_

1.138J/2.062J/18.376J, WAVE PROPAGATION

Fall, 2004 MIT
Notes by C. C. Mei
CHAPTER FOUR. WAVES IN WATER
1 Governing equations for waves on the sea surface
In this chapter we shall model the water as an inviscid and incompressible uid, and
consider waves of innitesimal amplitude so that the linearized approximation suces.
Recall in the rst chapter that when compressibility is included the velocity potential
dened by u = is governed by the wave equation:
1
2

2
= (1.1)
c
2
t
2
where c = dp/d is the speed of sound. Consider the ratio
1
2

c
2
t
2

2
/k
2

2

2
c

As will be shown later, the phase speed of the fastest wave is /k = gh where g is the
gravitational acceleration and h the sea depth. Now h is at most 4000 m in the ocean,
and the sound speed in water is c = 1400 m/sec
2
, so that the ratio above is at most
40000 1
= 1
1400
2
49
We therefore approximate (1.1) by

2
= 0 (1.2)
Let the free surface be z = (x, y, t). Then for a gently sloping free surface the
vertical velocity of the uid on the free surface must be equal to the vertical velocity of
the surface itself. i.e.,

= , z = 0. (1.3)
t z
Having to do with the velocity only, this is called the kinematic condition.
1
For small amplitude motion, the linearized momentum equation reads
u
= P ge
z
(1.4)
t
Now let the total pressure be split into static and dynamic parts
P = p
o
+ p (1.5)
where p
o
is the static pressure
p
o
= gz (1.6)
which satises
0 = p
o
+ ge
z
(1.7)
It follows that
u

t
=
t
= p (1.8)
so that

p =
t
(1.9)
which relates the dynamic pressure to the velocity potential.
Let us assume that the air above the sea surface is essentially stagnant. Because of
its very small density we ignore the dynamic eect of air and assume the air pressure to
be constant, which can be taken to be zero without loss of generality. If surface tension
is ignored, continuity of pressure requires that
p = p
o
+ p = 0, z = .
to the leading order of approximation, we have, therefore

g + = 0, z = 0. (1.10)
t
Being a statement on forces, this is called the dynamic boundary condition. The two
conditions (1.3) and (1.10) can be combined to give

2

+ g = 0, z = 0 (1.11)
t
2
z
2


_ _ _ _
_ _
_ _
If surface tension is also included then we adopt the model where there is a thin
lm covering the water surface with tension T per unit length. Consider a horizontal
rectangle dxdy on the free surface. The net vertical force from four sides is
_ _

2

2

T T dy +
_
T T
_
dx = T + dx dy
x

x
2
y
2
x

x+dx
x

x
y

y+dy y
Continuity of vertical force on an unit area of the surface requires
p

2

2

o
+ p + T + = 0.
x
2
y
2
Hence

2

2

g + T + = 0, z = 0. (1.12)
t x
2
y
2
which can be combined with the kinematic condition (1.3) to give

2
T
+ g
2
= 0, z = 0 (1.13)
t
2
z z
When viscosity is neglected, the normal uid velocity vanishes on the rigid seabed,
n = 0 (1.14)
Let the sea bed be z = h(x, y) then the unit normal is
(h
x
, .h
y
, 1)
n = _ (1.15)
1 + h
2
+ h
2
x y
Hence
h h
= , z = h(x, y) (1.16)
z x x y y
2 Progressive waves on a sea of constant depth
2.1 The velocity potential
Consider the simplest case of constant depth and sinusoidal waves with innitively long
crests parallel to the y axis. The motion is in the vertical plane (x, z). Let us seek a
solution representing a wavetrain advancing along the x direction with frequency and
wave number k,
= f(z)e
ikxit
(2.1)
3

_ _


_ _
In order to satisfy (1.2), (1.13) and (1.16) we need
f + k
2
f = 0, h < z < 0 (2.2)
T

2
f + gf

+ k
2
f

= 0, z = 0, (2.3)
f


= 0, z = h (2.4)
Clearly solution to (2.2) and (2.4) is
f(z) = B cosh k(z + h)
implying
= B cosh k(z + h)e
ikxit
(2.5)
In order to satisfy (2.3) we require
T

2
= gk + k
3
tanh kh (2.6)

which is the dispersion relation between and k. From (1.3) we get


= = (Bk sinh kh)e
ikxit
(2.7)
t z

z=0
= Ae
Upon integration,
ikxit
Bk sinh kh
ikxit
(2.8) = e
i
where A denotes the surface wave amplitude, it follows that
iA
B =
k sinh kh
and
iA
= cosh k(z + h)e
ikxit
k sinh kh
=
igA
1 +
Tk
2
cosh k(z + h)
e
ikxit
(2.9)
g cosh kh
4
_

_
_ _
_ _


_ _


_ _

2.2 The dispersion relation
Let us rst examine the dispersion relation (2.6), where three lengths are present : the
depth h, the wavelength = 2/k, and the length
m
= 2/k
m
with
g 2 T
k
m
= ,
m
= = 2 (2.10)
T k
m
g
For reference we note that on the air-water interface, T/ = 74 cm
3
/s
2
, g = 980 cm/s
2
,
so that
m
= 1.73cm. The depth of oceanographic interest ranges from O(10cm) to
thousand of meters. The wavelength ranges from a few centimeters to hundreds of
meters.
Let us introduce

2
= 2gk
m
= 2g
g
(2.11)
m
T
then (2.6) is normalized to

2
1 k k
2
= 1 + tanh kh (2.12)

2
2 k
m
k
2
m m
Consider rst waves of length of the order of
m
. For depths of oceanographic
interest, h , or kh 1, tanh kh 1. Hence

2
1 k k
2
= 1 + (2.13)

2
2 k
m
k
2
m m
or, in dimensional form,
Tk
3

2
= gk + (2.14)

The phase velocity is


g Tk
2
c = =
_
1 + (2.15)
k k g
Dening

m
k
c
m
= (2.16)
m
the preceding equation takes the normalized form
c

1 k
m
k
=
_
+ (2.17)
c
m
2 k k
m
Clearly
Tk
c , if k/k
m
1, or /
m
1 (2.18)

5
_
_
_
c
*
=c/ c
m
0
1
2

*
=/
m
0.25
0.5
0.75
1.25
1.5
1.75
1 2 3 4 5
Figure 1: Phase speed of capillary-gravity waves in water much deeper than
m
.
Thus for wavelengths much shorter than 1.7 cm, capillarity alone is important, These
are called the capillary waves. On the other hand
g
c , if k/k
m
1, or /
m
1 (2.19)
k
Thus for wavelength much longer than 1.73 cm, gravity alone is important; these are
called the gravity waves. Since in both limits, c becomes large, there must be a minimum
for some intermediate k. From
dc
2
g T
= + = 0
dk k
2

the minimum c occurs when
g
k = = k
m
, or =
m
(2.20)
T
The smallest value of c is c
m
. For the intermediate range where both capillarity and
gravity are of comparable importance; the dispersion relation is plotted in gure (1).
Next we consider longer gravity waves where the depth eects are essential.
= gk tanh kh (2.21)
6
_
_
_ _

c/

gh
0
1
0.25
0.5
0.75
kh
2 4 6 8 10
Figure 2: Phase speed of capillary-gravity waves in water of constant depth
For gravity waves on deep water, kh 1, tanh kh 1. Hence
g
gk, c (2.22)
k
Thus longer waves travel faster. These are also called short gravity waves. If however
the waves are very long or the depth very small so that kh 1, then tanh kh kh and
k gh, c gh (2.23)
Form intermediate values of kh, the phase speed decreases monotonically with increasing
kh. All long waves with kh 1 travel at the same maximum speed limited by the depth,

gh, hence there are non-dispersive. The dispersion relation is plotted in gure (??).
2.3 The ow eld
For arbitrary k/k
m
and kh, the velocities and dynamic pressure are easily found from
the potential (2.9) as follows
u =

x
=
gkA

_
1 +
Tk
2
g
_
cosh k(z + h)
cosh kh
e
ikxit
(2.24)
_ _
w =

z
=
igkA

1 +
Tk
2
g
sinh k(z + h)
cosh kh
e
ikxit
(2.25)
_ _
p =

t
= gA 1 +
Tk
2
g
cosh k(z + h)
cosh kh
e
ikxit
(2.26)
7
Note that all these quantities decay monotonically in depth.
In deep water, kh 1,
u =
gkA

_
1 +
Tk
2
g
_
e
kz
e
ikxit
(2.27)
_ _
w =

z
=
igkA

1 +
Tk
2
g
e
kz
e
ikxit
(2.28)
_ _
p =

t
= gA 1 +
Tk
2
g
e
kz
e
ikxit
(2.29)
All dynamical quantities diminish exponentially to zero as kz . Thus the uid
motion is limited to the surface layer of depth O(). Gravity and capillary-gravity waves
are therefore surface waves.
For pure gravity waves in shallow water, T = 0 and kh 1, we get
gkA
ikxit
u = e (2.30)

w = 0, (2.31)

p = = gAe
ikxit
= g (2.32)
t
Note that the horizontal velocity is uniform in depth while the vertical velocity is neg-
ligible. Thus the uid motion is essentially horizontal. The total pressure
P = p
o
+ p = g( z) (2.33)
is hydrostatic and increases linearly with depth from the free surface.
2.4 The particle orbit
In uid mechanics there are two ways of describing uid motion. In the Lagrangian
scheme, one follows the trajectory x, z of all uid particles as functions of time. Each
uid particle is identied by its static or initial position x
o
, z
o
. Therefore the instan-
taneous position at time t depends parametrically on x
o
, z
o
. In the Eulerian scheme,
the uid motion at any instant t is described by the velocity eld at all xed positions
x, z. As the uid moves, the point x, z is occupied by dierent uid particles at dierent
times. At a particular time t, a uid particle originally at (x
o
, z
o
) arrives at x, z, hence
its particle velocity must coincide with the uid velocity there,
dx dz
= u(x, z, t), = w(x, z, t) (2.34)
dt dt
8
_ _
_ _
_ _
Once u, w are known for all x, z, t, we can in principle integrate the above equations to
get the particle trajectory. This Euler-Lagrange problem is in general very dicult.
In small amplitude waves, the uid particle oscillates about its mean or initial posi-
tion by a small distance. Integration of (2.34) is relatively easy. Let
x(x
o
, z
o
, t) = x
o
+ x

(x
o
, z
o
, t), andz(x
o
, z
o
, t) = z
o
+ x

(x
o
, z
o
, t) (2.35)
then x

x, z

z in general. Equation (2.34) can be approximated by
dx

dz

= u(x
o
, z
o
, t), = w(x
o
, z
o
, t) (2.36)
dt dt
From (2.24) and (2.25), we get by integration,
x

=
gkA
i
2
1 +
Tk
2
g
cosh k(z
o
+ h)
cosh kh
e
ikxoit
=
gkA

2
_
1 +
Tk
2
g
_
cosh k(z
o
+ h)
cosh kh
sin(kx
o
t) (2.37)
(2.38)
_ _
z

=
gkA

2
1 +
Tk
2
g
sinh k(z
o
+ h)
cosh kh
e
ikxoit
_ _
=
gkA

2
1 +
Tk
2
g
sinh k(z
o
+ h)
cosh kh
cos(kx
o
t) (2.39)
(2.40)
Letting
_ _ _ _
a cosh k(z
o
+ h)
_ _ _
gkA Tk
2
_
_ _ = 1 + _ _ (2.41)
b

2
cosh kh g
sinh k(z
o
+ h)
we get
2 2
x z
+ = 1 (2.42)
a
2
b
2
The particle trajectory at any depth is an ellipse. Both horizontal (major) and vertical
(minor) axes of the ellipse decrease monotonically in depth. The minor axis diminishes
to zero at the seabed, hence the ellipse collapses to a horizontal line segment. In deep
water, the major and minor axes are equal
gkA Tk
2
a = b =

2
1 +
g
e
kzo
, (2.43)
therefore the orbits are circles with the radius diminishing exponentially with depth.
9

_ _

_ _

_ _
_ _
_ _ _ _
_ _ _ _
_ _
_ _ _
Also we can rewrite the trajectory as
gkA Tk
2
cosh k(z
o
+ h)
x = 1 + sin(t kx
o
) (2.44)

2
g cosh kh
gkA Tk
2
sinh k(z
o
+ h)
z = 1 + sin(t kx
o
) (2.45)

2
g cosh kh 2
When t kx
o
= 0, x

= 0 and z = b. A quarter period later, t k
o
= /2, x

= a
and z

= 0. Hence as time passes, the particle traces the elliptical orbit in the clockwise
direction.
2.5 Energy and Energy transport
Beneath a unit length of the free surface, the time-averaged kinetic energy density is
_
0

E
k
= dz u
2
+ w
2
(2.46)
2 h
whereas the instantaneous potential energy density is
_
1
E
p
=
1
g
2
+ T
(ds dx) 1
= g
2
+ T 1 +
2
1 = g
2
+ T
2
(2.47)
x
2 dx 2
x
2
Hence the time-average is
1 T

E
p
= g
2
+
2
(2.48)
2
x
2
Let us rewrite (2.24) and (2.25) in (2.48):
gkA
it
u = ' 1 +
Tk
2
cosh k(z + h)
e
ikx
e (2.49)
g cosh kh
igkA
it
w = ' 1 +
Tk
2
sinh k(z + h)
e
ikx
e (2.50)
g cosh kh
Then
_ _
2
_
_
0

=
gkA
1 +
Tk
2
_
2
1
dz cosh
2
k(z + h) + sinh
2
k(z + h) E
k
4 g cosh
2
kh h
_ _
2
_ _ _
2
_
gkA Tk
2
_
2
sinh 2kh gkA Tk
2
_
2
sinh kh
= 1 + = 1 +
4 g 2k cosh
2
kh 4 g k cosh kh
gA
2
Tk
2
_
2
gk tanh kh gA
2
Tk
2
= 1 + = 1 + (2.51)
4 g
2
4 g
10
_ _
_ _ _ _ _ _
_ _
_
_
_
_
after using the dispersion relation. On the other hand,

gA
2
Tk
2
E
p
= 1 + (2.52)
4 g
Hence the total energy density is
Tk
2
gA
2
k
2
gA
2

2

gA
2
E = E
k
+ E
p
= 1 + = 1 + = 1 +
m
(2.53)
2 g 2 k
2
2
2
m
Note that the total energy is equally divided between kinetic and potential energies; this
is called the equipartition of energy.
We leave it as an exercise to show that the power ux (rate of energy ux) across a
station x is
_
0
_
0
dE

= pu dz T
x

t
=
t

x
dz T
x

t
= Ec
g
(2.54)
dt h h
where c
g
is the speed of energy transport , or the group velocity

2
_ _ _ _
+ 3
2kh
2
k
m
2
k
+ 3 d 2kh
_
c c
_
_
_
_
_
2

m
(2.55) + + c = = =
g 2
k
m
2
k

2
dk 2 sinh 2kh
_
2
+ 1
sinh 2kh
+ 1
_
2

m
For pure gravity waves, k/k
m
1 so that
c 2kh
c
g
= 1 + (2.56)
2 sinh 2kh
where the phase velocity is
g
c = tanh kh (2.57)
k
In very deep water kh 1, we have
c 1 g
c
g
= = (2.58)
2 2 k
The shorter the waves the smaller the phase and group velocities. In shallow water
kh 1,
c
g
= c = gh (2.59)
Long waves are the fastest and no longer dispersive.
For capillary-gravity waves with kh 1, we have

2
_ _ _ _
+ 3
2
k
m
2
k
+ 3 2 g c
_
_
=
c
2
_
_
_
_
_
, k
m
2

m
(2.60) c = =
g 2
k
m
2
k

2
2
m
T
+ 1 + 1
_
2

m
11

_ _
where
g Tk
3
c = + (2.61)
k
Note that c
g
= c when k = k
m
, and
> >
c
g
< c, if k < k
m
(2.62)
c
In the limit of pure capillary waves of k k
m
, c
g
= 3c/2. For pure gravity waves
g
= c/2 as in (2.58).
3 Wave resistance of a two-dimensional obstacle
Ref: Lecture notes on Surface Wave Hydrodynamics Theodore T.Y. WU, Calif. Inst.Tech.
As an application of the information gathered so far, let us examine the wave resis-
tance on a two dimensional body steadily advancing parallel to the free surface. Let the
body speed be U from right to left and the sea depth be constant.
Due to two-dimensionality, waves generated must have crests parallel to the axis of
the body (y axis). After the steady state is reached, waves that keep up with the ship
must have the phase velocity equal to the body speed. In the coordinate system xed
on the body, the waves are stationary. Consider rst capillary -gravity waves in deep
water

= /
m
= O(1) and kh 1. Equating U = c we get from the normalized
dispersion relation
1 1
2
= c =

+ (3.1) U

2

2

where U

U/c
m
. Hence
2

2
2c

+ 1 = 0 = (


1
)(


2
)

which can be solved to give
_ _
_ _
1/2

1
2 4

_ _ = c

c

1 (3.2)
2
and
1

1
= (3.3)

2
12
_ _
_
_ _ _ _
Thus, as long as c

= U

> 1 two wave trains are present: the longer gravity wave
with length
1
, and the shorter capillary wave with length
2
. Since c < c = U and
c
g
1
g
2
> c = U , and energy must be sent from the body, the longer gravity waves must
follow, while the shorter capillary waves stay ahead of, the body.
Balancing the power supply by the body and the power ux in both wave trains, we
get

Rc = (c c
g
1
)E
1
+ (c c)E
2
(3.4)
g
2
c
Recalling that
g
1
2

+ 3
=
c 2
2

+ 1
we nd,
c
g
1 2
=
1 1/

1 1/

1 = 1 1 + =
2c
2
c 2
2

+ 1 2

+ 1/

2
For the longer wave we replace c
g
/c by c
g
1
/c

and

by
1
in the preceding
equation, and use (3.2), yielding
c
4
1
g
1
= 1 c

_
1/2
(3.5)
c

Similarly we can show that
_
c c
g
2 4
1 = 1 c

_
1/2
= 1
g
1
(3.6)
c

c

Since

gA
2
1 gA
2
1 1
1 2
E
1
=
1
1 +
2
=
1
+ = gA
2
1

2
c

, (3.7)
2
1

1
2

1
we get nally
_ _
1
_ _
R =
1
g
2
A
2 4
1
+
1
A
2
(U
4
1)
1/2
(3.8)
1
+
1
A
2
(c

1)
1/2
= g
2
A
2
2
2
2
2
Note that when U

= 1, the two waves become the same; no power input from the body
is needed to maintain the single innite train of waves; the wave resistance vanishes.
When U

< 1, no waves are generated; the disturbance is purely local and there is
also no wave resistance. To get the magnitude of R one must solve the boundary value
problem for the wave amplitudes A
1
, A
2
which are aected by the size (relative to the
wavelengths), shape and depth of submergence.
13
_ _
_ _
Figure 3: Dependence of wave resistance on speed for pure gravity waves
When the speed is suciently high, pure gravity waves are generated behind the
body. Power balance then requires that
c
g

gA
2
1 kh
R = 1 E = (3.9)
U 2 2 sinh 2kh
The wavelength generated by the moving body is given implicitly by
_ _
1/2
U tanh kh
= (3.10)
gh kh

When U gh the waves generated are very long, kh 1, c
g
c = gh, and the

wave resistance drops to zero. When U gh, the waves are very short, kh 1,
gA
2
R (3.11)
4
For intermediate speeds the dependence of wave resistance on speed is plotted in gure
(3).
14
_ _

_ _ _ _


4 Narrow-banded dispersive waves in general
In this section let us discuss the superposition of progressive sinusoidal waves with the
amplitudes spread over a narrow spectrum of wave numbers
_

_

(x, t) = [/(k)[ cos(kx t
A
)dk = ' /(k)e
ikxit
dk (4.1)
0 0

where/(k) is complex denotes the dimensionless amplitude spectrum of dimension

(length)
2
. The component waves are dispersive with a general nonlinear relation (k).
Let /(k) be dierent from zero only within a narrow band of wave numbers centered at
k
o
. Thus the integrand is of signicance only in a small neighborhood of k
o
. We then
approximate the integral by expanding for small k = k k
o
and denote
o
= (k
o
),

=

(k
o
), and

=

(k
o
),
o o
_ _

_
= ' e
ikoxiot
/(k) e
ikxi(o)t
dk
0
_ _

_ _ _ __
1
= ' e
ikoxiot

dk /(k) exp ikx i

k +
o
(k)
2
t +
o
0 2
= ' A(x, t)e
ikoxiot
(4.2)
where
_

A(x, t) = dk /(k) exp ikx i

k +
1
o
(k)
2
t + (4.3)
o
0 2
Although the integration is formally extends from 0 to , the eective range is only
from k
o
(k)
m
to k
o
+ (k)
m
, i,.e., the total range is O((k)
m
), where (k)
m
is the
bandwidth. Thus the total wave is almost a sinusoidal wavetrain with frequency
o
and
wave number k
o
, and amplitude A(x, t) whose local value is slowly varying in space and
time. A(x, t) is also called the envelope. How slow is its variation?
If we ignore terms of (k)
2
in the integrand, (4.3) reduces to
_

A(x, t) = dk /(k) exp [ik(x

t)] (4.4)
o
0
Clearly A = A(x

t). Thus the envelope itself is a wave traveling at the speed

.
o o
This speed is called the group velocity,
d

c
g
(k
o
) = (4.5)
dk

ko
15
Figure 4: Envelope of waves with a rectangular band of wavenumbers
Note that the characteristic length and time scales are (k
m
)
1
and (

k
m
)
1
respec-
o
tively, therefore much longer than those of the component waves : k
1
and
1
. In other
o o
words, (4.3) is adequate for the slow variation of A
e
in the spatial range of k
m
x = O(1)
and the time range of

k
m
t = O(1).
o
As a specic example we let the amplitude spectrum be a real constant within the
narrow band of k
o
, k
o
+ ,
_
ko+
= / e
ikxi(k)t
dk, k
o
(4.6)
ko
then
_

= k
o
/e
ikoxiot
de
iko(xcgt)
+

=
2/ sin (x c
g
t)
x c t
e
ikoxiot
= Ae
ikoxiot
(4.7)
g
where = k k
o
/k
o
and
2/ sin (x c
g
t)
A = (4.8)
(x c
g
t)
as plotted in gure (4).
By dierentiation, it can be veried that
A A
+ c
g
= 0, (4.9)
t x
16
_ _

Multiplying (4.9) by A ,

A

A
A + c
g
A = 0,
t x
and adding the result to its complex conjugate,

A A
A + c
g
A = 0,
t x
we get
[A[
2
[A[
2
+ c
g
= 0 (4.10)
t x
We have seen that for a monochromatic wave train the energy density is proportional
to [A[
2
. Thus the time rate of change of the local energy density is balanced by the net
ux of energy by the group velocity.
Now let us examine the more accurate approximation (4.3). By straightforward
dierentiation, we nd
_

i

(k
o
) A
= i

(k
o
)k (k)
2
/(k)e
iS
dk
t 0 2
_

A
= (ik)/(k)e
iS
dk
x 0

2
A
_
_ _
= (k)
2
/(k)e
iS
dk
x
2
0
where
S = k x

k t
1
o
(k)
2
t (4.11)
o
2
is the phase function. It can be easily veried that
A
+

A i

2
A
(4.12)
o
=
t
o
x 2 x
2
By keeping the quadratic term in the expansion, (4.12) is now valid for a larger spatial
range of (k)
2
x = O(1). In the coordinate system moving at the group velocity, =
x c
g
t, = t, we easily nd
A(, ) A A A(, ) A
= c
g
, =
t x x
so that (4.12) simplies to the Schr odinger equation:
A i

2
A
o
= (4.13)
2
2
17
_ _
_
By manipulations similar to those leading to (4.10), we get

[A[
2
i



A A
o
= A A (4.14)
2
Thus the local energy density is not conserved over a long distance of propagation.
Higher order eects of dispersion redistribute energy to other parts of the envelope.
For either a wave packet whose envelope has a nite length ( A() = 0), or for a
periodically modulated envelope (A(x) = A(x + L)), we can integrate (4.14) to give

[A[
2
d = 0 (4.15)

where the integration extends over the entire wave packet or the group period. Thus
the total energy in the entire wave packet or in a group period is conserved.
5 Radiation of surface waves forced by an oscillating
pressure
We demonstrate the reasoning which is typical in many similar radiation problems.
The governing equations are

2
=
xx
+
yy
= 0, < z < 0. (5.1)
with the kinematic boundary condition

z
=
t
, z = 0 (5.2)
and the dynamic boundary condition
p
a
+
t
+ g = 0 (5.3)

where p
a
is the prescribed air pressure. Eliminating the free surface displacement we
get
(p
a
)
t

tt
+ g
z
= , z = 0. (5.4)

Let us consider only sinusoidal time dependence:
p
a
= P (x)e
it
(5.5)
18
and assume
it
(x, z, t) = (x, z)e
it
, (x, t) = (x)e (5.6)
then the governing equations become

2
=
xx
+
yy
= 0, < z < 0. (5.7)

z
= i, z = 0 (5.8)
and

2
i

z
= P (x), z = 0. (5.9)
g g
Dene the Fourier transform and its inverse by
_

_

f () = dxe
ix
f (x), f (x) =
1
de
ix
f

(), (5.10)
2
We then get the transforms of (5.1) and (5.4)

= 0, z < 0 (5.11)
zz

subject to
i

z
= P

(), z = 0. (5.12)
g g
The solution nite at z for all is

||z
= Ae

To satisfy the free surface condition

2
iP

()P ()
[[A A =
g g
hence
iP

()
A =
g
[[
2
/g
or
1
de
_

iP

()
ix ||z g
= e
2 [[
2
/g
_

1
=
i
de
ix
e
||z
dx

e
ix

P (x

)
[[
1

2
/g
,
g 2
_

_

1
= dx

P (x ) d e
i(xx

)
e
||z
[[
1

2
/g
(5.13)
g 2
19

Let

2
k = (5.14)
g
we can rewrite (5.13) as
_

))
= dx

P (x )
1
d e
z
cos((x x
(5.15)
g 0 k
The nal formal solution is
=
i
g
e
it
_


dx

P (x

)
1

_

0
d e
z
cos((x x

))
k
(5.16)
If we chose
P (x

) = P
o
(x

) (5.17)
then
((x, z) =
iP
o
g
1

_

0
d e
z
cos(x)
k
(5.18)
is clearly the response to a concentrated surface pressure and the response to a pressure
distribution (5.16) can be written as a superposition of concentrated loads over the free
surface,
_

= dx

P (x )((x x , z). (5.19)

where
_

((x, z, t) =
iP
o
e
it
1
d e
z
cos(x)
(5.20)
g 0 k
In these results, e.g., (5.20), the Fourier integral is so far undened since the integrand
has a real pole at = k which is on the path of integration. To make it mathematically
dened we can chose the principal value, deform the contour from below or from above
the pole as shown in gure (5). This indeniteness is due to the assumption of quasi
Figure 5: Possible paths of integration
steady state where the inuence of the initial condition is no longer traceable. We must
now impose the radiation condition that waves must be outgoing as x . This
20
_ _
Figure 6: Closed contour in the upper half plane
condition can only be satised if we deform the contour from below. Denoting this
contour by , we now manipulate the integral to exhibit the behavior at innity, and to
verify the choice of path. For simplicity we focus attention on (. Due to symmetry, it
suces to consider x > 0. Rewriting,
1
((x, z, t) =
iP
o
e
it
(I
1
+ I
2
)
g 2
1
_
e
ix
it
=
iP
o
e d e
z
e
ix
+ (5.21)
g 2 k k
Consider the rst integral in (5.21). In order that the rst integral converges for
large [[, we close the contour by a large circular arc in the upper half plane, as shown
in gure (6), where > 0 along the arc. The term
ix ix x
e = e e
is exponentially small for positive x. Similarly, for the second integral we must chose
the contour by a large circular arc in the lower half plane as shown in gure (7).
Back to the rst integral in (5.21)
_
ix z
e e
I
1
= d (5.22)
k
The contour integral is
_
ix z
_
ix z
_
ix z
_
0
ix z
e e e e e e e e
d = d + d + d
k k C k i k
_
0
ix z
e e
= I
1
+0 + d
i k
21
_ _
Figure 7: Closed contour in the lower half plane
The contribution by the circular arc C vanishes by Jordans lemma. The left hand side
is
kz
LH S = 2ie
ikx
e (5.23)
by Cauchys residue theorem. By the change of variable = i, the right hand side
becomes
_
0
x iz
e e
RHS = I
1
+ i d
i k
Hence
e
kz
I
1
= 2ie
ikx
e + i
_

d
e
x iz
(5.24)
0 i k
Now consider I
2
_
ix z
e e
I
2
= d (5.25)
k
and the contour integral along the contour closed in the lower half plane,
_
ix z
_

e
ix z
e e e
d = I
2
+0 + d
k 0 k
Again no contribution comes from the circular arc C. Now the pole is outside the
contour hence LH S = 0. Let = i in the last integral we get
_

x iy
e e
I
2
= i d (5.26)
0 i k
Adding the results (5.24) and (5.26).,
_

iz
ie
x
e
iz
ie
x
e
kz
I
1
+ I
2
= 2ie
ikx
e + d
0 i k i k
x
= 2ie
ikx
e
kz
+2
_

d

2
e
+ k
2
( cos y + k sin y) (5.27)
0
22
_ _
_ _


Finally, the total potential is, on the side of x > 0,
1
((x, z, t) =

e
it
(I
1
+ I
2
) e
it
g 2i
_

e
x
1
ikx kz
=

e
it
e e + d

2
+ k
2
( cos z + k sin z) (5.28)
g 0
The rst term gives an outgoing waves. For a concentrated load with amplitude
P
o
, the displacement amplitude is P
o
/g. The integral above represent local eects
important only near the applied pressure. If the concentrated load is at x = x

, one
simply replaces x by x x

everywhere.
6 The Kelvin pattern of ship wave
The action of the ships propeller
Has a thrust pattern
To which the ship reacts by moving forward,
Which also results secondarily,
In the ships bow elevated waves,
And its depressed transverse stern wave,
Which wave disturbances of the water
Are separate from the propellers thrust waves.
R.Buckminster Fuller, Intuition- Metaphysical Mosaic. 1972.

Anyone ying over a moving ship must be intrigued by the beautiful pattern in the
ships wake. The theory behind it was rst completed by Lord Kelvin, who invented
the method of stationary phase for the task. Here we shall give a physical/geometrical
derivation of the key results (lecture notes by T. Y. Wu, Caltech).
Consider rst two coordinate systems. The rst r = (x, y, z) moves with ship at the
uniform horizontal velocity U. The second r

= (x , y , z) is xed on earth so that water
is stationary while the ship passes by at the velocity U. The two systems are related by
the Galilean transformation,
r = r + Ut (6.29)
23
A train of simple harmonic progressive wave
= 'A exp[i(k r

t)] (6.30)
in the moving coordinates should be expressed as
= 'A exp[ik (r Ut) it] = 'A exp[ik r i( k.U)t]
= 'A exp[ik r it] (6.31)
in the stationary coordinates. Therefore the apparent frequency in the moving coordi-
nates is
= k U (6.32)
The last result is essentially the famous Dopplers eect. To a stationary observer, the
whistle from an approaching train has an increasingly high pitch, while that from a
leaving train has a decreasing pitch.
If a ship moves in very deep water at the constant speed U in stationary water,
then relative to the ship, water appears to be washed downstream at the velocity U.
A stationary wave pattern is formed in the wake. Once disturbed by the passing ship,
a uid parcel on the ships path radiates waves in all directions and at all frequencies.
Wave of frequency spreads out radially at the phase speed of c = g/ according to
the dispersion relation. Only those parts of the waves that are stationary relative to the
ship will form the ship wake, and they must satisfy the condition
= 0, (6.33)
i.e.,
k
= k U, or c = = U (6.34)
k k
Referring to gure 8, let O, (x = 0) represents the point ship in the ship-bound
coordinates. The current is in the positive x direction. Any point x
1
is occupied by
a uid parcel Q
1
which was disturbed directly by the passing ship at time t
1
= x
1
/U
earlier. This disurbed parcel radiates waves of all frequencies radially. The phase of
wave at the frequency reaches the circle of radius ct
1
where c=g/ by the deep water
dispersion relation. Along the entire circle however only the point that satises (6.34)
24
Figure 8: Waves radiated from disturbed uid parcel
can contribute to the stationary wave pattern, as marked by P . Since OQ
1
= x
1
= Ut
1
,
Q
1
P = ct
1
and OP = Ut
1
k/k, where k is in the direction of Q

1
P . It follows that
OP Q
1
is a right triangle, and P lies on a semi circle with diameter OQ
1
. Accounting
for the radiated waves of all frequencies, hence all c, every point on the semi circle
can be a part of the stationary wave phase formed by signals emitted from Q
1
. Now
this argument must be rectied because wave energy only travels at the group velocity
which is just half of the phase velocity in deep water. Therefore stationary crests due
to signals from Q
1
can only lie on the semi-circle with the diameter O
1
Q
1
= OQ
1
/2.
Thus P
1
instead of P is one of the points forming a stationary crest in the ships wake,
as shown in gure 8.
t
Any other uid parcel Q
2
at x
2
must have been disturbed by the passing ship at time
2
= x
2
/U earlier. Its radiated signals contribute to the stationary wave pattern only
along the semi circle with diameter O
2
Q
2
= OQ
2
/2. Combining the eects of all uid
parcels along the +x axis, stationary wave pattern must be conned inside the wedge
which envelopes all these semi circles. The half apex angle
o
of the wedge, which denes
the wake, is given by
Ut/4
sin
o
= = 1/3, (6.35)
3Ut/4
hence
o
= sin
1
1/3 = 19.5

, see gure 9.
Now any point P inside the wedge is on two semicircles tangent to the boundary
of the wedge, i.e., there are two segments of the wave crests intersecting at P : one
perpendicular to PQ
1
and one to PQ
2
, as shown in gure 9.
Another way of picturing this is to examine an interior ray from the ship. In gure
25
Figure 9: Wedge angle of the ship wake
O O
Q
P1
M1
M2
P2
b
'
Figure 10: Geometrical relation to nd Points of dependence
(10), draw a semi circle with the diameter O

Q = OQ/2, then at the two intersections P

1
and P
2
with the ray are the two segments of the stationary wave crests. In other words,
signals originated from Q contribute to the stationary wave pattern only at the two
points P
1
and P
2
, as shown in gure 10. Point Q can be called the point of dependence
for points P
1
and P
2
on the crests.
For any interior point P there is a graphical way of nding the two points of depen-
dence Q
1
and Q
2
. Referring to gure 10, O

QP
1
and O

QP
2
are both right triangles.
Draw O
1
M
1
| QP
1
and O
2
M
2
| QP
2
where M
1
and M
2
lie on the ray inclined at the an-
gle . it is clear that OM
1
= OP
1
/2 and OM
2
= OP
2
/2, and M
1
O

P
1
and M
2
O

P
2
are both right triangles. Hence O

lies on two semi circles with diameters M
1
P
1
and
M
2
P
2
.
We now reverse the process, as shown in gure 11. For any point P on an interior
ray, let us mark the mid point M of OP and draw a semi circle with diameter MP .
The semi circle intersects the x axis at two points S
1
and S
2
. We then draw from P two
lines parallel to MS
1
and MS
2
, the two points of intersection Q
1
an Q
2
on the x axis
are just the two points of dependence.
26
_
_
Figure 11: Points of dependence
Let

PQ
1
O =

MS
1
O =
1
and

PQ
2
O =

MS
2
O =
2
. then
PS
i
PS
i
tan(
i
+ ) = = = 2 tan
i
i = 1, 2.
MS
i
PQ
i
/2
hence
tan
i
+tan
2 tan
i
=
1 tan
i
tan
which is a quadratic equation for
i
, with two solutions:
_ _

_
tan
1
_
1 1 8 tan
2

= (6.36)

4 tan _
tan
2
_
They are real and distinct if
1 8 tan
2
> 0 (6.37)
These two angles dene the local stationary wave crests crossing P, and they must
be perpendicular to PQ
1
and PQ
2
. There are no solutions if 1 8 tan
2
< 0, which
corresponds to sin > 1/3 or > 19.5

, i.e., outside the wake. At the boundary of the

wake, = 19, 5

and tan = 1/8, the two angles are equal

2

1
=
2
= tan
1
= 55

. (6.38)
2
By connecting these segments at all points in the wedge, one nds two systems of wave
crests, the diverging waves and the transverse waves, as shown in gure div-trans.
A beautiful photograph is shown in Figure 13
Knowing that waves are conned in a wedge, we can estimate the behavior of the
wave amplitude by balancing in order of magnitude work done by the wave drag R and
the steady rate of energy ux

RU = (Ec
g
)r ([A[
2
c
g
)r (6.39)
27
7
Figure 12: Diverging and transverse waves in a ship wake
hence
A r
1/2
(6.40)
This estimate is valid throughout the wedge except near the outer boundaries, where
A r
1/3
(6.41)
by a more rened analysis (Stoker, 1957, or Wehausen & Laitone, 1960).
Basic theory for two-dimensional Internal waves
in a stratied uid
[References]:
C.S. Yih, 1965, Dynamics of Inhomogeneous Fluids, MacMillan.
O. M. Phillips, 1977, Dynamics of the Upper Ocean, Cambridge U. Press.
P. G. Baines, 1995, Topographical Eects in Stratied Flows Cambridge U. Press.
M. J. Lighthill 1978, Waves in Fluids , Cambridge University Press.
28
Due to seasonal changes of temperature, the density of water or atmosphere can have
signicant variations in the vertical direction. Variation of salt content can also lead to
density stratication. Freshwater from rivers can rest on top of the sea water. Due to
the small diusivity, the density contrast remains for a long time.
Consider a calm and stratied uid with a static density distribution
o
(z) which
decreases with height (z). If a uid parcel is moved from the level z upward to z + , it
is surrounded by lighter uid of density (z + dz). The upward buoyancy force per unit
volume is
d
g((z + ) (z)) g
dz
and is negative. Applying Newtons law to the uid parcel of unit volume
d
2
d

dt
2
= g
dz

or
d
2

dt
2
+ N
2
= 0 (7.1)
where
N =
_

g

d
dz
_
1/2
(7.2)
is called the Brunt-V al ais a frequency. This elementary consideration shows that once
a uid is displaced from its equilibrium position, gravity and density gradient provides
restoring force to enable oscillations. In general there must be horizontal nonunifomities,
hence waves are possible.
We start from the exact equations for an inviscid and incompressible uid with
variable density.
For an incompressible uid the density remains constant as the uid moves,

t
+ q = 0 (7.3)
where q = (u, w) is the velocity vector in the vertical plane of (x, z). Conservation of
mass requires that
q = 0 (7.4)
The law of momentum conservation reads
(q
t
+ q q) = p ge
z
(7.5)
and e
z
is the unit vector in the upward vertical direction.
30






_




7.1 Linearized equations
Consider small disturbances
p = p + p , = (z) +

, q = (u

, w ) (7.6)
with


, p p

(7.7)
and u

, v , w are small. Linearizing by omitting quadratically small terms associated

with the uid motion, we get

+ w
d
= 0. (7.8)
t
dz
u + w = 0 (7.9)
x z
u

= p (7.10)
t x
w

= p
z
p g g

(7.11)
t z
In the last equation the static part must be in balance
0 = p
z
g, (7.12)
hence
z
p(z) = (z)dz. (7.13)
0
The remaining dynamically part must satisfy
w

= p g

(7.14)
t z
Upon eliminating p

from the two momentum equations we get
d

u + (u

w
x
)
t
= g

(7.15)
z x
dz
t
Eliminating

from (7.8) and (7.15) we get
d
u

+ (u

w
x
)
tt
= g

= g
d
w (7.16)
z xt
dz
tt
dz
x
Let us introduce the disturbance stream function :
u =
z
, w

=
x
(7.17)
31


It follows from (7.16) that
d
(
xx
+
zz
)
tt
= (g
xx

ztt
) (7.18)
dz
by virture of Eqns. (7.8) and (7.17). Note that
N =
g d
(7.19)
dz
is the Brunt-V al

ais a frequency. In the ocean, density gradient is usually very small (

N 5 10
3
rad/sec). Hence can be approximated by a constant reference value, say,
0
= (0) in (7.10) and (7.14) without much error in the inertia terms. However density
variation must be kept in the buoyancy term associated with gravity, which is the only
restoring force responsible for wave motion. This is called the Boussinesq approximation
and amounts to taking to be constant in (Eq:17.1) only. With it (7.18) reduces to
(
xx
+
zz
)
tt
+ N
2
(z)
xx
= 0. (7.20)
Note that because of linearity, u

and w

satisfy Eqn. (7.20) also, i.e.,
(w

+ w
zz
)
tt
+ N
2
w = 0 (7.21)
xx xx
etc.
7.2 Linearized Boundary conditions on the sea surface
Dynamic boundary condition : Total pressure is equal to the atmospheric pressure
(p + p

= 0. (7.22) )
z=
On the free surface z = , we have
_

p g (0)dz = g(0)
0
Therefore,
g + p

= 0, z = 0, (7.23)
implying
g
xxt
= p

(7.24)
xxt
, z = 0.
32

_ _
Kinematic condition :

t
= w, z = 0. (7.25)
The left-hand-side of (7.24) can be written as
g
xxt
= g w

xx
p
Using 7.10, the right-hand-side of 7.24 may be written,
xxt
= u

= w

xtt ztt
hence
w gw

= 0, on z = 0. (7.26)
ztt xx
Since w

=
x
, also satises the same boundary condition

ztt
g
xx
= 0, on z = 0. (7.27)
On the seabed, z = h(x) the normal velocity vanishes. For a horizontal bottom we
have
(x, h, t) = 0. (7.28)
8 Internal waves modes for nite N
Consider a horizontally propagating wave beneath the sea surface. Let
it
= F (z) e
ikx
e . (8.1)
From Eqn. (7.21),
d
2
F
_ _

2
k
2
F + N
2
k
2
F = 0
dz
2
or,
d
2
F N
2

2
+ k
2
F = 0 z < 0. (8.2)
dz
2

2
On the (horizontal) sea bottom
F = 0 z = h. (8.3)
33
ais a Figure 14: Typical variation of Brunt-V al frequency in the ocean. From O. M.
Phillips, 1977
From Eqn. (7.27),
dF k
2
g F = 0 z = 0. (8.4)
dz
2
Equations (8.2), (8.3) and (8.4) constitute an eigenvalue condition.
If
2
< N
2
, then F is oscillatory in z within the thermocline. Away from the
thermocline,
2
> N
2
, W must decay exponentially. Therefore, the thermocline is a
waveguide within which waves are trapped. Waves that have the greatest amplitude
beneath the free surface is called internal waves.

Since for internal waves, < N while N is very small in oceans, oceanic internal
waves have very low natural frequencies. For most wavelengths of practical interests
2
gk so that
F

= 0 on z = 0. (8.5)
This is called the rigid lid approximation, which will be adopted in the following.
With the rigid-lid approximation, if N=constant (if the total depth is relataively
small compared to the vertical scale of stratication, the solution for F is
_ _
N
2

2
F = A sin k(z + h) (8.6)

N
where

2

2
kh = n, n = 1, 2, 3... (8.7)

34
_
This is an eigen-value condition. For a xed wave number k, it gives the eigen-frequencies,
N

n
= _ (8.8)
_ _
2
1 +
n
kh
For a given wavenumber k, this dispersion relation gives the eigen-frequency
n
. For a
given frequency , it gives the eigen-wavenumbers k
n
,
n
k
n
= (8.9)
h N
2

2
For a simple lake with vertical banks and length L, 0 < x < L, we must impose the
conditions :
u

= 0, hence = 0, x = 0, L (8.10)
The solution is
_ _
N
2

2
nm

= A sin k
m
x exp(i
nm
t) sin
_
k
m
(z + h)
_
. (8.11)
nm
with
k
m
L = m, m = 1, 2, 3, ... (8.12)
The eigen-frequencies are:
N

nm
= _
_ _
2
(8.13)
1 +
nL
mh
9 Internal waves in a vertically unbounded uid
Consider N = constant (which is good if attention is limited to a small vertical extent),
and denote by (, ) the (x, z) components of the wave number vector

k Let the solution

be a plane wave in the vertical plane
=
0
e
i(x+zt)
Then

2
2
= N
2

2
+
2
or

= N (9.1)
k
35

_ _
_ _
_ _

k
2
=
2
+
2
(9.2)
This is the dispersion relation. Note that is

= cos

(9.3)
N
where

is the inclination of

k with repect to the x axis. For a given frequency, there

are two possible signs for . Since the above relation is also even in , there are four
possible inclinations for the wave crests and troughs with respect to the horizon; the
angle of inclination is
1
[

[ = cos

(9.4)
N
For > N, there is no wave.
To under the physics better we note rst that the phase velocity is

C =
k
2
(, ) (9.5)
while the group velocity components are
1
C
gx
= = N
k k
2
k
N
2
N
= 1 =
2
k k
2
k
3

C
gz
= = . (9.6)
k
3
Thus

C

g
= N . (9.7)
k
2
k
,
k
Therefore, the group velocity is perpendicular to the phase velocity,
C
g
C = 0. (9.8)
Since
_ _
N
C +

C
g
=
2
+
2
, 0 =
k
2
(k, 0) (9.9)
k
3
C and

the sum of

C
g
is a horizontal vector, as shown by any of the sketches in Figure
17. Note that when the phase velocity as an upward component, the group velocity has
a downward component, and vice versa. Now let us consider energy transport. from
36

cg
c
cg
cg
c
c
c
c
cg
cg
c
cg
Figure 15: Phase and group velocities
(7.10) we get
p

=
zt
=
o
e
i(x+zt)
x
hence the dynamic pressure is

o
e
i(x+zt)
p = i (9.10)
k
The uid velocity is easily calculated

i(x+zt)
(9.11) q = (u

, v ) = (
z
,
x
) = i(, )
o
e
The averaged rate of energy transport is therefore
E =
1

2
[[
2

(, ) (9.12)
2
which is in the same direction of the group velocity.
Energy must radiate outward from the oscillating source, hence the group velocity
vectors must all be outward. Since there are 4 directions for

k. There are four radial

beams parallel to c
g
, in four quadrants, forming St. Andrews Cross. The crests (phase
lines) in the beam in the rst quadrant must be in the south-easterly direction. Similarly
the crests in all four beams must be outward and toward the horizontal axis. Let be
the inclination of a beam with respect to the x axis, then = /2

in the rst
quadrant. The dispersion relation can be written as

= sin (9.13)
N
where is the inclination of a beam and not of the wavenumber vector.
37
Image by MIT OCW.
Figure 16: St Andrews Cross in a stratied uid. Top: /N = 0.7; bottom left
/N = 0.9; bottom right: /N = 1.11. From Mowbray & Rarity, 1965, JFM
Movie records indeed conrm these predictions. Within each of the four beams which
have widths comparable to the cylinder diameter, only one or two wave lengths can be
seen.
This unique property of anisotropy has been veried in dramatic experiments by
Mowbray and Stevenson. By oscillating a long cylinder at various frequencies verti-
cally in a stratied uid, equal phase lines are only found along four beams forming St
Andrews Cross, see Figure (16) for /N = 0.7, 0.9 and 1.11. It can be veried that
angles are [[ = 45

for /N = 0.7, and [[ = 64

for /N = 0.9. In the last photo,
/N = 1.11. There is no wave. These results are all in accord with the condition (9.4).
Comparison between measured and predicted angles is plotted in Figure (17) for a
wide range of /N
38
Figure 17: Comparison of measured and predicted angles of internal-wave beams. /N
vs. sin . From Mowbray & Rarity, 1965, JFM
10 Reection of internal waves at boundary
For another interesting feature, consider the reection of an internal wave from a slope.

Recall that

= cos
1
N
, i.e., for a xed frequency there are only two allowable
directions with respect to the horizon. Relative to the sloping bottom inclined at
o
the
inclinations of the incident and reected waves must be dierent, and are respectively

+
o
and


o
, see Figure 18.

Let be along, and be normal to the slope. Since the slope must be a streamline,
i
+
r
must vanish along = 0 and be proportional to e
i(t)
; the total stream
function must be of the form

i
e
i(k
(i)
t)
+
r
e
i(k
(r)
t t
t)
sin e
i(t)
.
In particular the wavenumber component along the slope must be equal,
k
(i)
= k
(r)
=
t t
Consider rst
o
>

, as shown in the left diagram in Figure(18). We must have

k
(i)
cos(

+
o
) = k
(r)
cos(


o
),
k
which implies rst that
(i)
,= k
(r)
. (10.1)
39


x
x
z
o
q
k

k

q0
q
x
'
'
(r)
(
i)

x
z
o
q
k
k

q0
q
'
'
(r)
(i)
Figure 18: Internal wave reected by in inclined surface. Left:
o
<

. Right:
o
> .
The incident wave and the reected wave have dierent wavelengths! Furthermore, the
reected wave is directed up-slope.If, however,
o
>

, as shown in the right diagram in

Figure(18), then
k
(i)
cos( +
o
) = k
(r)
cos( (
o
)) = k
(r)
cos(
o
),
The reected wave is directed down-slope instead (Phillips, 1977).
40