Kelvin-Helmholtz Instability

Nicolas Zapata
02/12/2016

Abstract

In the following project, a kind of system with a Kelvin-Helmholtz Instability is studied. In

particular, it is analyzed the evolution of an interface of two parallel and incompressible ows,
due to small perturbations. There were no consideration of the eects of viscosity and surface
tension, as well as it was kept all theoretical approximation to rst order. The obtained results
were later compared with previous numerical researches and some empirical observations. Finally,
it is concluded that although this case is not completely realistic, it serves for a good intuitive
understanding and as an introduction for more complex examples of the same instability.
1. Introduction

The study of ows instabilities are an important topic in uid mechanics. They appear because of
disturbances of the physical parameters of a specic system. Depending on the magnitude of such
disturbance, the system could archive a new stable state, or a turbulent (also known as chaotic)
regime [1]. The present report will focus only on the implications of small disturbances in a stationary
system and how it evolves to its new equilibrium state. In particular, an instability produced because
of perturbations by a shear ow in an interface of two dierent uids, will be analyzed. This is called
Kelvin-Helmholtz Instability. Its study is justied because the important role it has in the generation
of some wave-like patterns seen in nature [2]. An example of these phenomenon is shown in Figure 1.

Figure 1: Example of Kelvin-Hemlholtz Instability [3]

2. Scales and Restrictions

In this report, only a simple case of Kelvin-Helmholtz Instability will be studied. However, it will be
sucient in order to have an intuitive idea of how more complex examples of this phenomenon works.
We will analyze the time evolution in the interface of a system of two horizontal parallel, miscible
and incompressible dierent streams with uniform velocities. Likewise, the uids are considered in
a gravitational equilibrium i.e. the heaviest uid is below the lightest one (2 > 1 ). Therefore, no
instabilities due to buoyant forces or surface tensions are taken. Furthermore, the eect of viscosity is
completely neglected too (high Reynold numbers). Thus, Naiver-Stokes equation can be simplies to
a simple form of Euler's equation. In addition, no discontinuities of the pressure in the interface are
also considered. A qualitative description of these restrictions can be observed in Figure 2.

Figure 2: Considered System

In addition, based on experimental data, a consideration of some restrictions on the scales of the
system has to be made in order to obtain more emprically testable results. Taking as reference the
system of coordinates shown in Figure 2, the vertical and horizontal scales must be large enought
to neglect any boundary conditions due to obstructions and gradients of physical quatities. This is
considered because otherwise turbulent eects must be taken into account [1].
On the other hand, the report will only deal with instabilities of small perturbations. Hence, any
approximation will be kept rst order (linear).
Finally, we are taking all the mathematical formulation to be two-dimensional. This assumption
can be considered thanks to the result of Squire's Theorem. It states that the instabilities in twodimensional systems are larger than its equivalent in three-dimensions. Thus, there will be not losses
of generality in the results derived using a 2D approximation [1].
A more realistic analysis involving gradients of velocities and densities is not in the scope of this
project because of its mathematical complexity. In fact, this kind of instabilities rarely have an
analytical solution. Their complete study is purely numerical [4].
3. Analytical Study and Derivations

First of all, there is assumed that an instability y = (x, t) occurs in the interface at the time t = 0.
However, it is also considered that both ows far enough from the interface remain uniform and
unperturbed.
Now, since the velocities of both uids before the perturbation are uniform, it can be observed
that they are irrotational. However, this result together with the assumption that the disturbance is
small, allows applying Kelvin's Circulation Theorem for incompressible uids in order to determine
that both ows must remain irrotational. This theorem states that the circulation around a closed
curve moving together with an incompressible ow, which is subjected only to conservative forces,
must remain constant [4]. Thus, two scalar potential functions 1 and 2 can be dened for each uid,

such that
vi = i and 2 i = 0, for i = 1, 2. Their mathematical expression is shown in equation
(1).
i (x, y, t) = Ui x + i (x, y, t); i = 1, 2

(1)

The functions i are taken such that they satisfy Laplace's equation:2 i = 0. Also, since the
ows are taken to be uniform far from the surface we can take the following conditions:
1 0; when y

(2)

2 0; when y

(3)

Recall that Euler's equation for an irrotational ow and incompressible uid can be written as
expression (4).

+
t

v2
2

+ (gy) = 0

However, we are considering that

vi = i and assuming also
sions (5) and (6) can be derived.
)2
1
1 (

+
1 +
t
2
)2
2
1 (

+
2 +
t
2

t ( i )

(4)

i
t

, then expres-

P1
+ gy = C1
1

(5)

P2
+ gy = C2
2

(6)

Where C1 and C2 are a constants.

Since it has been considered that the perturbation take place at t = 0, we can take the condition
(7).
i (x, y, t = 0) = 0; i = 1, 2

(7)

Now, initially the interface is located at y = 0. Hence, evaluating equations (5) and (6) with t = 0,
y = 0 and using expression (7), the following results can be obtained:
1 2 P1
U +
= C1
2 1
1

(8)

1 2 P2
U +
= C2
2 2
2

(9)

On the other hand, since the pressure must be continuous at any time in the interface, it can be
taken that:
P1 = P2 ; at the interface

(10)

Hence, equating the values of P1 and P2 in expressions (8) and (9), it is obtained equation (11).
1
1
1 ( U12 C1 ) = 2 ( U22 C2 )
2
2

(11)

Observe now that the velocity in each ow will be given by the expression (12).

vi = i = (Ui + ui ) x
+ wi y; i = 1, 2

(12)

i
i
and wi = y
.
Where ui = x
But then, since we are taking only rst order approximations, equation (13) is obtained.
2

vi2 = (Ui + ui ) + wi2 Ui2 + 2Ui

i
; i = 1, 2
x

(13)

If this result is subtitute in equations (5) and (6), evaluated at y = (the interface), expressions
(14) and (15) can be derived.
1
1
1
P1
+ U22 + U1
+
+ g = C1
t
2
x
1

(14)

2
1
i
P2
+ U22 + U2
+
+ g = C2
t
2
x
2

(15)

After aplying condition (10) and using the equation (11), it can be obtained the expression (16).
1 (

1
2
1
2
+ U1
+ g) = 2 (
+ U2
+ g)
t
x
t
x

(16)

Consider now the function Y = y (x, t) = 0. Then, it can be assured that DY

Dt = 0. However,
using the denition of the substantial derivative for each uid, equation (17) can be derived.
0=

Ui

; i = 1, 2
y
t
x
x x

(17)

i
However, neglecting the quadratic term x
x , the conditions (18) and (19) can be obtained.
These relations determine that any uid particle at the interface has a time evolution equal to the one
produced in the location.

1
=
+ U1
y
t
x

(18)

=
+ U2
y
t
x

(19)

It is important at this point to recall what is the status of the problem. It is still necesarry to
obtain an analytical solution for the functions 1 , 2 and , restricted to the conditions (2), (3), (16),
(18) and (19). Since a wave-like non dissipative behavoir is spected, a solution of the form of equation
(20) is assumed. Note that it has the form of a normal modes solution with normal frecuency kc.

1
1
2 = 2 eik(xct)

(20)

Where k is an arbitrary positive real number,c is a complex number that needs to de determined
by the restrictions of the system and is a constant.
However, since the solutions of the functionsi must satisfy Laplace's equation together with restrictions (2) and (3), it is assumed that they take the from of relations (21) and (22).
1 = A1 eky eik(xct)

(21)

2 = A2 eky eik(xct)

(22)

Where A1 and A2 are constants.

Replacing the previous results into the restrictions (16), (18) and (19) respectively, the following
equations are obtained:
[
]
[
]
1 g + ikA1 eky (U1 c) = 2 g + ikA2 eky (U2 c)

(23)

U1 )
A1 eky = i(c

(24)

 2 c)
A2 eky = i(U

(25)

However, solving the previous equations for the value of c, expression (26) is obtained.
[
(
)2 ] 12
2 U2 + 1 U1
g 2 1
U2 U1
c=

2 1
2 + 1
k 2 + 1
2 + 1

(26)

4. Physical Interpretation and Empirical Applications

The result obtained in relation (26) allows having both steady and damping behavior of the instability
on the interface. The second one is obtained if the following condition is fullled:
[
(
)2 ] 12
(
)
2 U2 + 1 U1
g 2 1
U2 U1
2
+
2 1
c=
; g 22 21 < 2 1 k (U2 U1 )
2 + 1
k 2 + 1
2 + 1

(27)

The condition above determine a possible mixing of the uids after a large enough time. On the
other hand, if restriction (27) is not satised, a wave pattern, as the one shown in Figure 3, will appear
in the interface.

Figure 3: Wave-Like Behaviour of the Solution

Clearly, a more general solution will be a linear combination of both kind of behavior. Of course,
this result will be in more agreement with the empirical observations. This is because this kind of
instabilities, although in some cases obey the same geometrical pattern, it have been seen that in
nature they cannot be maintained for large periods of time [1].
In addition,it is important to mention previous numerical studies of Kelvin-Helmholtz instability
that took into account higher orders approximations. Figure 4 shows an evolution of an interface,
based on numerical results for second order approximations of the relations seen in section 3 [4].

Figure 4: Previous Numerical Results with Second Order Approximations [4]

5

As seen in the previous image, for short a period the system behaves in the same way that we
described in Figure 3. However, the nonlinear regime became more important as time passes. The
proximities of the interface tends to a vortex-type boundary layer, for large values of time.
Finally, Figures 1 (Introduction) and 5 show two real life application of this phenomenon. The rst
one is a Geophysical example called Kelvin-Helmholtz Cloud and the second related to the instabilities
produced in Jupiter's atmosphere

Figure 5: Kelvin-Hemlholtz Instability in Jupiter's Atmosphere [3]

.
The previous examples together with several oceanographic analysis of water-air waves generated
by this kind of instability [4], allows to validate, al least at an intuitive level, the results obtained in
this research. However, for a more realistic description of these phenomenons, it is necessary to include
the eects of viscosity, surface tension and turbulent ows in the system studied.
5. Conclusions

The analytical study made in this proyect gave a complete description of a simple case of the KelvinHelmholtz Instability. A validation of the theoretical results was archived after a comparation with
some previous studies and empirical observations. Although, this was accomplished only on an intuitive
level. A more detailed description of this phenomenon requiere some physical considerations that goes
beyond the scope of this proyect. This is because no analytical solution can be derived for this kind of
systems and complex numerical calculations are needed. However, the work done in this reseach allows
the reader to understand such cases and serve as an introduction of the broader topic of instabilities.
References

[1] Gramer, L. (2007).

. USA: NOAA.

Kelvin-Helmholtz Instabilities

[2] Cushman-Roisin, B. & Beckers, J.M. (2008).

Academic Press.
[3] Marra, V. (2014). Simulating
trieved from www.comsol.com
[4] Kundu, P. & Cohen, I. (2002).

. USA:

Introduction to Geophysical Fluid Dynamics

. Comsol. Re-

Kelvin-Helmholtz Instability and Climate Dynamics

. USA: Elsevier Science.

Fluid Mechanics