laminar Boundary Layer
1.328
Cf ≅
Re L
Turbulent Boundary Layer
0.008

Re L
103 3× 105
Turbulent Boundary Layer
C f for flat plate (JNN 2.3)
⎧⎪ Re x ~ 3×105
Transition at ⎨
⎩⎪Reδ ~ 600

Skin friction coefficient as a function of Re .

A look ahead: Turbulent Boundary Layers

Observe form the previous figure that the function Cf, laminar (Re ) for a laminar
boundary layer is different from the function Cf, turbulent (Re ) for a turbulent boundary
layer for flow over a flat plate.

Turbulent boundary layers will be discussed in proceeding Lecture.

10

4.8 Laminar Boundary Layers for Flow Over a Body of General

Geometry
The velocity profile given in BLBL is the exact velocity profile for a steady, laminar flow
over a flat plate. What is the velocity profile for a flow over any arbitrary body? In general
it is dp/dx = 0 and the boundary layer governing equations cannot be easily solved as was
the case for the BLBL. In this paragraph we will describe a typical approximative procedure
used to solve the problem of flow over a body of general geometry.

1. Solve P-Flow outside B ≡ B0

y
2. Solve boundary layer equations x
(with ∇P term) → get δ ∗ (x)
L ∇P ≠ 0
3. From B0 + δ ∗ → B
U ≠ const.
U

4. Repeat steps (1) to (3) until no

change
B0

• von Karman’s zeroth moment integral equation

τ0 d  2  dU
= U (x)θ(x) + δ ∗ (x)U (x) (4)
ρ dx dx

• Approximate solution method due to Polthausen for general geometry

(dp/dx = 0) using von Karman’s momentum integrals.

The basic idea is the following: we assume an approximate velocity profile (e.g. linear,

4th order polynomial, . . .) in terms of an unknown parameter δ(x). From the velocity

profile we can immediately calculate δ ∗ , θ and τo as functions of δ(x) and the P-Flow

velocity U (x).

Independently from the boundary layer approximation, we obtain the P-Flow solution
outside the boundary layer U (x), dU
dx
.
Upon substitution of δ ∗ , θ, τo , U (x), dU
dx
in von Karman’s moment integral equation(s)
we form an ODE for δ in terms of x.

11

• Example for a 4th order polynomial Polthausen velocity profile

Polthausen profiles - a family of profiles as a function of a single parameter Λ(x)
(shape function factor).

� Assume an approximate velocity profile, say a 4th order polynomial:

u (x, y) y  y 2  y 3  y 4
= a (x) + b (x) + c (x) + d (x) (5)
U (x, 0) δ δ δ δ
There can be no constant term in (5) for the no-slip BC to be satisfied y = 0,
i.e, u(x, 0) = 0.
We use three BC’s at y = δ

u ∂u ∂ 2u
= 1, = 0, = 0, at y = δ (6)
U ∂y ∂y 2
From (6) in (5), we re-write the coefficients a(x), b(x), c(x) and d(x) in terms
of Λ(x)
a = 2 + Λ/6, b = −Λ/2, c = −2 + Λ/2, d = 1 − Λ/6

� To specify the approximate velocity profile Uu(x,y)

(x,0)
in terms of a single unknown
parameter δ we use the x-momentum equation at y = 0, where u = v = 0
  
∂u ∂u ∂U ∂ 2 u  1 dU δ 2 dU δ 2 (x)
u +v =U + ν 2 ⇒b=− ⇒ Λ(x) =
↓ ∂x ↓ ∂y ∂x ∂y y=0 2 dx ν dx ν
0 0 


1 dp
− ρ dx
ν 2bU
2
δ


dU Λ > 0 : favorable pressure gradient
Observe: Λ ∝ ⇒
dx Λ < 0 : adverse pressure gradient
Putting everything together:

u (x, y) y   y 3  y 4

= 2 −2 + +
U (x, 0) δ δ δ
 
dU δ 2 1  y  1  y 2 1  y 3 1  y 4
+ − + −
dx ν 6 δ 2 δ 2 δ 6 δ

12
� Once the approximate velocity profile Uu(x,y)
(x,0)
is given in terms of a single unknown
∗
parameter δ(x), then δ , θ and τo are evaluated

∞   
∗ u 3 1  dU δ 2 
δ = 1− dy = δ −
U 10 120 dx ν
0

δ  
u u 37 1  dU δ 2  1  dU δ 2 2
θ = 1− dy = δ − −
U U 315 945 dx ν 9072 dx ν
0
  
∂u  μU 1  dU δ 2 
τo = μ = 2+
∂y y=0 δ 6 dx ν
Notes:
- Incipient flow (τo = 0) for Λ = −12. However, recall that once the flow is
separated the boundary layer theory is no longer valid.

- For dU
dx
= 0 → Λ = 0 Pohlhausen profile differs from Blasius LBL only by a
few percent.

� After we solve the P-Flow and determine U (x), dU

dx
we substitute everything into
von Karman’s momentum integral equation (4) to obtain

dδ 1 dU d2 U/dx2
= g(δ) + h(δ)
dx U dx dU/dx
where g, h are known rational polynomial functions of δ.
2
This is an ODE for δ = δ(x) where U, dU , d U are specified from the P-Flow
dx dx2
solution.

General procedure:

1. Make a reasonable approximation in the form of (5),

2. Apply sufficient BC’s at y = δ, and the x-momentum at y = 0 to reduce (5)
as a function a single unknown δ,
3. Determine U (x) from P-Flow, and
4. Finally substitute into Von Karman’s equation to form an ODE for δ(x).
Solve either analytically or numerically to determine the boundary layer
growth as a function of x.

13
 2.20 - Marine Hydrodynamics, Spring 2005
Lecture 18

2.20 - Marine Hydrodynamics

Lecture 18

4.9 Turbulent Flow – Reynolds Stress

Assume a flow v with a time scale T . Let τ denote a time scale τ << T . We can then
write for each component of the velocity

ui = ūi + ui (1)

where by definition
 τ
1
ūi = ui dt
τ 0
It immediately follows that

∂ ∂ui
u¯i = ui − ūi = ūi − ūi = 0, also ūi =
etc.

∂x
∂x

Substitute Eq. (1) into continuity and average over τ , i.e., take ( )

∂ui ∂ūi ∂ui ∂ūi

=
+ = 0, =⇒ =0
∂xi ∂xi ∂xi ∂xi

↓
0

∂ui ∂ūi ∂u ∂ui

but
= 0 =
+
i , =⇒
=0
∂xi ∂x ∂xi ∂xi
i
↓
0 , just shown

Substitute Eq. (1) into the momentum equations and take ( )

∂ui ∂ui 1 ∂τij 1 ∂p
+ uj = =− + ν∇2 ui
∂t ∂xj ρ ∂xj ρ ∂xi

⎧ 2
∂ui ∂ūi ∂ui ⎨
ν∇2 ui = ν∇ ui
=
+
; similarly
∂t ∂t ∂t


⎩ ∂p ∂ ∂p̄
∂xi
= ∂xi
(p̄ + p ) = ∂xi
etc.

∂ui

∂
∂ ūi ∂ūi ∂u ∂ 
uj ¯j + uj
= u
ui + ui ) = ūj
(¯ + uj + ūj i +uj u
∂xj ∂xj ∂xj ∂xj ∂xj ∂xj i
     
0 0

but from continuity we have

∂
 ∂
  ∂uj
uj u
i = u
u − ui
∂xj ∂xj j i ∂xj

0→by continuity

and thus we finally obtain

∂ūi ∂ūi 1 ∂p ∂  
+ ūj =− + ν∇2 ūi − uu
∂t
∂xj ρ ∂xi ∂xj i j
  
1 ∂
τ
ρ ∂xj ij

∂ūi ∂ūi 1 ∂ 
Reynolds averaged N-S equation:
+ ūj = τij − ρui uj
∂t ∂xj ρ ∂xj

Reynolds stress:
τRij ≡ −ρui uj

4.10 Turbulent Boundary Layer Over a Smooth Flat Plate

We have already seen that the function of the friction coefficient Cf (ReL ) differs for laminar
and turbulent flows. In this paragraph we will discuss the case of a turbulent boundary
layer.
Following a procedure similar to that for flow past a body of general geometry, we will
use an approximate velocity profile, obtain the P-Flow solution and eventually substitute
everything into von Karman’s momentum integral equation. The velocity profiles used in
practice are either empirical ((1/7)th power) or semi-empirical (logarithmic) laws.

u
y log
Uo

δ 1/7

U y
Uo u log
o δ

4.10.1 (1/7)th Power Velocity Profile Law

Let the velocity profile be determined by the following empirical law
ū  y 1/7
= (2)
Uo δ

where δ = δ(x) is to be determined.

From equation (2) we can obtain directly δ ∗ and θ
δ
δ∗ =
8
7 ∼
θ = δ = 0.0972 δ
72
However, we need to use an additional empirical law to determine the skin friction.
From Blasius’ law of friction for pipes we obtain an expression for τo
 −1/4
τo Uo δ
= 0.0227
ρUo2 ν

3
 From P-Flow for flow past a flat plate we have U (x) = U0 = const, and dp/dx = 0
Substituting δ ∗ , θ, τo , Uo into von Karman’s moment equation
 −1/4
τo d Uo δ 7 dδ
2
= (θ) =⇒ 0.0227 =
ρUo dx ν 72 dx

This is a 1s t order ODE for δ. One BC is required. We assume that the the flow is
tripped at x = 0, i.e., at x = 0 the flow is already turbulent. Further on, we assume
that the turbulent boundary layer starts at x = 0, i.e., δ(0) = 0. It follows that
 −1/5
Uo x δ ∼
δ (x) ∼
= 0.373x =⇒ = 0.373Re−1/5
x
ν x

Compare:

Laminar Boundary Layer Turbulent Boundary Layer (1/7th power law)

√
δ (x) ∝ x δ (x) ∝ x4/5
 νx  4 1/5
δ∗ ∼
= 1.72 Uo
δ ∗ ∼
= 0.047 νx
Uo

Once the profile has been determined we can evaluate the friction drag


D = 0.036 ρUo2 BL Re−1/5
L

Thus, the friction coefficient for turbulent (tripped and/or ReL > 5 × 105 ) flow over
a flat plate is
D
Cf = = 0.073Re−1/5
1 ρU 2 BL L

2 o

4.10.2 Logarithmic Velocity Profile Law

If the velocity profile is determined by the semi-empirical logarithmic velocity pro­
file law, following an approach similar to that for the 1/7th power law, we obtain
Schoenherr’s formula for the friction coefficient

0.242
 = log10 (ReL Cf )
Cf

 4.10.3 Summary of Boundary Layer Over a Flat Plate

Laminar BL (Blasius) Turbulent BL (1/7th power law)

δ δ
∝ Re−1/2
x
∝ Re−1/5
x
x x
√
δ ∗ = 1.72xRe−1/2
x
∝ x δ ∗ = 0.047xRe−1/5
x
∝ x4/5

τo = 0.0227ρUo2 Re−1/4
δ

τo = 0.332ρUo2 Re−1/2
x
τo = 0.02297ρUo2 Re−1/5
x

D = 0.664ρU02 (BL)Re−1/2
L
D = 0.03625ρU02 (BL)Re−1/5
L

D D
Cf ≡ = 1.328Re−1/2 Cf ≡ = 0.0725Re−1/5
ρUo2 (BL) L
ρUo2 (BL) L

For τo , the cross-over is at Rex ∼ 3.4 x 103 , i.e.,

Cf
(τo )laminar > (τo )turbulent for Rex < 3.4 × 103
(τo )laminar ∼ (τo )turbulent for Rex ∼ 3.4 × 103
(τo )laminar < (τo )turbulent for Rex > 3.4 × 103
C f L ~ RL−
1
2

C fT ~ RL−
1
5

~ 0.01
Therefore, for most prototype scales:
ln (RL)
(Cf )turbulent > (Cf )laminar
(τo )turbulent > (τo )laminar RL ~ 1.6 x 104

 2.20 - Marine Hydrodynamics, Spring 2005

Lecture 19

2.20 - Marine Hydrodynamics

Lecture 19

Turbulent Boundary Layers: Roughness Effects

So far, we have assumed a ‘hydraulically smooth’ surface. In practice, it is rarely so, due
to fouling, rust, rivets, etc. . . .

Viscous sublayer

Uo

δv
k = characteristic
roughness height

To account for roughness we first define an ‘equivalent sand roughness’ coefficient k (units:
[L]), a measure of the characteristic roughness height.

The parameter that determines the significance of the roughness k is the ratio
k
δ
k
We thus distinguish the following two cases, depending of the value of the ratio δ
on the
actual surface - e.g., ship hull.
1. Hydraulically smooth surface For k < δv << δ, where δv is the viscous sub-layer
thickness, k does not affect the turbulent boundary layer significantly.

k
<< 1 ⇒ Cf  Cf , smooth ⇒ Cf = Cf (ReL )
δ

1
2. Hydraulically rough surface For k >> δ >> δv , the flow will resemble what is
sketched in the following figure.

separation

k
δv

In terms of sand grains: each sand grain can be thought of as a bluff body. The flow,
thus separates downstream of each sand grain. Recalling that drag due to ‘separation’
= form drag >> viscous drag we can approximate the friction drag as the resultant
drag due to the separation behind each sand grain.
k k
>> 1 ⇒ Cf ≡ Cf , rough ⇒ Cf = Cf ( , ReL )
δ L ����
weak dependence

k/l ↑
Cf C D ≠ F (Re L )
k/l = constant

C f rough

RL
C f smooth

Cf , rough has only a weak dependence on ReL , since for bluff bodies CD =
 F (ReL )

In summary The important parameter is k/δ:

k
<< 1 : hydraulically smooth
δ (x)

k
>> 1 : rough
δ (x)
Therefore, for the same k, the smaller the δ, the more important the roughness k.

4.11.1 Corollaries

1. Exactly scaled models (e.g. hydraulic models of rivers, harbors, etc. . . )

k
Same relative roughness: ∼ const for model and prototype
L
� �
k kL k
= ∼ ReL 1/5
δ Lδ L

k
↑ for ReL ↑
δ

For Re model << Re prototype :

� � � �
k k
<
δ m δ p

(Cf )m < (Cf )p

2. Roughness Allowance. Often, the model is hydraulically smooth while the proto­
type is rough. In practice, the roughness of the prototype surface is accounted for
‘indirectly’.

Cf

Uk
= Rk = constant
ν

ΔCf remains
constant with Rl
RL
C f smooth

Uk
• For the same ship (Re same), different k gives different Rek = .
ν
• For a given Rek , the friction coefficient Cf is increased by almost a constant for
Uk
= Rek = const over a wide range of ReL .
ν
• If the model is hydraulically smooth, can we account for the roughness of the
prototype?
Notice that ΔCf = ΔCf (Rek ) has only a weak dependence on ReL . We can
therefore, run an experiment using hydraulically smooth model, and add ΔCf
to the final friction coefficient for the prototype

Cf (ReL ) = Cf smooth + ΔCf (R )

� ��ek�
not (ReL )
Gross estimate: For ships, we typically use ΔCf = 0.0004.

k Rek ∼ R ek
Reality: =� � = =⇒
δ δ/L ReL ReL 4/5
����
∼ReL −1/5
k
↓ as ReL ↑, i.e., ΔCf smaller for larger ReL .
δ

• Hughes’ Method Adjust for ReL dependence of Cfrough .

Cfrough = Cfsmooth (1 + γ) =⇒ ΔCf = γCfsmooth (ReL )

i.e., As ReL ↑, ΔCf ↓.

4
Chapter 5 - Model Testing.
5.1 Steady Flow Past General Bodies
- In general, CD = CD (Re ).

- For bluff bodies

Form drag >> Friction drag ⇒ CD ≈ const ≡ CP (within a regime)

Recall that the form drag (CP ) has only regime dependence on Reynold’s number,
i.e, its NOT a function of Reynold’s number within a regime.

- For streamlined bodies

CD (Re ) = Cf (Re ) + CP

5.1.1 Steps followed in model testing:

(a) Perform an experiment with a smooth model at ReM (ReM << ReS ) and obtain
the model drag CDM .
(b) Calculate CP M = CDM − Cf M (ReM ) = CP S = CP ; CDM measured, Cf M (ReM )
calculated.
(c) Calculate CDS = CP + Cf S (ReS )
(d) Add ΔCf for roughness if needed.

CP measured

CD predicted
Cf (Rship)
Cf (Rm)

calculated

Cf (Rship) calculated
R
Rm Rship

 Caution: In an experiment, the boundary layer must be in the same regime (i.e.,
turbulent) as the prototype. Therefore turbulence stimulator(s) must be added.

�
TBL �
TBL
LBL
CP turbulent regime

U MODEL

Laminar Cf Turbulent Cf

R Turbulent boundary layer

to be triggered here

5.1.2 Drag on a ship hull For bodies near the free surface, the Froude number Fr is
due to wave effects. Therefore CD = CD (Re , Fr ). In general the ra-
important, �
Re gL3
tio = . It is impossible to easily scale both Re and Fr . For example
Fr ν

Re Lm 1 νm gm

= constant and = ⇒ = 0.032 or = 1000!

Fr Lp 10 νp gp

This makes ship model testing seem unfeasible. Froude’s Hypothesis proves to be
invaluable for model testing
calculate measure indirectly
� �� � � �� �
CD (Re , Fr ) ≈ Cf (Re ) + CR (Fr )
� �� � � �� �
Cf for flat plate residual drag
of equivalent wetted area

In words, Froude’s Hypothesis assumes that the drag coefficient consists of two parts,
Cf that is a known function of Re , and CR , a residual drag that depends on Fr num­
ber only and not on Re . Since Cf (Re ) ∼ Cf (Re )flat plate , we need to run experiments
to (indirectly) get CR (Fr ).

Thus, for ship model testing we require Froude similitude to measure CR (Fr ), while
Cf (Re ) is estimated theoretically.

6
 5.1.3 OUTLINE OF PROCEDURE FOR FROUDE MODEL TESTING

(S ≡ ‘SHIP’ M ≡‘MODEL’; in general νS  = νM , and ρS =


ρM )
�
1. Given US , calculate: FrS = US / gLS = FrM

�
2. For Froude similitude, tow model at: UM = FrS gLM

3. Measure total resistance (drag) of model: Measure DM

DM
4. Calculate total drag coefficient for model: CDM = 2
0.5ρM UM SM
����
wetted area

0.075
5. Use ITTC line to calculate Cf (ReM ): Cf (ReM ) =
(log10 ReM − 2)2

6. Calculate residual drag of model: CRM = CDM − Cf (ReM )

7. Froude’s Hypothesis: CRM (ReM , Fr ) = CRM (Fr ) = CRS (Fr ) = CR (Fr )

0.075
8. Use ITTC line to calculate Cf (ReS ): Cf (ReS ) =
(log10 ReS − 2)2

9. Calculate total drag coefficient for ship: CDS = CR (Fr ) + Cf (ReS ) + ΔC

����f
∼
= 0.0004
typical value

� �
10. Calculate the total drag of ship: DS = CDS · 0.5ρS US2 SS
����
wetted area

11. Calculate the power for the ship: PS = DS US

12. Repeat for a series of US

7
 2.20 - Marine Hydrodynamics, Spring 2005
Lecture 20

2.20 - Marine Hydrodynamics

Lecture 20

Chapter 6 - Water Waves

6.1 Exact (Nonlinear) Governing Equations for
Surface Gravity Waves, Assuming Potential Flow

Free surface definition

y = η ( x, z , t ) or F ( x, y, z , t ) = 0
y

x
z

B(x, y, z,t) = 0

Unknown variables


Velocity field:
v (x, y, z, t) = ∇φ (x, y, z, t)
Position of free surface:
y = η (x, z, t) or F (x, y, z, t) = 0
Pressure field:
p (x, y, z, t)

Governing equations
Continuity:
∇2 φ = 0 y < η or F < 0
Bernoulli for P-Flow:
∂φ
∂t
+ 12 |∇φ|2 + p−pa
ρ
+ gy = 0; y < η or F < 0
Far way, no disturbance:
∂φ/∂t, ∇φ → 0 and p = pa − ρgy
 
atmospheric hydrostatic

1
Boundary Conditions

1. On an impervious boundary B (x, y, z, t) = 0, we have KBC:

∂φ  (x, t) · n
ˆ = ∇φ · n
v · n ˆ= =U ˆ (x, t) = Un on B = 0
∂n

Alternatively: a particle P on B remains on B, i.e., B is a material surface. For

example if P is on B at t = t0 , P stays on B for all t.

B(xP , t0 ) = 0, then B(xP (t), t) = 0 for all t,

so that, following P B is always 0.

DB ∂B
∴ = + (∇φ · ∇) B = 0 on B = 0
Dt ∂t

For example, for a flat bottom at y = −h ⇒ B = y + h = 0 ⇒

  
DB ∂φ ∂ ∂φ
= (y + h) = 0 ⇒ = 0 on B = y + h = 0
Dt ∂y ∂y ∂y
  
=1

2. On the free surface, y = η or F = y − η(x, z, t) = 0 we have KBC and DBC.

KBC: free surface is a material surface, no normal velocity relative to the free surface.
A particle on the free surface remains on the free surface for all times.
DF D ∂φ ∂η ∂φ ∂η ∂φ ∂η
=0= (y − η) = − − − on y = η
Dt Dt ∂y ∂t ∂x 
∂x ∂z 
∂z 
 still
vertical slope slope
unknown
velocity of f.s. of f.s.

DBC: p = pa on y = η or F = 0. Apply Bernoulli equation at y = η:

∂φ 1
+ |∇φ|2 + g η = pa on y = η
∂t 2   
still unknown
non-linear term

2
 6.2 Linearized (Airy) Wave Theory
Assume small wave amplitude compared to wavelength, i.e., small free surface slope
A
<< 1
λ

Wave height H crest

Wave amplitude A
H = A/2

SWL

Water depth h
trough wavelength Wave period T
λ

Consequently
φ η
, << 1
λ2 /T λ
We keep only linear terms in φ, η.

∂
For example: ()|y=η = ()y=0 + η ()| + . . . Taylor series
    ∂y  y=0
keep
discard

6.2.1 BVP In this paragraph we state the Boundary Value Problem for linear (Airy) waves.

∂ 2φ ∂φ
+g =0
∂t 2 ∂y
y=0
∇ 2φ = 0
y = -h

∂φ
=0
∂y

Finite depth h = const Infinite depth

GE: ∇2 φ = 0, −h < y < 0 ∇2 φ = 0, y < 0
∂φ
BKBC: ∂y = 0, y = −h ∇φ → 0, y → −∞
⎫
FSKBC: ∂φ ∂η ⎬
∂y = ∂t , y = 0 2
→ ∂∂t2φ + g ∂φ
∂y = 0
FSDBK: ∂φ
∂t + gη = 0, y = 0 ⎭

Introducing the notation {} for infinite depth we can rewrite the BVP:

Constant finite depth h {Infinite depth}

 2 
∇2 φ = 0, −h < y < 0 ∇ φ = 0, y < 0 (1)
∂φ
= 0, y = −h {∇φ → 0, y → −∞} (2)
∂y
2 
∂2φ ∂φ ∂ φ ∂φ
+ g = 0, y = 0 +g = 0, y = 0 (3)
∂t2 ∂y ∂t2 ∂y

Given φ calculate:
   
1 ∂φ  
1 ∂φ 
η (x, t) = − η (x, t) = − (4)
g ∂t y=0 g ∂t y=0

∂φ ∂φ
p − pa = −ρ − ρgy p − pa = −ρ − ρgy (5)
 ∂t   ∂t 
hydrostatic hydrostatic
dynamic dynamic

4
6.2.2 Solution Solution of 2D periodic plane progressive waves, applying separation of
variables.
We seek solutions to Equation (1) of the form eiωt with respect to time. Using the
KBC (2), after some algebra we find φ. Upon substitution in Equation (4) we can
also obtain η.

gA cosh k (y + h) gA
φ= sin (kx − ωt) φ= sin (kx − ωt) eky
ω cosh kh ω
⎧ ⎫
⎨ ⎬
= A cos (kx − ωt)
η  = A cos (kx − ωt)
η 
⎩ ⎭
using (4) using(4)

where A is the wave amplitude A = H/2.

Exercise Verify that the obtained values for φ and η satisfy Equations (1), (2), and
(4).
6.2.3 Review on plane progressive waves

(a) At t = 0 (say), η = A cos kx → periodic in x with wavelength: λ = 2π/k

Units of λ : [L]
λ
k = wavenumber = 2 /λ [L-1]
x K

(b) At x = 0 (say), η = A cos ωt → periodic in t with period: T = 2π/ω

Units of T : [T ]
T

ω = frequency = 2 /T [T-1], e.g. rad/sec

t
    
(c) η = A cos k x − ωk t ω L
Units of :
k T
ω ω 
Following a point with velocity , i.e., xp = t + const, the phase of η does
k k
ω λ
not change, i.e., = ≡ Vp ≡ phase velocity.
k T

5
6.2.4 Dispersion Relation
So far, any ω, k combination is allowed. However, recall that we still have not made
use of the FSBC Equation (3). Upon substitution of φ in Equation (3) we find that
the following relation between h, k, and ω must hold:

∂ 2φ ∂φ
+ g =0 −→ −ω 2 cosh kh + gk sinh kh = 0 ⇒ ω 2 = gk tanh kh
∂t2 ∂y ↑
φ= gA
ω
sin(kx−ωt)f (z)

• This is the Dispersion Relation

 
ω 2 = gk tanh kh ω 2 = gk (6)

Given h, the Dispersion Relation (6) provides a unique relation between ω and k,
i.e., ω = ω(k; h) or k = k(ω; h).
• Proof

C
ω2h
C ≡ = (kh) tanh (kh)
g 
kh

from (6)
1 tanh kh
C
= tanh kh
kh
kh → obtain unique solution for k
kh =f(c)

• Comments
- General As ω ↑ then k ↑, or equivalently as T ↑ then λ ↑.
  
λ ω g g
- Phase speed Vp ≡ = = tanh kh Vp =
T k k k
Therefore as T ↑ or as λ ↑, then Vp ↑, i.e., longer waves are ‘faster’ in terms of
phase speed.
- Water depth effect For waves the same k (or λ), at different water depths, as h ↑
then Vp ↑, i.e., for fixed k Vp is fastest in deep water.
- Frequency dispersion Observe that Vp = Vp (k) or Vp (ω). This means that waves of
different frequencies, have different phase speeds, i.e., frequency dispersion.

6
6.2.5 Solutions to the Dispersion Relation : ω 2 = gk tanh kh
Property of tanh kh:
long waves
shallow water
  
sinh kh 1 − e−2kh ∼ kh for kh << 1. In practice h < λ/20
tanh kh = = =
cosh kh 1 + e−2kh 1 for kh >∼ 3. In practice h > λ2
  
short waves
deep water

Shallow water waves Intermediate depth Deep water waves

or long waves or wavelength or short waves
kh << 1 Need to solve ω 2 = gk tanh kh kh >> 1
∼ h < λ/20 given ω, h for k ∼ h > λ/2
(given k, h for ω - easy!)
√
ω2 ∼
= gk · kh → ω = gh k (a) Use tables or graphs (e.g.JNN fig.6.3) ω 2 = gk
√ g 2
λ = gh T ω 2 = gk tanh kh = gk∞ λ= T
2π
k∞ λ Vp  
⇒ = = = tanh kh λ(in ft.) ≈ 5.12T 2 (in sec.)
k λ∞ Vp∞
(b) Use numerical approximation
(hand calculator, about 4 decimals )

i. Calculate C = ω 2 h/g
ii. If C > 2: ”deeper” ⇒
kh ≈ C(1 + 2e−2C − 12e−4C + . . .)
If C < 2: ”shallower” ⇒
√
kh ≈ C(1 + 0.169C + 0.031C 2 + . . .)
No frequency dispersion Frequency dispersion Frequency dispersion
 
 g g
Vp = gh Vp = tanh kh Vp = λ
k 2π

6.3 Characteristics of a Linear Plane Progressive Wave

λ η(x,t) = y
2π Vp
k= A
λ MWL x
2π
ω=
T

H = 2A
h

Define U ≡ ωA

Linear Solution:
Ag cosh k (y + h)
η = A cos (kx − ωt) ; φ= sin (kx − ωt) , where ω 2 = gk tanh kh
ω cosh kh
6.3.1 Velocity field

Velocity on free surface v (x, y = 0, t)

1 ∂η
u(x, 0, t) ≡ Uo = Aω cos (kx − ωt) v(x, 0, t) ≡ Vo = Aω sin (kx − ωt) =
tanh kh ∂t

Velocity field v (x, y, t)

∂φ Agk cosh k (y + h) ∂φ Agk sinh k (y + h)

u= = cos (kx − ωt) v= = sin (kx − ωt)
∂x ω cosh kh ∂y ω cosh kh
cosh k (y + h) sinh k (y + h)
= 
Aω cos (kx − ωt) ⇒ = 
Aω sin (kx − ωt) ⇒
sinh kh sinh kh
U U

⎧ ⎧
⎪
⎨ ⎪
∼ eky sinh k (y + h) ⎨ ∼ e
deep water ky
u cosh k (y + h) v deep water
= =
Uo cosh kh ⎪
⎩ ∼1 Vo sinh kh ⎪
⎩ ∼1+ y
shallow water h
shallow water

• u is in phase with η • v is out of phase with η

 Velocity field v (x, y)

Shallow water Intermediate water Deep water

 6.3.2 Pressure field

• Total pressure p = pd − ρgy.

• Dynamic pressure pd = −ρ ∂φ
∂t
.
• Dynamic pressure on free surface pd (x, y = 0, t) ≡ pdo

Pressure field

Shallow water Intermediate water Deep water

cosh k (y + h)
pd = ρgη pd = ρgA cos (kx − ωt) pd = ρgeky η
cosh kh
cosh k (y + h)
= ρg η
cosh kh
pd
same picture as Uuo
p do
pd (−h) pd (−h) 1 pd (−h)
= 1 (no decay) = = e−ky
p do pdo cosh kh p do
 
p= ρg(η − y) p = ρg ηeky − y
  
“hydrostatic” approximation

V p = gh V p = gλ
2π
y y y y y
y
x x
∇ ∇

kh << 1 kh >> 1

pd p ( − h) pd o pd o p (− h) pd o
p (− h) p (− h)

Pressure field in shallow water Pressure field in deep water

10

6.3.3 Particle Orbits (‘Lagrangian’ concept)

Let xp (t), yp (t) denote the position of particle P at time t.
Let (x̄; ȳ) denote the mean position of particle P.
The position P can be rewritten as xp (t) = x̄ + x (t), yp (t) = ȳ + y  (t), where
(x (t), y  (t)) denotes the departure of P from the mean position.

In the same manner let v ≡ v(¯ x, y,

¯
t) denote the velocity at the mean position and
vp ≡ v (xp , yp , t) denote the velocity at P.

P (x , y ) vp = v (x̄ + x , ȳ + y  , t) =⇒
P P TSE
(x' , y ' )
∂v ∂v
vp = v (¯
x, y,
¯ t) + (¯ ¯ t) x  +
x, y, ¯ t) y  + . . . ⇒
(x̄, y,
(x, y) ∂x ∂y
  
ignore - linear theory

vp ∼
= v

To estimate the position of P, we need to evaluate (x (t), y  (t)):

 cosh k (ȳ + h)
x = dt u (¯
x, y,
¯ t) = dt ωA cos (kx¯ − ωt) ⇒
sinh kh
cosh k (ȳ + h)
= −A sin (kx̄ − ωt)
sinh kh

 sinh k (ȳ + h)
y = dt v (¯
x, y,
¯ t) = dt ωA sin (kx̄ − ωt) ⇒
sinh kh
sinh k (ȳ + h)
= A cos (kx̄ − ωt)
sinh kh

Check: On ȳ = 0, y  = A cos (kx̄ − ωt) = η, i.e., the vertical motion of a free surface
particle (in linear theory) coincides with the vertical free surface motion.
It can be shown that the particle motion satisfies

x2 y 2 (xp − x̄)2 (yp − ȳ)2

+ = 1 ⇔ + =1
a2 b2 a2 b2
cosh k (ȳ + h) sinh k (ȳ + h)
where a = A and b = A , i.e., the particle orbits form
sinh kh sinh kh
closed ellipses with horizontal and vertical axes a and b.

11
 crest
Vp

ky
A
(a) deep water kh >> 1: a = b = Ae A
ky
circular orbits with radii Ae decreasing
exponentially with depth
trough
ky
Ae

A
Vp = gh
(b) shallow water kh << 1:

A y
a= = const. ; b = A(1+ )
kh h

decreases linearly

with depth A/kh

Vp

(c) Intermediate depth

P Vp
A
Q S Q S

R R R

λ
6.3.4 Summary of Plane Progressive Wave Characteristics

f (y) Deep water/ short waves Shallow water/ long waves

kh > π (say) kh << 1

cosh k(y+h)
= f1 (y) ∼ eky 1
cosh kh
e.g.pd

1
cosh k(y+h)
sinh kh
= f2 (y) ∼ eky kh
e.g.u, a

y
sinh k(y+h)
sinh kh
= f3 (y) ∼ eky 1+ h
e.g. v, b

13

C (x) = cos (kx − ωt) S (x) = sin (kx − ωt)

(in phase with η) (out of phase with η)

η
A
= C (x)

u v
Aω
= C (x) f2 (y) Aω
= S (x) f3 (y)

pd
ρgA
= C (x) f1 (y)

y� x�
A
= C (x) f3 (y) A
= −S (x) f2 (y)

a b
A
= f2 (y) A
= f3 (y)

b
a

14

DIFFRACTION OF OCEAN WAVES is clearly visible in this as they pass the end of the lower jetty. Variations in the way the
aerial photograph of Morro Bay, Calif. The waves are diffracted waves break are caused by contours of the shore and the bottom.

74

© 1959 SCIENTIFIC AMERICAN, INC

 OCEAN WAVES

Nlen have always been fascinated, and sOlnetinles awed, by the rhythlnic

nl0tions of the sea�s surface. A century of observation and experiment

h as revealed much about how these waves are generated and propagated

by �'illal'd Bascom

M
an is by nature a wave-watcher. actions of waves for the welfare of those equal to the height of the wave [see illus­
On a ship he finds himself star­ who live and work on the sea and along tration on next pagel. As each wave
ing vacantly at the constant its shores. passes, the water returns almost to its
swell that flexes its muscles just under Toss a pebble into a pond and watch original position. Gerstner observed that
the sea's surface; on an island he will the even train of waves go out. Waves the surface trace of a wave is approxi­
spend hours leaning against a palm tree at sea do not look at all like this. They mately a trochoid: the curve described
absently watching the rhythmic breakers are confused and irregular, with rough by a point on a circle as it rolls along the
on the beach. He would like to learn the diamond-shaped hillocks and crooked underside of a line. His work was ampli­
ways of the waves merely by watching valleys. They are so hopelessly complex fied by Sir George Airy later in the 19th
them, but he cannot, because they set that 2,000 years of observation by sea­ century, by Horace Lamb of England in
him dreaming. Try to count a hundred farers produced no explanation beyond the present century, and by others.
waves sometime and see. the obvious one that waves are somehow The first wave experimentalists were
\tVaves are not always so hypnotic. raised by the wind. The description of Ernst and Wilhelm Weber of Germany,
Sometimes they fill us with terror, for the sea surface remained in the province who in 1825 published a book on studies
they can be among the most destructive of the poet who found it "troubled, un­ employing a wave tank they had in­
forces in nature, rising up and over­ settled, restless. Purring with ripples un­ vented. Their tank was five feet long, a
whelming a ship at sea or destroying a der the caress of a breeze, flying into foot deep and an inch wide, and it had
town on the shore. Usually we think of scattered billows before the torment of glass sides. To make waves in the tank
waves as being caused by the wind, be­ a storm and flung as raging surf against they sucked up some of the fluid through
cause these waves are by far the most the land; heaving with tides breathed by a tube at one end of it and allowed the
common. But the most destructive waves a sleeping giant." fluid to drop back. Since the Weber
are generated by earthquakes and under­ The motions of the oceans were too brothers experimented not only with wa­
sea landslides. Other ocean waves, such complex for intuitive understanding. The ter and mercury but also with brandy,
as those caused by the gravitational at­ components had to be sorted out and their persistence in the face of tempta­
traction of the sun and the moon and by dealt with one at a time. So the first the­ tion has been an inspiration to all subse­
changes in barometric pressure, are oreticians cautiously permitted a perfect quent investigators. They discovered
much more subtle, often being imper­ train of waves, each exactly alike, to that waves are reflected without loss of
ceptible to the eye. Even such passive travel endlessly across an infinite ocean. energy, and they determined the shape
elements as the contour of the sea bot­ This was an abstraction, but it could at of the wave surface by quickly plunging
tom, the slope of the beach and the curve least be dealt with mathematically. in and withdrawing a chalk-dusted slate.
of the shoreline play their parts in wave Early observers noticed that passing By watching particles suspended in the
action. A wave becomes a breaker, for waves move floating objects back and water they confirmed the theory that wa­
example, because as it advances into in­ forth and up and down, but do not trans­ ter particles move in a circular orbit, the
creasingly shallow water it rises higher port them horizontally for any great dis­ size of which diminishes with depth. At
and higher until the wave front grows tance. From the motion of seaweeds the the bottom, they observed, these orbits
too steep and topples forward into foam motion of the water particles could be tend to be flattened.
and turbulence. Although the causes of deduced. But it was not until 1802 that As increasingly bolder workers con­
this beautiful spectacle are fairly well Franz Gerstner of Germany constructed tributed ideas in the 20th century, many
understood, we cannot say the same of the first wave theory. He showed that of the complexities of natural waves
many other aspects of wave activity. The water particles in a wave move in circu­ found their way into equations. How­
questions asked by the wave-watcher are lar orbits. That is, water at the crest ever, these gave only a crude, empirical
nonetheless being answered by intensive moves horizontally in the direction the answer to the question of how wind en­
studies of the sea and by the examination wave is going, while in the trough it ergy is h'ansferred to waves. The neces­
of waves in large experimental tanks. moves in the opposite direction. Thus sity for the prediction of waves and surf
The new knowledge has made it possible each water particle at the surface traces for amphibious operations in World \tVar
to measure the power and to forecast the a circular orbit, the diameter of which is II attracted the attention of Harald U.

75

© 1959 SCIENTIFIC AMERICAN, INC

Sverdrup and Walter Munk of the waves that overtake them. Thus as ener­ wavelengths and directions present at
Scripps Institution of Oceanography. As gy is added by the wind the smaller the same time, and their intersection
a result of their wartime studies of the waves continually give way to larger creates a random or a short-crested dia­
interaction of winds and waves they ones which can store the energy better. mond pattern. Under these conditions no
were the first investigators to give a rea­ But more small waves are continually meaningful dimensions can be assigned
sonably complete quantitative descrip­ formed, and in the zone where the wind to wave period and length. Height, how­
tion of how wind gets energy into the moves faster than the waves there is a ever, is important, at least to ships; sev­
waves. vVith this description wave stud­ wide spectrum of wavelengths. This is eral crests may coincide and add their
ies seemed to come of age, and a new era the generating area, and in a large storm heights to produce a very large wave.
of research was launched. it may cover thousands of square miles. Fortunately crests are much more likely
Let us follow waves as they are gen­ If storm winds apply more force than a to coincide with troughs and be canceled
erated at sea by the wind, travel for per­ wave can accept, the crest is merely out. There is no reason to believe that
haps thousands of miles across the ocean steepened and blown off, forming a the seventh wave, or some other arbitrar­
and finally break against the shore. The breaking wave at sea. This happens ily numbered wave, will be higher than
effectiveness of the wind in making when the wave crest becomes a wedge the rest; that is a myth of the sea.
waves is due to three factors: its average of less than 120 degrees and the height Since waves in a sea are so infinitely
velocity, the length of time it blows and of the wave is about a seventh of its variable, statistical methods must be em­
the extent of the open water across which length. Thus a long wave can accept ployed to analyze and describe them. A
it blows (called the fetch). more energy from the wind and rise simple way to describe height, for exam­
much higher than a short wave passing ple, is to speak of significant height-the
Waves and the Wind under the same wind. When the wind average height of the highest third of
produces waves of many lengths, the the waves. Another method, devised in
vVaves start up when the frictional shortest ones reach maximum height 1952 by Willard J. Pierson, Jr. , of New
drag of a breeze on a calm sea creates quickly and then are destroyed, while York University, employs equations like
ripples. As the wind continues to blow, the longer ones continue to grow. those that describe random noise in infor­
the steep side of each ripple presents a A simple, regular wave-train can be mation theory to predict the behavior of
surface against which the moving air can described by its period (the time it takes ocean waves. Pierson superposes the
press directly. Because winds are by na­ two successive crests to pass a point), by regular wave-trains of classical theory in
ture turbulent and gusty, wavelets of all its wavelength (the distance between such a way as to obtain a mathematically
sizes are at first created. The small, steep crests) and by its height (the vertical irregular pattem. The result is most con­
ones break, forming whitecaps, releasing distance between a trough and a suc­ veniently described in terms of energy
some of their energy in turbulence and ceeding crest). Usually, however, there spectra. This scheme assigns a value for
possibly contributing part of it to larger are several trains of waves with different the square of the wave height to each

WAVELENGTH

CROSS SECTION OF OCEAN WAVE traveling from left to right orbits of water particles in the wave. At the surface their diam.
shows wavelength as distance between successive crests. Tbe time eter equals the wave height. At a depth of half the wavelength
it takes two crests to pass a point is the \V3ve period. Circles are (left), orbital diameter is only 4 per cent of that at surface.

76
© 1959 SCIENTIFIC AMERICAN, INC
frequency and direction. Then, by de­ the 12,OOO-ton 5.5. Asc:anius reported an sea. It would have been disastrous to
termining the portion of the spectrum in extended storm in which the recording have steamed on any other course."
which most of the energy is concen­ barometer went off the low end of the From among a number of separately de­
trated, the average periods and lengths scale. vVhen the ship was in a trough on termined observations, that of the watch
can be obtained for use in wave fore­ an even keel, his observation post on the officer on the bridge was selected as the
casting. ship was 60 feet above the water level, most accurate. He declared that he "saw
Over a long fetch, and under a strong, and he was certain that some of the seas astern at a level above the main­
steady wind, the longer waves pre­ waves that obscured the horizon were at mast crow's-nest and at the moment of
dominate. It is in such areas of sea that least 10 feet higher than he was, ac­ observation the horizon was hidden from
the largest wind waves have been re­ counting for a total height of 70 feet or view by the waves approaching the
corded. The height of the waves in a more. Commodore Hayes of the 5.5. M a­ stern." On working out the geometry of
train does not, however, bear any simple iestic: reported in February, 1923, that the situation from the ship's plan, White­
relationship to their other two dimen­ his ship had experienced winds of hurri­ marsh found that this wave must have
sions: the period and the wavelength. cane force and waves of 80 feet in been at least 112 feet high [see illustra­
The mariner's rule of thumb relates wave height. Cornish examined the ship, close­ tion at the bottom of the next two pages J.
height to wind velocity and says that the ly interrogated the officers and concluded The period of these waves was clocked
height ordinarily will not be greater than that waves 60 to 90 feet high, with an at 14.8 seconds and their velocity at 55
half the wind speed. This means that an average height of 75 feet, had indeed knots.
80-mile-per-hour hurricane would pro­ been witnessed. As waves move out from under the
duce waves about 40 feet high. A wave reported by Lieutenant Com­ winds that raise them, their character
The question of just how large incli­ mander R. P. Whitemarsh in the Pro­ changes. The crests become lower and
vidual waves at sea can aetually be is ceedings of the U. S. Navallnslitute tops more rounded, the form more symmetri­
still unsettled, because observations arc all others. On February 7, 1933, the cal, and they move in trains of similar
difficult to make and substantiate from U.S.S. Ramapo, a Navy tanker 478 feet period and height. They are now called
shipboard in the midst of a violent long, was en route from Manila to San swell, or sometimes ground swell, and in
storm. Vaughan Cornish of England Diego when it encountered "a disturb­ this form they can travel for thous�nds
spent half a century collecting data on ance that was not localized like a of miles to distant shores. Happily for
waves, and concluded that storm waves typhoon . . . but permitted an unob­ mathematicians, swell coincides much
over 45 feet high are rather common. structed fetch of thousands of miles." more closely with classical theory than
Much higher waves have been fairly The barometer fell to 29.29 inches and do the waves in a rough sea, and this re­
well authenticated on at least two oc­ the wind gradually rose from 30 to 60 news their faith in the basic equations.
casions. knots over several days. "We were run­ Curiously enough, although each
In October, 1921, Captain Wilson of ning directly downwind and with the wave moves forward with a velocity

<ENERGY ADVANCE

I
,

,
�"'--------------�"
-;;:/;;-3-�-------=-=-�-�-������
-

I
J...
______ (-------- ------. .-_=" _....
. ,/..-- ; � .�----
...
.�---------.......-�;

WAVE ADVANCE

MOYING TRAIN OF WAVES advances at only half the speed of as wave 2 has. Meanwhile wave 1 has died, but wave 4 has formed
its indiddual waves. At top is a wave train in its first position. At at the rear of the train to replace it. Waves arriving at sho.·c
bottom the train, and its energy, have moved only half as far are thus remote descendants of waves originally generated.

77

© 1959 SCIENTIFIC AMERICAN, INC

DIFFERENT TRAINS OF WAVES, caused by winds of different ous truins, three of which are represented diagranlmalically here,
directious and strengths, make up the surface of a "sea." The vari· have a wide spectrum of wavelengths, heights and dit·ections. When

that corresponds to its length, the energy wavelength. Here their velocity is con­ it in relatively deep water continue to
of the group moves with a velocity only trolled by the depth of the water, and move rapidly. The wave thus bends to
half that of the individual waves. This is they are now called shallow-water waves. converge on the headland from all sides.
because the waves at the front of a group 'Vavelength decreases, height increases As it does, the energy is concentrated in
lose energy to those behind, and gradu­ and speed is reduced; only the period is less length of crest; consequently the
ally disappear while new waves form at unchanged. The shallow bottom greatly height of the crest is increased. This ac­
the rear of the group. Thus the composi­ modifies the waves. First, it refracts counts for the old sailors' saying: "The
tion of the group continually changes, them, that is, it bends the wave fronts to points draw the waves."
and the swells at a distance are but re­ approximate the shape of the underwater Another segment of the same swell
mote descendants of the waves created contours. Second, when the water be­ will enter an embayment and the wave
in the storm [see illust1"ation on preced­ comes critically shallow, the waves front will become elongated so that the
ing page]. One can measure the period at break [see illustration on page 84]. height of the waves at any point along
the shore and obtain from this a correct Even the most casual observer soon the shore is correspondingly low. This is
value for the wave velocity; however, notices the process of refraction. He sees why bays make quiet anchorages and
the energy of the wave train traveled that the larger waves always come in exposed promontories are subject to
from the storm at only half that speed. nearly parallel to the shoreline, even wave battering and erosion-all by the
Waves in a swell in the open ocean though a little way out at sea they seem same waves. One can deal quantitatively
are called surface waves, which are de­ to be approaching at an angle. This is with this characteristic of waves and can
fined as those moving in water deeper the result of wave refraction, and it has plot the advance of any wave across wa­
than half the wavelength. Here the bot­ considerable geological importance be­ ters of known depths. Engineers plan­
tom has little or no effect on the waves cause its effect is to distribute wave ning shoreline structures such as jetties
because the water-particle orbits dimin­ energy in such a way as to straighten or piers customarily draw refraction dia­
ish so rapidly with depth that at a depth coastlines. Near a headland thfl part of grams to determine in advance the effect
of half the wavelength the orbits are only the wave front that reaches shallow wa­ of waves of various periods and direc­
4 per cent as large as those at the surface. ter first is slowed down, and the parts of tion. These diagrams show successive
Surface waves move at a speed in miles
per hour roughly equal to 3.5 times the
period in seconds. Thus a wave with a
period of 10 seconds will travel about
35 miles per hour. This is the average
period of the swell reaching U. S. shores,
the period being somewhat longer in the
Pacific than the Atlantic. The simple re­
lationship between period and wave­
length (length=5.12T2) makes it easy
to calculate that a 10-second wave will
have a deep-water wavelength of about
512 feet. The longest period of swell
ever reported is 22.5 seconds, which cor­
responds to a wavelength of around 2,600
feet and a speed of 78 miles per hour.

Waves and the Shore

As the waves approach shore they WAVE 112 FEET HIGH, possibly the largest ever measured in the open sea, was en·
reach water shallower than half their countered in the Pacific in 1933 by the U.S.S. Ramapo, a Navy tanker. This diagram shows

78

© 1959 SCIENTIFIC AMERICAN, INC

they meet, the resulL is apparent confusion, represented at far the sea surface. The pattern becomes so complex that statistical
right by a topographic diagram drawn from actual photographs of methods must be used to analyze the waves and predict their height.

positions of the wave front, partitioned a lens to increase the wave heights to 3.5 orbits exist no more. The result is surf.
by orthogonals into zones representing times average at the point of damage. If the water continues to get shallow­
equal wave energy [see illustration on During World "Var II it was necessary er, the broken wave becomes a foam
next page J. The ratio of the distances be­ to determine the depth of water off ene­ line, a turbulent mass of aerated water.
tween such zones out at sea and at the my-held beaches against which am­ However, if the broken wave passes into
shore is the refraction coefficient, a con­ phibious landings were planned. Our deeper water, as it does after breaking
venient means of comparing energy re­ scientists reversed the normal procedure on a bar, it can form again with a lesser
lationships. for refraction studies; by analyzing a height that represents the loss of energy
Refraction studies must take into ac­ carefully timed series of aerial photo­ in breaking. Then it too will break as it
count surprisingly small underwater ir­ graphs for the changes in length (or moves into a depth critical to its new
regularities. For example, after the Long velocity) and direction of waves ap­ height.
Beach, Calif., breakwater had withstood proaching a beach, they were able to The depth of water beneath a break­
wave attack for years, a short segment of map the underwater topography. er, measured down from the still-water
it was suddenly wrecked by waves from The final transformation of normal level, is at the moment of breaking about
a moderate storm in 1930. The break­ swell by shoal or shallow water into a 1.3 times the height of the breaker. To
water was repaired, but in 1939 waves breaker is an exciting step. The waves estimate the height of a breaker even
breached it again. A refraction study by have been shortened and steepened in though it is well offshore, one walks from
Paul Horrer of the Scripps Institution of the final approach because the bottom the top of the beach down until the crest
Oceanography revealed that long-period has squeezed the circular orbital motion of the breaking wave is seen aligned
swell from exactly 165 degrees (south­ of the particles into a tilted ellipse; the with the horizon. The vertical distance
southeast), which was present on only particle velocity in the crest increases between the eye and the lowest point to
these two occasions, had been focused at and the waves peak up as they rush land­ which the water retreats on the face of
the breach by a small hump on the bot­ ward. Finally the front of the crest is un­ the beach is then equal to the height of
tom, 250 feet deep and more than seven supported and it collapses into the the wave.
miles out at sea. The hump had acted as trough. The wave has broken and the The steepness of the bottom influences

how the great wave was measured. An observer at A on the bridge line of sight to crest of wave, which had just come in line with
was looking toward the stern and saw the crow's·nest at B in his horizon. From geometry of situation, wave height was calculated.

79

© 1959 SCIENTIFIC AMERICAN, INC

the character of the breakers. When a the U. S. Army Corps of Engineers de­ Iighthouse-keeper's house, the floor of
large swell is forced by an abrupt under­ voted his career to studying the forces which is 91 feet above the water, and
water slope to give up its energy rapidly, of waves on engineering structures and fell back through the roof to wreck the
it forms plunging breakers-violent in 1904 reported some remarkable ex­ interior.
waves that curl far over, flinging the amples of their destructive power. At At Wick, Scotland, the end of the
crest into the trough ahead. Sometimes, Cherbourg, France, a breakwater was breakwater was capped by an 800-ton
the air trapped by the collapsing wave composed of large rocks and capped block of concrete that was secured to the
is compressed and explodes with a great with a wall 20 feet high. Storm waves foundation by iron rods 3.5 inches in
roar in a geyser of water [see illustration hurled 7,000-pound stones over the wall diameter. In a great storm in 1872 the
on opposite pagel. However, if the bot­ and moved 65-ton concrete blocks 60 designer of the breakwater watched in
tom slope is long and gentle, as at feet. At Tillamook Rock Light off the amazement from a nearby cliff as both
Waikiki in Hawaii, the crest forms a Oregon coast, where severe storms are cap and foundation, weighing a total of
spilling breaker, a line of foam that tum­ commonplace, a heavy steel grating now 1,350 tons, were removed as a unit and
bles down the front of the partly broken protects the lighthouse beacon, vvhich is deposited in the water that the wall was
wave as it continues to move shoreward. 139 feet above low water. This is neces­ supposed to protect. He rebuilt the
Since waves are a very effective mech­ sary because rocks hurled up by the structure and added a larger cap weigh­
anism for transporting energy against a waves have broken the beacon several ing 2,600 tons, which was treated simi­
coast, they are also effective in doing times. On one occasion a rock weighing larly by a storm a few years later. There
great damage. Captain D. D. Gaillard of 135 pounds was thrown well above the is no record of whether he kept his job

WA VE·REFRACTION DIAGRAM shows how energy of wave wide area B'. Horizontal lines are wave fronts; vertical lines
front at A is all concentrated by refraction at A' around small head· divide energy into equal units for purposes of investigation. Such
land area. Same energy at B enters a bay but is spread at beach over studies are vital preliminaries to design of shoreline structures.

80

© 1959 SCIENTIFIC AMERICAN, INC

and tried again. Gaillard's computations
show that the wave forces must have
been 6,340 pounds per square foot.

Tsunamis

Even more destructive than wind­

generated waves are those generated by
a sudden impulse such as an underwa­
ter earthquake, landslide or volcano. A
man-made variation of the sudden im­
pulse is the explosion of nuclear bombs
at the surface of the sea, which in recent
years have become large enough to be
reckoned with as possible callses of de­
structive waves.
The public knows sllch waves as tidal
waves, although they are in no way re­
lated to the tides and the implication has
long irritated oceanographers. It was
proposed that the difficulty could be re­
solved by adopting the Japanese word
tsunami. Some time later it was discov­
ered that Japanese oceanographers are
equally irritated by this word; in literal
translation tsunami means tidal wave!
However, tsunami has become the
favored usage for seismic sea waves.
Like the plunger in a wave channel,
the rapid motion or subsidence of a part
of the sea bottom can set a train of waves
in motion. Once started, these waves
travel great distances at high velocity
with little loss of energy. Although their
height in deep water is only a few feet,
on entering shallow water they are able
to rise to great heights to smash and inun­
date shore areas. Their height depends
almost entirely on the configuration of
the coastline and the nearby underwater W A VE·CREA TED "GEYSER" results when large breakers smash into a very steep beach.
They curl over and collapse, trapping and compressing air. This compressed air then e�·
contours. Tsunamis have periods of more
plodes as shown here, with spray from a 12·foot breaker leaping 50 feet into the air.
than 15 minutes and wavelengths of
several hundred miles. Since the depth
of water is very much less than half the Once when an alert was broadcast at dropping back into the water of nearly
wavelength, they are regarded as long­ Honolulu, thousands of people there a cubic mile of rock, but the waves were
or shallow-water waves, even in the dashed down to the beach to see what monumental. Their period close to the
13,000-foot average depth of the open luckily turned out to be a very small disturbance was two hours, and at great
ocean, and their velocity is limited by \vave. distances about one hour. vVaves at least
the depth to something like 4.50 miles Certain coasts near zones of unrest in 100 feet high swept away the town of
per hour. the earth's crust are particularly prone Merak, 33 miles from the volcano; on the
These fast waves of great destructive to such destructive waves, especially the opposite shore the waves carried the
potential give no warning except that shores of the Mediterranean, the Carib­ man-of-war Berow 1.8 miles inland and
the disturbance that causes them can be bean and the west coast of Asia. On the left it 30 feet above the level of the sea.
detected by a seismograph. The U. S. world-wide scale, they occur more fre­ Some 36,380 people died by the waves
Coast Guard operates a tsunami warning quently than is generally supposed: in a few hours. Tide gauges in South
network in the Pacific that tracks all nearly once a year. Africa (4,690 miles from Krakatoa),
earthquakes, and when triangulation in­ A well-known seismic sea wave, thor­ Cape Horn (7,820 miles) and Panama
dicates that a quake has occurred at sea, oughly documented by the Royal So­ (11,470 miles) clearly traced the prog­
it issues alerts. The network also has de­ ciety of London, originated with the ress of a train of about a dozen waves,
vices to detect changes in wave period eruption of the volcano Krakatoa in the and showed that their speed across the
which may indicate that seismic waves East Indies on August 27, 1883. It is not Indian Ocean had been between 350 and
are passing [see "Tsunamis," by Joseph certain whether the waves were caused 450 miles per hour.
Bernstein; SCIENTIFIC A M E RICAN , Aug­ by the submarine explosion, the violent A tsunami on April 1, 1946, originat­
ust, 1934]. Curiously the influence of the movements of the sea bottom, the rush ing with a landslide in the Aleutian sub­
system may not be entirely beneficial. of water into the great cavity, or the marine trench, produced similar effects,

81

© 1959 SCIENTIFIC AMERICAN, INC

HUNDRED-FOOT "TIDAL WAVE," or tsunami, wrought impres­ Atop the plateau a radio mast, its foundation 103 feet above sea,
sive destruction at Scotch Cap, Alaska, in 1946_ Reinforced concrete was also knocked down_ Lighthouse debris was on plateau_ Same
lighthouse that appears in top photograph was demolished, as tsunami, started by an Aleutian Island earthquake, hit Hawaiian
shown in lower photograph, which was made from a higher angle_ Islands, South America and islands 4,000 miles away in Oceania_

82

© 1959 SCIENTIFIC AMERICAN, INC

fortunately on less-populated shores. It
struck hard at the Hawaiian Islands, kill-
SYLVANIA ELECTRONIC SYSTEMs • • • IN SIMULATION
ing several hundred people and damag­
ing property worth millions of dollars.
At Hilo, Hawaii, the tsunami demon­
strated that such waves are virtually in­
visible at sea. The captain of a ship
standing off the port was astonished up­
on looking shoreward to see the harbor
and much of the city being demolished
by waves he had not noticed passing un­
der his ship. The same waves caused
considerable damage throughout the is­
lands of Oceania, 4,000 miles from
epicenter, and on the South American
coast, but they were most spectacular at
Scotch Cap in Alaska. There a two-story
reinforced-concrete lighthouse marked a
channel through the Aleutian Islands.
The building, the base of which was
32 feet above sea level, and a radio mast
100 feet above the sea were reduced to
bare foundations by a wave estimated to
be more than 100 feet high [see illustra­
tion on opposite page J.
Uncontrollable geologic disturbances
will cause many more seismic sea waves
in the future, and since the world's coast­
al population is continuously increasing,
the greatest wave disaster is yet to come.
'Vithin the next century we can expect
that somewhere a wave will at least
equal the one that swept the shores of
the Bay of Bengal in 1876, leaving
200,000 dead.

Tides and Other Waves

The rhythmic rise and fall of the sea

level on a coast indicate the passage of a
true wave we call a tide. This wave is
driven, as almost everyone knows, by the
gravitational influence of the sun and
the moon. As these bodies change their Creating electronic "realities"
relative positions the ocean waters are
attracted into a bulge that tends to re­
main facing the moon as the earth turns
in a low-cost manageable package
under it; a similar bulge travels around
the earth on the opposite side. The wave S IMULATING CONDITIONS of actual ially by current simulation projects.
period therefore usually corresponds to operations for flight trainers, radar In simulators or trainers, as in
half the lunar day. operators, or space travelers is an every major area of electronic sys­
vVhen the sun and the moon are experienced and growing facility of tems, Sylvania can assume full re­
aligned with the earth, the tides are Sylvania Electronic Systems. sponsibility from system analysis
large (spring tides); when the two bod­ Good examples are Sylvania's par­ and management through research
ies are at right angles with respect to the ticipation in the Navy's Universal and engineering, product design,
earth, the tides are small (neap tides). Digital Operational Flight Trainer production to field engineering.
and "ACTER," a new Sylvania Sylvania welcomes the opportu­
By using astronomical data it is possible
development which will add ECM nity to outline its special talents and
to predict the tides with considerable ac­
conditions to existing radar training capabilities to you or your organiza­
curacy. However, the height and time of
gear. tion personally. Simply address your
the tide at any place not on the open
Undoubtedly the simulation of inquiry to division headquarters,
coast are primarily a function of the space travel can be enhanced mater- address below .
shape and size of the connection to the
ocean.
Still another form of wave is a seiche,
a special case of wave reflection. All en­
Sylvania Electronic Systems
A Division of Sylvania Electric Products Inc.
:JI" SYLVAN IA
�;�;�;:� G�L
r
closed bodies of water rock with charac- 63 Second Avenue, Waltham, Mass., GENERAL TEL ELECTRONICS

83

© 1959 SCIENTIFIC AMERICAN, INC

teristics related to the size of the basin. The great advances both in wave the­ modeled in the laboratory. The trick is to
The motion is comparable to the slosh­ ory and in the actual measurement of use layers of two liquids that do not mix,
ing of water in the bathtub when one waves at sea have not reduced the need and create waves on the interface be­
gets out quickly. In an attempt to return for extensive laboratory studies. The so­ tween them. The speeds of the waves
to stability the water sways back and lution of the many complex engineering can be controlled by adjusting the densi­
forth with the natural period of the tub problems that involve ships, harbors, ties of the liquids.
(mine has a period of two seconds). beaches and shoreline structures re­ To reduce the uncertainties in ex­
Similarly a tsunami or a barometric pres­ quires that waves be simulated under trapolation from the model to prototype,
sure-change will often set the water in a ideal test conditions. Such model studies some of the new wave tanks are very
bay rocking as it passes. In fact, the tsu­ in advance of expensive construction large. The tank of the Beach Erosion
nami itself may reflect back and forth permit much greater confidence in the Board in Washington, D.C. (630 feet
across the ocean as a sort of super-seiche. designs. long and 20 feet deep, with a 500-horse­
In addition to seiches, tides, tsunamis power generator), can subject quarter­
and wind waves there are other waves in Experimental Tanks scale models of ocean breakwaters to
the sea. Some travel hundreds of feet be­ six-foot breakers. The new maneuvering
neath the surface along the thermocline, The traditional wave channel in tank now under construction at the
the interface between the cold deep wa­ which an endless train of identical small David Taylor Model Basin in Carderock,
ter and the relatively warm surface lay­ waves is created by an oscillating plung­ Md., measures 360 by 240 feet, is 35 feet
er. Of course these waves cannot be er is still in use, but some of the new deep along one side and will have wave
seen, but thermometers show that they wave tanks are much more sophisticated. generators on two sides that can inde­
are there, moving slowly along the In some the channel is covered, so that a pendently produce trains of variable
boundary between the warm layer and high velocity draft of air may simulate waves. Thus man can almost bring the
the denser cold water. Their study the wind in making waves. In others, like ocean indoors for study.
awaits proper instrumentation. Certain the large tank at the Stevens Institute of The future of wave research seems to
very low waves, with periods of several Technology, Hoboken, N.J. [see cover], lie in refinement of the tools for measur­
minutes, issue from storms at sea. These artificial llTegular waves approach the ing, statistically examining and repro­
long-period "forerunners" may be variability of those in the deep ocean. ducing in laboratories the familiar wind
caused by the barometric pulsation of In such tanks proposed ship designs, like waves and swell as well as the more re­
the entire storm against the ocean sur­ those of the America's Cup yacht Colum­ cently discovered varieties. It lies in
face. Since they travel at hundreds of bia, are tested at model size to see how completing the solution of the problem
miles an hour, they could presumably be they will behave at sea. of wave generation. It lies in the search
used as storm warnings or storm-center The ripple tank, now standard appa­ for forms of ocean waves not yet discov­
locators. Other waves, much longer than ratus for teaching physics, has its place ered-some of which may exist only on
tides, with periods of days or weeks and in shoreline engineering studies for con­ rare occasions. Nothing less than the
heights of less than an inch, have been veniently modeling diffraction and re­ complete understanding of all forms of
discovered by statistical methods and fraction. Even the fast tsunamis and the ocean waves must remain the objective
are now an object of study. very slow waves of the ocean can be of these studies.

w AVE BREAKS UP at the beach when swell moves into water in the crest have no room to complete their cycles; the wave form
shallower than haH the wavelength (1). The shallow bottom raises breaks (3). A foam line forms and water particles, instead of just
wave height and decreases length (2). At a water depth 1.3 times the wave form, move forward (4). The low remaining wave runs
the wave height, water supply is reduced and the particles of water up the face of the beach as a gentle wash called the uprush (5).

84
© 1959 SCIENTIFIC AMERICAN, INC
 Artist's draw i n g s h ows e n e m y s u b m a r i n e p o s i t i o n (wh ite p i p i n d icated by arrow) as it a p pears on B e n d i x S o n a r v i ew i n g scope
in h e l i copter. I t i s t h e fi r s t a i r borne system to prov i d e a visual p r e s e n t a t i o n w h i c h p i n po i n t s a target b e l ow t h e s u rface.

NOW NAVY HELICOPTERS, WITH BENDIX SONAR,

"SEE" ENEMY SUBS UNDERWATER
Prowling enemy submarines will meet a receiving transducer is lowered into the water
deadly countermeasure in Bendix* Sonar as the helicopter hovers. Known as " dunked"
equipped seagoing helicopters which have sonar, signals are transmitted through the
their own ways of hunting down their prey. water which echo back, somewhat like radar
This is one phase of the U. S. Navy's potent beams, from objects like submarines. The
and far reaching Anti-Submarine Warfare fact that the system is a long-range search
program. A pioneer in underwater detection and detection device adds greatly to its value.
and communications , our Bendix-Pacific Bendix produces a number of other sonic
Division in North Hollywood, C alifornia, devices. These include the famous depth
developed this first airborne scanning sonar s o u n d e r or " fi s h fin de r " ; sonic cleaners
system for use in Sikorsky HSS-2 helicopters. for the greatest efficiency in industrial and
It is the first airborne system to provide a medical cleaning and the Cavitron* for the
visual presentation which pinpoints a target ultrasonic machining of germanium, glass,
below the surface. In operation, a sending- quartz, silicon or ferrite. "REG. u.s. PAT. OFF.

A thousand products a million ideas

© 1959 SCIENTIFIC AMERICAN, INC

© 1959 SCIENTIFIC AMERICAN, INC
Adva n c e d l a n d m a ss s i m u l a t i o n c o n c e pt

d e m o n st rates Po m o n a D i v i s i o n c a p a b i l ity

i n rea l istic si m u lators for lower-cost tra i n i ng

of America's a i r a n d spacem e n .

I N D EV E LO P M E N T :
A M arquardt-conceived land mass s i mulation system that shrinks
the map scale factor to 1 : 3 , 0 0 0 , 0 0 0 , w h i l e enabl i n g the operator
to d i s t i n g u i s h lan dmarks smaller than a football field from any
altitude ! Application today : ground-trai n i n g ai rmen for i n ter­
contin ental m i s s i o n s at great sav i n gs i n cost - u s i n g a s i n gl e
8-foot map to real i stically si mulate the r a d a r reflectivity and
sh adow effects o f a 4 , 000-mile mission. Application tomorrow :
trai n i n g the free worl d ' s first space explorer for h i s safe return
to earth .

I N P R O D U CT I O N TO DAY :
Pomona D i v i s i o n ' s A N / G P S -T4 Radar S i gnal S i m u l ator, a rela­
tively low-cost tra i n i n g system that i s simple, compact, flexible
and real i s t i c ; produc i n g synthetic target a n d I F F video i n for­
m a t i o n f o r radar d i s p l ay. C u r r e n t m i s s i o n : t r a i n i n g U S A F
ai rcraft controllers to d i rect i n terceptor m i s s i o n s . Ready
a d a p t ab i l i ty w i l l p e r m i t t h e T 4 t o s i m u l at e m i s s i l e s , r a d a r
s u rface targets a n d sonar targets.

I N T H E U . S . D E F E N S E I N V E N T O RY:
T h e U S A F ' s AN / A PQ - T l , A N / A P G - T I A , A N / A P Q - T 2 and
T2A, and A N / A P Q - T 3 s i m u l a t o r - t r a i n i n g s y s t e m s , p l u s the
U S N ' s 1 5WV-2 A E W /CIC trainer system - all developed a n d
m a n u factu r e d b y the P o m o n a D i vi s i o n and its p r e d e c e s s o r
comp a n i e s . Numerous sub-systems, up-dati n g modification kits,
adapters, spares, and a worldwi de field service section add to the
organ ization's total experi ence.

P OM O NA D I VISION o f T h e Marquardt Corporation i s an

able and experienced organization with demons trated capab i l i t y
i n advanced research , d e s i g n , development and producti o n . The
result i s reliable hardware which prov i des solutions to the prob­
lems o f trai n i n g m i l i tary operational personnel.
D rawi n g on the Corporati o n ' s overall management-engineer­
i n g skills, additional fac i l i t i e s and financial support, Pomona
Division now offers i ndustry a n d the Armed Forces a unique and
proven ability to get the j ob done-del i veri n g v i tal s i m u l ator­
tra i n i n g system both on-time and at minimum cost.

?rg!Li!o�1t
O PE R AT I O N S AT :
V a n N u ys a n d
Po m o n a , C a l iforn i a
Ogd e n , Uta h
POMONA DI VISIO

S U B S I D I A RY : 2709 N o . G a r e y A v e n u e , P o m o n a , C a l i f .
Cooper Deve l o p m e n t Engineers and scientists capable 0 1 contri b u t ing
Corporat i o n , t o advances in s t at e-ol- t h e-art are invited to write
M o n rov i a , C a l iforn i a to : Dr. Wendell B . Sell, Vice-Presid ent . G-5 .

© 1959 SCIENTIFIC AMERICAN, INC

WHAT' S N EW IN RA RE EA RTH S
Prompt shipments in tonnage quantities . . . attractive low prices . . . available as oxides,
hydrates, chlorides, oxalates, metals and other forms

a report by LINDSAY

New . . . new . . . new. It's the old but 99.99% can be shipped in surpris­ There are many new uses for the
still magic word that keynotes in­ i n g l y l a r g e q u a n t i t i e s . It m ay b e rare earths. This obviously is the re­
dustrial planning for the products n e w s t o y o u t h a t L i n d s a y i s cur­ sult of research and development
that may be big profit winners in rently producing more than 100 rare work carried on during recent years .
the 1960's . earth, yttrium and thorium s alts for In glass and ceramics. In electronics _
R&D work, in laboratories from R&D work as well as for normal pro­ In commercial nuclear energy. I n
c o a s t to c o a s t, is h a p p i l y b l e s s e d duction operations. plastics. In glass polishing. A n d in
with huge budgets for the ever­ There's new low pricing of rare many other fields .
continuing search for what is new. earths. Costs have tumbled sharply We would be modest indeed if
New in processes, in methods, in during recent years . Partly because we failed to hint that much of the
materials. Whatever is new that will of rapidly increasing demand. Partly rapid expansion - first in res earch
contribute to the development of because of vastly expanded produc­ and then in actual industrial use of
new products to captivate the pub­ tion facilities and improved tech­ the rare e arths - h a s b een a t our
lic fancy and meet the critical needs n i q u e s in r e f i n i n g o u r m a t e r i a l s . gentle urging.
of industry. ( Which came first - the hen o r the The facts speak for themselves.
New plastics. New exotic metals. egg?) The important fact is that the Rare earths have come of age. They
New chemicals . . . and among them, rare earths are priced so low as to are important production materials:
the rare earths. make their use extremely interest­ in a broad cross-section of American
ing. We are talking, in many cases, industry.
NOT N EW . . . B U T about ¢, not $. So if you are thinking new . . . new
You may say the rare earths are not ideas, new processes, new materials,.
new. True. We've b een working with new products . . . look at the rare
t h e r a r e e a rths for m o r e t h a n 5 0 earths .
years . M any industries have known When you do, please remember
and used them for years . But they these two facts . They are readily
may be new to you. available in the grades, varieties,
What is new in rare earths will forms and quantities you will need..
command your attention and in­ And they are priced at surprisingly
trigue t h e interest of your people Rare earths now available in metal attractive levels .
concerned with research and devel­ form. Interested? You can obtain Our technical people will b e
opment work. rare e arth a n d yttrium i n m e t a l h appy t o b e helpful t o you. W e can
T h e re's n e w a v a i l a b i lity in the form, primarily as ingots a n d lumps. s u p p l y p e r t i n e n t d a t a a n d w il l
rare earths ! Commercial grades can They are presently available in ex­ promptly supply detailed informa­
be shipped promptly by the ton or perimental quantities and offer in­ tion if you will indicate your specific
carload. High purity grades up to teresting promise to many industries. area of interest.

P L E A S E A D D R E S S I N Q U I R I E S TO

® UNDSAY CH E M ICAL D IVISION

American Potash & Chemical Corporation
264 ANN ST R E E T . WEST CH I CAGO, I L L I N O I S

OFFIC E S : CHICAGO • LOS ANGELES • SAN F R AN C I S C O • NEW YORK . P O R T L A N D (ORE.) • ATLANTA • COLUMBUS (0.) • S H R E V E P O R 'r

88

© 1959 SCIENTIFIC AMERICAN, INC

 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

230

19
Waves and Ship Wave Patterns

Ships generate a very distinctive wave pattern, also known as the Kelvin wave pattern
in honor of William Thomson’s (Lord Kelvin, * – †) efforts in developing a
suitable theory (Thomson , Lord Kelvin). This chapter discusses basic properties of
waves and the Kelvin wave pattern utilizing results of linear wave theory. This may serve
as a foundation for our more detailed discussion of linear wave theory in subsequent
chapters.

Learning Objectives

j At the end of this chapter students will be able to: j

• discuss basic wave descriptors like wave height, wave length, and wave period.

• understand the phenomenon of wave dispersion

• distinguish phase velocity and group velocity

• describe ship wave patterns

19.1 Wave Length, Period, and Height

Free surface Before discussing the ship wave pattern itself, we will review some basic properties
waves of water waves. The surface separating the liquid water from the gaseous air is free of
parallel shear stresses when the fluids are in equilibrium. In physics, this is called a free
surface. Changes in pressure at the free surface, e.g. due to wind or a moving object,
create waves, which subsequently propagate across the water surface.
Wave profile Cutting through the water surface parallel to the direction of propagation reveals the
wave elevation 𝜁, which measures the deviation from the calm water level 𝑧 = 0. The
wave profile will resemble the solid curve shown in Figure ..
Wave length and The horizontal distance between two wave crests marks the wave length 𝐿𝑤 . The
height maximum vertical distance between wave trough and wave crest defines the wave
height 𝐻. Note that the vertical scale is enlarged in Figure .. Typically waves have
steepness ratios of 𝐻∕𝐿𝑤 less than .. In theory, waves will break once they reach a

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

19.1 Wave Length, Period, and Height 231

Figure 19.1 Definition of wave length 𝐿𝑤 and wave height 𝐻; the vertical scale is exaggerated

Figure 19.2 Surface elevation of a harmonic, long-crested wave

wave steepness of 𝐻∕𝐿𝑤 = 1∕7, but usually break before they reach this limit (Michell,
).
In simulations, we often repeat the same wave profile over and over again: a regular Regular,
j j
wave. Natural waves are irregular and have to be described by probabilistic methods. long-crested
waves
Furthermore, the wave profile is assumed to be the same in every plane parallel to the
direction of propagation. The crests and troughs of this long-crested wave stretch to
infinity in the transverse direction. In linear wave theory, the actual wave profile is
approximated by a cosine function.

𝜁(𝑧, 𝑡) = 𝜁𝑎 cos(𝑘𝑥 − 𝜔𝑡) (.)

This simplified, harmonic wave is represented in Figure . by the dashed curve. Note
that the wave crest of the real wave profile is higher than the wave trough is deep. The
wave trough is wider than the wave crest.
In Equation (.), 𝜃 = 𝑘𝑥−𝜔𝑡 defines the phase of a wave. Figure . shows the wave Wave amplitude
elevation 𝜁(𝑥, 𝑡) for one cycle 𝜃 ∈ [0, 2𝜋] from wave crest to wave crest. Since the cosine
function is limited to values between ±1, the wave elevation takes values between ±𝜁𝑎 .
Thus, the total range of values for a harmonic wave is two times the wave amplitude 𝜁𝑎 .
The wave amplitude is always positive and defines the maximum deviation of the wave
elevation from its mean value 𝑧 = 0.
Wave crests (maxima of 𝜁) occur at phases equal to even multiples of 𝜋, i.e. 0, 2𝜋, 4𝜋, Wave crests and
etc.: troughs
max(𝜁(𝜃)) = +𝜁𝑎 for 𝜃 = 2𝑘𝜋 with k=,,,…
Wave troughs (minima of 𝜁) occur when the phase 𝜃 is equal to odd multiples of 𝜋. For
harmonic waves, the wave height 𝐻 is equal to twice the wave amplitude.

𝐻 = 2𝜁𝑎 (.)

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

232 19 Waves and Ship Wave Patterns

Figure 19.3 Recording of surface elevation of a harmonic, long-crested wave at a fixed position
(𝑥 = 0)

Wave period Let us consider a wave probe which records the free surface elevation 𝜁(𝑡) as a function
of time at the fixed position 𝑥 = 0 (Figure .). The time that passes between the
recording of one wave crest and the next subsequent wave crest is the wave period 𝑇 .
Wave frequency The completion of a single cycle during one period 𝑇 requires a change in phase of Δ𝜃
equal to two 𝜋.
Δ𝜃 = 2𝜋 = (𝑘𝑥 − 𝜔𝑡)(𝑥=0,𝑡=𝑇 ) = − 𝜔𝑇
In other words, the wave frequency 𝜔 is reciprocal to the wave period 𝑇 .

2𝜋 2𝜋
𝜔 = 𝑇 = (.)
𝑇 𝜔
j As an angular frequency, the wave frequency measures how much of a single wave cycle j
is happening in 2𝜋 seconds. Its physical unit is [/s], however, we often denote it as
[rad/s] to distinguish it from a frequency measured in Hz = /s. A wave frequency of
𝜔 = 1 rad/s means a cycle is completed in 𝑇 = 2𝜋 s or approximately . seconds.
With a wave frequency of 𝜔 = 0.5 rad/s, only half of the cycle is completed within
𝜋 ≈ 6.283 seconds. The period of this wave would be 𝑇 = 4𝜋 s = 12.566 seconds.
Typical wave frequencies for ocean and ship generated waves are in the range of 0.1 rad/s
to 2.0 rad/s.
Wave length If we take a photograph with a very short exposure time, the wave is captured as a
function of space 𝑥. For convenience, we choose the time 𝑡 = 0. Cutting through the
wave in direction of wave progression 𝑥 reveals the wave length 𝐿𝑤 (see Figure .). If
the wave length is long, the wave is called long. As we will show later, long waves also
have a high wave period (small frequency). Short waves have small wave periods and
high wave frequencies.

Figure 19.4 Spatial extension of surface elevation of a linear, harmonic, long-crested wave
captured at time (𝑡 = 0)

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

19.2 Fundamental Observations 233

Figure 19.5 A snapshot of the wave elevation in a wave group

At any selected point in time 𝑡, the change in phase of Δ𝜃 from one wave crest to the Wave number
other is still equal to 2𝜋.

Δ𝜃 = 2𝜋 = (𝑘𝑥 − 𝜔𝑡)(𝑥=𝐿𝑤 ,𝑡=0) = 𝑘 𝐿𝑤

The wave number 𝑘 is the spatial equivalent to the wave frequency. Wave number and
wave length are reciprocal.
2𝜋 2𝜋
𝑘 = 𝐿𝑤 = (.)
𝐿𝑤 𝑘
The wave number expresses how much of the wave length fits into a distance of 2𝜋 meter.
Wave numbers are noted in the unit /m. Some prefer the unit rad/m to indicate the
j factor 2𝜋. A wave frequency of 𝑘 = 1 rad/m is equivalent to a wave length of 𝐿𝑤 = j
2𝜋 m ≈ 6.283 meter. With a wave number of 𝑘 = 0.1 rad/m, a wave is approximately
. meters long and a tenth of it fits into 2𝜋 meters. Since a wave length is always a
positive distance, wave numbers are positive as well.
A wave travels one wave length during one wave period. Therefore, the speed of Phase velocity
propagation of a water wave is
𝐿𝑤 𝜔
𝑐 = = (.)
𝑇 𝑘
If you move in lockstep with a wave crest or any other point along the wave, you have
to move with velocity 𝑐. Since you will have a constant phase relationship to the wave,
the velocity of wave propagation is known as phase velocity.

19.2 Fundamental Observations

When you throw a pebble into a calm pond, a group of waves is created. Figure . Wave groups
shows the vertical displacement of the water surface 𝜁 along the wave group. As the
waves propagate and expand in the form of concentric circles, three observations can
be made:

(i) Long waves propagate faster than short waves, i.e. if 𝐿𝑤1 > 𝐿𝑤2 then 𝑐1 > 𝑐2 .
(ii) At the front of the wave group, wave amplitude diminishes until the wave vanishes,
whereas at the end of the wave group, waves seem to emerge from nothing.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

234 19 Waves and Ship Wave Patterns

(iii) The waves seem to travel from the back to the front of the group, while the envelope
of the wave group moves slower than the individual waves.

Wave dispersion The first observation (i) is known as wave dispersion. Sound waves of varying frequency
all travel with a constant speed, i.e. the speed of sound. Electromagnetic waves of
different frequency propagate with the speed of light. In contrast, water waves with small
wave frequency (long waves) travel faster than short waves with high wave frequency.
Very long waves, like the ones created by tsunamis, may cross whole oceans in a few
hours. The phase velocity is also influenced by the water depth ℎ. We will derive
the exact relationship between wave frequency, water depth, and phase velocity in
Section ..
Wave energy Observation (ii) is related to the energy transport in waves. Chapter  presents a
detailed discussion of wave energy and wave propagation. For now, it may suffice to
state that waves contain two types of energy:

• kinetic energy – water particles underneath a wave move in elliptical paths. Since
water particles have mass and velocity, they possess kinetic energy.
• potential energy – the up and down of the free surface contains potential en-
ergy comparable to the energy of a pendulum. However, potential wave energy
propagates with the wave.

The water is initially at rest in front of the wave group. For the water particles to assume
j j
the motion characteristic to waves, potential energy is converted into kinetic energy at
the front of the wave. As a consequence, the wave amplitude diminishes. At the back of
the wave group, kinetic energy is converted into potential energy and waves emerge.
Once all kinetic energy has been converted into potential energy and transported away,
the water is at rest again.
Viscous effects Viscosity of the water will dissipate some wave energy during this process. This is
especially true for waves of very short length (< 0.1 m), so-called capillary waves. They
are also affected by surface tension. Ship and ocean waves have longer wave lengths,
and the effects of surface tension and viscosity may be neglected for many engineering
applications.
Group velocity The observations (ii) and (iii) are connected. Individual waves travel with their respective
phase velocity 𝑐. As the wave group approaches, water particles are set in motion, but
more or less come to rest in their initial position once the wave group has passed. This
means that only the potential energy propagates. As we will discuss in Chapter , wave
energy travels at the group velocity 𝑐𝐺 . On deep water it is just half of the phase velocity.
The envelope of a wave group reflects the reduced speed of the energy transport. For
that reason, you will see waves emerge from the back of the wave group, travel through
the group to the front, and finally diminish because their energy is used to set water
particles in motion. The relationship between group and phase velocity affects the
formation of the Kelvin wave pattern, which we will discuss in the following section.
Wave Next time you pass a quiet pond, throw two pebbles into the water a few yards apart.
superposition Watch how the two systems of radiated circular waves spread and interact. For the most
part, they will just pass each other and continue on undisturbed. A notable change will
occur only if the superposition of waves causes wave breaking. You may notice that

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

19.3 Kelvin Wave Pattern 235

Figure 19.6 Kelvin wave pattern in deep water

the waves appear to be circular after a short distance, no matter what the shapes of the
pebbles were.

j j
19.3 Kelvin Wave Pattern
Anything moving at or close enough to the water surface creates waves: You may drag
a stick through the water, watch a swan glide across a pond, or observe a koi swimming
just underneath the water surface. In all cases, you will observe the wave pattern that is
characteristic to all ships traveling the world’s oceans, lakes, and rivers.
Thomson (Lord Kelvin) studied the wave pattern created by a traveling point distur-
bance (Figure .). The solid curves represent lines of constant phase, like the position
of wave crests. The parametric equations in Figure . have been used to create the plot.
They represent an approximation of the rather complicated mathematics behind this
problem. From Figure . and observation, the following features may be discerned
for wave patterns:

• Waves appear only in a triangular region behind the vessel. Wave elevations Kelvin angle
rapidly diminish outside of the triangle. In deep water, the interior angle at the
tip is . degrees and is known as the Kelvin angle.
• The Kelvin wave pattern consists of two types of waves: divergent waves and Kelvin wave
transverse waves. We also call the divergent waves ‘diagonal waves’ and the pattern
transverse waves ‘following waves.’ Divergent waves appear at the sides of the
wave pattern and travel away from the ship’s path. Transverse waves follow the
ship with their crests and troughs aligned perpendicular to the ship’s course.
• Observed from a moving boat, the wave pattern appears to be stationary, which Waves follow the
means that the waves themselves propagate with ship speed in the direction of ship
the ship’s forward motion.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

236 19 Waves and Ship Wave Patterns

Figure 19.7 Change of Kelvin wave pattern with increasing velocity on deep water

Ship speed • The wave length 𝐿𝑤 depends on the ship’s speed and, in some cases, also on the
determines wave water depth (Figure .). The faster a vessel sails, the longer its waves become.
length
Wave breaking • Especially at higher speeds or with blunt bows, waves near the bow break, like
the waves rolling onto a beach.

Linear wave theory provides a relationship for wave length 𝐿𝑤 and wave phase velocity 𝑐.

𝑐2
𝐿𝑤 = 2𝜋 (.)
𝑔 tanh(𝑘ℎ)
On deep water, the product of wave number 𝑘 and water depth ℎ is larger than 𝜋, i.e.
j 𝑘ℎ > 𝜋 and tanh(𝑘 ℎ) ≈ 1. The transverse waves progress with the same speed as the j
disturbance, i.e. 𝑐 = 𝑣𝑆 or

𝑐 2 = 𝑣2𝑆 = 𝐹𝑟2 𝑔 𝐿𝑊𝐿 (.)

We substitute Equation (.) into (.) and obtain a formula which provides the length
of transverse waves as a function of Froude number for deep water.

𝐿𝑤 = 2𝜋 𝐹𝑟2 𝐿𝑊𝐿 for 𝑘ℎ > 𝜋 (.)

The photo in Figure . may exemplify that Lord Kelvin’s theoretical model is a pretty
good match to nature. The satellite photo shows a Kelvin wave pattern visualized by
clouds in the wake of Amsterdam Island in the Indian Ocean. The volcano on the island
creates a Kelvin pattern at the boundary of two layers of air with different temperatures.
As warm, moist air is pushed up into cooler air, water condenses and clouds form (wave
crests). When the air sinks, it warms up and the clouds dissolve again (wave troughs).
Ship wave A ship disturbs the water surface over its whole length rather than just at a single point.
pattern Figure . shows the simulation of the wave pattern behind a ship hull. The specific
shape of the vessel influences the height of the waves and the details of the wave pattern.
The basic pattern shape, however, does not change as long as the vessel is sailing on
deep water and is carried by its buoyancy. The nature of the wave pattern changes if
the vessel maneuvers from displacement to planing mode. A planing vessel no longer
produces a Kelvin wave pattern.
Wave Once waves have been generated by a moving vessel, they become independent of the
propagation vessel. When the vessel changes its course, newly generated waves will be oriented

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

19.3 Kelvin Wave Pattern 237

Figure 19.8 Kelvin wave pattern like cloud formation in the slipstream of Amsterdam Island in
j the southern Indian Ocean. Photo courtesy of NASA Earth Observatory j

according to the new course. The waves generated before the course change will
continue on their original path. If the vessel stops, wave generation ceases, but the
already existing waves continue to spread.
Waves contain and transport energy. Therefore, the creation of waves requires energy. Wave resistance
Wave energy grows quadratic with wave amplitude. Provided two waves have the same
wave length, a wave with twice the wave height contains four times the energy. In calm
water, the moving ship is the only possible source of energy for wave making. Kinetic
energy of the ship is constantly converted into wave energy. The loss of energy for the
ship results in its wave resistance. As naval architects we aim to minimize the wave
height in the wave pattern to minimize the loss of energy (see Chapter ).
If the water depth is less than half the wave length, the ocean bottom or river bed will Shallow water
influence the behavior of the waves. A first visible sign is a widening of the Kelvin effect
angle. Unfortunately, shallow water effects also result in increased wave resistance
(Schlichting, ).

References
Michell, J. (). The hightest wave in water. Philosophical Magazine Series , :–
.
Schlichting, O. (). Schiffswiderstand auf beschränkter Wassertiefe; Widerstand von

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

238 19 Waves and Ship Wave Patterns

Figure 19.9 Wave pattern of a ship at 𝐹𝑟 = 0.26

j j
Seeschiffen auf flachem Wasser. In Jahrbuch der Schiffbautechnischen Gesellschaft
(STG), volume , pages –. Springer Verlag, Berlin.
Thomson (Lord Kelvin), W. (). On ship waves, volume  of Cambridge Library Col-
lection – Physical Sciences, pages –. Cambridge University Press, Cambridge,
United Kingdom. First presented in .

Self Study Problems

. Define wave height and wave length in your own words.
. How does a real wave profile deviate from a harmonic signal (cosine function)?

. Which geometric properties of a Kelvin wave pattern change when the ship moves
into less deep water while maintaining its speed?
. Compute the wave length for a ship sailing with velocity 𝑣𝑆 = 24 kn?
. In deep water, a ship creates a wave pattern with a basic wave length of 𝐿𝑤 =
50.0 m. How fast is the ship? Compute the wave length and phase velocity of the
divergent waves!

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

239

20
Wave Theory

Waves are fascinating objects. At times, they demurely lap on a beach and at other
times they cause havoc and destruction. In this chapter we develop a mathematical
model for waves that will enable us to study many of their properties. Because waves
independently propagate from their source, we deliberately ignore how they are created.
It is helpful to first study the behavior of waves alone.

Learning Objectives
At the end of this chapter students will be able to:
j j
• formulate wave flow as a boundary value problem

• derive free surface boundary conditions

• interpret the physical meaning of boundary conditions in wave theory

20.1 Overview
Attempts to capture the physics of waves in mathematical models go way back in time. History
Craik () provides a discussion of the development of water wave theory over the
past three centuries. Many famous scientists and mathematicians contributed to the
field: Newton, Lagrange, Laplace, Green, Cauchy, and Poisson to name a few.
The mathematics of the problem are quite difficult because we deal with nonlinear Stokes and Airy
partial differential equations. Different theories have been developed for general, as wave theories
well as specialized, applications. The most commonly used were developed by Stokes
() who provided approximations of first, second, and higher order for the original
nonlinear problem. The approximations are known as Stokes’ wave theories of first,
second, third, and fifth order. We encountered Sir George Gabriel Stokes already in
the context of the Navier-Stokes equations. The first order, or linear solution, is also
known as Airy wave theory because Airy published his work a few years before Stokes
did (Airy, ). Sir George Biddell Airy (*–†), an English mathematician
and astronomer, became Astronomer Royal and established the prime meridian in
Greenwich, UK, which is still in use as the reference line for longitude.

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

240 20 Wave Theory

j Figure 20.1 Definition of coordinate system and domain boundaries for wave theory of j
long-crested waves

Linear wave Linear wave theory is one of the most important tools in seakeeping analyis and prob-
theory abilistic prediction of the short and long term behavior of marine vessels in waves.
Although it is limited to waves with a small wave height to wave length ratio 𝐻∕𝐿𝑤 ≪ 1,
linear wave theory provides important insight into the behavior of waves and the wave
making of a ship. Therefore, we will discuss linear wave theory in some detail.
Wave resistance The Australian mathematician Michell () developed a model for wave resistance
based on linear theory. It can be used to estimate wave resistance of fast, slender vessels,
especially multihull vessels. More on this in Section ..

20.2 Mathematical Model for Long-crested Waves

Long-crested In order to develop a two-dimensional mathematical model for waves, we assume that
waves their crests and troughs stretch in 𝑦-direction from negative infinity to positive infinity
(Figure .). These are commonly known as long-crested waves. This simplification
renders all flow properties independent of the 𝑦-coordinate and restricts wave flow
modeling to the 𝑥-𝑧-plane. The 𝑥-axis points in the direction of wave propagation and
the 𝑧-axis is pointing upward. We place the origin at the calm water level (𝑧 = 0). As
a consequence, points below the calm water level, i.e. in the water, will have negative
𝑧-coordinates (𝑧 < 0).

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

20.2 Mathematical Model for Long-crested Waves 241

Figure 20.2 Simplified two-dimensional fluid domain for long-crested waves

Figure . shows the simplified two-dimensional flow domain. The fluid domain 𝑉
is bordered at the top by the free surface 𝑆𝐹 . The ocean bottom 𝑆𝐵 forms the lower
boundary of the domain. We assume that the ocean bottom is parallel to the calm water
surface, i.e. the water depth ℎ is constant throughout the domain. Note that the water
depth ℎ is defined as a positive length.
As noted before, viscous effects are small for water waves of lengths typically created by Ignoring viscous
a moving ship. However, a notable amount of energy is dissipated in the case of breaking effects
j waves. If we exclude the phenomenon of wave breaking, the fluid flow in waves may be j
treated as inviscid for practical applications. Again, water is considered incompressible
(𝜌 = const.). In addition, we assume that the flow is irrotational (rot 𝑣 = 0), which
enables us to employ potential theory.
The Laplace equation and the Bernoulli equation represent the conservation of mass Conservation of
and conservation of momentum principles of fluid mechanics for potential flow. mass and
momentum
𝚫𝜙 = 0 Laplace eq. (.)
𝜕𝜙 𝜌
𝜌 + |∇ 𝜙|2 + 𝜌 𝑔 𝑧 + 𝑝 = 𝐶(𝑡) Bernoulli eq. for pot. flow (.)
𝜕𝑡 2
The wave flow is represented by its velocity potential 𝜙, which in general is a function
of space 𝑥 and time 𝑡. In order to define a unique solution of the Laplace equation,
boundary conditions are imposed at the borders of the flow region 𝑉 .
The domain 𝑉 is bordered by the ocean bottom 𝑆𝐵 and the free surface 𝑆𝐹 (see Fig- Domain
ure .). From our discussion of the two-dimensional flow around a cylinder in boundaries
Chapter , we know that an additional condition for the far field boundary 𝑆∞ at the
inlet and outlet of the flow is also required.

20.2.1 Ocean bottom boundary condition

Let us start with a simple boundary condition. Like the cylinder surface, the ocean No flow through
bottom 𝑆𝐵 is considered a fixed, impenetrable surface. Since we ignore friction, fluid ocean bottom

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

242 20 Wave Theory

motion parallel to the bottom is allowed. However, the normal velocity has to vanish
over 𝑆𝐵 .
𝜕𝜙
= 0 on 𝑆𝐵 (.)
𝜕𝑛
Since the ocean floor is parallel to the water surface, its outward oriented normal
vector points in the negative 𝑧-direction, i.e. 𝑛𝑇 = (0, 0, −1). Consequently, the normal
derivative is equal to the derivative in the negative 𝑧-direction.

𝜕𝜙 ( ) ⎛ 𝜙𝑥 ⎞ 𝜕𝜙
= 𝑛𝑇 ∇ 𝜙 = 0, 0, −1 ⎜ 𝜙𝑦 ⎟ = −
𝜕𝑛 ⎜ ⎟ 𝜕𝑧
⎝ 𝜙𝑧 ⎠
As a means to shorten the notation, we represent the partial derivatives with subscripts
𝜕𝜙 𝜕𝜙 𝜕𝜙
𝜙𝑥 = 𝜙𝑦 = 𝜙𝑧 = (.)
𝜕𝑥 𝜕𝑦 𝜕𝑧

The minus sign for the 𝑧 derivative is of no consequence, since we require the vertical
velocity component to vanish. Therefore, our final ocean bottom boundary condition
reads
𝜕𝜙
𝜙𝑧 = = 0 for 𝑥 on 𝑆𝐵 , 𝑧 = −ℎ (.)
𝜕𝑧
Physically, it is a kinematic boundary condition because it restricts how the water may
j move. Mathematically it is classified as a Neumann boundary condition because it j
applies to a derivative of the unknown potential 𝜙.

20.2.2 Free surface boundary conditions

The ocean bottom 𝑆𝐵 is easily described mathematically: 𝑆𝐵 consists of all points with
𝑧 = −ℎ. Similarly, the calm water surface is formed by points 𝑥 = (𝑥, 𝑦, 0). However,
the position of the water surface is a priori unknown when waves are present. The shape
of the free surface 𝑆𝐹 has to be computed as part of the solution of the flow problem.
As a consequence the boundary conditions for the free surface have to be satisfied at a
yet unknown surface 𝑆𝐹 .
Mathematical As a first step, a mathematical model is introduced for the shape of the free surface.
model for free We employ an implicit function as you may have seen in the definition of a circle of
surface
radius 𝑅: 𝑥2 + 𝑦2 − 𝑅2 = 0. The contour is formed by all points whose distance from
the origin is equal to 𝑅. The implicit function for the free surface captures all points
whose vertical distance from the calm water level (𝑧 = 0) is equal to the free surface
elevation 𝜁.
𝐹 (𝑥, 𝑧, 𝑡) = 𝑧 − 𝜁(𝑥, 𝑡) = 0 (.)
We call 𝜁 the wave elevation. Because we restricted the problem to long-crested waves,
the implicit function 𝐹 and the wave elevation 𝜁 do not depend on the transverse
coordinate 𝑦.
Limitations of The formulation (.) implies that 𝐹 and 𝜁 are analytic functions. In that case, a unique
free surface 𝜁 value must exist for all combinations (𝑥, 𝑡) which automatically excludes overturning
model
or breaking waves. As indicated in Figure ., in cases of wave breaking more than one

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

20.2 Mathematical Model for Long-crested Waves 243

Figure 20.3 The mathematical free surface model is valid for nonbreaking waves only

𝜁 value might occur for a pair (𝑥, 𝑡). As a consequence the function 𝑧 = 𝜁(𝑥, 𝑡) is not
unique and does not exist. For the discussion of breaking waves, other mathematical
models have to be employed.
At the ocean bottom, we specify a single boundary condition because its geometry is Two boundary
j j
known. The only unknown is the velocity potential. For the free surface we need two conditions
needed
boundary conditions because we have two unknowns: the potential 𝜙 and the wave
elevation 𝜁. The two free surface boundary conditions will be discussed in the following
subsections.

Kinematic free surface boundary condition

The free surface must be a stream surface at all times 𝑡. Consequently, no fluid particle
may cross the stream surface 𝑆𝐹 . Like in the case of the body boundary condition
for the cylinder flow, we restrict the movement of the fluid, therefore this will be a
kinematic free surface boundary condition. The explicit expression for the kinematic
boundary condition at the free surface may be derived in two different ways: (a) using a
physical argument or (b) a more mathematical argument.
Let us start with the longer but probably more comprehensible physical argument. If Physical
𝑆𝐹 is a stream surface, fluid particles at the surface may move tangential to the surface argument
but not normal to it. Therefore, the relative normal velocity between surface and water
particles must vanish. The velocity of a water particle is 𝑣 = ∇ 𝜙. The surface moves up
and down with local velocity
𝜕𝜁
𝑣𝑆 = 𝑘
𝐹 𝜕𝑡
where 𝑘 = (0, 0, 1)𝑇 is the unit vector in vertical direction. Thus, the relative velocity is

𝜕𝜁
𝑣 − 𝑣𝑆 = ∇𝜙 − 𝑘 (.)
𝐹 𝜕𝑡

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

244 20 Wave Theory

According to our stream surface requirement, the component of the relative velocity
in normal direction must vanish. The latter is obtained by forming the dot product of
normal vector 𝑛 and relative velocity:
( )
𝜕𝜁
𝑇
𝑛 ∇𝜙 − 𝑘 = 0 (.)
𝜕𝑡

Normal vector For implicit surfaces, the normal vector 𝑛 is identical to the gradient normalized to unit
length.
∇𝐹
𝑛 = (.)
|∇ 𝐹 |
This relationship follows from the characteristic of gradients to always point into the
direction of steepest ascent, which in turn is perpendicular to the lines of constant
function value.
We substitute the expression (.) into Equation (.) and multiply it with the length
|∇ 𝐹 | of the gradient:
( )
( )𝑇 𝜕𝜁
∇𝐹 ∇𝜙 − 𝑘 = 0
𝜕𝑡
Normal relative Expanding the dot products yields
velocity must
vanish ⎛ 𝜕𝜙 ⎞
⎜ 𝜕𝑥 ⎟
j ( )⎜ ⎟ j
𝜕𝐹 𝜕𝐹 𝜕𝐹 ⎜
, ,
𝜕𝜙 ⎟ − 𝜕𝜁 𝜕𝐹 = 0
𝜕𝑥 𝜕𝑦 𝜕𝑧 ⎜ 𝜕𝑦 ⎟ 𝜕𝑡 𝜕𝑧
⎜ 𝜕𝜙 ⎟
⎜ ⎟
⎝ 𝜕𝑧 ⎠

and we get

𝜕𝐹 𝜕𝜙 𝜕𝐹 𝜕𝜙 𝜕𝐹 𝜕𝜙 𝜕𝜁 𝜕𝐹
+ + − = 0 (.)
𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑧 𝜕𝑧 𝜕𝑡 𝜕𝑧

The partial derivatives of the implicit surface function 𝐹 = 𝑧 − 𝜁(𝑥, 𝑡) are

𝜕𝐹 𝜕𝜁 𝜕𝐹 𝜕𝜁
= − = −
𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦
(.)
𝜕𝐹 𝜕𝐹 𝜕𝜁
= +1 = −
𝜕𝑧 𝜕𝑡 𝜕𝑡
Finally, we substitute the derivatives (.) into the condition (.) that the normal
component of the relative velocity between water particles and free surface vanishes.
3D kinematic free For a three-dimensional wave field 𝐹 (𝑥, 𝑦, 𝑧, 𝑡) = 𝑧 − 𝜁(𝑥, 𝑦, 𝑡) = 0, the following is
surface condition obtained for the kinematic free surface condition:
𝜕𝜁 𝜕𝜁 𝜕𝜙 𝜕𝜁 𝜕𝜙 𝜕𝜙
− − − + = 0 on 𝑆𝐹 with 𝑧 = 𝜁(𝑥, 𝑦, 𝑡) (.)
𝜕𝑡 𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑧

2D kinematic free In the case of long-crested waves, the transverse derivatives 𝜕∕𝜕𝑦 vanish, and the kine-
surface condition

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

20.2 Mathematical Model for Long-crested Waves 245

matic boundary condition for long-crested waves requires that the following condition
is satisfied at the free surface.
𝜕𝜁 𝜕𝜁 𝜕𝜙 𝜕𝜙
− − + = 0 on 𝑆𝐹 with 𝑧 = 𝜁(𝑥, 𝑡) (.)
𝜕𝑡 𝜕𝑥 𝜕𝑥 𝜕𝑧

The equivalent mathematical argument will of course lead to the same kinematic bound- Mathematical
ary condition: The condition 𝐹 = 𝑧 − 𝜁(𝑥, 𝑡) = 0 must be valid at all times in order for argument
𝐹 to accurately describe the free surface 𝑆𝐹 . Consequently, its value 𝐹 = 0 cannot
change over time, and its total derivative with respect to time must vanish. Applying
the substantial derivative, fluid mechanics’ equivalent of the total derivative, to 𝐹 yields
in vector form:
D𝐹 𝜕𝐹
= + 𝑣𝑇 ∇ 𝐹 = 0
D𝑡 𝜕𝑡
and in components
𝜕𝐹 𝜕𝐹 𝜕𝐹 𝜕𝐹
+𝑢 +𝑣 +𝑤 = 0 (.)
𝜕𝑡 𝜕𝑥 𝜕𝑦 𝜕𝑧
We rewrite the latter equation by replacing the partial derivatives of 𝐹 with the results
from Equation (.) and by substituting partial derivatives of the potential 𝜙 for the
components 𝑢, 𝑣, and 𝑤 of the velocity vector. As expected, the result is the same as
Equation (.)
𝜕𝜁 𝜕𝜙 𝜕𝜁 𝜕𝜙 𝜕𝜁 𝜕𝜙
− − − + 1 = 0 (.)
j 𝜕𝑡 𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑧 j
⏟⏞⏞⏞⏞⏞⏞⏞⏞⏞⏟⏞⏞⏞⏞⏞⏞⏞⏞⏞⏟
nonlinear terms
In long-crested waves, the transverse velocity 𝑣 = 0 vanishes and we obtain again
Equation (.) as the kinematic boundary condition for the free surface.
Unfortunately, the resulting kinematic free surface conditions (.) and (.) are Nonlinear and
nonlinear because they contain products of the derivatives of the unknown surface implicit
elevation 𝜁 and the velocity potential 𝜙. As an added difficulty, the condition has to be
satisfied at a yet unknown location 𝑧 = 𝜁(𝑥, 𝑡).

Dynamic free surface boundary condition

As mentioned above, a second boundary condition is needed because we do not know

the actual location of the free surface. This second condition will be employed to
eliminate the unknown surface elevation 𝜁 from the kinematic boundary condition.
Besides kinematic boundary conditions that impose restrictions on how the fluid can Physical
move, we can also prescribe dynamic conditions which define how forces interact or do argument
not interact with the fluid. We are interested in the behavior of free waves, i.e. waves
that have been created by wind or a ship but are now progressing on their own. That
means, we will explicitly ignore the interaction of waves with the atmosphere. Pressure
forces on the surface vanish if pressure differences at the water surface vanish.
To cast this into a mathematical model we recall Bernoulli’s equation for potential flow: Mathematical
model
 Admittedly, this is a simplification which works for the linearized equations only. The exact boundary

conditions form a system of coupled equations that has to be solved simultaneously.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

246 20 Wave Theory

𝜕𝜙 1( )𝑇 𝑝
+ ∇ 𝜙 ∇ 𝜙 + 𝑔𝑧 + = 𝐶(𝑡) (.)
𝜕𝑡 2 𝜌
The fluid is at rest far away from the progressing waves, and the free surface is exposed
to atmospheric pressure 𝑝0 . This forms our constant for the right-hand side 𝐶(𝑡) = 𝑝0 ∕𝜌.
𝜕𝜙 1( )𝑇 𝑝 𝑝
+ ∇ 𝜙 ∇ 𝜙 + 𝑔𝑧 + = 0 at 𝑆𝐹 (.)
𝜕𝑡 2 𝜌 𝜌
The dynamic boundary condition requests that the pressure difference 𝑝0 − 𝑝 vanishes
at the free surface:
𝜕𝜙 1( )𝑇
+ ∇ 𝜙 ∇ 𝜙 + 𝑔𝑧 = 0 at 𝑆𝐹 with 𝑧 = 𝜁(𝑥, 𝑦, 𝑡) (.)
𝜕𝑡 2
Note that all terms in Equation (.) have to be evaluated at 𝑧 = 𝜁(𝑥, 𝑦, 𝑡). Thus,
solving Equation (.) for the unknown wave elevation does not provide a means to
immediately calculate 𝜁:

1 𝜕𝜙 1 ( )𝑇
𝜁(𝑥, 𝑦, 𝑡) = − − ∇ 𝜙 ∇ 𝜙 at 𝑆𝐹 with 𝑧 = 𝜁(𝑥, 𝑦, 𝑡) (.)
𝑔 𝜕𝑡 2𝑔
The wave elevation is also part of the right-hand side in (.) as argument of potential
𝜙.
Nonlinear Expansion of the dot product of the gradients of the potential reveals that the dynamic
j dynamic free free surface boundary condition is nonlinear, like the kinematic free surface boundary j
surface condition
condition.
[( ) ( ) ( ) ]
2 2 2
𝜕𝜙 1 𝜕𝜙 𝜕𝜙 𝜕𝜙
+ + + + 𝑔𝑧 = 0 at 𝑆𝐹 with 𝑧 = 𝜁(𝑥, 𝑦, 𝑡) (.)
𝜕𝑡 2 𝜕𝑥 𝜕𝑦 𝜕𝑧
⏟⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏟⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏟
nonlinear terms

The second nonlinear term vanishes for the two-dimensional problem of long-crested
waves and the dynamic free surface boundary condition reads:
[( ) ( ) ]
2 2
𝜕𝜙 1 𝜕𝜙 𝜕𝜙
+ + + 𝑔𝑧 = 0 at 𝑆𝐹 with 𝑧 = 𝜁(𝑥, 𝑡) (.)
𝜕𝑡 2 𝜕𝑥 𝜕𝑧
⏟⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏟⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏟
nonlinear terms

20.2.3 Far field condition

Radiation We have considered boundary conditions for the ocean bottom 𝑆𝐵 and the free surface
condition 𝑆𝐹 . What remains to be specified are conditions for the inlet and outlet surfaces 𝑆∞
in the far field. In the case of waves, the far field condition is often referred to as the
radiation condition.
Throw a stone A radiation condition commonly involves statements about the direction in which
into a pond waves are traveling and how their amplitude develops. Observe the waves created
after you throw a stone into a quiet pond: a set of circular waves emerges which travel

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

20.2 Mathematical Model for Long-crested Waves 247

away from the spot where the stone hit the water surface. The amplitudes of the waves
diminish as the rings expand until they finally fade away. In modeling this wave system,
the radiation condition would have to enforce the direction of wave propagation (away
from the stone) and that the amplitudes of waves vanish far away from their origin.
Friction forces do not exist in an inviscid fluid. Therefore, energy contained in the wave Radiation
flow cannot be dissipated into heat or eddies. In long-crested waves, no forces act in condition for
long-crested
the 𝑦-direction and thus energy cannot be distributed along the wave crests (parallel waves
to the 𝑦-axis). As a consequence, the wave height cannot vanish in the far field of our
two-dimensional domain. The only far field condition we impose is the requirement
that the waves progress with constant velocity 𝑐 in positive 𝑥-direction.

𝑐 = const. ≥ 0 at 𝑆∞ for |𝑥| → ∞ (.)

20.2.4 Nonlinear boundary value problem

Analogous to the cylinder flow problem, the Laplace equation together with the bound- 3D boundary
ary conditions form a boundary value problem whose solution contains the velocity value problem
potential 𝜙 and the wave elevation 𝜁.

𝚫𝜙 = 0 𝑥 = (𝑥, 𝑦, 𝑧)𝑇 ∈ 𝑉
𝜕𝜁 𝜕𝜁 𝜕𝜙 𝜕𝜁 𝜕𝜙 𝜕𝜙
− − − + = 0 𝑥 on 𝑆𝐹 , 𝑧 = 𝜁(𝑥, 𝑦, 𝑡)
j 𝜕𝑡 𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑧 j
𝜕𝜙 1
+ ||∇ 𝜙|| + 𝑔𝑧 = 0 (.)
2
𝑥 on 𝑆𝐹 , 𝑧 = 𝜁(𝑥, 𝑦, 𝑡)
𝜕𝑡 2
𝜕𝜙
= 0 𝑥 on 𝑆𝐵 , 𝑧 = −ℎ
𝜕𝑧
propagation, energy conservation on 𝑆∞

This reduces to the following set of equations for long-crested waves in the two- 2D boundary
dimensional case. value problem

𝚫𝜙 = 0 𝑥 = (𝑥, 𝑧)𝑇 ∈ 𝑉
𝜕𝜁 𝜕𝜁 𝜕𝜙 𝜕𝜙
− − + = 0 𝑥 on 𝑆𝐹 , 𝑧 = 𝜁(𝑥, 𝑡)
𝜕𝑡 𝜕𝑥 𝜕𝑥 𝜕𝑧
𝜕𝜙 1
+ ||∇ 𝜙|| + 𝑔𝑧 = 0 (.)
2
𝑥 on 𝑆𝐹 , 𝑧 = 𝜁(𝑥, 𝑡)
𝜕𝑡 2
𝜕𝜙
= 0 𝑥 on 𝑆𝐵 , 𝑧 = −ℎ
𝜕𝑧
𝑐 = const. ≥ 0 at 𝑆∞ for |𝑥| → ∞

A general analytic solution is not known for the exact two- and three-dimensional
boundary value problems because of the nonlinear free surface boundary conditions.
However, approximate solutions can be derived by using a Fourier series expansion
or similar methods. If required, applications use the Stokes’ wave theories with ap-
proximations from first to fifth order. Most important for practical applications is the
first order wave theory, which solves the linearized problem as outlined below and in
subsequent chapters.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

248 20 Wave Theory

20.3 Linearized Boundary Value Problem

Linearized The nonlinear free surface boundary condition can be linearized if the analysis is
boundary value restricted to waves with very small steepness 𝐻∕𝐿 ≪ 1. The linearization process is
problem
based on a perturbation series expansion of the exact solution. A detailed description
of the process is provided in the following chapter. For now we simply study the result
of the linearization and point out the major differences to the exact boundary value
problem.
Linearized Obviously, the Laplace equation (.) and boundary conditions at the sea bottom
boundary value 𝑆𝐵 (.) and the far field 𝑆∞ (.) are unchanged. As expected, the nonlinear
problem
terms have vanished on the left-hand side of the kinematic and dynamic free surface
conditions (.) and (.). In addition, however, the two free surface conditions
are now imposed on the calm water surface 𝑧 = 0 instead of on the actual free surface
𝑆𝐹 . This is a major simplification enabling an analytic solution which will be derived in
Chapter  on linear wave theory.

𝚫𝜙 = 0 𝑥 = (𝑥, 𝑧)𝑇 ∈ 𝑉 (.)

𝜕𝜁 𝜕𝜙
− + = 0 𝑥 with 𝑧 = 0 (.)
𝜕𝑡 𝜕𝑧
𝜕𝜙
+ 𝑔𝜁 = 0 𝑥 with 𝑧 = 0 (.)
𝜕𝑡
𝜕𝜙
= 0 𝑥 on 𝑆𝐵 , 𝑧 = −ℎ (.)
j 𝜕𝑧 j
𝑐 = const. ≥ 0 at 𝑆∞ for |𝑥| → ∞ (.)

A less obvious change in the linearized boundary value problem is that potential 𝜙 and
wave elevation 𝜁 are now interpreted as generally small disturbances of the fluid at rest.
Combined linear Now that the free surface conditions are enforced at the calm water level, they are no
free surface longer implicit. This allows us to solve the dynamic boundary condition (.) for the
condition
wave elevation 𝜁.
1 𝜕𝜙
𝜁(𝑥, 𝑡) = − 𝑥 with 𝑧 = 0 (.)
𝑔 𝜕𝑡
We take the time derivative of Equation (.) and substitute the result for the
time derivative of the wave elevation into the linearized kinematic boundary con-
dition (.):
𝜕2𝜙 𝜕𝜙
+𝑔 = 0 𝑥 with 𝑧 = 0 (.)
𝜕𝑡 2 𝜕𝑧

With the combined free surface condition for the free surface, the linearized boundary
value problem can be stated with the velocity potential 𝜙 as the only unknown:

𝚫𝜙 = 0 𝑥 = (𝑥, 𝑧)𝑇 ∈ 𝑉
𝜕2𝜙 𝜕𝜙
+𝑔 = 0 𝑥 with 𝑧 = 0
𝜕𝑡2 𝜕𝑧 (.)
𝜕𝜙
= 0 𝑥 on 𝑆𝐵 , 𝑧 = −ℎ
𝜕𝑧
𝑐 = const. ≥ 0 at 𝑆∞ for |𝑥| → ∞

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

20.3 Linearized Boundary Value Problem 249

Once we know the potential, the flow field in the fluid domain is given by the gra-
dient of the potential 𝑣 = ∇ 𝜙. The pressure follows from the linearized Bernoulli
equation (.):
𝜕𝜙
𝑝 − 𝑝∞ = −𝜌 − 𝜌𝑔𝑧 (.)
𝜕𝑡
and the wave elevation from Equation (.).

References
Airy, G. (). Tides and waves. In Smedley, E., Rose, H. J., and Rose, H. J., editors,
Encyclopaedia Metropolitana, volume V of Mixed Sciences, Vol. , pages –.
B. Fellowes et al., London.
Craik, A. (). The origins of water wave theory. Annual Review of Fluid Mechanics,
:–.
Michell, J. (). The wave resistance of a ship. Philosophical Magazine Series ,
():–.
Stokes, G. (). On the theory of oscillatory waves. Transactions of the Cambridge
Philosophical Society, :–.

j Self Study Problems j

. Why may we ignore viscosity for typical wind or ship generated free surface
waves?
. Why do we need two boundary conditions for the free surface but only one for
the ocean bottom?
. Derive the exact kinematic and dynamic free surface conditions for the three-
dimensional case, i.e. with 𝐹 (𝑥, 𝑦, 𝑧, 𝑡) = 𝑧 − 𝜁(𝑥, 𝑦, 𝑡) = 0.
. Name and explain in words the four boundary conditions of the exact boundary
value problem for long-crested waves:

𝜕𝜁 𝜕𝜁 𝜕𝜙 𝜕𝜙
− − + = 0 𝑥 on 𝑆𝐹 , 𝑧 = 𝜁(𝑥, 𝑡)
𝜕𝑡 𝜕𝑥 𝜕𝑥 𝜕𝑧
𝜕𝜙 1
+ ||∇ 𝜙|| + 𝑔𝑧
2
= 0 𝑥 on 𝑆𝐹 , 𝑧 = 𝜁(𝑥, 𝑡)
𝜕𝑡 2
𝜕𝜙
= 0 𝑥 on 𝑆𝐵 , 𝑧 = −𝑑
𝜕𝑧
𝑐 = const. > 0 at 𝑆∞ for |𝑥| → ∞

. Why is no analytic solution of the exact two-dimensional boundary value problem

(.) known?

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

250

21
Linearization of Free Surface Boundary
Conditions

In this chapter, we formally derive the linearized versions of the kinematic and dynamic
free surface conditions used in wave theory. The procedure is known as a perturbation
approach and is applicable to the linearization of nonlinear equations in general. As
results of the linearization process we obtain the kinematic and dynamic free surface
boundary conditions which are employed in linear wave theory.

Learning Objectives
j j
At the end of this chapter students will be able to:

• apply the perturbation approach

• perform linearizations of nonlinear equations
• understand the origin of linearized free surface conditions
• know the limitations of linear wave theory

21.1 Perturbation Approach

Nonlinear, For our potential wave flow problem, we derived two free surface boundary conditions.
implicit free Two boundary conditions are required because we have two unknown functions: the
surface
conditions
velocity potential 𝜙(𝑥, 𝑦, 𝑧, 𝑡) and the actual position 𝜁(𝑥, 𝑦, 𝑡) of the free surface relative
to the calm water level 𝑧 = 0. Unfortunately these boundary conditions are:

• nonlinear because they contain products of the unknown functions 𝜙 and 𝜁 or

their derivatives. As a consequence the equations cannot be resolved to eliminate
one unknown function from either of the two equations,
• implicit because all terms have to be evaluated at the unknown free surface 𝑆𝐹
with 𝑧 = 𝜁(𝑥, 𝑦, 𝑡).

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

21.1 Perturbation Approach 251

A general analytical solution of the nonlinear problem does not exist. Therefore, we
will linearize the free surface boundary conditions based on a perturbation approach.
Of course, this comes at a price: the solution of the linearized problem will be valid
only within certain limits, which we will discuss.
The perturbation approach follows the general concept of mathematical series like the Series expansion
Taylor series or the Fourier series. The idea is that an approximate, basic solution for the
unknown function is improved with additional terms that become smaller and smaller.
The basic solution is also called the ‘zero order’ solution and is assumed to be known.
The additional terms are of first order, second order, third order, and so on.
The unknown functions, in our case the velocity potential 𝜙 and the wave elevation
𝜁, are expressed as a power series expansion with respect to a small parameter 𝜀. The
perturbation series for our two unknowns are:

𝜙 = 𝜙(0) + 𝜀𝜙(1) + 𝜀2 𝜙(2) + 𝜀3 𝜙(3) + … (.)

𝜁 = 𝜁 (0)
+ 𝜀𝜁 (1) 2 (2)
+𝜀 𝜁 +𝜀 𝜁3 (3)
+… (.)

The superscripts (0), (1), (2), etc. express the order. 𝜙(𝓁) denotes a velocity potential
of 𝓁 𝑡ℎ -order and 𝜁 (𝓁) an 𝓁 𝑡ℎ -order wave elevation. The perturbation series may more
concisely be written as the sums:

∑
∞
𝜙(𝑥, 𝑦, 𝑧, 𝑡, 𝜀) = 𝜀𝓁 𝜙(𝓁) (𝑥, 𝑦, 𝑧, 𝑡),
j 𝓁=0 j
(.)
∑∞
𝜁(𝑥, 𝑦, 𝑡, 𝜀) = 𝜀𝓁 𝜁 (𝓁) (𝑥, 𝑦, 𝑡)
𝓁=0

The small parameter 𝜀 ≪ 1 (also called the perturbation parameter) can be interpreted
as a dimensionless wave steepness 𝜀 = 𝑘𝜁𝑎 = 𝐿2𝜋 𝐻2 ∼ 𝐿𝐻 . If this is too mathematical
𝑊 𝑊
for you, just think of replacing the exact solutions by a sum of potentials or wave
elevations that have fast decreasing contributions

𝜙(𝑥, 𝑦, 𝑧, 𝑡) = 𝜙(0) + 𝜙(1) + 𝜙(2) + … with 𝜙(0) > 𝜙(1) > 𝜙(2) > …
(.)
𝜁(𝑥, 𝑦, 𝑡) = 𝜁 (0) + 𝜁 (1) + 𝜁 (2) + … with 𝜁 (0) > 𝜁 (1) > 𝜁 (2) > …

Different levels of approximation (of the exact solution) are derived by truncating the
series expansion after a selected order. Series expansions with respect to 𝜀 lead to
solutions known as Stokes’ wave theories. Stokes’ wave theories up to fifth order are
used in offshore hydrodynamics.
In order to linearize the free surface boundary conditions we truncate the perturbation Linearization
series after the linear term

𝜙 = 𝜙(0) + 𝜀𝜙(1) + (2) (.)

𝜁 = 𝜁 (0)
+ 𝜀𝜁 (1)
+ (2) (.)

Furthermore, we assume the terms of second and higher order represented by (2) are
negligibly small, and we will omit them from here on.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

252 21 Linearization of Free Surface Boundary Conditions

21.2 Kinematic Free Surface Condition

Kinematic free Let us apply the perturbation approach to the kinematic free surface boundary condition
surface (.), which expresses the requirement that no particle leave the free surface.
boundary
condition 𝜕𝜁 𝜕𝜁 𝜕𝜙 𝜕𝜁 𝜕𝜙 𝜕𝜙
− − − + = 0 on 𝑆𝐹 with 𝑧 = 𝜁(𝑥, 𝑦, 𝑡) (.)
𝜕𝑡 𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑧
Another interpretation is that the velocity of a particle normal to the surface has to be
equal to the velocity of the surface in that direction.
Simplify notation Before we proceed, we rewrite the boundary condition using subscripts to indicate
partial differentiation.

− 𝜁𝑡 − 𝜁𝑥 𝜙𝑥 − 𝜁𝑦 𝜙𝑦 + 𝜙𝑧 = 0 on 𝑆𝐹 with 𝑧 = 𝜁(𝑥, 𝑦, 𝑡) (.)

Substitute Now we replace the unknown functions 𝜙 and 𝜁 with their linear perturbation approxi-
perturbations mations.
into boundary
condition ( ) ( ) ( )
− 𝜁 (0) + 𝜀𝜁 (1) 𝑡 − 𝜁 (0) + 𝜀𝜁 (1) 𝑥 𝜙(0) + 𝜀𝜙(1) 𝑥
( ) ( ) ( )
− 𝜁 (0) + 𝜀𝜁 (1) 𝑦 𝜙(0) + 𝜀𝜙(1) 𝑦 + 𝜙(0) + 𝜀𝜙(1) 𝑧 = 0
on 𝑆𝐹 with 𝑧 = 𝜁 (0) + 𝜀𝜁 (1) (.)

j Note that we apply the perturbation also to the position where we satisfy the boundary j
condition.
The differential operators are applicable to the individual terms, for instance:
( (0) )
𝜁 + 𝜀𝜁 (1) 𝑡 = 𝜁𝑡(0) + 𝜀𝜁𝑡(1)

Likewise for the other derivatives. The perturbation parameter 𝜀 is treated as a constant
factor. We first perform the differentiations within the parentheses and then expand
the products. Equation (.) transforms into

⁓≈ 0
− 𝜁𝑡(0) − 𝜀𝜁𝑡(1) − 𝜁𝑥(0) 𝜙(0) (0) (1) (1) (0)
𝜀2
𝜁𝑥(1)𝜙(1)

𝑥 − 𝜁𝑥 𝜀𝜙𝑥 − 𝜀𝜁𝑥 𝜙𝑥 −  𝑥
⁓≈ 0
− 𝜁𝑦(0) 𝜙(0) (0) (1) (1) (0)
𝜀2 𝜁𝑦(1)𝜙
(1)

𝑦 − 𝜁𝑦 𝜀𝜙𝑦 − 𝜀𝜁𝑦 𝜙𝑦 −  𝑦
+ 𝜙(0) (1)
𝑧 + 𝜀𝜙𝑧 = 0
on 𝑆𝐹 with 𝑧 = 𝜁 (0) + 𝜀𝜁 (1) (.)

Neglect higher The products of derivatives of the first order potential 𝜙(1) and the first order wave
order elevation 𝜁 (1) result in second order terms as indicated by the factor 𝜀2 . Since we
contributions
neglected second order contribution already in our perturbation approximation, it is
only logical to neglect the newly created second order terms as well.
Equation (.) is now linear if we consider the zero order functions and their derivatives
known. There are no longer products of the unknown first order functions 𝜙(1) and 𝜁 (1)
or their derivatives with each other. This is also reflected in the fact that there are only
terms left that have a factor 𝜀 of one.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

21.2 Kinematic Free Surface Condition 253

We are not done yet, however. The boundary condition still has to be satisfied at the Not done yet
unknown position 𝑧 = 𝜁 (0) + 𝜀𝜁 (1) . This affects the derivatives of both zero and first
order potentials. In order to remedy this problem, the potentials are developed in a
Taylor series with respect to the zero order wave elevation. The 𝑥-derivative of the zero
order potential may serve as an example.

𝜕𝜙(0) || 1( ) 𝜕 2 𝜙(0) ||
𝜙(0)
𝑥 (𝑥, 𝑦, 𝑧, 𝑡) = | + 𝑧 − 𝜁 (0) |
𝜕𝑥 ||𝑧=𝜁 (0) 1! 𝜕𝑥𝜕𝑧 ||𝑧=𝜁 (0)

1( )2 𝜕 3 𝜙(0) ||
+ 𝑧 − 𝜁 (0) | + (3) (.)
2! 𝜕𝑥𝜕𝑧2 ||𝑧=𝜁 (0)

Of course, all terms of order two and higher will be neglected.

𝜕𝜙(0) || ( ) 𝜕 2 𝜙(0) ||
𝜙(0)
𝑥 (𝑥, 𝑦, 𝑧, 𝑡) = | + 𝑧 − 𝜁 (0) | (.)
𝜕𝑥 ||𝑧=𝜁 (0) 𝜕𝑥𝜕𝑧 ||𝑧=𝜁 (0)

At the free surface 𝑆𝐹 we have 𝑧 = 𝜁. Together with the linear perturbation we obtain
( )
𝑧 = 𝜁 = 𝜁 (0) + 𝜀𝜁 (1) or 𝑧 − 𝜁 (0) = 𝜀𝜁 (1) (.)

Applying this result to the Taylor series expansion of the zero order potential yields Taylor series
approximation
j j
𝜕𝜙(0) || 𝜕 2 𝜙(0) ||
𝜙(0)
𝑥 (𝑥, 𝑦, 𝜁 , 𝑡) = | + 𝜀𝜁 (1) | (.)
𝜕𝑥 ||𝑧=𝜁 (0) 𝜕𝑥𝜕𝑧 ||𝑧=𝜁 (0)

The same Taylor series expansion is applied to all other potentials and derivatives in
Equation (.):

𝜕𝜙(0) || 𝜕 2 𝜙(0) ||
𝜙(0)
𝑦 (𝑥, 𝑦, 𝜁 , 𝑡) = | + 𝜀𝜁 (1) | (.)
𝜕𝑦 ||𝑧=𝜁 (0) 𝜕𝑦𝜕𝑧 ||𝑧=𝜁 (0)
𝜕𝜙(0) || 𝜕 2 𝜙(0) ||
𝜙(0)
𝑧 (𝑥, 𝑦, 𝜁 , 𝑡) = | + 𝜀𝜁 (1) | (.)
𝜕𝑧 ||𝑧=𝜁 (0) 𝜕𝑧2 ||𝑧=𝜁 (0)
𝜕𝜙(1) || 2 (1) |
(1) 𝜕 𝜙 |
𝜙(1) (𝑥, 𝑦, 𝜁 , 𝑡) = | + 𝜀𝜁 | (.)
𝑥 𝜕𝑥 ||𝑧=𝜁 (0) 𝜕𝑥𝜕𝑧 ||𝑧=𝜁 (0)
𝜕𝜙(1) || 𝜕 2 𝜙(1) ||
𝜙(1)
𝑦 (𝑥, 𝑦, 𝜁 , 𝑡) = | + 𝜀𝜁 (1) | (.)
𝜕𝑦 ||𝑧=𝜁 (0) 𝜕𝑦𝜕𝑧 ||𝑧=𝜁 (0)
𝜕𝜙(1) || 2 (1) |
(1) 𝜕 𝜙 |
𝜙(1) (𝑥, 𝑦, 𝜁 , 𝑡) = | + 𝜀𝜁 | (.)
𝑧 𝜕𝑧 ||𝑧=𝜁 (0) 𝜕𝑧2 ||𝑧=𝜁 (0)

The first and second order derivatives of the potentials on the right-hand side are now
evaluated at the known zero order wave elevation 𝜁 (0) .
The right-hand sides of Equations (.) through (.) replace the corresponding
terms in Equation (.). We again revert to subscripts for derivatives and omit the

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

254 21 Linearization of Free Surface Boundary Conditions

function arguments, but note that all potential derivatives are evaluated at 𝑧 = 𝜁 (0)
from here on.
( )
− 𝜁𝑡(0) − 𝜀𝜁𝑡(1) − 𝜁𝑥(0) 𝜙(0)
𝑥 + 𝜀𝜁 𝜙𝑥𝑧
(1) (0)

( ) ( )
− 𝜁𝑥(0) 𝜀 𝜙(1)
𝑥 + 𝜀𝜁 (1) (1)
𝜙 𝑥𝑧 − 𝜀𝜁 (1)
𝑥 𝜙 (0)
𝑥 + 𝜀𝜁 (1) (0)
𝜙 𝑥𝑧
( ) ( )
− 𝜁𝑦(0) 𝜙(0) (1) (0) (0)
𝑦 + 𝜀𝜁 𝜙𝑦𝑧 − 𝜁𝑦 𝜀 𝜙𝑦 + 𝜀𝜁 𝜙𝑦𝑧
(1) (1) (1)

( ) ( )
− 𝜀𝜁𝑦(1) 𝜙(0)𝑦 + 𝜀𝜁 (1) (0)
𝜙 𝑦𝑧 + 𝜙 (0)
𝑧 + 𝜀𝜁 (1) (0)
𝜙 𝑧𝑧
( )
+ 𝜀 𝜙(1)𝑧 + 𝜀𝜁 𝜙𝑧𝑧
(1) (1)
= 0 on 𝑆𝐹 with 𝑧 = 𝜁 (0) (.)

As a result, the kinematic boundary condition is now enforced at the known zero order
wave elevation instead of the unknown exact wave elevation.
Linearized Expanding the products results in five additional second order terms which we conse-
kinematic free quently neglect. Finally, we have morphed the nonlinear kinematic boundary condition
surface
boundary
into its linearized form.
condition
− 𝜁𝑡(0) − 𝜀𝜁𝑡(1) − 𝜁𝑥(0) 𝜙(0) (0) (1) (0) (0) (1) (1) (0)
𝑥 + 𝜁𝑥 𝜀𝜁 𝜙𝑥𝑧 − 𝜁𝑥 𝜀𝜙𝑥 − 𝜀𝜁𝑥 𝜙𝑥
− 𝜁𝑦(0) 𝜙(0) (0) (1) (0) (0) (1) (1) (0)
𝑦 + 𝜁𝑦 𝜀𝜁 𝜙𝑦𝑧 − 𝜁𝑦 𝜀𝜙𝑦 − 𝜀𝜁𝑦 𝜙𝑦
+ 𝜙(0) (1) (0) (1)
𝑧 + 𝜀𝜁 𝜙𝑧𝑧 + 𝜀𝜙𝑧 = 0 on 𝑆𝐹 with 𝑧 = 𝜁 (0) (.)
j j
Only the first order wave elevation 𝜁 (1) and the first order velocity potential 𝜙(1) are
unknown in Equation (.). The equation is linear because it no longer contains
products or powers of the unknown functions or their derivatives. At most a first order
function is multiplied with known zero order functions.
Zero order What is left is the selection of an appropriate zero order solution! However, before we
solution do so, the same linearization process is applied to the dynamic free surface boundary
condition. The zero order solution is introduced in Section ..

21.3 Dynamic Free Surface Condition

The dynamic free surface condition followed from the unsteady Bernoulli equation and
expresses the requirement that the pressure is to be constant over the free surface 𝑆𝐹 .
𝜕𝜙 1
+ |∇ 𝜙|2 + 𝑔𝑧 = 0 at 𝑆𝐹 with 𝑧 = 𝜁(𝑥, 𝑦, 𝑡) (.)
𝜕𝑡 2
The same equation reads as follows if again subscripts are used to represent partial
differentiation.
[ ]
1 2
𝜙𝑡 + 𝜙𝑥 + 𝜙2𝑦 + 𝜙2𝑧 + 𝑔𝑧 = 0 at 𝑆𝐹 with 𝑧 = 𝜁(𝑥, 𝑦, 𝑡) (.)
2

Subsequently, the same process is followed as was used for the kinematic free surface
boundary condition. Only major way points are provided here and the step by step
execution is left as a self study problem (see page ).

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

21.3 Dynamic Free Surface Condition 255

. Introduce the linear perturbations Linear

perturbation
𝜙 = 𝜙(0) + 𝜀𝜙(1) (.)
𝜁 = 𝜁 (0) + 𝜀𝜁 (1) (.)

into Equation (.).

. Take derivatives and expand all products.

. Delete all terms proportional to 𝜀2 (second order, negligibly small)! As a result, Neglect higher
we obtain for the dynamic free surface condition the following expression: order terms

[ ]
1 ( (0) )2 ( (0) )2 ( (0) )2
𝜙(0) (1)
𝑡 + 𝜀𝜙𝑡 + 𝜙𝑥 + 𝜙𝑦 + 𝜙𝑧
2
+ 𝜀𝜙𝑥 𝜙𝑥 + 𝜀𝜙(0)
(0) (1) (1) (0) (1)
𝑦 𝜙𝑦 + 𝜀𝜙𝑧 𝜙𝑧 + 𝑔𝜁
(0)
+ 𝑔𝜀𝜁 (1) = 0
at 𝑆𝐹 with 𝑧 = 𝜁 (0) + 𝜀𝜁 (1) (.)

. The derivatives of the potential still have to be computed at the unknown free Taylor series
surface location 𝑧 = 𝜁 (0) + 𝜀𝜁 (1) . The derivatives are, once again, approximated expansion
by the truncated Taylor series approximations (.) through (.) and corre-
sponding Taylor series for the time derivatives 𝜙(0) (1)
𝑡 and 𝜙𝑡 of the zero and first
order potential respectively. The dynamic free surface condition is now satisfied
j at the zero order wave elevation 𝑧 = 𝜁 (0) . j

. After renewed expansion of products and the omission of second order terms we Linearized
have as linearized dynamic free surface boundary condition: dynamic free
surface condition
[ ]
1 ( (0) )2 ( (0) )2 ( (0) )2
𝜙(0) (1) (0) (1)
𝑡 + 𝜀𝜁 𝜙𝑡𝑧 + 𝜀𝜙𝑡 + 𝜙𝑥 + 𝜙𝑦 + 𝜙𝑧
2
+ 𝜙(0)
𝑥 𝜀𝜁 (1) (0)
𝜙 (0) (1) (0) (0) (1) (0)
𝑥𝑧 + 𝜙𝑦 𝜀𝜁 𝜙𝑦𝑧 + 𝜙𝑧 𝜀𝜁 𝜙𝑧𝑧
+ 𝜀𝜙(0) (1) (0) (1) (0) (1)
𝑥 𝜙𝑥 + 𝜀𝜙𝑥 𝜙𝑦 + 𝜀𝜙𝑥 𝜙𝑥 + 𝑔𝜁
(0)
+ 𝑔𝜀𝜁 (1) = 0
at 𝑆𝐹 with 𝑧 = 𝜁 (0) (.)

Admittedly, the linearized three-dimensional kinematic and dynamic boundary condi-

tions for the free surface, (.) and (.) respectively, look even more complicated
than their nonlinear equivalents. However, the linearized forms have two distinct
advantages:

• The unknown first order wave elevation and potential are multiplied with known
constants or zero order functions only.

• The conditions are applied at the known location of the zero order wave elevation
instead of the exact position 𝜁(𝑥, 𝑦, 𝑡).

Finally, a suitable pair of zero order wave elevation and velocity has to be selected to
formulate the linearized boundary value problem already introduced in Section ..

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

256 21 Linearization of Free Surface Boundary Conditions

21.4 Linearized Free Surface Conditions for Waves

Zero order The zero order potential 𝜙(0) and zero order wave elevation 𝜁 (0) were introduced as
solution must be basic solutions to the problem which we assume to know. In fact, we will select a basic
physical
solution. A valid basic solution has to be physical, i.e. it does not violate the underlying
conditions of the problem.
Simplest zero It was mentioned earlier that linear wave theory is only valid for waves with small
order solution steepness
wave height 𝐻
= ≪ 1 (.)
wave length 𝐿𝑤
For diminishing wave steepness the disturbance of the free surface eventually becomes
invisible. If the surface is not moving, the fluid underneath will not move either (we
ignore currents). Thus, a suitable set of zero order solutions is the trivial case of no
wave elevation and no fluid velocity, or in other words calm water. We select for the
zero order potential:

𝜙(0) (𝑥, 𝑦, 𝑧, 𝑡) ≡ 0 (.)

and for the zero order wave elevation:

𝜁 (0) (𝑥, 𝑦, 𝑡) ≡ 0 (.)

j Obviously, derivatives of the zero order solutions vanish as well. j

Simplifying the First, the zero order solutions are substituted into the linearized kinematic free surface
kinematic condition.
boundary
condition
=0
⌃ = 0(0) (1) ⁓ = 0(0) (1⁓ = 0 (1)(0)
⁓= 0
− 𝜁𝑡(0)
 − 𝜀𝜁 (1) − 𝜁 (0) (0)
 
(0)
𝑡 𝜙 +
𝜁𝑥𝜀𝜁𝜙𝑥𝑧 − 

𝜁𝑥𝜀𝜙𝑥 ) − 
𝜀𝜁
𝑥 𝜙𝑥


𝑥 𝑥

⌃ = 0(0) (1) ⌃ = 0 (1) (0)
⁓ = 0(0) (1)
 ⌃= 0

− 𝜁𝑦(0) (0) (0)
 
𝜙 𝑦 + 𝜙
𝜁𝑦 𝜀𝜁 𝑦𝑧 − 𝜁𝑦 𝜀𝜙𝑦 − 𝜀𝜁𝑦 𝜙𝑦

  
=0
⁓= 0
+ 𝜙(0) (1)(0)
𝜙𝑧𝑧 + 𝜀𝜙(1) on 𝑆𝐹 with 𝑧 = 𝜁 (0) (.)


𝑧 + 𝑧 = 0
 𝜀𝜁

Astonishingly, only two terms remain:

− 𝜀𝜁𝑡(1) + 𝜀𝜙(1)
𝑧 = 0 on 𝑆𝐹 with 𝑧 = 𝜁 (0) = 0 (.)

Kinematic free At this point we may drop the perturbation parameter and omit the superscript for
surface condition the first order classification. The first order wave elevation and the first order velocity
for linear wave
theory
potential are the only nonvanishing components of the solution left.
𝜕𝜁 𝜕𝜙
− + = 0 on 𝑆𝐹 with 𝑧 = 0 (.)
𝜕𝑡 𝜕𝑧

Physical Let us take a step back after this long and tedious mathematical derivation. The deriva-
interpretation tive of the potential with respect to 𝑧 is the vertical component 𝑤 of the fluid velocity
𝑣. The time derivative of the wave elevation 𝜕𝜁∕𝜕𝑡 is the velocity of the free surface in

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

21.4 Linearized Free Surface Conditions for Waves 257

𝑧-direction. As originally claimed, the kinematic free surface condition enforces that
the relative velocity between free surface and fluid particles vanishes. In the case of
the linear kinematic free surface condition, the slope of the free surface is ignored. In
fact, it is treated as if it is flat. Therefore, the linear kinematic condition is valid only for
waves with small steepness 𝐻∕𝐿𝑤 .
Applying the selected zero order solution and its derivatives to the dynamic free surface Simplifying the
condition (.) yields: dynamic
boundary
=0 =0 [ ] condition
=(0 ) =(0 ) =0
1 (  )
2 2 2
𝜙(0) (1) (0) (1) (0)
⌃ (0)⌃ (0)⌃
𝑡 + 𝜀𝜁 𝜙𝑡𝑧 + 𝜀𝜙𝑡 + 𝜙 + 𝜙 + 𝜙
   
2 𝑥 𝑦 𝑧

⁓ = 0 (0) (1)
 ⁓ = 0 (0) (1) ⁓= 0

𝜙(0) (1) (0) 𝜙(0) (0)
  𝜙
+𝑥𝜀𝜁 𝜙𝑥𝑧 +  𝜙𝑦𝜀𝜁 𝑦𝑧 + 
𝜙𝑧𝜀𝜁 𝑧𝑧

⁓ = 0 (0)(1)
(0)(1) ⁓= 0  ⁓ = 0 (1)
𝜀𝜙(0)
(1)

  

+ 𝑥 𝜙𝑥 + 
𝜀𝜙 𝑥 𝜙𝑦 + 
𝜀𝜙 
𝑥 𝜙𝑥 + 𝑔𝜀𝜁 = 0

at 𝑆𝐹 with 𝑧 = 𝜁 (0) = 0 (.)

What remains is
𝜀𝜙(1)
𝑡 + 𝑔𝜀𝜁
(1)
= 0 at 𝑆𝐹 with 𝑧 = 0 (.)

As for the linear kinematic free surface condition, the perturbation parameter and the Dynamic free
superscript for first order may be dropped knowing that the first order wave elevation surface condition
j for linear wave j
and potential are the only substantial parts of the solution. theory
𝜕𝜙
+ 𝑔𝜁 = 0 at 𝑆𝐹 with 𝑧 = 0 (.)
𝜕𝑡
Note that in linear wave theory the free surface conditions are satisfied at the calm
water level 𝑧 = 0.
In Section (.), we introduced the linearized Bernoulli equation (.) Physical
interpretation
𝜕𝜙
𝑝 − 𝑝∞ = −𝜌 − 𝜌𝑔𝑧 (.)
𝜕𝑡
The dynamic boundary condition requires that the pressure is equal everywhere at the
free surface with 𝑧 = 𝜁. Therefore, the pressure difference must vanish 𝑝 − 𝑝∞ = 0 for
𝑧 = 𝜁.
𝜕𝜙
0 = −𝜌 − 𝜌 𝑔𝜁 (.)
𝜕𝑡
Dividing the linear Bernoulli equation applied to the free surface by the negative fluid
density −𝜌 results in the linear dynamic free surface condition (.). Here we neglect
the influence of the nonlinear dynamic pressure term 1∕2𝜌|𝑣|2 , which we assume is
small.
The first order wave elevation is derived from the dynamic free surface boundary First order wave
condition (.). elevation
𝜕𝜙 ||
𝜁(𝑥, 𝑦, 𝑡) = −𝑔 (.)
𝜕𝑡 ||𝑧=0
In long-crested waves, the wave elevation is only a function of 𝑥 and time 𝑡.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

258 21 Linearization of Free Surface Boundary Conditions

Self Study Problems

. Linearize the following differential equation, assuming the zero order basis solu-
tion is linear 𝑓 (0) = 𝑎 ⋅ 𝑧. with 𝑎 being a constant factor.

𝜕𝑓 𝜕𝑓 𝜕2𝑓
+ 𝜙 = 0
𝜕𝑥 𝜕𝑧 𝜕𝑧2
Hint: in this case the Taylor series expansion around a specific 𝑧 value is not
needed.
. Linearize the dynamic free surface condition by executing in full detail the steps
 through  on page .

. Assume steady flow (all time derivatives vanish) and simplify the linearized
kinematic and dynamic free surface boundary conditions (.) and (.) with
respect to the following zero order solutions (Neumann-Kelvin linearization):

𝜙(0) = −𝑈 𝑥 parallel flow in negative 𝑥-direction

𝜁 (0) = 0 calm water

j j

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

259

22
Linear Wave Theory

After thoroughly explaining the mathematical model for long-crested waves, it is time
to actually solve the boundary value problem. In this chapter the linearized boundary
value problem is solved for the two-dimensional case of long-crested waves progressing
in positive 𝑥-direction. Again, the separation of variables is applied to derive the now
time dependent velocity potential. Equipped with the potential, we will study the
behavior and properties of waves.

Learning Objectives

j At the end of this chapter students will be able to: j

• apply potential theory to water waves
• solve potential flow problems
• compute the wave number
• distinguish exact formulation from deep water approximation

22.1 Solution of Linear Boundary Value Problem

The linearized boundary value problem for long-crested waves has been introduced in Mathematical
Section .. Chapter  provides a detailed account of the linearization process. A model
recaptured
solution to the boundary value problem (.) (repeated below) will yield a velocity
potential that describes the unsteady flow in waves with small wave steepness.

Laplace equation (conservation of mass)

𝜕2𝜙 𝜕2𝜙
𝚫𝜙(𝑥, 𝑡) = + = 0 𝑥 = (𝑥, 𝑧)𝑇 ∈ 𝑉 (.)
𝜕𝑥 2 𝜕𝑧2
generalized linear free surface boundary condition

𝜕2𝜙 𝜕𝜙
+𝑔 = 0 𝑥 with 𝑧 = 0 (.)
𝜕𝑡2 𝜕𝑧

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

260 22 Linear Wave Theory

Figure 22.1 Simplified two-dimensional fluid domain for long-crested regular waves

ocean bottom boundary condition

𝜕𝜙
= 0 𝑥 on 𝑆𝐵 , 𝑧 = −ℎ (.)
𝜕𝑧
far field boundary condition

j 𝑐 = const. ≥ 0 at 𝑆∞ for |𝑥| → ∞ (.) j

The coordinate system, fluid domain, and its boundaries are depicted in Figure ..
The origin is at the calm water level 𝑧 = 0. With the 𝑧-axis pointing upwards, points in
the fluid domain 𝑉 have negative 𝑧-coordinates 𝑧 ≤ 0. Note that the linear free surface
boundary condition (.) is satisfied at the calm water level 𝑧 = 0 and not at the yet
unknown displaced free surface 𝑆𝐹 with 𝑧 = 𝜁(𝑥, 𝑡).
Separation of The velocity potential is a function of 𝑥, 𝑧, and time 𝑡. Therefore, we seek a solution
variables which represents the potential as a product of three functions 𝑋(𝑥), 𝑍(𝑧), and 𝑇 (𝑡).
Each of these is a function of just one independent variable.

𝜙(𝑥, 𝑧, 𝑡) = 𝑋(𝑥) 𝑍(𝑧) 𝑇 (𝑡) (.)

This approach is called separation of variables because it will convert the partial differ-
ential Laplace equation into a set of three simultaneous, ordinary differential equations.
The second order spatial derivatives of the potential (.) are needed for the Laplace
equation.
𝜕2𝜙 ( )
𝜕2 𝜕2𝑋
= 𝑋𝑍𝑇 = 𝑍 𝑇 = 𝑋𝑥𝑥 𝑍 𝑇
𝜕𝑥 𝜕𝑥2 𝜕𝑥2 (.)
𝜕2𝜙 ( )
𝜕2 𝜕2𝑍
= 𝑋𝑍𝑇 = 𝑋 𝑇 = 𝑋 𝑍𝑧𝑧 𝑇
𝜕𝑧 𝜕𝑧2 𝜕𝑧2
With substitution of the derivatives (.) into (.) and using subscripts to denote
differentiation, the Laplace equation becomes:

𝑋𝑥𝑥 𝑍 𝑇 + 𝑋 𝑍𝑧𝑧 𝑇 = 0 for 𝑥 = (𝑥, 𝑧)𝑇 in 𝑉 (.)

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

22.1 Solution of Linear Boundary Value Problem 261

Remember that the zero order potential 𝜙(0) ≡ 0 describes the fluid at rest. Therefore,
it is safe to assume that 𝑋 𝑍 ≠ 0, and we may divide Equation (.) by 𝑋𝑍:
[ ]
𝑋𝑥𝑥 𝑍𝑧𝑧
+ 𝑇 = 0 (.)
𝑋 𝑍

For this to be true at all times, the term in brackets must vanish, i.e.
𝑋𝑥𝑥 𝑍
+ 𝑧𝑧 = 0 (.)
𝑋 𝑍
The first fraction is a function of 𝑥 alone. The second fraction only depends on 𝑧. As a
consequence, the fractions will be equal for all (𝑥, 𝑧) only if they are constant in (𝑥, 𝑧)
and one is the negative of the other.

𝑋𝑥𝑥 𝑍
= − 𝑧𝑧 = const. = 𝐶 (.)
𝑋 𝑍
Equation (.) may be split into two ordinary differential equations that are connected Ordinary
by the same constant 𝐶 differential
𝑋𝑥𝑥 equations
for 𝑋(𝑥): = 𝐶
𝑋 (.)
𝑍
for 𝑍(𝑧): − 𝑧𝑧 = 𝐶
𝑍
j We select the constant to be 𝐶 = −𝑘2 with 𝑘 being the wave number. Equations (.) j
are multiplied by 𝑋 and 𝑍 respectively, and all terms are moved to the left-hand side.

𝑋𝑥𝑥 + 𝑘2 𝑋 = 0 (.)
2
𝑍𝑧𝑧 − 𝑘 𝑍 = 0 (.)

The complimentary solution of the ordinary differential equation in 𝑥 (.) is a Solution for 𝑋
function whose second order derivative is equal to the negative of the function itself,
except for a constant factor. The sine and cosine functions are of this type. However, we
do not know which of the two. Since combinations of sine and cosine are also plausible,
we define
𝑋 = 𝐶1 ei𝑘𝑥 (.)
The exponential function combines sine and cosine with Euler’s formula for complex
numbers ei𝑥 = cos 𝑥 + i sin 𝑥. The complex notation is a useful mathematical tool which
simplifies the treatment of our equations significantly. However, the velocity field is
real and only the real part of the resulting velocity potential will bear physical meaning.
Differentiating Equation (.) twice and substituting it back shows that this is indeed
a solution of ODE (.). Boundary conditions will be used later to solve for the new
constant 𝐶1 which generally differs from constant 𝐶 above.
The second ODE (.) is satisfied by functions whose second order derivative is Solution for 𝑍
equal to the function itself. This time hyperbolic sine and cosine functions will suffice.
Selecting the hyperbolic cosine, a general solution to (.) is given by

𝑍 = 𝐶2 cosh(𝑘𝑧 + 𝛼2 ) (.)

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

262 22 Linear Wave Theory

Solution for 𝑇 Finally, a solution is needed for the time dependent function 𝑇 . Equation (.) is of no
use because it would only yield the trivial solution 𝑇 ≡ 0. Instead, the product (.) 𝜙 =
𝑋 𝑌 𝑍 is introduced into the generalized linear free surface boundary condition (.).

𝜙𝑡𝑡 + 𝑔 𝜙𝑧 = 𝑋 𝑍 𝑇𝑡𝑡 + 𝑔 𝑋 𝑍𝑧 𝑇 = 0 for 𝑧 = 0 (.)

Division by the product 𝑋 𝑍 yields

𝑍𝑧 ||
𝑇𝑡𝑡 + 𝑔 𝑇 = 0 (.)
𝑍 ||𝑧=0
The expression 𝑍𝑍 ∕𝑍 has to be evaluated at 𝑧 = 0 because Equation (.) is derived
from the free surface boundary condition. 𝑧 is the only variable in the expression 𝑍𝑍 ∕𝑍.
Consequently, the term 𝑔𝑍𝑧 ∕𝑍|𝑧=0 is constant. We select
𝑍𝑧 ||
𝑔 = 𝜔2 (.)
𝑍 ||𝑧=0
𝜔 is the wave frequency.
The ordinary differential equation for 𝑇 is of the same type as the ODE (.) for 𝑋.

𝑇𝑡𝑡 + 𝜔2 𝑇 = 0 (.)

An appropriate complimentary solution is

𝑇 = 𝐶3 e−i𝜔𝑡 (.)
j j
The minus sign is deliberate in the argument of the exponential function. We come
back to this when we discuss the far field condition.
So far, the solution for the velocity potential is using the results from Equations (.),
(.), and (.):

𝜙(𝑥, 𝑧, 𝑡) = 𝑋(𝑥) 𝑍(𝑧) 𝑇 (𝑡) = 𝐶1 𝐶2 𝐶3 cosh(𝑘𝑧 + 𝛼2 ) ei𝑘𝑥 e−i𝜔𝑡 (.)

The product of constants is yet another constant. We substitute 𝐶4 = 𝐶1 𝐶2 𝐶3 and get

𝜙(𝑥, 𝑧, 𝑡) = 𝐶4 cosh(𝑘𝑧 + 𝛼2 ) ei𝑘𝑥 e−i𝜔𝑡 (.)

Constants The remaining constants 𝐶4 and 𝛼2 in (.) are selected so that the solution 𝜙 =
𝑋 𝑌 𝑍 satisfies the remaining boundary conditions of the problem.
The ocean bottom boundary condition requires that the normal velocity vanishes at
the ocean floor.
𝜕𝜙
= 0 for 𝑧 = −ℎ (.)
𝜕𝑧
Introducing the partial solution (.) yields:
[ ]
𝜕
𝐶4 cosh(𝑘𝑧 + 𝛼2 ) ei𝑘𝑥 e−i𝜔𝑡 = 0 for 𝑧 = −ℎ (.)
𝜕𝑧

Since only the hyperbolic cosine is a function of 𝑧 and 𝐶4 ei𝑘𝑥 e−i𝜔𝑡 ≠ 0 this simplifies to
[ ]
𝜕
cosh(𝑘𝑧 + 𝛼2 ) = 0 for 𝑧 = −ℎ (.)
𝜕𝑧

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

22.1 Solution of Linear Boundary Value Problem 263

4 Figure 22.2 The hyperbolic sine and cosine functions

1
sinh(x),cosh( x)

3
sinh(x)
cosh(x)
4
3 2 1 0 1 2 3
x

Differentiation results in

j 𝑘 sinh(𝑘𝑧 + 𝛼2 ) = 0 for 𝑧 = −ℎ (.) j

The hyperbolic sine only vanishes if its argument is zero (see Figure .). Substituting
the negative water depth for 𝑧, we get
𝑘(−ℎ) + 𝛼2 = 0 (.)
Therefore, the constant 𝛼2 is equal to
𝛼2 = 𝑘ℎ (.)

This leaves the constant 𝐶4 to be determined. Except for the far field boundary condition,
we have explicitly used all boundary conditions listed in the linearized boundary value
problem (.) through (.). The far field condition is of no use to determine 𝐶4 . One
has to remember that the generalized free surface condition (.) actually consists of
two boundary conditions, i.e. the kinematic and the dynamic free surface condition.
The linearized dynamic free surface condition (.) requires that the pressure at the
free surface remains constant.
𝜕𝜙
+ 𝑔𝜁 = 0 at 𝑆𝐹 with 𝑧 = 0 (.)
𝜕𝑡
Solving (.) for the unknown wave elevation 𝜁 and substituting the partial solution
for the potential (.) and the new found constant 𝛼2 from Equation (.) yields:
1 𝜕𝜙 || 1
𝜁(𝑥, 𝑡) = − = − 𝐶4 cosh(𝑘 ⋅ 0 + 𝑘ℎ) ei𝑘𝑥 (−i𝜔)e−i𝜔𝑡
𝑔 𝜕𝑡 ||𝑧=0 𝑔
i𝜔
= 𝐶 cosh(𝑘ℎ) ei(𝑘𝑥−𝜔𝑡) (.)
𝑔 4

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

264 22 Linear Wave Theory

Wave amplitude The exponential function represents a general harmonic oscillation of unit amplitude.
ei(𝑘𝑥−𝜔𝑡) = cos(𝑘𝑥 − 𝜔𝑡) + i sin(𝑘𝑥 − 𝜔𝑡)
Consequently, the factor preceding the harmonic function may be interpreted as an
amplitude 𝜁𝑎 , which must be real valued since the surface elevation is real.
i𝜔
𝜁𝑎 = 𝐶 cosh(𝑘ℎ) (.)
𝑔 4
Accordingly, the constant 𝐶4 must be complex and is given by
i𝑔
𝐶4 = − 𝜁 (.)
𝜔 cosh(𝑘ℎ) 𝑎

Solution for This completes the solution of the linearized boundary value problem. We found that
wave elevation the first order wave elevation is a cosine function with amplitude 𝜁𝑎 . In complex notation
the wave profile is
̄ 𝑡) = 𝜁𝑎 ei(𝑘𝑥−𝜔𝑡)
𝜁(𝑥, (.)
However, only its real part ℜ(𝜁) ̄ is of interest to us:
( )
𝜁(𝑥, 𝑡) = ℜ 𝜁(𝑥,̄ 𝑡) = 𝜁𝑎 cos(𝑘𝑥 − 𝜔𝑡) (.)

Solution for The second part of the solution is the first order velocity potential for a long-crested
velocity wave in two dimensions. Equation (.) with the results for constants 𝛼2 (.) and
potential
𝐶4 (.) becomes
j j
̄ 𝑧, 𝑡) = − 𝜁𝑎 i 𝑔 cosh(𝑘𝑧 + 𝑘ℎ) ei𝑘𝑥 e−i𝜔𝑡
𝜙(𝑥, (.)
𝜔 cosh(𝑘ℎ)
In order to compute the velocity field, we need the real valued potential 𝜙.
( )
̄ 𝑧, 𝑡)
𝜙(𝑥, 𝑧, 𝑡) = ℜ 𝜙(𝑥,
𝜁 𝑔 cosh(𝑘𝑧 + 𝑘ℎ) ( i𝑘𝑥 −i𝜔𝑡 )
= − 𝑎 ℜ ie e
𝜔 cosh(𝑘ℎ)
𝜁 𝑔 cosh(𝑘𝑧 + 𝑘ℎ) ( )
= − 𝑎 ℜ i cos(𝑘𝑥 − 𝜔𝑡) + i2 sin(𝑘𝑥 − 𝜔𝑡)
𝜔 cosh(𝑘ℎ)
which results in
𝜁𝑎 𝑔 cosh(𝑘𝑧 + 𝑘ℎ)
𝜙(𝑥, 𝑧, 𝑡) = sin(𝑘𝑥 − 𝜔𝑡) (.)
𝜔 cosh(𝑘ℎ)
Note that the velocity potential (.) is valid for all water depths ℎ as long as the
general assumptions for linear wave theory are satisfied.
Wave parameters A wave on constant water depth ℎ is defined by two parameters:

• wave frequency 𝜔, and

• wave amplitude 𝜁𝑎

Wave frequency and wave number are connected via the dispersion relation (see below).
Therefore, the wave number 𝑘 may be used instead of the wave frequency 𝜔 to define
the wave.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

22.2 Far Field Condition Revisited 265

22.2 Far Field Condition Revisited

In the solution of the boundary value problem we hinted that the choice of e−i𝜔𝑡 (see Direction of
page ) as a basis for the time dependent part of the potential is deliberate to satisfy propagation
the far field condition.
𝑐 = const. ≥ 0 at 𝑆∞ for |𝑥| → ∞ (.)
The far field condition enforces that the wave is moving with constant velocity in positive
𝑥-direction.
The velocity of a wave is determined by how fast the crest-trough-crest contour travels. Constant phase
An observer moving as fast as the wave crest experiences a stationary wave field as we
have observed for our ship wave system. Traveling with a specific point of the wave
contour, like for instance a crest, requires a constant phase 𝜃 = 𝑘𝑥 − 𝜔𝑡 = const.
Therefore, the velocity with which a wave progresses is called phase velocity. If the
phase is to be constant in the far field, it cannot change over time and its derivative
must vanish.
d𝜃 d( )
= 𝑘𝑥 − 𝜔𝑡 = 0 (.)
d𝑡 d𝑡
Wave frequency 𝜔 and wave number 𝑘 are constants for the wave. Therefore, the phase Phase velocity
velocity of a regular wave is
d𝑥 𝜔 𝐿
𝑐 = = = 𝑤 ≥ 0 (.)
d𝑡 𝑘 𝑇
j Since both wave length 𝐿𝑤 and wave period 𝑇 are positive and constant, the phase j
velocity 𝑐 is positive and constant as well. The far field condition is satisfied.
If we had opted for e+i𝜔𝑡 instead of e−i𝜔𝑡 as the basic time function (see Equation (.)),
the condition of constant phase (.) would have resulted in 𝑐 = −𝐿𝑤 ∕𝑇 and the
wave would move in negative 𝑥-direction.

22.3 Dispersion Relation

Back on page , we stated that wave number and wave frequency are connected. The Link between
linearized kinematic free surface condition (.) provides this connection. frequency and
wave number
𝜕𝜁 𝜕𝜙
−+ = 0 on 𝑆𝐹 with 𝑧 = 0 (.)
𝜕𝑡 𝜕𝑧
Based on our linear solution for wave elevation (.) and potential (.), we form
the derivatives:
𝜕𝜁
= 𝜔𝜁𝑎 sin(𝑘𝑥 − 𝜔𝑡)
𝜕𝑡

𝜕𝜙 𝜁 𝑔 sinh(𝑘𝑧 + 𝑘ℎ)
= 𝑘 𝑎 sin(𝑘𝑥 − 𝜔𝑡)
𝜕𝑧 𝜔 cosh(𝑘ℎ)
Both derivatives are evaluated at 𝑧 = 0 and substituted into the kinematic free surface
condition above.
𝜁 𝑔 sinh(𝑘ℎ)
−𝜔𝜁𝑎 sin(𝑘𝑥 − 𝜔𝑡) + 𝑘 𝑎 sin(𝑘𝑥 − 𝜔𝑡) = 0
𝜔 cosh(𝑘ℎ)

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

266 22 Linear Wave Theory

The ratio of hyperbolic sine and cosine is equal to the hyperbolic tangent.
[ 𝑔 ]
𝜁𝑎 sin(𝑘𝑥 − 𝜔𝑡) −𝜔 + 𝑘 tanh(𝑘ℎ) = 0 (.)
𝜔
The latter equation is satisfied when the sine vanishes for 𝜃 = 𝑛𝜋 with 𝑛 = 0, 1, 2, … ,
or, much more generally, if the term in brackets vanishes.
[ 𝑔 ]
−𝜔 + 𝑘 tanh(𝑘ℎ) = 0 (.)
𝜔
Dispersion This may be transformed into
relation
𝜔2 = 𝑘 𝑔 tanh(𝑘ℎ) (.)
which is known as the dispersion relation. Since wave frequencies are considered
positive, one may also write
√
𝜔 = 𝑘 𝑔 tanh(𝑘ℎ) (.)
The relationship between wave frequency and wave number is a nonlinear, transcenden-
tal equation. It is influenced by the gravitational acceleration 𝑔 and the water depth ℎ.
Thus, waves will behave differently depending on the water depth. If one would consider
wave experiments on the Moon, with one sixth of Earth’s gravity, quite different wave
behavior could be expected. Obviously, the wave frequency can be readily determined
for any combination of water depth and wave number. In the next chapter, you will
j learn how to compute the wave number 𝑘 for a given water depth and wave frequency j
using the dispersion relation.
Phase velocity As stated above, the surface elevation moves with phase velocity 𝑐 = 𝜔∕𝑘 = 𝐿𝑤 ∕𝑇 .
Replacing the wave frequency 𝜔 with the dispersion relation (.) yields the phase
velocity as a function of wave length. The water depth ℎ serves as a parameter.
√ √
𝜔 𝑘 𝑔 tanh(𝑘ℎ) 𝑔 tanh(𝑘ℎ)
𝑐 = = = (.)
𝑘 𝑘 𝑘
Figure . visualizes this relationship. The primary horizontal axis is the wave number
𝑘. The lower axis shows the corresponding wave length 𝐿𝑤 . Both axes use a logarithmic
scale. Wave number is increasing from left to right. The reciprocal wave length grows
from right to left, and the vertical axis represents the phase velocity 𝑐. Note that the
vertical axis is logarithmic in the left part of Figure . but linear in the right subfigure
which shows values appropriate for model scale.
Dispersion Several important observations can be made:

• The phase velocity of waves changes with wave length (or wave frequency). As
mentioned, this effect is called dispersion. Water waves are therefore quite
different from sound and electromagnetic waves, which have constant phase
velocities that are independent of wave length (speed of sound, speed of light).
• Waves become faster with increasing wave length (decreasing wave number).
Long waves progress astonishingly fast. A wave of  m length reaches a
velocity of almost  m/s. A wave of  m length travels in deep water as fast
as a commercial air liner.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

22.4 Deep Water Approximation 267

j j

Figure 22.3 Wave phase velocity as function of wave number and water depth based on linear wave theory

• Depending on the water depth ℎ, waves reach a limiting phase velocity. The
lower the water depth, the lower the maximum attainable phase velocity. In very
shallow water, dispersion vanishes and waves of different wave length have the
same phase velocity.

22.4 Deep Water Approximation

Ship trial conditions usually include the requirement of ‘deep water.’ In this context,
water is deep if it does not affect wave motion at the free surface and other parts of the
flow around a ship hull. We still have to clarify which water depth is considered deep
enough.
Figure . shows the hyperbolic tangent function as it appears in the dispersion relation.
For large arguments 𝑘ℎ ⟶ ∞, the hyperbolic tangent asymptotically approaches the

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

268 22 Linear Wave Theory

1.2
lim [tanh(kh)]=1
kh
1.0

kh tanh(kh)
0.8
3.000000.99505
3.141590.99627
tanh(kh)

0.6 4.000000.99933
5.000000.99991
10.000001. 00000
0.4

0.2

0.0
0 1 2 3 4 5
kh

Figure 22.4 The positive arm of the hyperbolic tangent function

value one. For a value of 𝑘ℎ = 𝜋, the deviation is only .% from the asymptotic
limit . For 𝑘ℎ = 5, the difference is less than .% and for 𝑘ℎ = 10, the tangent is
smaller than one by less than 1.0 ⋅ 10−8 .
j j
Deep water For practical purposes, water is considered deep if:
condition
𝑘ℎ ≥ 𝜋

Under this premise, we obtain the following relationship between wave length and
water depth:
2𝜋
𝑘ℎ = ℎ ≥ 𝜋
𝐿𝑤
which is equivalent to
𝐿𝑤
ℎ ≥ or 𝐿𝑤 ≤ 2 ℎ (.)
2
Thus, if the water depth is larger than half the wave length, the influence of the ocean
floor on a wave becomes negligible.
Deep water wave If 𝑘ℎ is large enough, the hyperbolic tangent will be close to one (tanh(𝑘ℎ) ≈ 1) and the
number dispersion relation (.) becomes
√ 𝜔2
𝜔 = 𝑘0 𝑔 or 𝑘0 = (.)
𝑔

𝑘0 is the deep water wave number.

Deep water With the deep water wave number, phase velocity (.) simplifies to
phase velocity
√
𝑔
𝑐0 = (.)
𝑘0

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

22.4 Deep Water Approximation 269

Therefore, the relationships between wave length and phase velocity are
√
𝑔 𝐿𝑤 2𝜋 2
𝑐0 = or 𝐿𝑤 = 𝑐 on deep water (𝑘ℎ ≥ 𝜋) (.)
2𝜋 𝑔 0

The assumption of deep water simplifies the velocity potential as well. The only term Deep water
in the velocity potential (.) (or (.) for the complex form) that depends on the velocity
potential
water depth is the ratio of hyperbolic cosines.
( )
cosh 𝑘(𝑧 + ℎ)
cosh(𝑘ℎ)
The hyperbolic functions may be replaced by a special combination of exponential
functions (Abramowitz and Stegun, , p. ).
1 𝑥
sinh(𝑥) = (e − e−𝑥 ) (.)
2
1
cosh(𝑥) = (e𝑥 + e−𝑥 ) (.)
2
Using the latter for our ratio of hyperbolic cosines yields:
( )
cosh 𝑘(𝑧 + ℎ) e𝑘𝑧+𝑘ℎ + e−(𝑘𝑧+𝑘ℎ) e𝑘𝑧 e𝑘ℎ + e−𝑘𝑧 e−𝑘ℎ
= = (.)
cosh(𝑘ℎ) e𝑘ℎ + e−𝑘ℎ e𝑘ℎ + e−𝑘ℎ

j If the product of wave number and water depth becomes very large, i.e. 𝑘ℎ → ∞, then j
the exponential of (−𝑘ℎ) will vanish e−𝑘ℎ → 0. Therefore, the limit of the ratio of
hyperbolic cosines for deep water is equal to the exponential function with argument
𝑘𝑧.
( ) ⌃0
cosh 𝑘(𝑧 + ℎ) e𝑘𝑧 e𝑘ℎ + e−𝑘𝑧e−𝑘ℎ
 
e𝑘𝑧 e𝑘ℎ
lim = lim = lim = e𝑘0 𝑧 (.)
𝑘ℎ→∞ cosh(𝑘ℎ) 𝑘ℎ→∞ 0 𝑘ℎ→∞ e𝑘ℎ
e−𝑘ℎ
e𝑘ℎ +  ⌃


Finally, the linear velocity potential for waves in deep water becomes

̄ 𝑧, 𝑡) = − 𝜁𝑎 i 𝑔 e𝑘0 𝑧 ei𝑘0 𝑥 e−i𝜔𝑡

𝜙(𝑥, (complex) (.)
𝜔
𝜁𝑎 𝑔 𝑘 𝑧
𝜙(𝑥, 𝑧, 𝑡) = e 0 sin(𝑘0 𝑥 − 𝜔𝑡) (real) (.)
𝜔
This simplification was very important in the pre-computer age. With modern desktop
computers, it is better to work with the exact solution for the velocity potential (.),
which is valid for all water depths.

References
Abramowitz, M. and Stegun, I., editors (). Handbook of mathematical functions
with formulas, graphs, and mathematical tables. Number  in Applied Mathemat-
ics Series. National Bureau of Standards, Washington, DC. Tenth printing, with
corrections.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

270 22 Linear Wave Theory

Self Study Problems

. Summarize the assumptions that have been made in deriving the linearized
boundary value problem (.) through (.).
. Show that the first order velocity potential for a harmonic wave

𝜁𝑎 𝑔 cosh(𝑘𝑧 + 𝑘ℎ)
𝜙(𝑥, 𝑧, 𝑡) = sin(𝑘𝑥 − 𝜔𝑡)
𝜔 cosh(𝑘ℎ)
satisfies the Laplace equation (.) and boundary conditions (.), (.), and
(.).

. At which water depth ℎ is a wave of length 𝐿𝑤 = 80 m considered a deep water

wave?
. Compute wave number and wave frequency for the wave of length 𝐿𝑤 = 80 m in
deep water.
. You have two waves of different wave length: 𝐿𝑤1 = 100 m and 𝐿𝑤2 = 120 m.
Which wave has the greater phase velocity (assume deep water)? Is this still true
for a water depth of 20 m? Use Figure . to justify your answer.
. A ship is producing a transverse wave of wave length 𝐿𝑤 = 120 m. Assuming
deep water, what is the ship’s speed? Use Figure . to check if your answer is
j reasonable. j

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

271

23
Wave Properties

Based on the linear wave theory solution, basic kinematics and dynamics of water
waves are investigated in this and the following chapter. First, a method is introduced
to solve the nonlinear dispersion relation. With the wave number known, wave par-
ticle velocities, accelerations, motions, and unsteady pressure are discussed. A more
comprehensive analysis of water wave properties may be found in offshore engineering
books by Dean and Dalrymple () and by Clauss et al. ().

Learning Objectives

j At the end of this chapter students will be able to: j

• understand wave kinematics and dynamics

• solve the dispersion relation

• apply potential flow theory results

• compute wave properties

23.1 Linear Wave Theory Results

This section briefly summarizes the most important results of Chapter  for the reader
who skipped the finer details of linear wave theory. The results form a basis for the
discussion of wave properties.
The solution, which satisfies the Laplace equation and the linearized boundary con- Regular waves
ditions for long-crested progressing waves, predicts a sinusoidal wave elevation with
amplitude 𝜁𝑎 . Waves with a sinusoidal profile are commonly referred to as regular waves
of wave frequency 𝜔.

̄ 𝑡) = 𝜁𝑎 ei(𝑘𝑥−𝜔𝑡)
𝜁(𝑥, (complex)
(.)
𝜁(𝑥, 𝑡) = 𝜁𝑎 cos(𝑘𝑥 − 𝜔𝑡) (real)

Although the discussion here uses almost exclusively the real valued equations, it should
be pointed out that the complex notation has many advantages in computational tools.

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

272 23 Wave Properties

For example, the superposition of regular waves of different frequencies, amplitudes,

and phases into more realistic looking wave patterns is much easier in the complex
than the real domain.
Velocity The basis for all analysis of kinematics and kinetics of regular waves is the velocity
potential potential:
𝜁𝑎 𝑔 cosh(𝑘𝑧 + 𝑘ℎ) i𝑘𝑥 −i𝜔𝑡
̄ 𝑧, 𝑡) = −i
𝜙(𝑥, e e (complex)
𝜔 cosh(𝑘ℎ)
(.)
𝜁 𝑔 cosh(𝑘𝑧 + 𝑘ℎ)
𝜙(𝑥, 𝑧, 𝑡) = 𝑎 sin(𝑘𝑥 − 𝜔𝑡) (real)
𝜔 cosh(𝑘ℎ)
Equation (.) represents the exact solution of the linearized boundary value problem
and is valid for all water depths ℎ.
Dispersion From the kinematic boundary condition follows the dispersion relation, which expresses
relation the relationship between wave frequency 𝜔, wave number 𝑘, and water depth ℎ.
√
𝜔 = 𝑘 𝑔 tanh(𝑘ℎ) (.)

Deep water Influence of the sea bottom on the wave motion becomes negligibly small for water
approximation depths larger than half the wave length, i.e. ℎ > 𝐿𝑤 ∕2. As shown in Section ., the
velocity potential simplifies to the following expression under the assumption that
𝑘ℎ > 𝜋.
j ̄ 𝑧, 𝑡) = − 𝜁𝑎 i 𝑔 e𝑘0 𝑧 ei𝑘0 𝑥 e−i𝜔𝑡
𝜙(𝑥, (complex) j
𝜔 (.)
𝜁 𝑔
𝜙(𝑥, 𝑧, 𝑡) = 𝑎 e𝑘0 𝑧 sin(𝑘0 𝑥 − 𝜔𝑡) (real)
𝜔
The corresponding deep water dispersion relation states

𝜔2 = 𝑘0 𝑔 (.)

Based on the assumptions made during the solution process, the equations above
are valid for fluids with negligible viscosity and constant density as long as the wave
steepness 𝐻∕𝐿𝑤 ≪ 1 is small.

23.2 Wave Number

Waves are often specified by providing values for wave period 𝑇 and wave amplitude 𝜁𝑎 .
From the wave period follows the wave frequency 𝜔 = 2𝜋∕𝑇 . For deep water (𝑘ℎ > 𝜋),
the wave number 𝑘0 follows immediately from Equation (.).

𝜔2
𝑘0 = (.)
𝑔
A wave of period 𝑇 = 8 s has a frequency of

2𝜋 6.2832
𝜔 = ≈ = 0.7854 s−1
𝑇 8s

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

23.2 Wave Number 273

Therefore, its deep water wave number is

𝜔2 0.78542 s−2
𝑘0 = ≈ = 0.06288 m−1
𝑔 9.81 m s−2
Its associated wave length is 𝐿𝑤 = 2𝜋∕𝑘0 = 99.923 m.
In the case of restricted water depth ℎ, the nonlinear dispersion relation (.) must
be solved. Prior to the age of programmable calculators and desktop computers this
was a cumbersome task, but with today’s computer support it is no longer a hurdle.
The method of choice to solve Equation (.) for the wave number is the Newton- Newton-
Raphson method (see for example Akai, ). Like all methods for nonlinear problems, Raphson
method
it is iterative, i.e. it starts with an initial, approximate solution and improves the solution
up to the desired accuracy. The Newton-Raphson method follows from a Taylor series
expansion around a known solution 𝑘𝑖 . The series is truncated after the linear term and
higher order terms are ignored.
( ) d𝑓 |
𝑓 (𝑘𝑖+1 ) = 𝑓 (𝑘𝑖 ) + 𝑘𝑖+1 − 𝑘𝑖 | (.)
d𝑘 ||𝑘=𝑘𝑖

The function 𝑓 in Equation (.) represents the nonlinear equation to be solved written
in an homogeneous form. We square Equation (.) and shift all terms to the left
side.
j 𝜔2 j
0 = − 𝑘 tanh(𝑘ℎ)
𝑔

Thus, the function 𝑓 (𝑘) and its derivative are

𝜔2
𝑓 (𝑘) = − 𝑘 tanh(𝑘ℎ) (.)
𝑔
d𝑓 ℎ
= −𝑘 − tanh(𝑘ℎ) (.)
d𝑘 cosh2 (𝑘ℎ)

The solution 𝑘 we seek satisfies the condition 𝑓 (𝑘) = 0. Consequently, the left-hand
side of the truncated Taylor series expansion vanishes if we assume that 𝑘 = 𝑘𝑖+1 is the
desired solution.
( ) d𝑓 |
𝑓 (𝑘𝑖+1 ) = 0 = 𝑓 (𝑘𝑖 ) + 𝑘𝑖+1 − 𝑘𝑖 | (.)
d𝑘 ||𝑘=𝑘𝑖

The remainder of Equation (.) is solved for 𝑘𝑖+1 . Iterative

procedure
𝑓 (𝑘𝑖 )
𝑘𝑖+1 = 𝑘𝑖 − (.)
d𝑓 |
|
d𝑘 |𝑘=𝑘𝑖

Equation (.) represents an iterative procedure. Repeated application will converge

toward the real solution in a few steps as long as the derivative in the denominator does
not vanish, i.e. d𝑓 ∕d𝑘 ≠ 0. The deep water wave number 𝑘0 is readily available as a

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

274 23 Wave Properties

starting value. The iterations are stopped when the difference between two subsequent
approximations is smaller than a tiny threshold value 𝜀.

|𝑘𝑖+1 − 𝑘𝑖 | ≤ 𝜀 (.)

Application The following example may serve as an illustration of the procedure. We want to know
example the wave number 𝑘 and wave length 𝐿𝑤 of a wave with period 𝑇 = 8 s when it travels in
water of depth ℎ = 20 m. Consequently, the wave frequency is 𝜔 = 0.7854 s−1 and the
deep water wave number is 𝑘0 = 0.0628797 m−1 . Equipped with 𝑘0 as the starting value,
we compute 𝑓 (𝑘0 ) = 0.0094068 m−1 and the derivative at 𝑘0 as d𝑓 ∕d𝑘 = −1.1985276.
These values are substituted into Equation (.) to derive 𝑘1 = 0.0707284 m−1 . For
the next iteration 𝑖 = 1, the values of 𝑓 (𝑘1 ) and its derivative are computed. Application
of Equation (.) results in 𝑘2 and so on. After 𝑖 = 2 no changes in the first seven
digits can be seen, and 𝑘 = 0.0707624 m−1 is considered the result for the wave number
with sufficient precision for all practical purposes. Values used in the iteration are
summarized in the table below:

𝑖 𝑘𝑖 𝑓 (𝑘𝑖 ) d𝑓 ∕d𝑘 |𝑘𝑖

[−] [m−1 ] [m−1 ] [−]
 0.0628797 0.0094068 −1.1985276
 0.0707284 0.0000404 −1.1864226
 0.0707624 0.0000000 −1.1863489
j j
 0.0707624 −0.0000000 −1.1863489
 0.0707624 0.0000000 −1.1863489
… … … …

As one can see, the iteration converges very fast. Usually less than a handful of iterations
will yield a wave number with high accuracy. The wave length 𝐿𝑤 derived from the
new wave number is 𝐿𝑤 = 2𝜋∕𝑘 = 88.793 m which is considerably less than the wave
length on deep water (𝐿𝑤 = 99.923 m). Note that the length of a wave decreases with
decreasing water depth for constant wave period.
Solutions of nonlinear equations should be thoroughly checked.

(i) Does the solution satisfy the equation? In our case the solution clearly satisfies
𝑓 (𝑘) = 𝑓 (0.0707624) = 0 within numerical precision.
(ii) You may also perform a graphical check by plotting the function 𝑓 (𝑘) in the
vicinity of a solution. Figure . shows the function 𝑓 (𝑘) using the input data
provided in the example. The intersection of 𝑓 (𝑘) with the 𝑘-axis falls on the value
𝑘 = 0.0707624 m−1 as expected.

Therefore, the solution 𝑘 = 0.0707624 m−1 makes sense.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

23.3 Water Particle Velocity and Acceleration 275

0.06
water depth h =20.00m
[1/m]

wave period T =8.000s

0.04 wave frequency =0.7854 0s 1
2
grav. acceleration g =9.807ms
implicit dispersion relation f(k)

0.02

0.00
1
solution f(k)=0 for k =0.0707786m
0.02 2
f(k)= g k tanh(kh)

0.04
0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.10
wave number k [1/m]

Figure 23.1 Graphical verification of a solution of the nonlinear dispersion relation

23.3 Water Particle Velocity and Acceleration

j j
Equipped with

• water depth ℎ,

• wave frequency 𝜔,

• wave number 𝑘,

• and wave amplitude 𝜁𝑎

the real-valued velocity potential 𝜙 from Equation (.) provides the tool to investi-
gate water wave kinematics.
As we explored in our discussion of potential flow and specifically the flow around a Water particle
cylinder, flow velocities follow from the spatial derivatives of the velocity potential. The velocity
velocity vector 𝑣 = (𝑢, 𝑤)𝑇 in the two-dimensional wave flow problem has components
in 𝑥- and 𝑧-directions.
𝜕𝜙
𝑢 = horizontal particle velocity
𝜕𝑥
[ ]
𝜕 𝜁𝑎 𝑔 cosh(𝑘𝑧 + 𝑘ℎ)
= sin(𝑘𝑥 − 𝜔𝑡)
𝜕𝑥 𝜔 cosh(𝑘ℎ)
𝜁𝑎 𝑔 cosh(𝑘𝑧 + 𝑘ℎ)
= 𝑘 cos(𝑘𝑥 − 𝜔𝑡) (.)
𝜔 cosh(𝑘ℎ)

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

276 23 Wave Properties

𝜕𝜙
𝑤 = vertical particle velocity
𝜕𝑧
[ ]
𝜕 𝜁𝑎 𝑔 cosh(𝑘𝑧 + 𝑘ℎ)
= sin(𝑘𝑥 − 𝜔𝑡)
𝜕𝑧 𝜔 cosh(𝑘ℎ)
𝜁𝑎 𝑔 sinh(𝑘𝑧 + 𝑘ℎ)
= 𝑘 sin(𝑘𝑥 − 𝜔𝑡) (.)
𝜔 cosh(𝑘ℎ)
If the dispersion relation (.) is brought into the form (.)

𝜔2
𝑘𝑔 = (.)
tanh(𝑘ℎ)
and substituted into the equation for the horizontal velocity 𝑢, one gets:

𝜔2 𝜁𝑎 cosh(𝑘𝑧 + 𝑘ℎ)
𝑢 = cos(𝑘𝑥 − 𝜔𝑡)
tanh(𝑘ℎ) 𝜔 cosh(𝑘ℎ)
cosh(𝑘𝑧 + 𝑘ℎ)
= 𝜁𝑎 𝜔 cos(𝑘𝑥 − 𝜔𝑡)
tanh(𝑘ℎ) cosh(𝑘ℎ)
Exploiting that tanh(𝑘ℎ) = sinh(𝑘ℎ)∕ cosh(𝑘ℎ) yields:

cosh(𝑘𝑧 + 𝑘ℎ)
𝑢 = 𝜁𝑎 𝜔 cos(𝑘𝑥 − 𝜔𝑡) (.)
sinh(𝑘ℎ)
j j

The analog result for the vertical particle velocity 𝑤 is

sinh(𝑘𝑧 + 𝑘ℎ)
𝑤 = 𝜁𝑎 𝜔 sin(𝑘𝑥 − 𝜔𝑡) (.)
sinh(𝑘ℎ)
Both velocity components 𝑢 and 𝑤 are harmonic functions of time and 𝑥-coordinate.
The factor preceding the sine and cosine functions may be interpreted as velocity
amplitudes 𝑢𝑎 and 𝑤𝑎 which are a function of submergence 𝑧.
Water particle The water particle acceleration is of great importance in the computation of wave forces
acceleration and vessel motions. Equations for the horizontal and vertical accelerations are obtained
by taking the time derivatives of Equations (.) and (.). Observe that the time
derivative of 𝜃 = 𝑘𝑥 − 𝜔𝑡 introduces an additional factor −𝜔.
.
𝑢 =
𝜕𝑢
= 𝜁𝑎 𝜔2
cosh(𝑘𝑧 + 𝑘ℎ)
sin(𝑘𝑥 − 𝜔𝑡) (.)
𝜕𝑡 sinh(𝑘ℎ)

. 𝜕𝑤 sinh(𝑘𝑧 + 𝑘ℎ)
𝑤 = = −𝜁𝑎 𝜔2 cos(𝑘𝑥 − 𝜔𝑡) (.)
𝜕𝑡 sinh(𝑘ℎ)

Phase Figure . shows a snapshot of several wave properties at time (𝑡 = 0). From top to
relationships bottom, we see wave elevation 𝜁, particle velocities 𝑢 and 𝑤, and particle accelerations
. .
𝑢 and 𝑤. The graph at the bottom of Figure . shows the dynamic pressure, which we
will discuss in the next section. Each curve represents the spatial distribution of the
respective property over . wave lengths, calculated at the calm water level (𝑧 = 0).

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

23.3 Water Particle Velocity and Acceleration 277

j j

Figure 23.2 Distribution of wave properties over 1.5 wavelength at the calm water level (𝑧 = 0)
for a wave with wave period 𝑇 = 10 s

Wave elevation 𝜁 and horizontal particle velocity 𝑢 are in phase, i.e. wave elevation has Velocity
its positive maximum at the same time as the horizontal particle velocity. Under a wave amplitudes
crest, particles move with maximum velocity in direction of wave propagation, whereas
under a wave trough, the particles move with maximum velocity against the direction
of wave propagation. The amplitude of the wave elevation is obviously equal to 𝜁𝑎 . At

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

278 23 Wave Properties

𝑧 = 0 we obtain an amplitude 𝑢𝑎 of the velocity of

cosh(𝑘𝑧 + 𝑘ℎ) ||
𝑢 𝑎 = 𝜁𝑎 𝜔
sinh(𝑘ℎ) ||𝑧=0
𝜁𝑎 𝜔
= for 𝑧 = 0
tanh(𝑘ℎ)
For a deep water wave, the numerator becomes equal to one and the maximum hori-
zontal particle velocity is 𝑢𝑎 = 𝜁𝑎 𝜔.
The vertical particle velocity 𝑤 is phase shifted by a quarter cycle (Δ𝜃 = −𝜋∕2). It
is zero when the wave elevation is at its maximum or minimum (Figure .). The
maximum value 𝑤𝑎 at 𝑧 = 0 is equal to
2𝜋𝜁𝑎
𝑤𝑎 = 𝜁𝑎 𝜔 = for 𝑧 = 0 (.)
𝑇
There is no difference in vertical velocity amplitude 𝑤𝑎 whether the wave is in deep
water or restricted water depths. An expansion of wave amplitude multiplied by wave
frequency on the right-hand side reveals that the maximum vertical particle velocity is
equal to the ratio of the circumference of a circle with radius 𝜁𝑎 and the wave period 𝑇 .
Acceleration
.
The horizontal particle acceleration 𝑢 is phase shifted by Δ𝜃 = −𝜋∕2 like the vertical
amplitudes particle velocity (Figure ., fourth graph). Its maximum value is for 𝑧 = 0:

j . 𝜁𝑎 𝜔2 j
𝑢𝑎 = for 𝑧 = 0 (.)
tanh(𝑘ℎ)
For the amplitude of the vertical particle acceleration we get at the calm water level
(𝑧 = 0):
.
𝑤𝑎 = 𝜁𝑎 𝜔2 for 𝑧 = 0 (.)
.
𝑤 is in opposite phase to the wave elevation. Particles experience the largest accelera-
tions downward under a crest and upward under a wave trough.
Vertical Figure . emphasizes the change of particle velocities with increasing submergence.
distribution of We learned from Figure . that the horizontal particle velocity has its positive maxi-
velocities
mum (in the direction of wave propagation) underneath the wave crest. The vertical
particle velocity has its positive (upward) maximum where the wave elevation vanishes
in front of the approaching wave crest.
The horizontal velocity vanishes where the vertical velocity has an extreme value and vice
versa. Velocities are largest at the calm water level and rapidly decline with increasing
depth. The decline is governed by the expressions
cosh(𝑘𝑧 + 𝑘ℎ) sinh(𝑘𝑧 + 𝑘ℎ)
and
sinh(𝑘ℎ) sinh(𝑘ℎ)
for horizontal and vertical particle velocities respectively. On deep water both velocities
decline with e𝑘𝑧 .
The vertical particle velocity vanishes at the sea bottom (𝑧 = −ℎ) as requested by the
bottom boundary condition (.). However, in restricted water depth the horizontal
particle velocity does not vanish at the sea bottom. Every passing wave moves water

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

23.4 Dynamic Pressure 279

Figure 23.3 Snapshot (𝑡 = 0) of the velocity field for a wave in restricted water depth

back and forth parallel to the bottom. Just watch the sand rolling back and forth on the
sea bottom next time you visit a beach.
Observe that the origin of the 𝑧-axis has been laid into the calm water level. Since Mind the vertical
j the 𝑧-axis is positive upward, values of 𝑧 are negative for all points in the water. The axis
j
attentive reader will notice that the velocity distributions 𝑢(𝑧) and 𝑤(𝑧) in Figure .
have been drawn up to the calm water level irrespective of whether the position is under
a wave crest or a wave trough. It shall serve as a reminder that in linear wave theory
wave amplitudes are assumed to be very small compared with the wave length.

23.4 Dynamic Pressure

The pressure distribution in a wave is obtained from the Bernoulli equation (see Sec-
tions .). Since we linearized the boundary value problem, it is opportune to use the
linearized form of the Bernoulli equation.
𝜕𝜙
𝑝 − 𝑝∞ = −𝜌 − 𝜌𝑔𝑧 (.)
𝜕𝑡
The dynamic pressure in the wave field is equal to the unsteady portion of the pressure:
𝜕𝜙
𝑝dyn = −𝜌
𝜕𝑡 [ ]
𝜕 𝜁𝑎 𝑔 cosh(𝑘𝑧 + 𝑘ℎ)
= −𝜌 sin(𝑘𝑥 − 𝜔𝑡)
𝜕𝑡 𝜔 cosh(𝑘ℎ)
𝜁𝑎 𝑔 cosh(𝑘𝑧 + 𝑘ℎ)
= 𝜌𝜔 cos(𝑘𝑥 − 𝜔𝑡)
𝜔 cosh(𝑘ℎ)
cosh(𝑘𝑧 + 𝑘ℎ)
= 𝜌 𝑔 𝜁𝑎 cos(𝑘𝑥 − 𝜔𝑡) (.)
cosh(𝑘ℎ)

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

280 23 Wave Properties

Figure 23.4 Amplitude of dynamic pressure over

0
depth 𝑧

10

20

[m]
depth z 30

40
water depth h =50.00m
wave period T =10.00s
1
wave number k =0.04154m
2
grav. acceleration g =9.807ms
50
0.0 0.2 0.4 0.6 0.8 1.0
pdyna
dynamic pressure amplitude []
g a
j j

As the bottom graph in Figure . shows, dynamic pressure is in phase with the wave
elevation. The maximum value is equivalent to the hydrostatic pressure underneath
a water column of height 𝜁𝑎 . The positive maximum is found at the calm water level
(𝑧 = 0) under wave crests (𝜃 = 𝑛𝜋, with 𝑛 = 0, 2, 4, … ).

𝑝dyn𝑎 = 𝜌 𝑔 𝜁𝑎 for 𝑧 = 0 (.)

The dynamic pressure 𝑝dyn at the sea bottom 𝑧 = −ℎ fluctuates between the values

𝜌 𝑔 𝜁𝑎 𝜌 𝑔 𝜁𝑎
− < 𝑝dyn < + for 𝑧 = −ℎ.
cosh(𝑘ℎ) cosh(𝑘ℎ)

Like wave particle velocities and accelerations, dynamic pressure amplitude declines
rapidly with depth 𝑧 (see Figure .). Dynamic pressure is positive underneath wave
crests and negative underneath wave troughs.

23.5 Water Particle Motions

Particle paths Integration of the wave particle velocities Equations (.) and (.) (or their deep
water equivalents) over time 𝑡 results in the paths of a water particle with initial position
(𝑥, 𝑧). The photo in Figure . shows the unsteady movement of small, neutrally
buoyant particles underneath a wave. The picture was taken in a small wave flume at the

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

23.5 Water Particle Motions 281

Figure 23.5 Photo of water particle trajectories. Photo courtesy of Dr. Walter L. Kuehnlein,
sea2ice Ltd. & Co. KG, www.seaice.com

Technical University of Berlin. Its walls are made out of acrylic. The photo’s exposure
j j
time was chosen to be equal to one wave period. Wave steepness 𝐻∕𝐿𝑤 is small and, as
linear wave theory predicts, paths of particles are closed. Therefore, water particles are
not transported along with the wave elevation. On average, they stay at their original
location (𝑥, 𝑧). The vertical axis of the elliptical paths declines visibly with depths. The
horizontal extensions decline very little because the wave length is about  times the
water depth. As a consequence, shallow water effects are strong.
In steep waves with high 𝐻∕𝐿𝑤 ratio, particle paths are no longer closed. They look
like a spiral and during each wave cycle particles are moved a small distance further
in the direction of wave propagation. This mass transport contributes considerably to
coastal storm surge in heavy storms like hurricanes.
Figure . depicts water particle trajectories for deep water (left side) and restricted Deep water
water depth (on the right). The × mark the position of water particles when they are trajectories
at rest. For deep water the trajectories are circles. The diameter is largest at the calm
water level and equal to twice the wave amplitude. With increasing submergence the
diameters of trajectories decline rapidly and vanish almost completely at a depth of half
the wave length.
On the right side of Figure ., trajectories are plotted for the same wave period and Trajectories on
amplitude but consider a finite water depth of ℎ = 50 m. Due to the influence of the restricted water
depth
ocean floor, wave length has shrunk by %. The ratio of water depth to wave length is
about ℎ∕𝐿𝑤 ≈ 0.2. The trajectories have become ellipses. The eccentricity is still small
close to the water surface but increases with increasing submergence. At the ocean
floor, the vertical axis of all trajectories vanishes and particles simply move back and
forth parallel to the bottom (see also Figure .). This is of course a consequence of
the bottom boundary condition which prohibits flow through the ocean floor.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

282 23 Wave Properties

horizontal part. displacement x [m] horizontal part. displacement x [m]

10 5 0 5 10 10 5 0 5 10

0 0

10 10
vertical particle displacement z [m]

restricted water depth h

20 20
deep water

30 30
j j
depth z,

40 40
a = 2.500 m a =2 .500m

T = 12.00 s T =1 2.00s
Lw = 224.760 m Lw =204.784m
h h =5 0.000m
50 50

Figure 23.6 Water particle trajectories over one wave period 𝑇 for deep water (left) and
restricted water (right)

References
Akai, T. (). Applied numerical methods for engineers. John Wiley & Sons, Inc.,
New York, NY.
Clauss, G., Lehmann, E., and Östergaard, C. (). Offshore structures, volume :
Conceptual design and hydrodynamics. Springer Verlag, London.
Dean, R. and Dalrymple, R. (). Water wave mechanics for engineers and scientists,
volume  of Advanced Series on Ocean Engineering. World Scientific, Hackensack,
NJ.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

23.5 Water Particle Motions 283

Self Study Problems

. Compute the wave number 𝑘 and wave length 𝐿𝑤 for a wave with period 𝑇 = 6 s
in ℎ = 12 m water depth.
. Will the wave length become shorter or longer when the wave propagates from
deep into shallow water?
. Compute the wave length for the wave with frequency 𝜔 = 0.95 s−1 in water of
depth ℎ = 25 m.

. Find the wave length 𝐿𝑤 of a wave of frequency 𝜔 = 0.95 s−1 in deep water.
. Use the real-valued deep water potential
𝜁𝑎 𝑔 𝑘 𝑧
𝜙(𝑥, 𝑧, 𝑡) = e 0 sin(𝑘0 𝑥 − 𝜔𝑡) (.)
𝜔
to derive equations for the water particle velocities, accelerations, and dynamic
pressure for the case 𝑘ℎ > 𝜋.
. Use the equations from Problem  to plot maxima of velocities 𝑢𝑎 (𝑧), 𝑤𝑎 (𝑧),
. .
accelerations 𝑢𝑎 (𝑧), 𝑤𝑎 (𝑧), and dynamic pressure 𝑝dyn (𝑧) from 𝑧 = −𝐿𝑤 ∕2 to
𝑧 = 0 for a wave with period 𝑇 = 5 s and amplitude 𝜁𝑎 = 1 m.

j . Derive the equations for the complex water particle velocities and accelerations j
from the complex linear wave theory velocity potential (.).

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

284

24
Wave Energy and Wave Propagation

The preceding chapter discusses wave kinematics and the distribution of wave particle
velocities, accelerations, and dynamic pressure throughout the water column under-
neath a long crested wave. In this chapter, we will use these results to define the energy
contained in a regular wave and how the wave and its energy propagate. The results will
form the basis for a brief discussion of ship wave resistance in the following chapter.

Learning Objectives
At the end of this chapter students will be able to:
j j
• distinguish between the kinematic energy and potential energy of waves

• describe and discuss how wave energy propagates

• compute group velocity and energy density of waves

24.1 Wave Propagation

Wave elevation Over the time of a single wave period 𝑇 a wave travels the distance of one wave length
𝐿𝑤 . We used this to define the phase velocity 𝑐 = 𝜔∕𝑘 in Equation (.). See also
Section .. Figure . depicts a wave in  time instances during one wave period
(𝜔𝑡 ∈ [0, 2𝜋]). The wave crest, marked with a small triangle ▿, travels exactly one wave
length from 𝑘𝑥 = 0 to 𝑘𝑥 = 2𝜋.
Water particles It is important to understand that the wave elevation is propagating but not the water
particles. In the center of Figure . at 𝑘𝑥 = 2𝜋, a particle (small dot) is tracked
over one wave period. On average, it stays in the same position. Its horizontal and
vertical excursions are equal to the wave amplitude ±𝜁𝑎 . If you look closely, you will
notice that the dot marking the water particle is not always exactly on the plotted wave
contour. This is most evident at the locations of maximum horizontal excursion. The
apparent discrepancy is a consequence of the linearization of the free surface boundary
conditions. The harmonic wave elevation 𝜁 = 𝜁𝑎 cos(𝑘𝑥 − 𝜔𝑡) is only truly valid for
vanishing wave amplitude or more precisely if 2𝜁𝑎 ∕𝐿𝑤 ≪ 1.

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

24.1 Wave Propagation 285

wave crest water particle

0

2
6

4
6

5
6
[]
ttime

7
6

8
6

9
6

j j
10
6

11
6

0 3 2 5 3
2 2 2
position kx []

Figure 24.1 Propagation of wave profile and the movement of a water particle over one wave period

The larger the wave steepness 𝐻∕𝐿𝑤 , the less accurate is the sinusoidal wave profile. Mass transport in
In fact, as mentioned previously, in steep waves the particle paths are no longer closed steep waves
loops, and a slow but steady mass transport occurs in the direction of wave propagation.
Once the water depth ℎ is less than half the wave length, the sea bottom begins to affect Effects of finite
the way a wave propagates. Did you ever wonder why waves always roll onto a beach water depths
with their crests and troughs parallel to the shoreline? As a wave moves from deeper
into shallower water, its wave length becomes shorter. Figure . shows the wave
length 𝐿𝑤 as a function of water depth ℎ for a selection of wave periods 𝑇 . The wave
length stays constant in deep water where 2ℎ > 𝐿𝑤 .
A decreasing wave length causes the wave number 𝑘 to increase. As a consequence, the
wave phase velocity 𝑐 = 𝜔∕𝑘 must decrease, i.e. a wave slows down as it moves into

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

286 24 Wave Energy and Wave Propagation

j j
Figure 24.2 Wave length 𝐿𝑤 as a function of water depth ℎ for constant wave period 𝑇

Figure 24.3 Effect of gradually decreasing water depth ℎ on wave propagation and direction

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

24.2 Wave Energy 287

shallower water. Figure . shows waves approaching a beach. Further offshore, in
deeper water, waves move in the direction of prevailing winds. In the figure, waves
travel at an oblique angle with respect to the beach. The parts of waves closer to shore
will be affected first by the decreasing water depth. Those parts will slow down while
the parts further offshore are still traveling at the original phase velocity. The difference
in phase velocity causes the wave to turn and eventually its crest becomes parallel to
the shoreline. As it further approaches the beach, the front of a wave continues to
slow down. The wave becomes higher and steeper, and it will eventually break if the
steepness exceeds 𝐻∕𝐿𝑤 > 1∕7 (Michell, ).

24.2 Wave Energy

Whoever has sailed through a storm at sea or watched waves crashing into a sea wall
has experienced the destructive force of waves first hand. In fact, waves carry so much
energy that it is a potential source of renewable power. The Bureau of Ocean Energy
Management states on its web pages (https://www.boem.gov/Ocean-Wave-Energy/)
that in the United States alone the recoverable wave power along the U.S. shelf edge
would amount to more than  TWh/yr (i.e. tera-watt hours per year), which is
enough to power millions of homes.
Waves contain two forms of energy: Contributions to
wave energy
j j
(i) Kinetic energy: In the previous chapter we discussed the motion of water particles
in waves. Obviously, if a mass of fluid moves, it possesses kinetic energy.

(ii) Potential energy: The water surface at a selected point moves up and down in a
way similar to the movement of a mass hanging from a spring. In fact, when a wave
crest passes, fluid elements have to be lifted above their equilibrium condition.
The necessary work done by the flow is converted into potential energy in the
gravity field.

Below, both forms of energy are computed and, as before, it is assumed that the wave is
regular and its wave steepness is small enough to apply linear wave theory.

24.2.1 Kinetic wave energy

Figure . shows a slice of ocean 𝑉 which is one wave length 𝐿𝑤 long, has a width of Kinetic energy
𝑏 in 𝑦-direction, and stretches from the ocean floor at 𝑧 = −ℎ to the calm water level
𝑧 = 0. The wave propagates in 𝑥-direction and the wave crests are parallel to the 𝑦-axis
(long crested wave). The volume is cut into small volumes d𝑉 of size

d𝑉 = d𝑥 d𝑧 𝑏 (.)

The kinetic energy contained in the volume d𝑉 is

1
d𝐸kin = 𝜌 |𝑣|2 d𝑉 (.)
2

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

288 24 Wave Energy and Wave Propagation

Figure 24.4 Kinetic energy 𝐸kin in the control volume 𝑉 spanning one wave length 𝐿𝑤 in
𝑥-direction

Obviously, the total kinetic energy in the fluid space 𝑉 is found by integrating d𝐸kin
across the volume.
1
𝐸kin = 𝜌 |𝑣|2 d𝑉 (.)
∫ 2
𝑉
j j
Since the flow is planar in the 𝑥-𝑧-plane, the volume integral reduces to a D integral
over wave length and water depth. The components of the velocity vector 𝑣𝑇 = (𝑢, 𝑤)
are given by Equations (.) and (.).

0 𝐿𝑤
1 ( 2 )
𝐸kin = 𝜌𝑏 𝑢 + 𝑤2 d𝑥d𝑧
2 ∫ ∫
−ℎ 0
( )2 0 𝐿𝑤
1 𝜁𝑎 𝜔 [
= 𝜌𝑏 cosh2 (𝑘𝑧 + 𝑘ℎ) cos2 (𝜃)
2 sinh(𝑘ℎ) ∫ ∫
−ℎ 0
]
+ sinh2 (𝑘𝑧 + 𝑘ℎ) sin2 (𝜃) d𝑥d𝑧 (.)

Simplifying the As a first step, the integrand is simplified using the following identities for the hyperbolic
integrand cosine and sine functions.
( ) ( )
1 𝛼 1 𝛼
cosh(𝛼) = e + e−𝛼 and sinh(𝛼) = e − e−𝛼
2 2

from which follow the relationships

( ) ( )
1 2𝛼 1 2𝛼
cosh2 (𝛼) = e + e−2𝛼 + 2 and sinh2 (𝛼) = e + e−2𝛼 − 2
4 4

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

24.2 Wave Energy 289

Setting 𝛼 = 𝑘𝑧 + 𝑘ℎ, the integrand in Equation (.) becomes

cosh2 (𝛼) cos2 (𝜃) + sinh2 (𝛼) sin2 (𝜃)

1 [( 2𝛼 ) ( ) ]
= e + e−2𝛼 + 2 cos2 (𝜃) + e2𝛼 + e−2𝛼 − 2 sin2 (𝜃)
4
1 [( 2𝛼 )( ) ( )]
= e + e−2𝛼 cos2 (𝜃) + sin2 (𝜃) + 2 cos2 (𝜃) − sin2 (𝜃)
4
1 [( 2𝛼 ) ( )]
= e + e−2𝛼 + 2 cos2 (𝜃) − sin2 (𝜃)
4

For the second part, we used the identity 1 = cos2 (𝜃) + sin2 (𝜃).
Further, with cos(2𝜃) = cos2 (𝜃) − sin2 (𝜃) we finally get a much simpler form of the
integrand:

cosh2 (𝛼) cos2 (𝜃) + sinh2 (𝛼) sin2 (𝜃)

[ ]
1
= cosh(2𝑘𝑧 + 2𝑘ℎ) + cos(2𝑘𝑥 − 2𝜔𝑡) (.)
2

This expression captures the magnitude of velocity |𝑣| as a function of 𝑥 and 𝑧.

Equation (.) is rewritten based on the result (.): Integration

j ( )2 ⎧ 0 𝐿𝑤 j
𝜁𝑎 𝜔 ⎪
𝐸kin = 𝜌 𝑏
sinh(𝑘ℎ) ⎨ ∫ ∫ cosh(2𝑘𝑧 + 2𝑘ℎ) d𝑥d𝑧
⎪−ℎ 0
⎩
0 𝐿𝑤 ⎫
⎪
+ cos(2𝑘𝑥 − 2𝜔𝑡) d𝑥d𝑧⎬ (.)
∫ ∫ ⎪
−ℎ 0 ⎭

The remaining two integrals are of standard form.

The integrand of the first integral in Equation (.) is independent of 𝑥. The 𝑥-
integration simply yields a factor of one wave length and the integral becomes:

0 𝐿𝑤 0

cosh(2𝑘𝑧 + 2𝑘ℎ) d𝑥d𝑧 = 𝐿𝑤 cosh(2𝑘𝑧 + 2𝑘ℎ) d𝑧

∫ ∫ ∫
−ℎ 0 −ℎ
[ ]0
1
= 𝐿𝑤 sinh(2𝑘𝑧 + 2𝑘ℎ) (.)
2𝑘 −ℎ

𝐿𝑤
= sinh(2𝑘ℎ)
2𝑘

The second integrand in Equation (.) is only a function of 𝑥 and time 𝑡. Integration
over the water depth yields the factor ℎ. However, the remaining integral over the wave
length vanishes.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

290 24 Wave Energy and Wave Propagation

0 𝐿𝑤 𝐿𝑤 [ ]𝐿𝑤
sin(2𝑘𝑥 − 2𝜔𝑡)
cos(2𝑘𝑥 − 2𝜔𝑡) d𝑥d𝑧 = ℎ cos(2𝑘𝑥 − 2𝜔𝑡) d𝑥 = ℎ
∫ ∫ ∫ 2𝑘 0
−ℎ 0 0
[ ]𝐿𝑤
sin(2𝑘𝑥) cos(2𝜔𝑡) − cos(2𝑘𝑥) sin(2𝜔𝑡)
= ℎ (.)
2𝑘 0

= 0

Kinetic energy in With the results from Equations (.) and (.), we obtain from Equation (.) the
control volume kinetic wave energy in the control volume 𝑉 :
𝑉
( )2
𝜁𝑎 𝜔 1
𝐸kin = 𝜌 𝑏 𝐿𝑤 sinh(2𝑘ℎ)
2 sinh(𝑘ℎ) 2𝑘

Utilizing that sinh(2𝑘ℎ) = 2 sinh(𝑘ℎ) cosh(𝑘ℎ) results in

( 2 )
1 𝜔 cosh(𝑘ℎ)
𝐸kin = 𝜌 𝑏 𝐿𝑤 𝜁𝑎2 (.)
4 𝑘 sinh(𝑘ℎ)

The ratio of hyperbolic cosine and sine is equivalent to the reciprocal hyperbolic tangent:

j cosh(𝑘ℎ) 1 j
= (.)
sinh(𝑘ℎ) tanh(𝑘ℎ)
The resulting expression may be simplified based on the dispersion relation (.).

𝜔2
= 𝑔 (.)
𝑘 tanh(𝑘ℎ)
Thus, the kinetic energy contained in our control volume 𝑉 is
1
𝐸kin = 𝜌 𝑔 𝜁𝑎2 𝑏 𝐿𝑤 (.)
4

Kinetic wave The kinetic energy per unit ocean surface is known as kinetic wave energy density. We
energy density divide Equation (.) by the area of the free surface 𝑏 𝐿𝑤 .
𝐸kin 1
𝐸 kin = = 𝜌 𝑔 𝜁𝑎2 (.)
𝑏 𝐿𝑤 4
Note that the energy density is independent of water depth and wave frequency. It is
solely a function of the squared wave amplitude.

24.2.2 Potential wave energy

Potential energy Potential energy in a gravity field is determined by the gravitational acceleration, eleva-
tion above a datum level, and mass of the object. When calculating the potential wave
energy, we consider the work done when a volume of water is lifted above the calm

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

24.2 Wave Energy 291

Figure 24.5 Change in potential energy 𝐸pot when a fluid element is lifted above the calm
water level

water level. Figure . shows a piece of the free surface. We consider a small volume
d𝑉 of length d𝑥, width 𝑏, and height equal to the local wave elevation 𝜁. Its mass will be
d𝑚 = 𝜌 d𝑉 = 𝜌 𝑏 𝜁 d𝑥 (.)
The center of a mass element d𝑚 is lifted above the calm water surface by half of its
height. The work necessary for the lift is equal to the change in potential energy.
𝜁 1
d𝑚 = 𝜌 𝑔 𝑏 𝜁 2 d𝑥
d𝐸pot = 𝑔 (.)
2 2
The same work has to be done to depress the free surface. Outside the range of a
progressing wave elevation, overall mass distribution remains unchanged and water
j j
fills the control volume. Therefore, changes in potential energy occur at the free surface
only.
We integrate Equation (.) over one wave length 𝑥 ∈ [0, 𝐿𝑤 ]. Potential wave
energy
2𝜋
1
𝐸pot = 𝜌𝑔𝑏 𝜁 2 d𝑥
2 ∫
0
2𝜋
1
= 𝜌𝑔𝑏 𝜁 2 cos2 (𝑘𝑥 − 𝜔𝑡) d𝑥 (.)
2 ∫ 𝑎
0
[ ]𝐿𝑤
1 𝑥 1
= 𝜌 𝑔 𝑏 𝜁𝑎2 − sin(2𝑘𝑥 − 2𝜔𝑡)
2 2 4𝑘 0
( )
1 2 𝐿 𝑤
= 𝜌 𝑔 𝑏 𝜁𝑎 −0
2 2

Thus, the potential energy distributed over one wave length is equal to the kinetic
energy.
1
𝐸pot = 𝜌 𝑔 𝜁𝑎2 𝑏 𝐿𝑤 (.)
4

The potential wave energy density is Potential wave

energy density
𝐸pot 1
𝐸 pot = = 𝜌 𝑔 𝜁𝑎2 (.)
𝑏 𝐿𝑤 4

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

292 24 Wave Energy and Wave Propagation

1.5

Ekin + Epot
mean total wave
energy density Etot
[−]

1.0 potential wave

energy density
2
a

Epot
E
g 1
2
energy density

0.5

kinetic wave energy density Ekin

0.0
0 /2 3/2 2
phase [rad]
j j
Figure 24.6 Wave energy density distribution of a regular wave over one wave cycle according
to linear wave theory

24.2.3 Total wave energy density

The total energy density is equal to the sum of kinetic and potential wave energy density.
According to linear wave theory, the energy density of a regular wave amounts to

1
𝐸 = 𝐸 pot + 𝐸 kin = 𝜌 𝑔 𝜁𝑎2 (.)
2
This result is significant as it relates to the statistical energy density of regular waves
𝑆(𝜔)d𝜔 = 𝜁𝑎2 ∕2. 𝑆(𝜔) is the value of the spectral energy density spectrum also known
as wave spectrum (Dean and Dalrymple, ). The latter plays an important role in
the statistical treatment of irregular sea states which are modeled as a superposition
of many regular components with varying frequency and phase. More on this may be
found in offshore engineering references like Barltrop and Adams ().
For our discussion of wave resistance, we may draw the conclusion that doubling the
wave amplitude in the ship’s wave pattern will quadruple its energy density. Low wave
resistance is achieved if wave amplitudes in the wave pattern stay as small as possible.
Distribution of Figure . shows the distribution of wave energy density for a wave over one period.
wave energy Kinetic energy density is constant across a wave cycle because the amplitude of the wave
density
particle velocity is only varying with submergence. Equation (.) reveals that the

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

24.3 Energy Transport and Group Velocity 293

potential wave energy density follows a cos2 distribution over one cycle. The potential
wave energy density is at a maximum for wave crest and trough. It vanishes at the zero
crossings of the wave elevation. However, as we calculated in (.), its mean value is
equal to the kinetic energy density distribution.

24.3 Energy Transport and Group Velocity

Waves transport energy in the direction of wave propagation. However, since mass
transport does not exist in linear waves due to the closed particle paths, only the
potential wave energy portion is transported by the wave. In this section we determine
how fast wave energy propagates.
First, we study the idealized simulation depicted in Figure .. It follows an idea taken
from Clauss et al. (). Imagine a wave maker which transfers  energy density
units into the water during each wave cycle. The horizontal axis in Figure . marks
the distance from the wave maker which acts at location 𝑥 = 0. The vertical scale
depicts time measured from top to bottom in wave periods.
Beginning at the top of Figure ., we track the energy density distribution over  Simulation from
wave maker cycles. The wave maker begins by introducing  energy density units, cycle 0 to 1
one half of which is converted into kinetic energy; the other half becomes potential
energy. During the first cycle, the potential energy propagates one wave length forward
j along with the wave elevation. The kinetic wave energy remains at the wave maker j
which adds another  energy density units. At the end of cycle  a total of 
energy density units are at the wave maker while  units have been propagated to
location 𝑥 = 𝐿𝑤 .
At station 𝑥 = 𝐿𝑤 , half of the  units must be used to get the water particles moving Simulation from
(kinetic energy) and subsequently remain at 𝑥 = 𝐿𝑤 . The other half ( units) is cycle 1 to 2
propagated to station 𝑥 = 2𝐿𝑤 . At the same time, half of the  energy density units
from 𝑥 = 0 have moved to station 𝑥 = 𝐿𝑤 . The remaining  units at 𝑥 = 0 are
augmented by another  units from the wave maker. Thus, the energy distribution
after two wave maker cycles is

position 𝑥∕𝐿𝑤   
energy density units   

The lowest row in Figure . shows the energy density distribution after  wave maker
cycles. Note that we are slowly approaching equilibrium at the wave maker, in which as
much energy is propagated as is added. With increasing distance from the wave maker,
energy density declines rapidly. Since the energy density is tied to the wave amplitude,
the latter must diminish as well with increasing distance from the wave maker.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

294 24 Wave Energy and Wave Propagation

Birk — “fshy” — // — : — page  — #
j

j
Trim Size: mm × mm Single Column Tight

Figure 24.7 Schematic propagation of wave energy for a deep water wave based on linear wave theory

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

24.3 Energy Transport and Group Velocity 295

j j

Figure 24.8 Wave elevation profiles after a few selected cycles of wave making (deep water)

Although the wave maker adds an equal amount of energy in each cycle, it takes six Wave elevation
cycles to raise the energy density to within % of its asymptotic value. Figure . depicts
the resulting wave elevations for wave maker cycles , , , and . The unavoidable
build-up of wave amplitude in the tank makes model testing in regular waves difficult
as the model will experience the change in amplitude as a wave of different frequency.
Figure . shows the wave energy density distributions corresponding to the wave Wave energy
elevations in Figure .. A small amount of the wave energy has progressed as many density
distribution
wave lengths as there are wave cycles. However, the mean value of the wave energy
density has traveled only half that distance. This implies that the energy propagates
with half the phase velocity.
𝑐
𝑐𝐺 = for deep water waves (.)
2

The propagation speed of wave energy was first derived by Sir Gabriel Stokes and based
on a physical argument (Darrigol, ). Later Strutt, Baron Rayleigh () provided

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

296 24 Wave Energy and Wave Propagation

Figure 24.9 Distribution of energy density after a few selected cycles of wave making (deep
water)

a formal proof by computing the work done by the dynamic pressure in a cross section
j of the fluid. In general, the energy propagation velocity of water waves depends also on j
the water depth.
[ ]
𝑐 2𝑘ℎ
𝑐𝐺 = 1+ (.)
2 sinh(2𝑘ℎ)

Group velocity The velocity at which wave energy is transported corresponds to the propagation speed
of wave groups. In continuation of our wave making simulation from Figure ., a
wave maker is operated for  cycles and shut down. Figure . depicts the resulting
group of regular waves after , , and  wave periods. After  cycles, the actual
front of the wave group has traveled  wave lengths, but the amplitudes at the front are
very small and cut off on the plots. The dashed lines at the top and bottom of the groups
represent group envelopes, which interpolate wave crests and troughs, respectively,
A gray triangle marks the maximum peak of the group envelop. During the  wave
periods between individual plots, the peak of the wave group has traveled exactly  wave
lengths because the energy transport velocity (.) is just half the phase velocity on
deep water. Therefore, the energy transport velocity is also known as group velocity 𝑐𝐺 .
The plots also track a wave labeled #. As expected, it progresses with phase velocity
 wave lengths during  wave periods.

References
Barltrop, N. and Adams, A. (). Dynamics of fixed marine structures. Butterworth-
Heinemann Ltd, Oxford, United Kingdom, third edition.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

24.3 Energy Transport and Group Velocity 297

j j

Figure 24.10 Propagation of a group of regular waves over 20 wave periods

Clauss, G., Lehmann, E., and Östergaard, C. (). Offshore structures, volume :
Conceptual design and hydrodynamics. Springer Verlag, London.
Darrigol, O. (). The spirited horse, the engineer, and the mathematician: water
waves in nineteenth-century hydrodynamics. Archive for History of Exact Sciences,
():–. doi ./s---.
Dean, R. and Dalrymple, R. (). Water wave mechanics for engineers and scientists,
volume  of Advanced Series on Ocean Engineering. World Scientific, Hackensack,
NJ.
Michell, J. (). The hightest wave in water. Philosophical Magazine Series , :–
.
Strutt, Baron Rayleigh, J. (). On progressive waves. In Proceedings of the London
Mathematical Society, volume IX, pages –.

j
 5 Random Seas

Let us begin by stating the term random is synonymous with unpredictable. A truly
random phenomenon should by definition defy mathematical analysis. One might
question then if random waves can be analyzed. By making certain assumptions,
waves that are random in the time domain can be shown to have rather predictable
properties in the frequency domain. In this chapter, a brief history of random wave
analysis is first presented. This is followed by discussions and illustrations of the
various statistical methods of wave analysis.

5.1 Introduction
We can safely assume that seafarers down through the ages have been aware of the
randomness or unpredictability of ocean waves. The occurrence of “rogue waves”
has been documented again and again. These are extremely high waves that occur in
the open ocean without warning. The earliest attempts to deal mathematically with
random waves were confined to averaging observed wave heights and periods. The
data were obtained by visual means in a laboratory setting (as by Weber and Weber,
1825, according to St. Denis, 1969), in lakes, onboard a ship (as by Abercromby,
1888), or in coastal waters. When log-keeping came into being, wave height and
period observations were recorded along with wind speeds. The more sophisticated
mariners also recorded wavelength estimations obtained by comparing the wave-
lengths with lengths of their vessels. The accuracy of the wave height estimations
from visual observations from ships is discussed by Cornish (1910). He reports that
the estimated wave height observations at the time of his writing were about 90%
accurate provided that the observer was well practiced. More examples of wave
height and wavelength estimations made during open-ocean voyages are given by
Cornish (1934). In addition, an excellent historical perspective on wave measure-
ment and analysis is presented by St. Denis (1969).
There are many instruments that have been created to measure the properties
of waves. The earliest laboratory devices were either vertical staffs or tank walls
with painted height scales. These were later replaced by capacitance probes and
digital wave staffs. At-sea wave studies are now conducted by using such devices
as digital wave staffs, floats equipped with accelerometers, and submerged pressure
gauges. All these systems are designed for studies of the wave climates at specific

113
114 Random Seas

sites. Wider perspectives of ocean waves are obtained from both air and space crafts
equipped with both photographic and microwave imagery systems. For a discus-
sion of these techniques, the reader is referred to the collection of papers resulting
from the symposium on Measuring Ocean Waves from Space at the Johns Hopkins
University’s Applied Physics Laboratory in 1986 (see Beal, 1987). Excellent discus-
sions of the various wave measurement techniques are found in the books by Tucker
(1991) and Young (1999). Also in those references are clear and concise discussions
of the various statistical wave analysis methods.
Statistical averaging of wave heights and periods continued into the twentieth
century. Early in that century, oceanographers and applied mathematicians began
to consider the probabilities of occurrence of specific properties of ocean waves.
Many of the probabilistic studies were based on the work of John W. Strutt. In
1880, Strutt (then Baron Rayleigh, and later Lord Rayleigh) developed the well-
known Rayleigh probability distribution by considering the superposition of “a large
number of vibrations of the same pitch and of random phase” (see Strutt, 1880).
Strutt’s (Rayleigh’s) analysis was analogous to superimposing linear water waves
of the same height and frequency but randomly differing in phase. Approximately
70 years later, random phasing was used in the analysis of ocean waves by Pierson
(1952), Longuet-Higgins (1952), and others, and in the analysis of ship motions by
St. Denis and Pierson (1953).
In the mid-twentieth century, the method of time-series analysis was introduced
to physical oceanography. The work of Taylor (1921) (in which the correlation tech-
nique was used in the analyses of diffusion and turbulence), Rice (1944) (in the study
of noise), and Tuckey (1949) (in spectral analysis) inspired a number of investigators
to apply time-series techniques in the analysis of random ocean waves. In one year,
papers by Darbyshire (1952), Longuet-Higgins (1952), Neumann (1952), Pierson
(1952), and Putz (1952) were published, all containing analyses of ocean wave spec-
tra. According to Ewing (1971), the first detailed measurements of waves were made
by Barber and Ursell in 1945 (see Barber and Ursell, 1948, for details of their study).
Those investigators were able to estimate wave spectra. Later, empirical wave spec-
tral density formulas of Phillips (1958), Pierson and Moskowitz (1964), Hasselman
et al. (1973), and others appeared, each accounting for certain physical conditions.
The first analytical wave spectral density expression is attributed to Bretschneider
(1959). The works of both Pierson and Moskowitz (1964) and Bretschneider (1959)
are discussed in Section 5.7 of this book. The reader is also referred to the book
edited by Le Méhauté and Hanes (1990) and that written by Sorensen (1993) for
excellent discussions of these and other spectral representations. At the end of
the twentieth century, McCormick (1998b, 1999) applied the probability formula
of Weibull (1951) to water-wave spectral analysis with some success. This method is
also described and discussed herein.
In this chapter, the statistical analysis of ocean waves is developed. First, the
wave height (H) and period (T) values are determined from a wave trace, and a
bivariate (H-T) nomograph is constructed. Based on this data sample, various sta-
tistical averages, the probability, and the probability density are defined. Also, the
wave spectral density for discrete data is defined and applied to the H-T data. The
statistical functions are then applied to continuous data. Finally, the directional-
ity of random or irregular seas is discussed. All discussions contained herein are
introductory in nature. The reader is encouraged to consult the references for more
expansive discussions on wave measurement and analysis.
5.2 Statistical Analysis of Measured Waves 115

20 20
H1 H2 H3 H4 H5 H6 H7 H8 H9 H10 H11 H12 H13H14 H15 H16H17 H18

10 10
η(cm) η(cm)
0 0

−10 −10

T1 T2 T 3 T4 T5 T6 T7 T8 T9 T10 T11 T12 T13 T14 T15 T16 T17 T18

−20 −20
0 10 20 0 10 20
t(sec) t(sec)
a. Measured Wave Time-Trace b. Wave Properties from Zero Up-Crossing
Method
Figure 5.1. Measured Wave Properties. In (a) the trace was obtained in a wave-tank simu-
lation of a wind-generated sea. In (b) the zero up-crossing method of wave identification
is applied. As can be seen, some waves are not identified by the zero up-crossing method
because they are totally above or below the horizontal axis. For large data sets, these missed
waves are considered to be statistically insignificant.

5.2 Statistical Analysis of Measured Waves

The properties of waves in the open ocean are typically determined by analyzing
signals from floating wave gauges. These signals are either recorded onboard the
floats or telemetered to ships, aircraft, or shore facilities for analysis. Such signals
might be transmitted for twenty minutes of each hour, twenty hours each day for
several years. A signal from an at-sea measurement system would be similar to that
shown in Figure 5.1a, which shows the displacement (␩) of the free surface from the
SWL as a function of time (t). The trace shown in Figure 5.1a was obtained in a
laboratory simulation of a wind-generated sea. The wave pattern was produced by a
wave maker creating a number of regular waves (having specific heights and periods
but differing phases) over a period of time. In other words, these waves were phys-
ically superimposed on each other to produce an irregular wave pattern. Although
the trace appears to be random in time, there is a definite energy distribution in the
frequency domain of the input signal to the wave maker. Because the nature of a
wind-generated sea is somewhat well defined in the frequency (or period) domain,
the seas are often referred to as irregular seas instead of random seas.
There are several methods of identification of waves in an irregular trace such
as that shown in Figure 5.1a. One technique that is widely used by those performing
“hand and eye” analyses of wave traces is the zero up-crossing method. This method
is illustrated in Figure 5.1b, where the waves in the trace in Figure 5.1a are identified.
The method is as follows: Each time the wave trace crosses the time axis with a
positive slope (that is, a zero up-crossing), assume that one wave has passed and
another is just appearing. The zero up-crossing then is at the “front” of the trailing
wave and the “back” of the leading wave. The wave period is defined as the time
between consecutive up-crossings, whereas the wave height is defined as the vertical
distance between the maximum displacement and the minimum displacement of the
trace between consecutive up-crossings. The reader might question the accuracy of
the method as small waves that are entirely above or below the time axis are missed
116 Random Seas

Table 5.1. Example of wave data obtained from at-sea measurements

H(m) – measured T(sec) – measured H(m) – rounded off T(sec) – rounded off

0.56 3.12 0.5 3.0

1.27 6.36 1.5 6.5
0.78 5.77 1.0 6.0
....... ....... ....... .......

by the method. Statistically speaking, there are millions of waves that pass a point
in the open ocean each year. The sample is so large that the number of small waves
missed do not significantly affect the statistics.
The tabulated wave data would resemble the wave height and period data
shown in Table 5.1. Those data are taken with some specified accuracy. The ques-
tion might be raised as to how accurate we should be in these measurements. That
decision is strictly up to the individual reducing the data. To illustrate, the data in
Table 5.1 first appear with an accuracy of two places past the decimal point for both
the heights and periods. If this accuracy is maintained, the amount of data to be
analyzed is somewhat overwhelming. Instead of maintaining this accuracy, round
off the data to 0.5 m for the wave heights and 1.0 sec for the wave period, as illus-
trated in the table. Those rounded-off data are used to create a nomograph, such as
that in Figure 5.2. The 100 data points in the figure are used to illustrate the deriva-
tions and applications of the statistical wave properties discussed herein. The total
sample of wave heights and periods in Figure 5.2 is


i=11 
j=7
N= nTi = n Hj = 100 (5.1)
i=1 j=1

In eq. 5.1, i and j are indices of the wave periods and heights, respectively. To illus-
trate, the j = 4 wave height is 2 m, and there are 11 waves with this rounded-off wave
height value, that is, n H4 = 11.
We can now define some of the basic statistical properties of the wave data.
First, the cumulative frequency of occurrence of the wave height sample in Fig-
ure 5.2 is defined as

J
n Hj
P(H ≤ HJ ) = (5.2)
N
j=1

This is simply the percentage or fraction of the sample wave heights that are less
than or equal to the wave height, HJ . For an infinite sample, the cumulative fre-
quency of occurrence becomes the cumulative probability of occurrence, which is the
probability that any wave height, H, will be less than or equal to the wave height,
HJ . Mathematically, this probability is defined as

J
n Hj
P(H ≤ HJ ) = lim P(H ≤ HJ ) = lim (5.3)
N→∞ N→∞ N
j=1

From this, a deterministic sea is one for which

P(H ≤ ∞) = 1 (5.4)
5.2 Statistical Analysis of Measured Waves 117

i 1 2 3 4 5 6 7 8 9 10

nTi 3 5 6 9 11 16 10 12 11 8 nHj j

3.5 1 1 1 0 3 7

3.0 1 1 2 1 1 1 7 6

2.5 1 1 2 3 1 1 0 9 5

H(m) 2.0 1 1 1 3 2 1 1 1 11 4

1.5 1 2 1 2 2 3 4 3 3 3 24 3

1.0 2 2 2 2 4 4 5 3 3 1 29 2

0.5 1 2 3 2 3 2 2 1 1 17 1

0.0
3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0 100 (= N)
T (s)
Figure 5.2. Wave Height – Period Nomograph. In this figure, i and j are the indices of the
respective period (T) and wave height (H), and nTi and nHj are the numbers of observation
for the respective period and wave height values. The total sample is N.

because all wave heights must be less than infinite. At the other extreme, an impos-
sible sea is one for which
P(H ≤ 0) = 0 (5.5)
because no wave heights can be less than zero, and a zero wave height is unde-
fined. For a finite number of data points, as in Figure 5.2, the cumulative frequency
of occurrence can be considered to be an approximate cumulative probability of
occurrence.
EXAMPLE 5.1: CUMULATIVE PROBABILITY OF OCCURRENCE The probability that
any measured height in Figure 5.2 is less than or equal to 2 m (J = 4, H4 = 2 m)
is approximately
(n H1 + n H2 + n H3 + n H4 )
P(H ≤ 2 m)  P(H ≤ 2 m) = = 0.81
N
The term “approximately” is used because an infinite sample is required to have
a true probability. The probability is approximately equal to the cumulative
frequency of occurrence of eq. 5.2 for this finite sample. Following this example,
a chart of the probability as a function of wave height is constructed (see Fig-
ure 5.3).
118 Random Seas

1.0

P (H ≤ Hj) 0.5

0
0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
Hj (m)

Figure 5.3. Cumulative Frequency of Occurrence for the Data in Figure 5.2. This can be con-
sidered to be an approximate probability of occurrence for the wave data.

We define the probability density function as the slope of the probability curve.
For continuous wave height data, the probability density function is

d[P(H ≤ HJ )] [P(H ≤ HJ )]
p(HJ ) = = lim (5.6)
dH H→0 H
Using the last term without passing to the limit, we can obtain an approximate prob-
ability density function for a discrete data set, such as that in Figure 5.2. The expres-
sion of that approximation is


J
n Hj − n Hj−1 (n HJ /N)
p(HJ )  = (5.7)
N(H) H
j=1

The most-probable wave height, Hmp , is that wave height corresponding to the max-
imum value of the probability density. This wave height is then determined from

d[ p(HJ )] 
Hmp = 0 (5.8)
dH
if the wave height distribution is continuous. As is demonstrated in Sections 5.5 and
5.6, the most probable value of a random variable, from either experimental studies
or field measurements, can be used to determine the form of the probability density
function.

EXAMPLE 5.2: PROBABILITY DENSITY AND MOST-PROBABLE WAVE HEIGHT As in

Example 5.1, we focus our attention on the 2-m wave height (J = 4) in Fig-
ure 5.2. The value of the probability density, from eq. 5.7, is

(n H4 /N) (11/100)
p(H4 ) = p(2 m) = = = 0.22 m−1
H 0.5
The probability distribution (the plot of the probability density) for the wave
height data in Figure 5.2 is presented in Figure 5.4. From the condition of
eq. 5.8, the most probable wave height in Figure 5.4 is Hmp = 1.0 m.
5.2 Statistical Analysis of Measured Waves 119

0.75

0.50

Figure 5.4. Probability Density for the P(Hj)

Wave Height Data in Figure 5.2.
0.25

0
0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
Hj (m)

In addition to the cumulative probability of occurrence and the probability den-

sity, other statistical properties of use in random wave analysis are listed in the
following.

A. Average Wave Period and Wave Height

The average wave period over the period range Ti < T ≤ TI is

%
I
nTi Ti %
I
N
p(Ti )Ti T
i i
Ti I = = (5.9a)
%
I
nTi %
I
N
p(Ti )T
i i

where the expression for p(Ti ) is obtained by simply replacing the wave height (H)
by the period (T) and the indices (j, J) by (i, I) in eq. 5.7. When the expression in
eq. 5.9a is applied to the total sample (N) of wave periods, then the expression for
the average wave period is obtained, that is,
i%
max i%
max
nTi Ti
N
p(Ti )Ti T
i=1 i=1
Tavg = = (5.9b)
1 1
In a similar manner, the average wave height over the range Hj < H ≤ HJ is

%
J n H %
J
p(Hj )Hj H
Hj j
N
j j
Hj J = = (5.10a)
%
J n
Hj %
J
N
p(Hj )H
j j

The average wave height of the entire sample in Figure 5.2 is then

%
jmax n Hj Hj %
jmax
N
p(Hj )Hj H
j=1 j=1
Havg = = (5.10b)
1 1
120 Random Seas

The average expressions of the wave properties are used in the design of wave-
energy conversion systems and other dynamic ocean systems.

B. Mean-Square and Root-Mean-Square Wave Heights

The mean-square wave height is obtained from an averaging formula that is similar
to that of eq. 5.10b. The expression for this wave property is

jmax
n Hj Hj2 
jmax
H2 = = p(Hj )Hj2 H (5.11)
N
j=1 j=1

where the unity value in the denominator (the value of the probability of occurrence
over the entire sample) is assumed here and henceforth. The square root of this is
called the root-mean-square wave height, that is,

Hr ms = H2 (5.12)
The root-mean-square wave height is considered to be a measured wave property.
It is extensively used in many statistical formulas including various expressions for
the probability density function for continuous wave data.

C. Variance of the Wave Heights

The variance of the wave heights is a measure of the spread of the data about the
average wave height, and is mathematically defined by

jmax
n Hj (Hj − Havg )2 
jmax
(H − Havg )2 = = p(Hj )(Hj − Havg )2 H (5.13)
N
j=1 j=−1

In the following example, the averages of the wave height and period and the mean-
square, root-mean-square, and variance of the wave heights are demonstrated.
EXAMPLE 5.3: AVERAGE WAVE PERIOD AND STATISTICAL WAVE HEIGHTS For the
data in Figure 5.2, the values of the respective average, mean-square, root-
mean-square wave heights, and the wave height variance are
Havg = 1.50 m
H2 = 2.87 m2
Hr ms = 1.69 m
(H − H)2 = 0.64 m2
These values are obtained from eqs. 5.10b, 5.11, 5.12, and 5.13, respectively. The
average wave period obtained from eq. 5.9b is Tavg = 6.13 sec.

D. Significant Wave Height and Period

As previously stated in this chapter, most of the open-ocean wave height data prior
to the mid-twentieth century resulted from visual observations. Although Cornish
(1910) estimated the accuracy of such observations to be about 90%, Sverdrup and
Munk (1947) had different thoughts on the matter. According to Sverdrup and
Munk, the visually observed “average” wave heights more closely correspond to the
5.2 Statistical Analysis of Measured Waves 121

i 1 2 3 4 5 6 7 8 9 10
nTi 3 5 6 9 11 16 19 12 11 8 nHj j

3.5 1 1 1 3 7

3.0 1 1 2 1 1 1 7 6

2.5 1 1 2 3 1 1 0 9 5

2.0 1 1 1 3 2 1 1 1 11 4
H(m)

1.5 1 2 1 2 2 3 4 3 3 3 24 3

1.0 2 2 2 2 4 4 5 3 3 2 29 2

0.5 1 2 3 2 3 2 2 1 1 17 1

0.0
3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0 100 (= N)
T(s)
Figure 5.5. Method for Determining the One-Third Highest Waves. By convention, the count
begins in the upper right corner of the nomograph. For the 100-wave sample, the highest one-
third of the sample is approximately 33 waves. Although there are 24 1.5-m waves, only 3 on
the right side are used in the count by our convention.

average height of the one-third highest waves. They reason that the visual observer
is selective in that one tends to neglect the smaller waves. Again, because so much
of the wave height data available at the time of their writing resulted from visual
observations, Sverdrup and Munk suggested that the average height of the one-third
highest waves be a standard of measure, and called that wave height the significant
wave height. The average period of the one-third highest waves is called the signifi-
cant wave period.
Consider again the wave height and period data of Figure 5.2. To determine
the significant wave height and significant wave period, we must first identify the
highest one-third waves in our data set. Because the sample (N) for the data in
Figure 5.2 is 100, the highest 33 waves are used. By convention, our count of the
highest waves begins with the data point in the upper right corner of the nomograph,
as illustrated in Figure 5.5. The indices corresponding to the data are i = 9 and j =
7. We count from right to left, and terminate our count after counting 33 waves,
that is, we terminate the count for index values of i = 10 and j = 3. The significant
wave height and period values for the data of Figures 5.2 and 5.5 are given in Exam-
ple 5.4.
EXAMPLE 5.4: SIGNIFICANT WAVE PROPERTIES The highest one-third wave heights
for the data in Figure 5.2 are identified in Figure 5.5. The average height of these
waves is the significant wave height, whereas the average period correspond-
ing to the one-third highest wave is the significant wave period. The respective
approximate values for these are
Hs = 2.44 m, Ts = 6.58 sec
122 Random Seas

The term “approximate” is used here as 33/100 is approximately 1/3. By compar-

ing our results with the average wave height and period values of Example 5.3,
we see that the significant wave height is about 1.63 times the average height,
whereas the average period is about 1.07 times the average period. For these
data then, the significant and average wave periods are approximately equal.

To avoid “hand and eye” measurements of wave data, as described in this sec-
tion, we seek similarities between our measured data and well-established statis-
tical properties of continuous wave data. In other words, after collecting as much
wave data as we can, we try to fit established formulas and curves to the data to
generalize our results. For example, if the sea is assumed to be ergodic (having sta-
tistical properties that are invariant in time and space), then the statistically aver-
aged wave periods and heights at a specific site in the ocean should be the same as
those at other sites. When we take into account the generation of these waves by
the wind, then we must qualify the ergodic hypothesis by specifying both the wind
speed and duration of the storm generating the seas. That is, for a given wind con-
dition (speed and duration) at a point, the statistical averages of the wave heights
and periods will be the same as those measured at some other time and place in the
ocean having the same wind conditions. Wave properties that are invariant in time
are said to be statistically stationary, whereas those that are invariant in space are
said to be statistically homogeneous. We assume that the waves are ergodic in our
discussions.

5.3 Continuous Probability Distributions

As stated in the previous section, it is not practical to measure wave properties
at every site of interest in the open ocean. For this reason, we try to match our
finite statistical wave height and period data with some widely used expressions for
either the probability of occurrence or the probability density. These established
expressions are continuous functions of the property in question. This means that
the sample (N) of random data is now considered to be infinite. With this assump-
tion, we can express eqs. 5.9, 5.10, 5.11, and 5.13 as integrals by passing to the limit
as N → ∞. The respective results are shown in the following.

A. Average Wave Period and Wave Height

For continuous wave data, the summations in eqs. 5.9a and 5.9b evolve into inte-
gral expressions of the average wave period and mean wave period, respectively.
The mathematical expression for the average wave period for the period range of
Ti < T ≤ TI is

&TI
p(T)TdT
Ti
Ti I = (5.14a)
&TI
p(T)dT
Ti

where p(T) is the period probability density function. When the lower limits of the
integrals of eq. 5.14a are 0 and the upper limits are ∞, the resulting expression is
5.3 Continuous Probability Distributions 123

that of the average wave period, that is,

∞
Tavg = p(T)TdT (5.14b)
0

Similarly, the average wave height over the wave height range of Hj < H ≤
HJ is

&HJ
p(H)HdH
Hj
Tj J = (5.15a)
&HJ
p(h)dH
Hj

where p(H) is the probability density of the wave heights. Over the semi-infinite
range 0 < H ≤ ∞, eq. 5.15a becomes
∞
Havg = p(H)HdH (5.15b)
0

which is the expression for the mean wave height.

The mathematical expressions of the probability density functions in eqs. 5.14
and 5.15 must be specified. Several of these expressions are introduced and dis-
cussed in the next three sections.

B. Mean-Square Wave Height

The result of passing to the limit of the sum defining the mean-square wave height
in eq. 5.11 is
∞
H2 = p(H)H2 dH (5.16)
0

C. Variance of the Wave Heights

The variance for a continuous probability distribution of wave heights from eq. 5.13
is
∞
(H − Havg )2 = p(H)(H − Havg )2 dH (5.17)
0

where, again, the expression for the wave height probability function, p(H), must be
specified. This is done later in this chapter.

D. Significant Wave Height

Using the average wave height expression in eq. 5.15a, the significant wave height
for an infinite sample of wave heights can be determined by performing the
124 Random Seas

integration from the smallest wave height of the highest one-third heights. Rep-
resent this wave height by H33% . The expression for the significant wave height is
then
&∞
p(H)HdH
H33%
Hs = (5.18)
&∞
p(H)dH
H33%

It is shown later in this chapter how one determines the value of H33% for a data
sample.
Our task is to find suitable probability distributions for the wave height data. Of
special interest to oceanographers and ocean engineers is the Rayleigh probability
distribution for the wave heights. The use of this distribution is of particular value
in determining wave loads of fixed and floating structures in the open ocean.

5.4 Rayleigh Probability Distribution of Wave Heights

The reader is again reminded that the probability distribution is a plot of the prob-
ability density function. There are a number of probability density functions that
have resulted from statistical analyses of various phenomena. As previously men-
tioned, the Rayleigh probability density function (see Strutt, 1880) resulted from
a study of the random vibrations created by superimposing linear, monochromatic
vibrations having the same frequency but differing in phase, the phase angle being
randomly chosen. In this section, the probability density of Lord Rayleigh (J. W.
Strutt) is used to represent the statistical distributions of wave heights in random
seas. The Rayleigh distribution applied to ocean waves is discussed in the classical
paper by Longuet-Higgins (1952).
For wave heights measured in the open ocean, the plot of the probability density
function (probability distribution) was found to resemble the Rayleigh probability
distribution. The expression of the Rayleigh probability density function for wave
heights is

2
2H − H
p(H) = 2 e Hr ms (5.19)
Hr ms
where Hrms is the root-mean-square wave height, defined by eq. 5.12. Note that the
value of Hrms is considered to be a measured quantity.
As written in eq. 5.6, the probability density function is the derivative of the
probability. Hence, the expression for the Rayleigh probability of occurrence can
be obtained by integrating eq. 5.19 between two wave height limits. That is, the
probability that a measured wave height (H) will be between H = Hj and H = HJ is
obtained by integrating eq. 5.19 between these limits. The resulting expression for
the probability of occurrence is
HJ
Hj 2
HJ
2
− −
P(Hj < H ≤ HJ ) = p(H)dH = e Hr ms −e Hr ms (5.20)
Hj

The application of the Rayleigh probability relationships is demonstrated in the fol-

lowing example.
5.4 Rayleigh Probability Distribution of Wave Heights 125

1.0

Rayleigh
Measured
P (H ≤ Hj) 0.5

0
0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
Hj (m)

Figure 5.6. Cummulative Frequency of Occurrence for the Data in Figure 5.2. The Rayleigh
probability formula in eq. 5.20 is used to obtain the theoretical curve.

EXAMPLE 5.5: RAYLEIGH PROBABILITY DENSITY AND PROBABILITY FUNCTIONS To

apply the probability density expression of eq. 5.19 and the expression for the
probability of occurrence of eq. 5.20 to the data in Figure 5.2, all that is needed
is the value for the root-mean-square wave height, the value of which is Hrms =
1.69 m from Example 5.3. The expression for the Rayleigh probability density
function is then

p(H) = 0.700He−0.350H , m−1

2

The corresponding Rayleigh cumulative probability of occurrence (the proba-

bility that H lies between 0 and HJ ) is
HJ
p(H)dH = 1 − e−0.350HJ
2
P(0 < H ≤ HJ ) =
0

This is also called the probability of non-exceedance by some wave analysts.

Results obtained from the application of this probability expression to the data
in Figure 5.3 are presented in Figure 5.6. The probability distribution, p(H), is
presented in Figure 5.7 with the values found in Figure 5.4.

A. Average Wave Height

The combination of the Rayleigh probability function of eq. 5.19 and the generic
equation for the average wave height of eq. 5.15a yields the following expression for
the average wave height in Hj < H ≤ HJ :
   2
−
Hj 2
−
HJ √     
␲ Hj
Hj e Hr ms
− HJ e Hr ms
+ Hr ms 2
erf HJ
Hr ms
− erf Hr ms
Hj J = (5.21a)
P (Hj < H ≤ HJ )
where erf( ) is called the error function. Values of this function can be obtained from
the book edited by Abramowitz and Stegun (1965). From that reference, a rather
accurate approximate formula is found, and is presented following eq. 5.34. In the
denominator of eq. 5.21a is the probability of occurrence, the expression for which
is in eq. 5.20.
126 Random Seas

0.75

0.50

P(Hj), (1/m)

0.25 Measured

Rayleigh
0
0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
Hj (m)

Figure 5.7. Probability Density for the Data in Figure 5.2. The Rayleigh curve is obtained from
eq. 5.19.

When the expression in eq. 5.21a is applied to the entire range of wave heights,
that is, from Hj = 0 to HJ = ∞, the expression for the average wave height corre-
sponding to a Rayleigh wave height distribution results, that is,
√
␲
Havg = Hr ms (5.21b)
2
As before, the root-mean-square wave height is normally considered to be a mea-
sured quantity in these equations.

B. Probability of Exceedance
The probability that a measured wave height will be greater than, say, HK is called
the probability of exceedance. From eq. 5.20, this probability is found by integrating
from HK to ∞. That probability is then expressed by

HK 2
−
P(HK < H ≤ ∞) = e Hr ms = f (5.22)

where f is the fraction of waves having heights greater than HK . One application
of the formula is in the determination of the significant wave height, discussed in
Section 5.2D. From the expression in eq. 5.22, we can determine the value of HK for
the percentage of waves in question. The expression for that wave height is
  
1
HK = Hr ms ln (5.23)
f

C. Significant Wave Height

The significant wave height is defined in Section 5.2D as the average height of the
one-third highest waves, and is mathematically represented by eq. 5.1 for a contin-
uous probability distribution of wave heights. To determine the value of the signif-
icant wave height, we need to know the smallest wave height value of the highest
one-third waves, which is also the lower limit of the integral in eq. 5.18. This lower
5.5 Weibull Probability Distribution of Wave Heights 127

limit is obtained from eq. 5.23, where the fraction is f = 1/3 and the subscript is K =
33%. The result is H33% = 1.048Hr ms . Replace the lower limits of the integrals of
eq. 5.18 to obtain the expression for the significant wave height for a Rayleigh wave
height probability distribution. The result is
Hs = 1.416Hr ms (5.24)

D. Extreme Wave Height

Equation 5.23 can be used to determine the extreme wave height of a sample of wave
heights assumed to have a Rayleigh distribution. Suppose that we determine (from
measurements) that there are NM waves passing a site in the open ocean in M years.
For that sample, we wish to determine the value of the height of the highest wave. In
other words, our interest is in the height of the f = 1/NM wave because there is only
one wave that can be the highest. That extreme (maximum) wave height is obtained
from eq. 5.23 by using this frequency value, that is,
 2 
Hmax = Hr ms ln(NM ) = √ Havg ln(NM ) (5.25)
␲
Here, the results of eq. 5.21b have been incorporated. For the data in Figure 5.2, NM
= 100 and Hrms = 1.69 m from Example 5.3. These values in eq. 5.25 yield Hmax =
3.63 m. This extreme value is close to the maximum wave height in the Figure 5.2
data, which is 3.5 m.
EXAMPLE 5.6: SIGNIFICANT AND EXTREME WAVE HEIGHTS At a site in the open
ocean, a wave staff is placed to determine the statistical wave properties. Mea-
surements are taken for 20 min each hour, 20 hours each day for 365 days. In
eq. 5.25, then, M = 1. The average measured wave properties are those found
in Example 5.3. From that example, we find
Havg = 1.5 m, Hr ms = 1.69 m, Tavg = 6.13 sec
The significant wave height expression of eq. 5.24 then yields Hs = 2.39 m.
To determine the extreme wave height, the total sample of waves passing the
wave staff in 1 year must be determined. To make that determination, simply
divide the total number of seconds in 1 year (365 days × 24 hours × 3600 sec =
3.1536 × 107 sec) by the average wave period. The result is N1  5.14 × 106 .
Substitute the wave property values and the N1 value into eq. 5.25 to obtain
Hmax = 6.63 m.
The reader can see that much information can be obtained if the wave height
probability distribution resembles that predicted by a Rayleigh probability density
function. This distribution is most valid in the open ocean. When in a coastal area,
the wave height distribution might be somewhat different than that predicted by
eq. 5.19. In that situation, a more general probability density function is desir-
able. One such distribution is that of Weibull (1951), which is discussed in the next
section.

5.5 Weibull Probability Distribution of Wave Heights

The Weibull probability density function is a parametric relationship that has wide
application. In fact, most of the commonly used probability density functions can
5.7 Wave Spectral Density 133

wave components are present. We shall determine if the Gaussian assumption

leading to eq. 5.35 is satisfactory for these data. Using the amplitude (a+ and
a− ) data in Table 5.2, the root-mean-square value of the wave height is found
to be
'

( j=N
(  n j 2 a 2j √
Hr ms = ) = 141.35  11.9 cm
N
j=1

where, for a given amplitude value, H± = 2a± . There are 49 measured wave
heights in Table 5.2, so the sample or ensemble (Na ) equals 49.
The application of eq. 5.32 to the data in Table 5.2 yields ␩r ms = 4.54 cm,
which is approximately 0.382 Hr ms . A comparison of this result with the expres-
sion in eq. 5.35b shows a difference in the two expressions of approximately
7%. Hence, for the data in Table 5.2, the Gaussian assumption leading to the
expression in eq. 5.35 is satisfactory.

5.7 Wave Spectral Density

This section is devoted to the relationship between the energies of both wind-
generated seas and long-term seas and the statistical properties of the wave fields.
Long-term seas are those for which data are recorded and averaged over time peri-
ods of months or years. Wind-generated seas have received far more attention than
long-term seas. The reason is that extreme wave heights in wind-generated seas are
used in the probabilistic design of ocean structures for survivability. The phenomena
associated with wind-wave generation is rather complex. In Chapters 3 and 4, the
waves are treated as two-dimensional phenomena, and the heights of the monochro-
matic waves are assumed to be independent of the wave period. This is not the case
in a wind-generated sea. The wind eddies that convect in the boundary layer adja-
cent to the free surface each have energies that are associated with their convection
velocities. Pressure variations on the free surface result from the passage of these
eddies. Convecting pressure pulses can deform the compliant free surface, thereby
creating the first waves, called ripples. There is a transverse curvature of the first
waves created that is partially the result of the increasing transverse curvature of
the rotational axes of the convecting wind eddies, as illustrated in Figure 5.10. The
interaction of such eddies and compliant surfaces is discussed by McCormick and
Mouring (1995). The pressure-induced deformations of the free surface then have
definite geometrical patterns. These geometrical patterns change due to wind-wave
interactions in which the wind stress acts on the free surface, and wind wakes are
formed on the leeward side of the waves, as illustrated in Figure 3.1d. The gener-
ation of wind-generated seas is also discussed in Section 1.1. The classification of
wind-generated seas is discussed in Section 1.2.
The literature for both wind-generated seas and long-term seas is extensive.
The first definitive analysis of wind-generated waves is attributed to Phillips (1957).
Discussions of this analytical method are found in the books by Kinsman (1965)
and Phillips (1966). More recent analyses and discussion of wind-generated waves
are given by Donelan (1990), in the book edited by Le Méhauté and Hanes (1990),
and by Tucker and Pitt (2001). The subject of wind stress on waves is discussed by
Wu (1980). The statistics of long-term seas are discussed by Isaacson and MacKenzie
134 Random Seas

Burst
z

Uz

u(z)
x

Line Vortex
Figure 5.10. Sketch of Boundary-Layer Vortex Distortion and Migration. The vortices form
at various heights above the calm-water surface. Because of the distortion in the line vortex,
additional energy is absorbed at points along the line, and the vortex grows (strengthens)
at these points. As the vortex grows, it is subjected to greater wind speeds and lower wind
pressures. The changing wind properties further distort the vortex and cause it to grow and
migrate in both the horizontal and vertical directions over the calm-water surface. The time-
scale for the phenomenon depends on the height of the initial vortex formation because that
scale depends on the vertical velocity distribution of the wind.

(1981), Hogben (1990), and McCormick (1998b, 1999). Wind-wave spectra are ana-
lyzed in Section 5.8, whereas long-term spectra are analyzed in Section 5.9.
In the following subsections, general discussions of measured wave spectra and
empirical spectral expressions are presented.

A. Point Spectra from Discrete Wave Data

To analyze the energy content of a random sea from data obtained from point mea-
surements, we first assume that the sea is composed of a large number of linear
waves that are superimposed upon each other. Because of this assumption, let us
refer to this assumed sea as irregular rather than random. For a component wave of
an irregular sea, eq. 3.67 is used to obtain the energy per free-surface area (called
the energy intensity, E) per unit weight of water, that is,

(E/␭b) E H2
E≡ = = (5.36)
␳g ␳g 8
where ␭b is the surface area and ␳ g is the weight-density of the water.
The expression in eq. 5.36 can be applied to the data in Figure 5.2 to determine
the average energy intensity distribution over the wave periods. For a period Ti in
that figure, we can write
j(i) max
 n j(i) Hj2(i)
E(Ti ) = ␳g = ␳gS(Ti )T (5.37)
j(i) min
8N

where T is the period increment (equal to 0.5 sec in Figure 5.2), j(i) is the wave
height index corresponding to the ith period, and N is the total sample (equal to 100
for the Figure 5.2 data). To illustrate, for i = 2 in Figure 5.2, T2 = 4.0 sec, whereas
j(2)min = 1 and j(2)max = 3. The function S(Ti ) is called the wave spectral density. The
purpose of this function is to show the period dependence of the average energy
5.7 Wave Spectral Density 135

intensity of a random sea. From eq. 5.37, the mathematical expression of the spectral
density for discrete data is
j(i) max
 n j(i) Hj2(i)
S(Ti ) = (5.38)
j(i) min
8NT

The average energy intensity of all of the waves in the sample is simply the sum of
the terms in eq. 5.37 over all of the measured periods, that is,


i max i max j
 n j(i) Hj(i)2 
jmax
(i) max
n j Hj2
E(Ti ) = ␳g = ␳g
8N 8N
i=1 i=1 j(i) min j=1


i max
= ␳g S(Ti )T (5.39)
i=1

Note that the double summation of the second term is equivalent to the single sum-
mation of the third term as both sum over the entire sample. We can also rewrite
the summation of the third term in terms of the mean-square wave height and the
probability density function using the results of eq. 5.11. The result is

jmax

i max
H2 = Hr2ms = p(Hj )Hj2 H = 8 S(Ti )T (5.40)
j=1 i=1

The applications of the spectral density formula of eq. 5.38 and the mean-square
wave height expression of eq. 5.40 are illustrated in the following example by using
the data in Figure 5.2.
EXAMPLE 5.9: WAVE SPECTRAL DENSITY FROM DISCRETE WAVE DATA The applica-
tion of the spectral density expression of eq. 5.38 to the data in Figure 5.2 yields
the following results:
S(T1 ) = S(3.5 sec) = [2(1.0 m)2 + 1(1.5 m)2 ]/[8(100)0.5 sec]  0.0106 m2 /s
S(T2 ) = S(4.0 s)  0.0169 m2 /s
S(T3 ) = S(4.5 s)  0.0219 m2 /s
S(T4 ) = S(5.0 s)  0.0438 m2 /s
S(T5 ) = S(5.5 s)  0.0706 m2 /s
S(T6 ) = S(6.0 s)  0.1125 m2 /s
S(T7 ) = S(6.5 s)  0.1788 m2 /s
S(T8 ) = S(7.0 s)  0.1044 m2 /s
S(T9 ) = S(7.5 s)  0.1038 m2 /s
S(T10 ) = S(8.0 s)  0.0550 m2 /s
When these values are multiplied by eight times the period increment (T =
0.5 sec) and summed as in eq. 5.40, the result is 2.87 m2 . This value is the same
as that of the mean-square wave height in Example 5.3, as expected.
The values of the spectral density are shown in Figure 5.11 with their corre-
sponding period values. The plot of the S(T) is called the wave spectrum. This plot
represents the distribution of the energy content of the sea over a range of mea-
sured wave periods. The most energetic waves in this sample occur at a period of
6.5 sec. This period is called the modal period. The modal period is one to avoid in
136 Random Seas

0.2

S(T ), (m2/s)
0.1

0.0
0 2 4 6 8 10
T, (s)
Figure 5.11. Wave Spectrum for the Data in Figure 5.2 and Example 5.9.

the engineering design of a compliant offshore structure. On the other hand, the
design period for an ocean wave energy conversion system would be the modal
period.

B. Empirical Expressions of the Point Spectral Density

An enormous amount of wave data now exist. These data have resulted from point
measurements at various open-ocean and confined-water sites. Because of the abun-
dance of the data, a number of empirical expressions for the wave spectral density
have been formulated. Some of the works leading to these empirical expressions are
referred to in Section 5.1.
An empirical expression of the spectral density is a continuous functional rela-
tionship. To obtain such a function, the expression for the mean-square wave height
of eq. 5.40 is transformed by first passing to the limit as T → 0 and H → 0,
and then integrating the resulting expression over the wave period from T = 0 to
T = ∞. The result is
∞ ∞
H2 = Hr2ms = p(H)H dH =2
8␩2 =8 S(T)dT (5.41a)
0 0

where the results of eq. 5.35a have been incorporated. A more general form of this
equation is
∞
H2 = Hr2ms =C S(T)dT (5.41b)
0
5.8 Wind-Wave Spectra 137

where the constant, C, is determined from field data. The integral is sometimes
referred to as the zero moment because many physical oceanographers and ocean
engineers refer to
∞
T m S(T)dT
0

as the mth moment. We shall avoid referring to these expressions as “moments” as

the term could cause confusion in later chapters dealing with physical moments on
fixed and floating structures.
For fully developed seas (seas for which the statistical properties are both uni-
form over the free surface and constant in time), the generic equation used by a
number of investigators is

S(T) = AT me−BT
n
(5.42)

where the coefficients A and B depend on the statistical wave properties. In turn,
these properties depend on those associated with the wind for wind-generated seas.
The wind properties are the nominal wind speed (U), the fetch (F), and the duration
(td ). The fetch is the length of the sea over which the winds blow, and the duration is
the time-life of the wind event. In the empirical formulas, m and n result from curve-
fitting of the data. Four wind-wave spectra that are represented by eq. 5.42 are the
following: Neumann (1952), for which m = 4 and n = 2; Bretschneider (1959), for
which m = 3 and n = 4; Pierson and Moskowitz (1964), for which m = 3 and n = 4;
and the long-term formula of McCormick (1998b), for which m = 6 and n = 7. The
choice of the exponent m is rather critical in representing the condition of the sea, as
discussed by Pierson and Moskowitz (1964). The value of the exponent must be such
that it is neither small enough to result in prematurely breaking-wave conditions
(Phillips, 1958) nor large enough to introduce viscous dissipation (Hamada, 1964, as
reported by Pierson and Moskowitz, 1964).
The spectral density expression of eq. 5.42 can be combined with eq. 5.41b to
obtain a relationship for the mean-square and root-mean-square wave heights. The
result is
 
A m+1
H2 = Hr2ms = C m+1
 (5.43)
nB n n

5.8 Wind-Wave Spectra

Because the formulas of Bretschneider (1959) and of Pierson and Moskowitz (1964)
have the same exponent values, that version of eq. 5.42 is used herein for the sake
of illustration. The Bretschneider and Pierson-Moskowitz spectra have been widely
used since their original appearances in the literature. The International Ship Struc-
tures Committee (ISSC) has modified the Bretschneider formula, whereas the Inter-
national Towing Tank Committee (ITTC) has modified the Pierson-Moskowitz for-
mula. For discussions of these respective modifications, see ISSC (1967) and ITTC
(1972). An additional modification of the Bretschneider spectrum was performed
by Mitsuyasu (1970), according to Goda (1985, 1990). Goda refers to the modi-
fied spectrum as the Bretschneider-Mitsuyasu spectrum. The 1990 paper by Goda
138 Random Seas

appears as an appendix in the book edited by Pilarczyk (1990). Based on measure-

ments in the German Bight of the North Sea, Hasselmann et al. (1973) modified the
Pierson-Moskowitz formula to account for limited fetch. The resulting spectral for-
mula is called the JONSWAP spectrum, the acronym derived from the Joint North
Sea Wave Project. Hogben (1990) and Komen et al. (1994) give excellent discussions
of the various wave spectral density formulas.
The Bretschneider (1959) and Pierson-Moskowitz (1964) spectra expressions
have the following form in the wave period domain:

S(T) = AT 3 e−BT
4
(5.44a)
We can also write the spectral expressions in the wave frequency domain ( f = 1/T),
where S(T)dT = −S( f )d f . This relationship results from the fact that the energy
of the wave field over the frequency domain must be equal to that over the period
domain. The negative sign results from the derivative d f/dT. Solving for the spectral
density in the frequency domain, we then obtain
dT A −B
S( f ) = −S(T) = T 2 S(T) = 5 e f 4 (5.44b)
df f
The preference in this book is to work with the period-domain spectra.
It must be noted here that the wave spectral density functions of Bretschneider
(1959) and of Pierson and Moskowitz (1964) are defined differently, that difference
being in the value of C in eq. 5.41b. Bretschneider (1959, 1963) specifies C = 1,
whereas Pierson and Moskowitz (1964) specify C = 1/8, as in eq. 5.39.
The coefficients A and B are directly related to the wave properties. The expres-
sions for these coefficients distinguish the various wave spectral density formu-
las. Bretschneider (1959, 1963) obtains expressions for the coefficients by assuming
Rayleigh probability distributions of the wave heights and wavelengths. The Pierson
and Moskowitz (1964) coefficients are based the similarity hypothesis of Kitaigorod-
skii (1962) and on at-sea measurements of the wave properties. Before discussing
the specific spectral expressions, we shall determine relationships for the various
statistical periods.

A. Statistical Wave Periods

Up to this point, most of the discussion of random waves has been focused on sta-
tistical wave heights. With the introduction of the concept of the wave spectral den-
sity, much information concerning the statistical wave periods can be obtained. In
the analyses that follow, the expressions in eq. 5.44 are used. The peak value of the
period-based spectral density of eq. 5.44 is found by setting the derivative dS/dT
equal to zero. The period that yields that peak value is called the modal period, and
is expressed as
 1/4
3
To = (5.45)
4B
This formula is important to the design engineer because it represents the wave
period for which the energy of a wind-generated sea is a maximum. Sarpkaya and
Isaacson (1981) note that the modal period is not that for which the frequency-based
spectral density [S( f ) in eq. 5.44b] is a maximum. They refer to that period as the
peak period, Tp . This period is determined by setting the derivative of the expression
5.8 Wind-Wave Spectra 139

in eq. 5.44b equal to zero. The result is

 1/4  1/4
1 5 5
Tp = = = To (5.46)
fp 4B 3
The last equality is obtained by solving eq. 5.45 for B and combining the result with
the third term in eq. 5.46.
The statistical averages of the wave period can be obtained by using the spectral
expression in eq. 5.44a. The average period is found from
&∞
TS(T)dT

0  14 3.6256
Tavg = =  (5.47)
&∞ 4B1/4 4B1/4
S(T)dT
0

where, again, the gamma function is discussed both by Abramowitz and Stegun
(1965). In a similar manner, the expressions for the mean-square period and the
root-mean-square period are obtained. The results are
&∞
T 2 S(T)dT
 
0  32 1 ␲
T2 = Tr2ms = = √ = (5.48)
&∞ B 2 B
S(T)dT
0

By eliminating B in the combination of eqs. 5.47 and 5.48, we obtain the relationship
between the average period and the root-mean-square period, which is
Tavg  0.96282Tr ms (5.49)
The significant wave period, Ts , is defined as the average period of the highest
one-third waves. According to Sarpkaya and Isaacson (1981), Bretschneider (1977)
proposed the following relationship between the significant period and the peak
period of eq. 5.46:
  14
4
Ts = Tp (5.50)
5
With this relationship, we can now relate the significant, peak, modal, mean, and
root-mean-square wave periods as
Ts  0.946Tp  1.075To  1.104Tavg  1.063Tr ms (5.51)
The reader will find these relationships to be of value when dealing with the various
spectral density formulas that are used in ocean engineering studies. The reader
will notice that the significant wave period and the other periods in eq. 5.51 do not
vary by more than 10%. When the concept of the significant wave period was first
introduced, it was common practice to equate the significant period to the mean
period and other periods of interest in engineering studies.
EXAMPLE 5.10: STATISTICAL WAVE PERIODS In Example 5.3, we find that the aver-
age period for the data in Figure 5.2 is 6.13 sec. The average period for the
highest 33 waves of those data is approximately the significant wave period, Ts .
Referring to the data in Figure 5.5, the value of the significant wave period is
approximately 6.58 sec, which is 1.073 times the average period value. This value
differs by less than 3% from the coefficient value of 1.104 in eq. 5.51.
140 Random Seas

B. The Bretschneider Spectrum

Following Bretschneider (1959, 1963), assume that the wave heights have a Rayleigh
probability distribution, as expressed in eq. 5.19. By doing so, we can relate average
wave height and the significant wave height to the root-mean-square wave height,
as in eq. 5.21b and eq. 5.24, respectively. These relationships are
√
␲
Havg = Hr ms  0.886Hr ms (5.52)
2
and
Hs  1.416Hr ms (5.53)
Bretschneider (1959, 1963) applies his analysis to deep-water waves, where the
wavelength-period relationships are
gT 2 d␭0 gT
␭0  ,  (5.54)
2␲ dT ␲
from eq. 3.36. The subscript “0” signifies deep-water conditions. Assume that the
wavelengths also have a Rayleigh probability distribution in deep water. The prob-
ability density function for the wavelengths is


2␭0 − ␭ ␭0 2
p(␭0 ) = 2 e 0r ms (5.55)
␭0r ms
which is similar to the wave height expression in eq. 5.19. The root-mean-square
wavelength in eq. 5.55 is related to the average wavelength, the mean-square, and
root-mean-square periods, according to
2 g
␭0r ms = √ ␭0avg = 3/2 Tr2ms (5.56)
␲ ␲
We now require that p(T)dT = p(␭0 ) d␭0 . By combining this relationship with the
results of eqs. 5.54 through 5.56, we obtain the probability density function for the
deep-water wave periods, which is

4
4
d␭0 T3 ␲ T 3 −0.675 Tavg
T
= ␲ 4 e− 4 Tr ms  2.70 4 e
T
p(T) = p(␭0 ) (5.57)
dT Tr ms Tavg
where the results of eq. 5.49 have been used to obtain the last approximation. With
the probability density function for the wave period, we can rewrite the expression
for the average wave period of eq. 5.47 as
&∞ &∞
TSB(T)dT TSB(T)dT ∞
0 0
Tavg = = = Tp(T)dT (5.58)
&∞ Hr2ms
SB(T)dT 0
0

The subscript “B” identifies the spectral density function as that of Bretschneider
(1959, 1963). From the second equality, we see that C = 1 in eq. 5.41b for the
Bretschneider spectral density function. Hence, the integral of the Bretschneider
spectral function is eight times that of eq. 5.41a, by definition. Bretschneider defined
the spectrum in this manner. From the last equality in eq. 5.58, the relationship
between the spectral density and the probability density function is obtained. By
5.8 Wind-Wave Spectra 141

replacing the probability density function of the integrand of the last integral in
eq. 5.58 by the expression in eq. 5.57, the Bretschneider spectral density formula is
obtained. The result is
2  4
Havg −0.675 Tavg
T
SB(T) = Hr2ms p(T) = 1.27Havg
2
p(T) = 3.437 4
T3e (5.59)
Tavg
A comparison of the spectral density expressions in eqs. 5.44a and 5.59 shows that
the expression for the A coefficient is
2
Havg
AB = 3.437 4
(5.60)
Tavg
while that of the B coefficient is
   
1 1
BB = 0.75  0.675 (5.61)
T04 4
Tavg

As stated in Section 5.7B, when the spectral density expression of Bretschneider

is integrated over 0 < T < ∞, the result is the mean-square of the wave height,
that is,
∞ ∞
AB
SB(T)dT =  1.27Havg
2
= Hr2ms p(T)dT = Hr2ms (5.62)
4BB
0 0

C. The Pierson-Moskowitz Wind-Wave Spectrum

The wave spectral density expression of Pierson and Moskowitz (1964) is based on
the assumption of Kitaigorodskii (1962) that the wave spectrum is a function of four
variables. Those are the fetch (F), gravitational acceleration (g), frictional velocity
of the wind (U+ ), and the frequency ( f ). Pierson and Moskowitz (1964) choose to
use the wind speed (U19.5 ) measured by a manometer at z = 19.5 m in their formula
in place of the frictional velocity. Their spectral formula contains constants that sat-
isfy the measured data of Moskowitz (1964) in open-ocean, fully developed seas.
The fetch is then assumed to be infinite. The Pierson-Moskowitz spectral density
function is
 4
0.00810 2 3 − (2␲)
0.74 g
T
SPM = 4
g T e 4 U19.5
(5.63)
(2␲)
where the subscript PM refers to this particular spectral density formula. The con-
stant in eq. 5.41b is C = 1/8. A comparison of the expressions of eqs. 5.43 and 5.63
results in the following expressions for the Pierson-Moskowitz coefficients:
0.00810g 2
APM = (5.64)
(2␲)4
and
 4
0.74 g
BPM = (5.65)
(2␲)4 U19.5
The integral of the Pierson-Moskowitz wave spectral density expression
[SPM (T) in eq. 5.63] is equal to the variance of the free-surface displacement, ␩.
142 Random Seas

60

Logarithmic
50

40
Figure 5.12. Relationship Between the Ave-
U19.5, (m/s) Approximate rage Wind Speeds at z = 19.5 m and z =
30 10 m. The logarithmic and approximate
curves result from eq. 5.67b. The verti-
cal dashed line is the upper limit of the
20 JONSWAP data, as reported by Carter
(1982).

10

0
0 10 20 30 40 50
U10 , (m/s)

For a Gaussian sea, the average free-surface displacement is zero, and the variance
is simply equal to the mean-square of the free-surface displacement. By using the
results in eq. 5.35a, the integral of the Pierson-Moskowitz spectrum is
∞
APM (U19.5 )4 Hr2ms
SPM (T)dT =  2.74x10−3 = = ␩2 (5.66)
4BPM g2 8
0

Compare this expression with the discrete data expression of eq. 5.40. From the
comparison, we see that the Pierson-Moskowitz spectrum and the discrete-data
spectrum have similar definitions. However, a comparison of the expressions in
eqs. 5.40 and 5.62 shows that the integral of the Bretschneider spectrum is eight
times that of the Pierson-Moskowitz spectrum, and also approximately eight times
the sum of the discrete-data spectrum.
Following Pierson (1964), the relationship between the wind speed Uz (mea-
sured at any z) and that measured at z = 10 m is given by the following form of
logarithmic law:
 √ 
0.80 + 0.114U10  z 
Uz = U10 1 + ln (5.67a)
0.4 × 103/2 10
The reference height of 10 m is that used by Bretschneider (1959) and others. The
relationship between the nominal wind speeds at z = 19.5 m and z = 10 m is then
*  +
U19.5 = U10 1 + 0.0528 0.80 + 0.114U10  1.075U10 (5.67b)

The approximation is used by Carter (1982) in an analysis of the JONSWAP data,

and is valid to approximately U10  30 m/s. The approximate expression is also used
later in this chapter. Curves obtained from the exact and approximate expressions
in eq. 5.67b are presented in Figure 5.12. The approximation introduces an error
of approximately 6.7% for U10 = 50 m/s. The logarithmic expression is presented
5.8 Wind-Wave Spectra 143

0.2
Experimental Data
Bretschneider
Pierson-Moskowitz

S (T ), (m2/s)
Figure 5.13. Comparison of the Bretsch-
neider, Pierson-Moskowitz Spectra and 0.1
the Long-Term Spectral Data of Fig-
ure 5.11.

0.0
0 2 4 6 8 10
T, (s)

because it appears in the paper of Moskowitz (1964). However, Kinsman (1965)

expresses doubt about the validity of the logarithmic law for wind over water.

EXAMPLE 5.11: COMPARISON OF OPEN-OCEAN WIND-WAVE SPECTRA The wave

spectral density values corresponding to the data in Table 5.2 are plotted in
Figure 5.13. Let us see how well the spectrum of these values compares with
the spectra predicted by the Bretschneider formula (eq. 5.59) and the Pierson-
Moskowitz formula (eq. 5.63). From the results of eqs. 5.62 and 5.66, the two
empirical spectral formulas are related by

AB 2APM
= = Hr2ms (5.68)
4BB BPM
To apply these two formulas to the data, we must use the experimental value
of the mean-square wave height, which is 2.87 m2 from Examples 5.3 and 5.8.
First, the wind speed in eq. 5.66 can be determined. That value is approximately
10.6 m/s. Combine this value with the Pierson-Moskowitz spectral density
expression of eq. 5.63. Next, the mean wave height value of 1.50 m and the
mean period value of 6.13 sec (both values from Example 5.3) are combined
with the Bretschneider spectral density expression of eq. 5.59, and the result-
ing expression is multiplied by eight. The results are presented in Figure 5.13.
In that figure, we see that the discrete spectrum is sharply peaked at a 6.5-sec
period. The empirical spectra have peak values of less than half of the dis-
crete data peak value. The peak values of the empirical spectra occur at the
modal period (T0 ), defined by eq. 5.45. For the Pierson-Moskowitz spectrum,
the modal period value is 6.81 sec. For the Bretschneider spectrum, the value is
6.29 sec. The reader must keep in mind that comparisons of the spectra in this
example are for the purpose of illustration. The wave height and period data in
Figure 5.2 are representative of long-term seas but not of wind-generated seas.
144 Random Seas

7
Moskowitz Data
Bretschneider
6
Pierson-Moskowitz

5
(g/U 3) S × 104

4
Figure 5.14. Non-Dimensional Bretschnei-
der and Pierson-Moskowitz Spectra and the
3 Observed Wind-Wave Data of Moskowitz
(1964).

0
0 10 20
(g/U)T

For open-ocean wind-generated seas, spectral results obtained by using the

Bretschneider and Pierson-Moskowitz formulas compare well with measured spec-
tra. This fact is demonstrated by the results in Figure 5.14, where the spectral
data of Moskowitz (1964) and the results of eq. 5.59 (subject to the condition of
eq. 5.68) and eq. 5.63 are presented. The reader should note that the parameters of
the Pierson-Moskowitz spectral density formula (APM and BPM ) are based on the
Moskowitz (1964) data. The good agreement of the Bretschneider spectral results
and the Moskowitz data indicates that the statistics of open-ocean wind-generated
seas are Rayleigh in nature.

D. The JONSWAP Spectra – the Fetch-Limited Sea

The discussions to this point have been concerned with open-ocean wind-generated
seas, that is, where the fetch is assumed to be infinite. If our engineering interest is in
a site that is near the coast of a land mass, then the wave statistics can be significantly
altered by the limited fetch. To establish fetch-limited statistics, the Joint North Sea
Wave Project (JONSWAP) was conducted in 1969, and reported by Hasselmann
et al. (1973). As discussed by Komen et al. (1994), Hasselmann et al. (1973) modified
the form of the Pierson-Moskowitz spectral representation (eq. 5.63) for an open-
ocean fully developed sea to obtain a four-parameter spectral representation of a
fetch-limited sea. The result is the five-parameter JONSWAP wave spectral density
formula, which is
 2
1 Tp
− −1
2
2␴a,b T
SJ ON (T) = AJ ON T 3 e−BJ ON T ␥ e
4
(5.69)
5.8 Wind-Wave Spectra 145

where
 −0.22
0.076 Fg
AJ ON = g 2
(5.70)
(2␲)4 2
U10

and
1.25
BJ ON = (5.71)
Tp4

In the spectral formula, the five parameters are AJON , Tp , ␥ , ␴ a , and ␴ b . The first two
are called the scale parameters, and the last three are the shape parameters. The term
␥ is called the peak enhancement parameter, whereas ␴ a and ␴ b modify the width of
the spectrum and thereby also enhance the peak. As reported by Hasselmann et al.
(1976), the mean values of the shape parameters are ␥ = 3.3 and ␴a = 0.07, which
apply to the high-period side of the peak period, Tp , in eq. 5.46 and ␴ b = 0.09, which
applies to the low-period side.
The mathematical form of the JONSWAP spectral formula of eq. 5.69 presents
a problem in obtaining explicit relationships. Relationships among the root-mean-
square wave height, peak spectral period, nominal wind speed, and fetch are
obtained using numerical techniques. These relationships are presented in the next
section.
For more in-depth discussions of the JONSWAP spectral density, the reader is
urged to consult the publications of Goda (1985), Hogben (1990), Tucker (1991),
Komen et al. (1994), and McCormick (1999).

E. Wave Property Relationships from Empirical Wind-Wave Spectra

In the following subsections, the empirical spectra for three situations are discussed.
Those are the fully developed open-ocean seas, the fetch-limited seas, and the
duration-limited seas.

(1) Fully Developed, Open-Ocean Seas

For wind-generated seas in the open ocean, results of the formula of Pierson
and Moskowitz (1964) can be used to obtain the relationships between average
wave properties (wave height and period) and the nominal wind speed. Following
Bretschneider (1959), who assumes that the Rayleigh probability density function
is valid for both the wave heights and wavelengths, the average wave heights and
periods can be related to the nominal velocities (U19.5 and U10 ). The wind-speed
approximation of eq. 5.67b is used to relate the nominal speeds at z = 10 m and z =
19.5 m. The resulting wave height and period expressions are as follows.
Root-Mean-Square Wave Height: The expression for the root-mean-square
wave height can be obtained directly from eq. 5.66. The result is
2
U19.5 U2
Hr ms  0.148  0.171 10 (5.72)
g g
Average Wave Height: To obtain the mean wave height expression, combine
the expressions of eq. 5.72 and the Rayleigh result of eq. 5.21b. This combination
146 Random Seas

10 15

10
Hs (T), (m)
5 To, (s)
Eq. 5.74 Eq. 5.77
5

0 0
0 5 10 15 20 25 0 5 10 15 20 25
U19.5, (m/s) U19.5, (m/s)
a. Significant Wave Height b. Modal Wave Period
Figure 5.15. Comparison of Moskowitz (1964) Data and Analytical Curves. The significant
wave height (HS ) and the modal period (To ) are shown as functions of the wind speed mea-
sured at 19.5 m above the SWL.

results in
√
␲ U2 U2
Havg = Hr ms  0.131 19.5  0.152 10 (5.73)
2 g g
Significant Wave Height: Again, assuming that the Rayleigh distribution of
wave heights is valid, eqs. 5.24 and 5.72 can be combined to obtain the expression
for the significant wave height. The resulting expression is
2
U19.5 U2
Hs  1.416Hr ms  0.210  0.243 10 (5.74)
g g
Results of this expression are presented in Figure 5.15a with the observed data
reported by Moskowitz (1964).
Average Period: To obtain the relationship between the average wave period
and the nominal velocity, combine eqs. 5.47 and 5.65 by eliminating the BPM coeffi-
cient. The result is
U19.5 U10
Tavg  6.14  6.60 (5.75)
g g
Root-Mean-Square Period: From the combination of eqs. 5.48 and 5.65, the
expression for the root-mean-square wave height is
U19.5 U10
Tr ms  6.38  6.86 (5.76)
g g
Modal Period: Equations 5.45, 5.46, and 5.65 are combined to obtain the expres-
sions for the modal and peak spectral periods, which is
U19.5 U10
To  0.88Tp  6.30  6.77 (5.77)
g g
Results obtained from eq. 5.77 are presented in Figure 5.15, with the modal period
values corresponding to the Moskowitz (1964) data.
5.8 Wind-Wave Spectra 147

The reader can see that the results obtained from the empirical equations and
the Moskowitz (1964) data in Figure 5.15b compare quite well. These good compar-
isons reinforce the conclusion that the probability distributions of the wave heights
and the wavelengths for a fully developed, open-ocean sea are well represented by
the Rayleigh probability density function, as assumed by Bretschneider (1959).

(2) Fetch-Limited Seas

Further analyses of the data leading to the JONSWAP spectral formula of eq. 5.69
yield averaged heights and periods for fetch-limited seas, that is, the fetch length (F)
over which the wind blows is limited by the presence of a land mass located upwind
from the site in question. Two such analyses are those of Hasselman et al. (1976)
and Carter (1982). The Carter results are presented here.
The Carter (1982) equations are presented in terms of the fetch (F) in kilome-
ters, wind duration (tD ) in hours, and the nominal wind speed (U10 ) in meters per
second. The wave properties can be determined by either the fetch or the duration,
depending on value of the critical duration, defined by
  0.7
U10 F
t DF ≡ 2.315 g 2 (5.78)
g U10

For the fetch-limited sea, Carter (1982) states that

F0.7
t D > t DF = 1.167 0.4
(5.79)
U10

where, again, the fetch is in kilometers, the duration in hours, and the wind speed in
meters per second. The gravitational constant (g) used is 9.81 m/s2 . For this condi-
tion, the following expressions apply:

Significant Wave Height:

  0.5
2
U10 gF √
Hs  0.0511 2
 0.0163U10 F (5.80)
g U10

Peak Spectral Period:

  0.3
U10 gF
Tp  2.80 2
 0.566U10
0.4 0.3
F (5.81)
g U10

Average Wave Period:

  0.3
U10 gF
Tavg  0.857Tp  2.40 2
 0.485U10
0.4 0.3
F (5.82)
g U10

where the relationship between the two periods is obtained from eq. 5.51.
Rather than using the mean period as in eq. 5.82, Carter (1982) presents an
expression for the zero up-crossing period (Tz ), which he states is approximately
equal to 0.710Tp .
148 Random Seas

(3) Duration-Limited Seas

For the duration-limited sea, the condition reported by Carter (1982) is
F0.7
t D ≤ t DF = 1.167 0.4
(5.83)
U10
For this condition, the following expressions apply:
Significant Wave Height:
  5/7
2
U10 gt D 9/7 5/7
Hs  0.0280  0.0146U10 t D (5.84)
g U10

Peak Spectral Period:

  3/7
U10 gt D 4/7 3/7
Tp  1.99  0.540U10 t D (5.85)
g U10
Average Wave Period:
  3/7
U10 gt D 4/7 3/7
Tavg  0.857Tp  1.70  0.461U10 t D (5.86)
g U10
By using the results of Sections (1), (2), and (3), the limiting conditions of
the JONSWAP relationships can be determined. That is, by assuming that the
JONSWAP and Pierson-Moskowitz are identical at limiting values of either the
fetch or duration for a given wind speed, we can obtain the limiting expressions for
F and tD for a developing sea. We shall do this by equating the various significant
wave height expressions presented in the subsections. For the fetch-limited case, the
Pierson-Moskowitz expression of eq. 5.74 and the JONSWAP expression of eq. 5.80
are equated to obtain the minimum fetch for a fully developed sea (Ffds ). The result
is
F f ds = 2.32U10
2
(5.87)
in kilometers. Hence, when F ≥ F f ds , the sea is fully developed. For the duration-
limited case, the minimum duration for a fully developed sea is obtained by equating
the expressions in eqs. 5.74 and 5.84 and solving for the duration. The result is
t Df ds = 2.10U10 (5.88)
The analysis of the situation described in Example 5.12 illustrates the use of
the equations in hours. So when tD ≥ tDfds , the sea is fully developed. Following
the example of Carter (1982), a decision tree for the selection of the appropriate
expressions is given in Table 5.3.
EXAMPLE 5.12: FETCH-LIMITED AND DURATION-LIMITED SEAS Prior to the com-
mencement of a beach nourishment project near Ocean City, Maryland, on the
Atlantic coast of the contiguous United States, offshore operations at several
borrow sites are terminated at sea states above certain values. Borrow sites are
large sand bars that are the sand resources. The reader should consult the papers
by Grosskopf and Kraus (1994) and Dean (2003) for excellent discussions of
beach nourishment. For this particular operation, the maximum possible sea
states at the offshore sites are based on a 5-hour storm having a nominal wind
5.8 Wind-Wave Spectra 149

Table 5.3. Decision tree for wave height and period equations

t D > t DF ?
(hours)
↓
Yes ← ↔ → No
(Fetch Limited) (Duration Limited)
↓ ↓
F > F f ds ? t D > t Df ds ?
(km) (hours)
↓ ↓
No ← → Yes ← → No
(Fetch Limited) (Fully Developed) (Duration Limited)
↓ ↓ ↓
Hs (m): eq. 5.80 eq. 5.74 eq. 5.84
↓ ↓ ↓
Tp (sec): eq. 5.81 eq. 5.77 eq. 5.85
↓ ↓ ↓
Tavg (sec): eq. 5.82 eq. 5.75 eq. 5.86

speed of 15 m/s. Hence, U10 = 15 m/s and tD = 5 hours. The borrow sites are
located 1 km, 5 km, and 10 km east of the Ocean City shoreline. Assume all sites
are in deep water. For winds out of the west, the JONSWAP equations are to
be used, and for winds out of the east, the Pierson-Moskowitz equations apply.
For the winds out of the west, we must first determine the value of the criti-
cal duration, defined in eq. 5.78 for each borrow site. Applying the expression in
that equation to the various sites results in the following critical duration values:
0.40 hour(F = 1 km)
t DF  0.395F0.7  1.22 hour(F = 5 km)
1.98 hour(F = 10 km)
Because our 5-hour duration is greater than these values, as in eq. 5.79, the seas
are fetch-limited at each site. Following the decision tree in Table 5.3, we must
now determine if the sea is fully developed. From eq. 5.87, the minimum fetch
for a fully developed sea is 522 km for our 15 m/s wind speed. So at each site,
the sea is developing when the winds are from the west. The significant wave
height values at each site are determined from eq. 5.80. Those values are

√ 0.244 m(F = 1 km)

Hs  0.244 F  0.546 m(F = 5 km)
0.772 m(F = 10 km)
The application of eq. 5.82 to each site yields the following average period val-
ues:
1.137 sec(F = 1 km)
Tavg  1.137F0.3  1.843 sec(F = 5 km)
2.269 sec(F = 10 km)
When the winds are out of the east at Ocean City, the seas will be fully devel-
oped if the fetch exceeds 522 km, from eq. 5.87. Assume that this is the case
for the 10 km borrow site. However, note that the duration does not exceed
31.5 hours – the condition of eq. 5.88 for a duration-limited sea. From Table 5.3,
150 Random Seas

for the fetch-limited sea where the fetch is the minimum for a fully developed
sea, we can use the Pierson-Moskowitz results of eqs. 5.74 and 5.75 to obtain
the respective values of the significant wave height and average period. Those
results are

Hs  5.57 m and Tavg  10.09 sec

Comparing the statistical wave properties resulting from the two wind con-
ditions, we conclude that operations can continue at any of the sites for the
15 m/s wind out of the west, but should terminate long before the 5-hour dura-
tion when the winds are out of the east.

In lieu of a decision tree, such as that presented in Table 5.3, the U.S. Army
(1984) Shore Protection Manual contains curves that result from the JONSWAP
equations. The reader should consult that reference to see if the curves are more to
their liking.

F. Directional Properties of Wind-Generated Waves

The expressions for the spectral density up to this point are those that describe point
spectra, that is, the expressions describe the energy of the wave climate at a single
point in the ocean without consideration of the directions of travel of the component
waves. If three or more closely spaced wave gauges are simultaneously used to mea-
sure the wave properties, then the directionality of the sea can be determined, that
is, the energy distribution as a function of direction can be determined by correlat-
ing the signals from the gauges. The directionality of the waves can be important in
a number of engineering applications. For example, McCormick (1978) takes into
account the directionality of the sea in determining the wave energy resource for
arrays of wave energy conversion devices.
Wind waves are directional due to the nature of their creation. This fact is
demonstrated by the following simplistic analysis of the waves created by pressure
fluctuations on the free surface of the water caused by the passage of eddies in the
wind’s turbulent boundary layer, adjacent to the surface. Consider such a boundary
layer adjacent to the free surface of (initially) calm water. The randomly occur-
ring eddies, convecting in the nominal direction of the wind, are at various heights
above the free surface. Because of the velocity distribution in the boundary layer, as
illustrated in Figure 5.16, the convection velocity increases with distance above the
surface. In addition, the energies associated with the various eddies also depend on
their vertical position.
Let us consider one of these eddies. As described by McCormick and Mouring
(1995) and McCormick, Bhattacharyya, and Mouring (1997) and others, the eddies
first appear as rectilinear line vortices, as sketched in Figure 5.10. Certain segments
of the line vortex acquire energy more rapidly than other segments. As the energy of
a segment increases, the vortex grows and subsequently migrates away from the free
surface. The time scale of the events just described depends on the elevation of the
formation of the vortex in question. The evidence of the events on the water surface
is in the form of pressure fluctuations. The magnitudes of the pressure fluctuations
depend on the position of the vortex and, hence, on the speed at which the vortex is
convected.
Consider two surface-pressure pulses on the calm-water surface that are con-
vecting at velocities u1 and u2 , where u2 > u1 , as illustrated in Figure 5.17. The pulses
 3 Linear Surface Waves

Ocean waves are caused by the motions of celestial bodies, seismic disturbances,
moving bodies, and winds. The waves produced by these phenomena differ in size
and character, and the consequences of each must be dealt with differently.
The gravitational attractions of both the moon and the sun cause the largest
water waves, called the tides. The predictable tidal wave can be treated as a shallow-
water wave because its length is much greater than the water depth. Extreme tides,
called spring tides, occur when the attractive forces of both the moon and sun are
aligned and in the same direction. These tides can cause flooding of lowlands if
dikes or levees are not present. The tides can also be exploited by converting their
energies into useable energy forms. This is normally accomplished by creating tidal
barriers equipped with hydroturbines, taking advantage of the tidal-induced water
level changes on opposite sides of the barriers. An excellent book on tidal energy
conversion is that written by Charlier (1982).
Both sub-marine earthquakes and volcanic eruptions can produce a long, high-
energy wave called a tsunami (the Japanese word for a tidal wave, although the
wave referred to is not tidal in nature). This type of wave can pass a ship in the
open ocean and not be noticed by the ship’s crew because of the small wave height-
to-wavelength ratio (called the wave steepness). As the tsunami approaches a land
mass, the energy of the wave is transformed from mostly kinetic to mostly potential.
This causes the wave steepness to increase significantly, and the resulting high wave
can be devastating to coastal areas. An excellent discussion of the nature and conse-
quences of tsunami is found in the book of Professor Robert L. Wiegel (1964). The
book is a classic in the ocean engineering literature.
The speed of a ship traveling in a restricted waterway is normally limited
because of the potential damage caused by ship waves. These waves travel in a group
away from the line of motion of the ship. The energy of the ship-wave group depends
on the speed and geometry of the ship. When ship waves travel into shallow water,
they can become rather steep without breaking and, as a result, retain much of their
energy. The steep waves can cause excessive motions of small moored vessels that, in
turn, can result in mooring failures. Furthermore, when the steep ship waves break
near the shore, erosion can occur.
Wind-generated waves and their consequences are the primary focus of this
book. As discussed in the next section, winds produce waves that vary in length
from the short capillary wave to the long swell. Wind waves can be classified as

44
3.1 Wind-Wave Generation 45

linear waves (having sinusoidal profiles), nonlinear waves (having nonsymmetric

profiles with respect to the still-water level), and breaking waves. Each of these
waves has a special significance in ocean engineering, as discussed herein. The wind
scale and the sea state of wind waves are presented and discussed in Chapter 1.

3.1 Wind-Wave Generation

Let us first consider the effects of a slight breeze of velocity U0 over the free sur-
face (air-sea interface) of calm water of an open ocean. As discussed in Section 2.5,
because the air velocity is relatively small, the air flow is laminar. This laminar air
flow simply drags the water particles on the free surface in the direction of the flow
due to the viscosities of both the air and water. This air flow produces no wave, as
illustrated in Figure 3.1a.
When the speed of the air increases such that the flow in the boundary layer
adjacent to the free surface is turbulent, then the pressure fluctuations on the free
surface beneath the turbulent air flow deform the free surface, and small waves are
created. These waves are called capillary waves, and have a profile similar to that
sketched in Figure 3.1b. In that figure, we see that the crest of the wave is broad,
while the trough of the wave is narrow. The cause of this rather odd wave profile
is the surface tension, which is the dominant force. Capillary waves travel in the
direction of the air flow because of the shear stress on the free surface. For most
engineering problems involving water waves, capillary waves are of little significance
because of their low energy content.
As the air speed increases, the energy in the air turbulence increases as does that
of the surface shear stress. The air flow can now be referred to as a wind rather than
a breeze. The water converts the energies transferred to it by wind turbulence and
shear stress into longer waves having sinusoidal profiles, as sketched in Figure 3.1c.
These sinusoidal waves are called linear waves because they can be analyzed using
linearized equations, as discussed in the next section. One of the linear properties of
the waves is that of superposition, that is, linear waves of different heights (H) and
lengths (␭) can be combined to form other wave profiles. As discussed in Chapter
5, the most basic analysis of random wave phenomena is that which exploits the
property of superposition.
When the wind speed increases further, both the height of the wave and the
wavelength increase due to the horizontal pressure gradient resulting from the sep-
aration of the air flow on the leeward side of the wave. The altered profile has
a narrow crest and a broad trough, as sketched in Figure 3.1d. We note that the
mean-water level (MWL) – the mean way between the crest and trough – is above
the still-water level (SWL). The reason for this is that volumes of water above and
below the SWL must be equal. Waves having a profile similar to that in Figure 3.1d
are called nonlinear waves because their properties can no longer be well predicted
using linearized equations. To analyze nonlinear waves, we use the expansion theory
of Stokes (1847, 1880), presented in Chapter 4.
A further increase in the wind speed can produce breaking waves, as sketched in
Figure 3.1e. A break is defined as the condition where the horizontal water-particle
velocity at the crest equals the wave velocity (c). The profile of a breaking wave
has a pointed crest and a broad trough. After the wave breaks, the wind can shear
off the crest, producing foamy white water that spills down the leeward side of the
wave. This white-water spill is called a white cap. Breaking waves are discussed in
Section 4.3.
46
Figure 3.1. Scenario of Wind-Wave Generation with Increasing Wind Velocity, U. Some of the notation used in this chapter is defined
in (c). The capillary wave profile in (c) is influenced by the surface tension. The separation of the air flow at the crests in (d) and the
resulting wakes contribute to the crest profile. In (e), the breaking wave profile is sharp-crested. The broken wave has foam on the
leeward side of the crest.
3.2 Airy’s Linear Wave Theory 47

Figure 3.2. Sketch and Notation for the Linear Wave Analysis.

There are a number of hypotheses concerning the generation and growth of

wind waves. These are well discussed and illustrated by St. Denis (1969). In the
remainder of this chapter, the analysis of linear waves is presented. In Chapters 4
and 5, the respective nonlinear waves and random waves are discussed and analyzed.

3.2 Airy’s Linear Wave Theory

The first meaningful analysis of surface waves was performed by George B. Airy
in the middle of the nineteenth century; see Airy (1845). His analysis is known as
either the Airy wave theory or the linear wave theory. The latter name results from
the nature of Airy’s analysis, which involves the solution of the linear equation of
continuity for an irrotational flow (eq. 2.41) and the application of linearized bound-
ary conditions. Although the theory is somewhat basic, the kinematic wave proper-
ties derived from the theory agree quite well with those actually observed, provided
that the wave steepness (H/␭) is small. In this section, we outline the linear wave the-
ory and discuss the predicted behavior of linear waves. Thorough discussions of the
theory are found in the classical books by Kinsman (1984), Lamb (1945), Lighthill
(1979), Phillips (1966), and Stoker (1957).
Before embarking on the analysis of linear waves, it is helpful to consider the
physical properties of a traveling surface wave, as sketched in Figure 3.2. In this fig-
ure, we see that the origin of the Cartesian coordinate system is on the SWL, which
is the calm-water position. The water depth, h, is measured from the sea floor to the
SWL. The wave has a height, H (measured from the trough to crest), and a wave-
length, ␭ (measured from crest to crest), and travels in the x-direction at a celerity
or phase velocity, c. The vertical free-surface displacement, ␩(x, t), is measured from
the SWL and, as indicated, is a function of both time, t, and distance, x. Note: Books
that are directed at the civil engineering aspects of ocean engineering normally rep-
resent the water depth by d and the wavelength by L.
48 Linear Surface Waves

z n 1 c

4 H 2 Figure 3.3. Elliptical Path of a Surface

x SWL
Particle of a Linear Wave. The numbers
3 correspond to the passing of the wave
crest (1), the nodes (2 and 4), and the
Free Surface Paricle Path trough (3). The unit vectors n and N
are the outward normal vectors on the
N free surface and sea floor, respectively.

We begin our analysis by assuming that the flow beneath the free surface is irro-
tational, where irrotational flow is discussed in Section 2.3. By making this assump-
tion, the velocity of the water particles can be represented by the velocity potential,
␾, defined in eq. 2.38. The equation of continuity for an incompressible, irrotational
flow is expressed by Laplace’s equation, eq. 2.41. That equation is a specialized form
of the wave equation, which is a second-order linear equation describing most phys-
ical wave phenomena, including light and sound. In our analysis, the general solu-
tion of Laplace’s equation is obtained first, and the solution is then subjected to
linearized boundary conditions. Those boundary conditions are the following:
a. Kinematic Free-Surface Condition. This boundary condition requires that the
same water particles comprise the free surface at all times. Referring to Fig-
ure 3.2, this kinematic free-surface condition is mathematically represented by
V|z=␩ = Vn (3.1)
Physically, this equation expresses the condition that the velocity of a particle
on the free surface must equal the velocity of the free surface itself. Because
irrotational flow is assumed, eq. 3.1 can also be written as
∂␾ 
V|z=␩ = ∇␾|z=␩ = n (3.2)
∂n z=␩
where n is the outward unit normal vector on the free surface, as sketched in Fig-
ure 3.3. Assume that the free-surface displacement, ␩(x,t), is small compared to
the wavelength, ␭, so that the boundary condition described in eq. 3.2 can be
approximated by
∂␩ ∂␾ 
V|z=␩  k k (3.3)
∂t ∂z z=0
Note: The boundary condition in eq. 3.3 is mixed. That is, the left term is applied
at z = ␩ while the right term is applied at z ≈ 0. The space and time dependen-
cies of the free-surface displacement are implied in the notation ␩. For the most
part, this notation will be used in the remainder of this chapter.
b. Sea-Floor Condition. Here, it is required that the water particles adjacent to
the floor (or bed) cannot cross that solid boundary. Mathematically, if the bed
is at z = −h, then
∂␾ 
V · N|z=−h = =0 (3.4)
∂ N z=−h
where N is the outward normal unit vector at the sea floor, sketched in Fig-
ure 3.3.
c. Dynamic Free-Surface Condition. For this condition, the pressure on the free
surface is zero (gauge) at any position, x, and any time, t. Because the flow is
3.2 Airy’s Linear Wave Theory 49

assumed to be irrotational, Bernoulli’s equation (eq. 2.70) can be applied to

the flow to obtain the pressure at any point in the water. Let us first examine
Bernoulli’s equation when applied to the free surface during a calm-water con-
dition. Because there is no motion, the first two terms of eq. 2.70 equate to zero.
On the free surface, the coordinate z is zero because the origin of the coor-
dinate system is on the SWL. Furthermore, because the pressure on the free
surface (here, the SWL) is zero, the time function, f (t), in eq. 2.70 must also be
equal to zero. Still water can then be considered to be a wave with a zero height
and an infinite wavelength. Because f (t) = 0 for this special wave, the equality
must hold for all waves.

Now consider a wave having a finite height and length, where the height is much
smaller than the wavelength (H  ␭), and apply Bernoulli’s equation to the free
surface. The result is
 
∂␾ 1 2 
+ g␩ + V =0 (3.5)
∂t 2 z=␩

Equation 3.5 is nonlinear because of the velocity-squared term. For small values of
the wave steepness, the nonlinear velocity term in eq. 3.5 can be neglected, that is,
the magnitude of the term is of second order when compared to the other two terms.
The linearized dynamic free-surface condition can be expressed mathematically by

1 ∂␾ 
␩= − (3.6)
g ∂t z=␩
Physically, the linearization of eq. 3.5 resulting in the expression of eq. 3.6 is based
on the assumption that the vee-squared kinetic energy of the fluid particles is much
less than the other mechanical energies of the fluid.
The reader might ask what threshold values of wave properties are required to
justify the linearization of eq. 3.5. The following example, a comparison of small and
large waves, is presented to answer that question.

EXAMPLE 3.1: LINEARIZATION Both in the laboratory and in the field, it has been
observed that water particles travel in nearly circular paths if the ratio H/␭
(the wave steepness) is small, and if the wave is not influenced by the bed. The
condition on the bed is termed the deep-water condition. Such a circular path
is a special case of the elliptical path sketched in Figure 3.3 for a particle on
the free surface. As can be seen in that figure, the minor axis of the elliptical
path is equal to the wave height. Therefore, when the wave is in deep water,
the minor and major axes of the path are equal, and the wave height is then the
diameter of a circular path. For the circular path, when two consecutive crests
pass over a time period T, the particle travels a distance ␲H with an average
speed V = ␲ H/T.
Using this information, consider two waves having lengths of 100 m and
periods of 8.0 sec but with respective heights of 10 m and 1 m. The particles
on the steeper wave (for which H = 10 m) have a speed of 3.93 m/s whereas
those on the less-steep wave (for which H = 1 m) have a speed of 0.393 m/s.
With these values, compare the last two terms of eq. 3.5. For the 10-m wave,
the maximum values of the respective terms are 49.0 (m/s)2 and 7.71 (m/s)2 . For
the 1-m wave, the values of these terms are, respectively, 4.90 (m/s)2 and 0.0771
(m/s)2 . The differences in the values of these two terms is less than an order of
50 Linear Surface Waves

magnitude for the 10-m wave but greater than an order of magnitude for the 1-
m wave. Hence, the linearization is justified for the smaller wave but not for the
larger wave. In general, the limits of applicability of the linearized wave theory
depend on the wave height, period, and water depth. An excellent discussion
of these limits is given by Le Méhauté (1969). In Chapter 4, the limits of the
various wave theories are discussed.
The free-surface conditions of eqs. 3.3 and 3.6 are now combined by eliminating
␩ to obtain the linearized free-surface condition, which is
 2 
1 ∂ ␾ ∂␾ 
+ =0 (3.7)
g ∂t 2 ∂z z=␩0
Because the flow in the wave is assumed to be irrotational, the equation of continuity
is expressed by Laplace’s equation, eq. 2.41. That equation is
∇ 2␾ = 0 (3.8)
where the Laplacian operator,∇ 2 , is defined in eq. 2.42. Equation 3.8 is an elliptic
partial differential equation and, as such, can be solved using a product solution.
There are two forms of the product solution that can be used. The first is of the
form
␾ = X(x)Z(z)T(t) (3.9)
and results in a standing wave, as shown in the book by McCormick (1973) and
others. The second product solution is for a traveling wave, and is
␾ = X(x ± ct)Z(z) = X(␰)Z(z) (3.10)
The coordinate system for eq. 3.9 is fixed at a point, whereas that of eq. 3.10 moves
with the wave. For the traveling wave, substitute the expressions of eq. 3.10 into that
of eq. 3.8, and separate terms of the same variables to obtain
1 d2 X 1 d2 Z
= − = −k2 (3.11)
X d␰ 2 Z dz2
where k is a constant. The negative sign arises because the free-surface profile is
sinusoidal in the ␰-direction. An expression similar to that in eq. 3.11 is obtained
from the combination of eqs. 3.9 and 3.8.
The general ␰- and z-solutions of the ordinary differential equations in eq. 3.11
are, respectively,
X(␰) = C␰ sin(k␰ + ␣) (3.12)
and
Z(z) = Cz cosh(kz + ␤) (3.13)
where C␰ , Cz , ␣, and ␤ are arbitrary constants. Because the origins of the horizontal
coordinates ␰ and x are arbitrary, the constant ␣ can be assigned a zero value without
loss of generality. Equation 3.12 can then be written as
X(␰) = C␰ sin(k␰) (3.14)
To determine the constant ␤, apply the sea-floor condition of eq. 3.4 to the velocity
potential expression resulting from the combination of eqs. 3.10 and 3.13. Assuming
3.2 Airy’s Linear Wave Theory 51

that the sea floor is uniformly flat and horizontal, only the z-term is affected by the
boundary condition. That is, the sea-floor condition results in the equation
dZ 
= kCz sinh(−kh + ␤) = 0 (3.15)
dz z=−h
from which ␤ = kh. The expression in eq. 3.13 is then
Z(z) = Cz cosh[k(x + h)] (3.16)
The combination of the expressions of eqs. 3.14 and 3.16 with that of eq. 3.10 results
in
␾ = C␾ cosh[k(z + h)] sin(k␰) (3.17)
where C␾ = C␰ Cz .
Now consider the nature of the horizontal coordinate, ␰. From eq. 3.10, that
coordinate is defined as
␰ = x ± ct (3.18)
The origin of the coordinate corresponds to
x = ∓ ct (3.19)
From the result in eq. 3.19, we see that the value of x decreases as the time, t,
increases for the upper sign (−). Hence, the wave must travel in the negative x-
direction at a celerity or phase speed, c. The waves corresponding to the upper signs
in eqs. 3.18 and 3.19 are then called left-running waves. Following the same line of
reasoning, the lower signs in those equations correspond to right-running waves. The
respective coordinates for the right- and left-running waves are then
␰ + = (x − ct) (3.20a)
and
␰ − = (x + ct) (3.20b)
Returning to the expression for the velocity potential in eq. 3.17, the arbitrariness
of the coefficient C␾ in the equation can now be removed. To do this, combine the
expressions in eq. 3.6 (the linearized free-surface condition) and eq. 3.17 by elim-
inating the velocity potential, ␾. This combination yields the following expression
for the free-surface displacement of a sinusoidal wave:
ckC␾± H
␩± = cosh(kh) cos(k␰ ± ) = cos(k␰ ± ) (3.21)
g 2
where the last equality results from our knowledge that the wave is sinusoidal in
both time and space. In the right term of eq. 3.21, H is the wave height. Comparing
the terms of the last equality of eq. 3.21, we obtain the expression for the coefficient
C␾, which is
H g 1
C␾± = ± (3.22)
2 kc cosh(kh)
Now combine this expression and that of eq. 3.20 with that of eq. 3.17 to obtain the
final expression for the velocity potential of a traveling wave, that is,
H g cosh[k (z + h)]
␾± = ± sin[k (x ∓ ct)] (3.23)
2 kc cosh(kh)
52 Linear Surface Waves

The velocity potential yields the velocity components of the particles in the irrota-
tional flow beneath traveling waves. The velocity potential expression in eq. 3.23 is
the primary result of the Airy wave theory. The other properties of the linear waves
can now be derived.

3.3 Traveling or Progressive Waves

By convention, right-running waves are normally considered in two-dimensional
wave mechanics problems. This convention is followed herein. Consider the right-
running wave sketched in Figure 3.2. The free-surface displacement caused by the
wave is
H
␩= cos[k(x − ct)] (3.24)
2
from the combination of equations 3.20a and 3.21. Because the free-surface profile
is sinusoidal in both time and space, the maximum displacement from the SWL, or
the crest, occurs when

k(x − ct) = 0, ±2␲, ±, ±4␲, . . . . (3.25)

Consider first the case for which t = 0. The distance in the x-direction between two
successive crests is the wavelength, ␭. From eq. 3.25, we obtain
2␲
k= (3.26)
␭
The wave parameter, k, is called the wave number.
Next, consider the expression of eq. 3.25 when x = 0. The time lapse between
successive crests is the wave period, T. From the results in eq. 3.25, we obtain
2␲
kc = = 2␲ f ≡ ␻ (3.27)
T
In this equation, f is the wave frequency in units of Hertz (Hz), and ␻ is the circu-
lar wave frequency in units of radians per second. Combining eqs. 3.26 and 3.27 by
eliminating the wave number yields the expression for the celerity or phase velocity,
␭
c= (3.28)
T
Returning to the velocity potential of eq. 3.23, the expression in that equation can
now be written in terms of the circular wave frequency as
H g cosh[k(z + h)]
␾= sin(kx − ␻t) (3.29)
2 ␻ cosh(kh)
Combine this expression with that of eq. 3.7 by eliminating the velocity potential.
After some simplification and the introduction of the expression in eq. 3.27, one
obtains the following relationship for the circular wave frequency:

␻ = gk tanh(kh) (3.30)

Waves are dispersive in that the celerity depends on the frequency and length of
the wave and, in addition, on the water depth. Equations 3.26 and 3.30 can be
3.3 Traveling or Progressive Waves 53

Table 3.1. Linear theory parameters

h/␭ kh tanh(kh)

0.01 = 1/100 0.06283.. 0.06275..

0.04 = 1/25 0.25313.. 0.24616..
0.05 = 1/20 0.31415.. 0.30421..
0.10 = 1/10 0.62831.. 0.55689..
0.50 = 1/2 3.14159.. 0.99627..
1.00 = 1/1 6.28318.. 0.99999..

combined to obtain the expression for the wavelength, which is a form of the dis-
persion equation, that is,
 
2␲ 2␲g gT 2 2␲h
␭= = 2 tanh(kh) = tanh = cT (3.31)
k ␻ 2␲ ␭
The equation for the wavelength is transcendental because ␭ cannot be isolated.
That is, the solution of ␭ cannot be obtained analytically, so the values of the wave-
length must be determined using a numerical technique. One such technique is
that called successive approximations or successive substitutions, as discussed by
Carnahan (1969) and others. McCormick (1973) demonstrates the application of
this numerical technique to wave mechanics problems. The significance of eq. 3.31
is illustrated in the following examples.

EXAMPLE 3.2: WAVELENGTH VARIATION WITH WATER DEPTH From a knowledge of

hyperbolic functions, we know that tanh(kh) approaches unity as kh approaches
infinity. This behavior of kh would be observed if either h → ∞ or ␭ → 0. At
the other extreme, where the values of kh are very small, tanh (kh) → kh. Phys-
ically, this extreme corresponds to either relatively small water depths or rel-
atively large wavelengths. The term “relative” is used because the behavior
depends on the ratio h/␭. Some practical values of this ratio are presented in
Table 3.1 along with the corresponding values of kh and tanh(kh) for the sake
of illustration.
The results in Table 3.1 show that the difference in the values of kh and
tanh(kh) is about 3.16% or less for h/␭ ≤ 1/20, whereas for h/␭ ≥ 1/2 the dif-
ference between tanh(kh) and unity is less than 0.4%. The significance of these
approximation ranges is now discussed.

Because of the approximations presented in Example 3.2, it is a common prac-

tice in ocean engineering to partition the infinite range of h/␭ into three regions,
those being:

a. Shallow Water: h/␭ ≤ 1/20 so that tanh(kh) ≈ kh

b. Intermediate Water: 1/20 < h/␭ < 1/2 so that tanh(kh) = tanh(2␲h/␭)
c. Deep Water: h/␭ ≥ 1/2 so that tanh(kh) ≈ 1

In intermediate water, the transcendental relationship of eq. 3.31 applies as

it is written, and must be solved numerically. The numerical method of succes-
sive approximations, as applied to eq. 3.31, is demonstrated in Example 3.3. Note:
Coastal engineers working in the field can use the shallow-water approximation to
gain a fairly good idea of the length of waves in the coastal zone. Outside the coastal
54 Linear Surface Waves

100
T = 5 seconds FA (λ) = λ
h = 10 meters
meters

50 FA [FB (λ1)] = Figure 3.4. Graphical Illustration of a

λ2 = 21.7m
Successive Approximation Solution for
the Wavelength.
FB (λ1)
λ6 = 36.6m
FB (λ)

0
0 50
λ, meters λ1 = 100m

zone, offshore engineers find the deep-water approximation to be very useful in field
calculations of the wavelength.
EXAMPLE 3.3: WAVELENGTH SOLUTION BY SUCCESSIVE APPROXIMATIONS A wave
having a period (T) of 5 sec travels in water that is 10 m deep (h = 10 m).
The wavelength value determined from eq. 3.31 is obtained using the method
of successive approximations. The method can best be illustrated as follows:
Define two functions in the equation, those being
FA(␭) = ␭ (3.32)
and
 
gT 2 2␲h
FB(␭) = tanh (3.33)
2␲ ␭
Plot these functions as shown in Figure 3.4. Obviously, the solution of eq. 3.31
is obtained when
FA(␭) = FB(␭) (3.34)
or where the lines in the figure cross. In fact, depending on the degree of accu-
racy desired, one could use the graphical solution. However, to increase the
degree of accuracy we use successive approximations. To find the wavelength
value numerically, assume a starting value of ␭ equal, say, ␭1 = 10h = 100 m, as
illustrated in Figure 3.4. The initial value of FA (␭) is then 100 m. If this is the
solution, then FB (␭1 ) should also have the value of ␭1 . However, for this first
approximation the value of the function is 21.7 m. To test the accuracy, let us
define the difference function as
 F = |FB(␭i ) − FA(␭i )| = |FB(␭i ) − ␭i | (3.35)
The value of this function for the first approximation (i = 1) is 78.3 m. For
this example, let the desired value of  F be 0.1 m. The second approximation
(i = 2) is obtained by letting FA(␭2 ) = ␭2 = 21.7 m and evaluating FB (␭2 ). The
new value of that function is 38.8 m, and the value of the difference function is
17.1 m. By following this procedure, the desired accuracy is achieved in the sixth
approximation (i = 6), and the value of ␭6 is 36.6 m. A flow chart for successive
3.3 Traveling or Progressive Waves 55

INPUT DEFINE FA (λ) = λ

T g λ1 = 100m I gT 2 2π
h π FB (λ) = tanh
2π h

no
I λ = FA (λ) = FB (λ) FB (λ) = FA (λ) ≤ 0.1m?

yes
Print
END
λ

Figure 3.5. Flow Chart for the Successive Approximation Solution for the Wavelength Value.

approximation is in Figure 3.5. The only restrictions on the initial value of ␭ are
that it must be both greater than zero and finite.
From the graph in Figure 3.4, we can see that the number of itera-
tions required to achieve the desired accuracy is reduced as the value of ␭1
approaches the actual solution. However, from the results of this example it
is apparent that the process rapidly converges on the solution.
Now apply the deep- and shallow-water approximations to eq. 3.31. First, for
deep water the wavelength approximation is
gT 2
␭ ≡ ␭0   c0 T (3.36)
2␲
where the subscript 0 is, by convention, used to identify deep-water properties. Phys-
ically, the result in eq. 3.36 shows that the deep water wavelength and celerity are
functions of the wave period only, both increasing as the period increases.
The shallow-water approximation of eq. 3.31 is
gT 2 gT 2 2␲h gT 2 h
␭ kh = =  cT (3.37)
2␲ 2␲ ␭ ␭
from which

␭= ghT  cT (3.38)
From this relationship, we see that the shallow-water wavelength is then a function
of both depth and period, decreasing as both decrease. However, the shallow-water
celerity is a function of depth and independent of period. From these results, we
see that waves both shorten and slow down as they approach the shoreline. The
approximations of both eqs. 3.36 and 3.38 are applied to the conditions in Example
3.3 in the following example.
EXAMPLE 3.4: DEEP- AND SHALLOW-WATER WAVELENGTH APPROXIMATIONS In
Example 3.3, the wave period is 5 sec, and the water depth is 10 m. For this
period value, the deep-water approximation in eq. 3.36 yields a wavelength
value of 39.0 m. The shallow-water wavelength approximation from eq. 3.38
for the given period and water depth is 49.5 m. To an accuracy of 0.1 m, the
numerical solution of eq. 3.31 yields 36.6 m. For the conditions in Example 3.3,
the deep-water approximation is rather good as a first estimate whereas the
shallow-water approximation is relatively poor.
A knowledge of traveling waves is needed for engineering analyses of both off-
shore and coastal structures. The linear wave theory for traveling waves discussed
in this section is useful in the analysis of the motions of floating bodies, as discussed
56 Linear Surface Waves

c −c
η+ η−
H
x Figure 3.6. Illustration of the Method of Images
as Applied to Perfect Reflection.

λ λ

a. Incident Wave b. Mirror Image

in Chapter 10. However, as waves travel into shallow water the steepness of the
waves increases and the wave profiles become nonlinear. For nonlinear waves, the
analytical methods are discussed in Chapters 4 and 6.

3.4 Standing Waves

In Section 3.2, it is stated that there are two forms of the product solution of the
equation of continuity, expressed as Laplace’s equation, eq. 2.41. The first is the
standing wave solution of eq. 3.9 and the second is the traveling wave solution of
eq. 3.10. The velocity potential for the traveling wave (both right- and left-running),
as derived from the linear theory, is presented in eq. 3.23. We can use that equation
to determine the velocity potential for a standing wave by taking advantage of the
property of superposition, discussed in Section 2.3D, that is, the velocity potential of
any two (or more) linear waves can be added together to obtain an additional wave
pattern.
Referring to Figure 3.6, consider two traveling waves having the same heights
and periods passing each other in the x-z plane, one wave being right-running and
the other left-running. The vector celerities of the respective waves are c+ and c− .
Because these vectors are equal in magnitude and opposite in direction, they cancel
each other when the two wave patterns are superimposed. The resulting wave pat-
tern is one having zero celerity, or a standing wave. By matching the signs in eq. 3.23,
the respective velocity potentials (␾+ and ␾− ) are obtained. Again, because of the
superposition property, we can add these potentials to obtain the potential for a
standing wave,
−Hg cosh[k(z + h)]
 = ␾+ + ␾− = cos(kx) sin(␻t) (3.39)
␻ cosh(kh)
In a similar manner, the free-surface displacements of the right-running wave,
␩+ of eq. 3.24, and the left-running wave (␩− ) can be added to obtain the following:
H
␩S = ␩+ + ␩− = H cos(kx) cos(␻t) = cos(kx) cos(␻t) (3.40)
2
after using trigonometric identities and simplifying (see Figure 3.6). In eq. 3.40, we
see that the standing wave height, H, is twice that of the traveling wave height, H,

H = 2H (3.41)

as illustrated in Figure 3.7.

3.4 Standing Waves 57

Figure 3.7. Sketch of a Standing Wave Profile Resulting from a ηs

Perfect Reflection. H = 2H
x

In both eqs. 3.39 and 3.40, we note that the spacial (x) and temporal (t) functions
are independent of each other. Because of this, the positions of the nodes (where
␩ = 0) and the antinodes (positions of the crests and troughs) are fixed in space.
Now take the origin of the spacial coordinate system to be as shown in Figure 3.7.
The nodes occur when
2␲ ␲ 3␲
kx = x= , ,... (3.42)
␭ 2 2
So the nodal points are at
␭ 3␭
x= , ,... (3.43)
4 4
where ␭ is the wavelength of both the traveling and standing waves.
The velocity potential expression in eq. 3.39 can be used in the Cauchy-Riemann
expressions of eqs. 2.51 and 2.52 to determine the stream function for the standing
wave pattern. The result is
Hg sinh[k(z + h)]
= sin(kx) sin(␻t) (3.44)
2␻ cosh(kh)
As discussed in Section 2.3C, the stream function can be used for both two-
dimensional rotational or irrotational flows to establish the flow streamlines. In fact,
any constant value of  (the notation for a standing wave stream function) will
define a specific streamline. Furthermore, because no flow can cross a streamline,
any streamline can be considered to be a flow boundary. In Example 2.3, we see that
a zero value of the stream function defines the stagnation streamline and the surface
of a circular cylinder in a uniform flow. Consider the streamlines corresponding to
 = 0 when t > 0. The zero value of the stream function occurs where
kx = 0, ␲, 2␲. . . . (3.45)
and when z = −h , i.e., on the sea floor. The condition in eq. 3.45 corresponds to
␭
x = 0, , ␭, . . . (3.46)
2
as sketched in Figure 3.8. To illustrate, consider the following example.
EXAMPLE 3.5: STANDING WAVES AT A SEAWALL Traveling waves having a height
of 0.5 m and a period of 5 sec perfectly reflect from a vertical seawall where the
water depth is 2 m. So, H = 0.5 m, T = 5 sec, and h = 2 m. These values of
period and water depth in eq. 3.31 yield a wavelength of approximately 20.9 m.
58 Linear Surface Waves

λ
Nodal Points
ηs
SWL H x

h Ψ>0 Streamline

Ψ=0

Figure 3.8. Sketch of Streamlines within a Standing Wave.

The water depth-to-wavelength ratio for the stated conditions is h/␭ = 0.096, or
approximately 1/10. From the discussion in Section 3.3, this value corresponds
to intermediate water. After the wave reflects from the wall, the resulting stand-
ing wave height is twice that of the incident wave, so H = 1.0 m according to
eq. 3.41. The steepness of the incident traveling wave (H/␭) is 0.0239, whereas
that of the standing wave (H/␭) is 0.0478. The resulting stream function from
eq. 3.44 is
 = 0.335 sinh(0.601 + 0.301z) sin(0.301x) sin(1.26t) (3.47)
From this equation, the geometries of the streamlines corresponding to con-
stant values of  are determined. The resulting streamlines are similar to those
sketched in Figure 3.8, where the time is t = ␲/2␻. In that figure, we see that
the fluid is partitioned into flow cells, the cells being half a wavelength in width,
or 10.45 m for the stated conditions. The boundaries of each cell correspond to
 = 0, as does the sea bed. Antinodes (crests and troughs) occur at the sides of
the cells and nodes are at the center of the cells.
Because there can be no flow across streamlines, any liquid spilled into the
water when standing waves are present should (theoretically) remain in the cell
in which the liquid is poured, provided that there are no other currents in the
cell. Contaminants should then be containable adjacent to sea walls.
From the discussion, we see that the reflection of traveling waves can result in
a standing wave, the height of which is twice that of the incident traveling wave, as
in eq. 3.41. However, the wavelength is the same for both the traveling and standing
waves because the wavelength depends only on the water depth and wave period.
One very important consequence of wave reflection from seawalls or quays is the
doubling of the wave steepness, as illustrated in Example 3.5. That is, because the
wavelength is unaffected by reflection, the wave steepnesses of the standing wave
and the incident traveling wave are related by
H H
=2 (3.48)
␭ ␭
This condition follows directly from eq. 3.41. If a ship or a boat is moored to the sea-
wall, the increased wave steepness can cause excessive motions of the vessel, with
possible damage to the hull or the moorings. For this reason, when storms are fore-
cast large ships are usually taken from their berths to sea, whereas the smaller boats
3.5 Water Particle Motions 59

are taken to more sheltered waters. Standing wave formation is a major concern in
harbor design.

3.5 Water Particle Motions

In Section 2.3, the relationships between the velocity potential, stream function, and
velocity components of the water particles in a two-dimensional irrotational flow
are introduced as the Cauchy-Riemann equations, eqs. 2.51 and 2.52. In Section 3.2,
Airy’s theory yields the expression for the velocity potential. For right-running trav-
eling waves, the velocity potential expression is in eq. 3.29, whereas for a standing
wave, the potential is given by eq. 3.39.
To obtain the expressions for the two-dimensional water particle velocity com-
ponents in traveling waves in irrotational flows, simply combine eqs. 2.51 and 2.52,
respectively, with eq. 3.29 to obtain the following:
Hgk cosh[k(z + h)]
u= cos(kx − ␻t) (3.49)
2␻ cosh(kh)
H␻ cosh[k(z + h)]
= cos(kx − ␻t)
2 sinh(kh)
in the horizontal direction, and
Hgk sinh[k(z + h)]
w= sin(kx − ␻t) (3.50)
2␻ cosh(kh)
H␻ cosh[k(z + h)]
= sin(kx − ␻t)
2 sinh(kh)
in the vertical direction, where the last term in each equation results from eq. 3.31.
For standing waves, the expressions for the velocity components are obtained
from the combinations of eqs. 2.51 and 2.52 with eq. 3.39. The respective results are
H␻ cosh[k(z + h)]
U= sin(kx) sin(␻t) (3.51)
2 sinh(kh)
in the horizontal direction, and
H␻ sinh[k(z + h)]
W=− cos(kx) sin(␻t) (3.52)
sin(kh)
in the vertical direction.
In the remainder of the present chapter, we concentrate on the motions of par-
ticles within right-running traveling waves. The horizontal and vertical particle dis-
placements (␰, ␨ ) about a fixed mean point (xo , zo ) are found by replacing (x,z) by
(xo + ␰, zo + ␨ ), expanding the results in Maclaurin series in ␰ and ␨ and, finally,
integrating the respective velocity expressions over time. Referring to Figure 3.9,
the resulting displacement expressions are, to the first order,


H cosh[k(zo + h)]
␰= udt|xo,zo = − sin(kxo − ␻t) (3.53)
2 sinh(kh)
in the horizontal direction, and

H sinh[k(zo + h)]
␨ = wdt|xo,zo = cos(kxo − ␻t) (3.54)
2 sinh(kh)
60 Linear Surface Waves

z c

x
r
ξk

x o , zo
ξi
Figure 3.9. Position Vector for a Particle Beneath a Traveling Wave.

in the vertical direction. The subscript “o” identifies the mean position of the par-
ticle. The integration constants in eqs. 3.53 and 3.54 are neglected by adjustment of
the time origin. The location vector of the particle from the mean position is
r = ␰i + ␨ k (3.55)
H/2
= {− cosh[k(zo + h)] sin(kxo − ␻t)i + sinh[k(zo + h)] cos(kxo − ␻t)k}
sinh(kh)
Referring to the graphs of the hyperbolic functions in Figure 3.10, the expression in
eq. 3.55 results in the following case.

CASE 1. DEEP WATER (1/2 ≤ h/λ < ∞, or π ≤ kh < ∞) For this range, sinh(kh) ≈
cosh(kh) ≈ ekh /2 and tanh(kh) ≈ 1. With these approximations, the expression
in eq. 3.55 is approximately
H0 k0 zo
r=e [− sin(k0 xo − ␻t)i + cos(k0 xo − ␻t)k] (3.56)
2
Here, the subscript “0” identifies deep-water properties and, again, “o” iden-
tifies the mean position of the particle. Because the values of zo are negative
beneath the SWL, the expression for r describes circular particle paths hav-
ing diameters that decrease with depth. Plots of these paths are sketched in
Figure 3.11. The maximum diameter corresponds to a mean position on the
SWL, where zo = 0.

CASE 2. INTERMEDIATE WATER (1/20 < h/λ < 1/2, or 2π/20 < kh < π)
For this range
of kh, the equation for the location vector given in eq. 3.55 must be used as

Shallow Water Intermediate Water Deep Water

15

KEY
10
cosh(kh)
sinh(kh) Figure 3.10. Behaviors of Hyperbolic
tanh(kh)
Functions.
5

1
0
0 2π 1 2 3π 4
20 2π h
kh =
λ
3.6 The Wave Group 61

H0
x
zo

H0ek0 zo c
λ0
h 2 H c
x
H
h x
h

Figure 3.11 Figure 3.12 Figure 3.13

Figures 3.11, 3.12 and 3.13. Particle Paths Predicted by Airy’s Linear Wave Theory. The deep-
water paths in Figure 3.11 are circular with diameters that decrease exponentially with depth.
In the figure, zo ≤ 0. In Figure 3.12, the intermediate water paths are elliptical with the major
and minor axes decreasing with depth. The shallow-water paths in Figure 3.13 are elliptical
with minor axes that decrease with depth and vertically uniform major axes.

stated. Plots resulting from the equation are sketched in Figure 3.12. In that
figure, we see that the particle paths are ellipses with major and minor axes that
decrease with depth. At zo = −h, the minor axis is zero, and the particle motions
are adjacent to the bed.

CASE 3. SHALLOW WATER (0 < h/λ ≤ 1/20, or 0 < kh ≤ 2π/20) Again, referring to the
plots of the hyperbolic functions in Figure 3.10, the shallow-water range of kh
justifies the approximations sinh(kh) ≈ tanh(kh) ≈ kh and cosh(kh) ≈ 1. With
these approximations, eq. 3.55 simplifies to

H
r= {− sin(kxo − ␻t)i + [k(zo + h)] cos(kxo − ␻t)k} (3.57)
2kh
This equation also results in elliptic particle paths; however, in this case the
major axes are uniform and the minor axes decrease with depth. These elliptic
paths are sketched in Figure 3.13.
The particle paths described by eqs. 3.55 through 3.57 are good approxi-
mations of those actually observed when the wave steepness (H/␭) is small, or
when the wave profile is approximately sinusoidal. As the steepness increases,
or as the profile becomes nonlinear, there is a net convection of the particles in
the direction of wave travel, as is discussed in Chapter 4.

3.6 The Wave Group

It is demonstrated in Section 2.3D that various flow patterns in irrotational flows
can be superimposed upon each other to obtain other flow patterns. Because the
linear wave theory is based on the assumption of irrotational flow, we can superim-
pose regular wave patterns (those having specified heights and periods) to obtain
irregular wave patterns (those for which the heights and periods vary).
One such pattern is called the wave group, which is formed by superimposing
two waves having the same heights and slightly different periods. When many waves
of differing heights and periods are superimposed, the resulting pattern can appear
62 Linear Surface Waves

c cg

2H
x

Crest-Trough Envelope
Figure 3.14. Sketch of a Wave Group. In deep and intermediate waters, the waves form at the
left side of the group and travel to the right side, where they disappear because the celerity
(c) is greater than the group velocity (cg ). In shallow water, the wave profile is uniform over
the group length because the celerity and the group velocity are equal.

to be random in the time, as discussed in Chapter 5. In this section, our interest is in

the simple wave group.
Consider two waves of height H and periods that differ by T. The period dif-
ference results in small differences in both the period (␻) and the wave number
(k), as expected. The free-surface displacements of the two waves are described
by
H
␩1 = cos(kx − ␻t) (3.58)
2
and
H
␩2 =cos[(k + k)x − (␻ + ␻)t] (3.59)
2
The superposition of these waves results in an irregular wave pattern described by
 
k ␻
␩ = ␩1 + ␩2 = H cos x− t cos[(k + k)x − (␻ + ␻)t] (3.60)
2 2
where the expression is obtained with the help of trigonometric identities. Because it
is assumed that all of the incremental quantities are relatively small, that is, k  k
and ␻  ␻, the expression in eq. 3.60 can be approximated by
   
k ␻ k ␻
␩ ≈ H cos(kx − ␻t) cos x− t = 2␩1 cos x− t (3.61)
2 2 2 2
The wave pattern is called a wave group, and consists of a wave component, ␩1 , and
an overriding function that modifies the wave height. An example of a wave group is
sketched in Figure 3.14. The waves within the group are described by the expression
in eq. 3.24, and have a celerity or phase velocity (c) obtained from eq. 3.31.
Now imagine that we are riding with the group such that the angle of the over-
riding cosine is held constant. The speed at which we must travel to obtain this effect
is found by taking the time derivative of this constant angle to obtain
dx ␻ d␻
cg = = lim = (3.62)
dt k→0 k dk
√
This velocity is called the group velocity. From eq. 3.30, we find ␻ = [kg tanh(kh)].
The combination of this expression with that of eq. 3.62 yields
 
c 2kh
cg = 1+ (3.63)
2 sinh(2kh)
To gain an idea of the significance of the expression in eq. 3.63, consider the group
velocity at the extremes of the depth. In deep water, sinh(2kh) → 0.5e2kh , and the
3.6 The Wave Group 63

L ᐉ
c cg

Figure 3.15. Schematic Sketch of a Deep-

Water Wave Group for Example 3.6.
x
λ0

group velocity is approximately

c0
cg0 = lim (cg ) = (3.64a)
h→∞ 2
where, as before, the subscript “0” refers to the deep-water conditions. Physically,
the component waves of the group travel twice as fast as the group itself. Hence, in
deep water, waves appear at the back of the group, travel to the front of the group,
and then disappear.
In shallow water, sinh(2kh) → 2kh, and eq. 3.63 is approximately
cg = lim (cg ) = c (3.64b)
h→0

From this result, we see that the component waves of the group remain in the same
positions with respect to the front and back of the group in shallow water.
EXAMPLE 3.6: DEEP-WATER WAVE GROUP A 5-sec wave traveling in deep water is
found to have a wavelength of approximately 39.0 m, according to the expres-
sion in eq. 3.36. The celerity of the wave is c0 = ␭0 /T = 7.8 m/s and, from
eq. 3.64, the group velocity is cg0 = c0 /2 = 3.9 m/s. Consider a group composed
of ten of these deep-water waves. We wish to determine how far the wave group
travels when a component wave appears at the rear of the group and travels to
the front of the group. In other words, our interest is in how far the group trav-
els during the life of a component wave. Referring to the sketch in Figure 3.15,
the length of the group is L = 10␭0 = 390 m, whereas the length of group travel
is  = cg 0 t. The component wave must travel the sum of the group length and
the distance that the group travels, or L +  = c0 t. By eliminating the time t, the
length equations can be combined. The resulting value of the group travel dis-
tance is  = L = 390 m. That is, the group travels a distance equal to its length,
and the component waves travel twice the group length.
The engineering consequences of wave grouping are often debated. Most of
the attention devoted to the subject has been in the area of structural design in the
coastal zone. One of the advocates of including the effects of wave grouping is Per
Bruun (1985), who states that the stability of rubble mound structures is sensitive
to wave groups. However, J. W. van der Meer (1988) counters that wave groups
play only a minor role in the stability problem. More recently, Medina, Fassardi,
and Hudspeth (1990) present a comprehensive discussion on the wave group effects
on the stability of rubble mound breakwaters. The reader should bear in mind that
the wave height in the center of the group is twice that of the component waves,
according to eq. 3.61. This can cause a problem in shallow water where, as discussed
in Chapter 4, the wave breaks when the height is approximately 0.9 times the water
depth. For this reason, the center waves can be quite high in the coastal zone before
breaking. This is one argument for accounting for wave groupiness in the design of
coastal structures.
64 Linear Surface Waves

b
z y

x
η(x, t)
Figure 3.16. Notation for the Wave Energy and Energy
SWL Flux Analyses.
V
wj
h δz c
δx ui

3.7 Wave Energy and Power

All physical waves (water, acoustical, optical, etc.) represent energy in transition.
That is, energy is being transmitted by waves from one region to another. Free-
surface water waves are produced by the motions of bodies, such as ships, earth-
quakes, and winds, as discussed previously in this chapter. In this section, we derive
the expressions for both the average energy and the average energy flux or power
of water waves.
We begin our analysis by considering the potential energy of an elemental mass
of water displaced above the SWL, as sketched in Figure 3.16. The mass of that
element is

␦m = ␳ ␩(␦x)b

where b is the width of the wave crest under consideration, and the free-surface
displacement (␩) is positioned at the vertical centerline of the horizontally and ver-
tically symmetric element. The length of the element, ␦x, is assumed to be small
enough to allow us to both neglect the curvature of the free surface over this length
and to assume that the center of mass is a distance ␩/2 above the SWL. The elemen-
tal potential energy of this elemental mass is then

␩ 1
␦Ep = g(␦m) = ␳g␩2 (␦x)b
2 2
The expression for the total potential energy of the wave is found by combining this
expression with that of eq. 3.24 and integrating the resulting expression over one
wavelength. The resulting expression for the total potential energy is then

␭
␳g H 2 ␳g H 2 ␭b
Ep = cos2 (kx − ␻t)dxb = (3.65)
8 16
0

Next, we consider the kinetic energy of the submerged elemental mass of water
sketched in Figure 3.16. The elemental kinetic energy is

1
␦Ek = ␳ (u2 + w 2 )b␦x␦z
2
where, for a linear wave, the respective particle velocity components (u, w) are
obtained from eqs. 3.49 and 3.50. The combination of these equations with the ␦Ek
3.7 Wave Energy and Power 65

expression, and the subsequent integration over both the water depth and wave-
length, results in the total kinetic energy of the wave:

 0 ␭
1 ␳g H 2 ␭b
Ek = ␳ (u2 + w2 )dxdzb = (3.66)
2 16
−h 0

Comparing the results of eqs. 3.65 and 3.66, we see that the total energy of a linear
wave is equally divided between potential and kinetic energies. Hence, the total
energy of the wave is the sum of the expressions in eqs. 3.65 and 3.66:

␳g H 2 ␭b
E = Ep + Ek = (3.67)
8
From this result, we see that a doubling of the wave height (H) results in a four-fold
increase in the wave energy, whereas the energy is a linear function of the wave-
length. The depth effects on the wave energy are demonstrated in the following
example.

EXAMPLE 3.7: DEEP- AND SHALLOW-WATER WAVE ENERGY The combination of

wave-energy expression of eq. 3.67 and the deep-water wavelength expression
of eq. 3.36 results in

␳g 2 H 2 T 2 b
E0 = (3.68)
16␲
Similarly, the combination of the shallow-water wavelength expression of
eq. 3.38 and the energy expression of eq. 3.67 yields
√
␳g 3/2 H 2 hTb
E= (3.69)
8
Comparing the results of eqs. 3.68 and 3.69, we see that the wave energy
is proportional to the square of the wave period in deep water, but directly
proportional to the period in shallow water.
To gain an idea of the magnitudes of the deep- and shallow-water energies,
apply eqs. 3.68 and 3.69 to waves having a 1-m height and a 5-sec period. For
the deep-water wave, the total energy per unit meter of crest (E0 /b) is 49,300
N-m/m from eq. 3.68. For the shallow-water wave, applying eq. 3.69 to a wave
in 1 m of water yields an energy per unit meter of crest of 19,800 N-m/m. From
these results, we see that waves of equal height and period in deep and shallow
waters will have significantly different energies. This is true because the wave-
length decreases with depth, as illustrated in Example 3.4.

As stated previously, a wave represents energy in transition. Of interest then is

the rate of energy transmission in the direction of wave propagation, referred to
as the energy flux. To obtain a mathematical expression for the energy flux, begin
with the equation of energy conservation for an irrotational flow, Bernoulli’s equa-
tion (eq. 2.70). When this equation is applied to a surface wave, two simplifications
are made. First, as in the derivation of eq. 3.5, the time function f (t) in eq. 2.70
is zero. Second, as in the derivation of the linearized free-surface condition of eq. 3.6,
66 Linear Surface Waves

the nonlinear kinetic energy term (␳ V2 /2) is negligible for waves of very small steep-
ness, that is, H/␭  1. Bernoulli’s equation applied to linear waves is then
∂␾
␳ + pgz + p = 0 (3.70)
∂t
The first term represents both the unsteady kinetic energy per unit volume and the
dynamic pressure, with the latter interpretation of interest here. The second term
of eq. 3.70 represents both the potential energy per unit volume and the hydrostatic
pressure. The third term in that equation is the flow energy per unit volume or the
total pressure in the water. The only energy term in eq. 3.70 that is explicitly time-
dependent is the kinetic energy term. Hence, the energy flux, which is the time-rate
of change of energy per unit area normal to the flow direction, is simply the product
of the dynamic pressure and the fluid velocity:
∂␾ ∂␾
␳ V=␳ ∇␾ (3.71)
∂t ∂t
where the mathematical expression for the velocity potential, ␾, is given in eq. 3.29.
To obtain the total energy flux in the direction of wave travel, combine eqs. 3.71 and
3.29 and integrate the resulting expression over both the normal area (bh) and the
wave period (T). The resulting expression can actually be considered to be the wave
power in the direction of wave travel. Hence, we represent the resulting expression
by the wave-power expression
T 0  
1 ∂␾ ␳g H2 cb 2kh
P = ␳b ∇␾dzdt = +1 i (3.72a)
T ∂t 16 sinh(2kh)
0 −h

In this expression, we recognize the formula for the group velocity, cg , given in
eq. 3.63. Eq. 3.72 can then be rewritten in the more popular form,
␳g H 2 cg b
P= i (3.72b)
8
Physically, the wave power is convected in the direction of wave travel at the velocity
of the wave group.
To obtain the deep-water approximation for the wave-power expression, com-
bine eqs. 3.72b and 3.64 to obtain
␳g H 2 c0 b ␳g 2 H 2 Tb
P= i= i (3.73)
16 32␲
Comparing this expression with the deep-water energy expression of eq. 3.68, we
see that the energy is proportional to the square of the period, whereas the power is
directly proportional to the period. In shallow water, the results in eqs. 3.72b, 3.64,
and 3.38 are combined to obtain
√
␳g H3/2 H2 hb
P= i (3.74)
8
In shallow water, the wave power is independent of the period.
EXAMPLE 3.8: DEEP- AND SHALLOW-WATER WAVE POWER To obtain an idea of
the magnitude of the wave power, consider the 1-m, 5-sec wave of Example 3.7
in deep water. The value of the power per unit meter of crest obtained from
eq. 3.73 is P/b = 4,930 watts/meter (W/m) or 4.93 kilowatts/meter (kW/m).
3.7 Wave Energy and Power 67

If the wave height increases to 2 m while the period is kept at 5 sec, then the
power per unit length increases to 19.7 kW/m. That is, by doubling the wave
height and maintaining the period, the power is increased by a factor of 4. When
the shallow-water power expression of eq. 3.74 is applied to the 1-m, 5-sec wave
in 1 m of water, as in Example 3.7, the resulting power per unit crest width is
3.96 kW/m, which is approximately 80% of the deep-water power associated
with the same wave height and period.

EXAMPLE 3.9: WAVE POWER CONVERSION In Example 2.9, the scaling laws are
applied to a pneumatic wave-energy conversion system. The system is sketched
in Figure 2.24, and its operation is described in that example. Now, assume that
the prototype system of Example 2.9 operates in 2 m of water (h = 2.0 m)
where the wave height (H) is 1.0 m and the wave period (T) is 6 sec. From
the example, the prototype power (Pp ) is 128 kW. McCormick (2007) and oth-
ers state that the maximum efficiency (⑀) for the system sketched in Figure 2.24
is 50%; that is, a maximum of 50% of the incident wave energy or power can be
converted to the energy or power of the oscillating water column. For this rea-
son, 256 kW of incident wave power is required to obtain the desired prototype
power under ideal conditions. Note that the efficiency applies to the hydraulics
and pneumatics of the system and not to the turbine, generator, and other power
take-off components. Our goal is to determine the width (bwc ) of the water col-
umn needed to produce the 128 kW of usable power, assuming that there are
no point-absorbing effects, as described in Chapter 10.
First, the wave properties in 2 m of water must be determined for the 6-
sec wave. The transcendental wavelength expression of eq. 3.31 yields a value
of 25.6 m for ␭. The approximate wavelength expression of eq. 3.38 yields a
value of 26.6 m, a difference of less than 4%. Hence, we assume that the wave
conditions are those corresponding to shallow water so that the wave-power
expression of eq. 3.74 can be used. The equation that yields the water column
width (bwc ) is then
√
␳g 3/2 H 2 hbwc
Pp = 128,000 Watts = ⑀ P = ⑀ (3.75)
8

where ⑀ = 0.5, ␳ = 1,030 kg/m3 , g = 9.81 m/s2 , H = 1 m, and h = 2 m. The width

value is then 45.8 m.
The quarter-scale model of the prototype system yields a power of 1 kW
from Example 2.9. The width of the model water column is 11.4 m. This model
operates in a tank having a 0.5-m depth, where the respective wave height and
period are 0.25 m and 3.0 sec.

The area of wave-energy conversion is receiving considerable attention as waves

are a viable alternative energy resource for both island and isolated coastal commu-
nities. The reason for this interest can be understood by considering the follow-
ing: The average power required by a citizen of the contiguous United States is
1 kW. The island communities that are candidates for wave-energy conversion have
a requirement of about 100 W per person. The prototype of Examples 2.9 and 3.9
would then provide the average power for 1,280 island citizens. This converted wave
power could be used directly in the production of either electricity or potable water.