Professional Documents
Culture Documents
WAVES IN ARTERIES 1
I-campus project
School-wide Program on Fluid Mechanics
MODULE ON WAVES IN FLUIDS
T. R. Akylas & C. C. Mei
To describe a problem in mathematical terms, one must make use of the basic laws
that govern the elements of the problem. In continuum mechanics, these are the conser-
vation laws for mass and momentum. In addition, empirical constitutive laws are often
needed to relate certain unknown variables such as the equations of state, and relations
between stress and strain rates, etc.
To derive the conservation law one may consider an inÞnitesimal element (a line
segment, area or volume element), yielding a differential equation directly. Alternately,
one may consider a control volume (or area, or line segment) of arbitrary size in the
medium of interest. The law is Þrst obtained in integral form; a differential equation is
then derived by using the arbitrariness of the control volume. The two approaches are
completely equivalent.
where C is deÞned by s s
S dp Eh
C= = (1.9)
ρ dS 2ρa
and has the dimension of velocity. In view of (1.6), equations (1.1) and (1.8) are a pair
of nonlinear equations for the two unknowns u and S.
For inÞnitesimal amplitudes we can linearize these equations. Let a = ao + a0 with
a0 ¿ ao then the (1.1) becomes, to the leading order,
∂a0 ao ∂u
+ =0 (1.10)
∂t 2 ∂x
∂u ∂p
ρo =− (1.11)
∂t ∂x
1.1. WAVES IN ARTERIES 4
∂ 2 a0 2 0
2∂ a
= co (1.14)
∂t2 ∂x2
where s
Eh
co = (1.15)
2ρao
Alternately one can eliminate a to get the equation for u
∂2u 2
2∂ u
= co (1.16)
∂t2 ∂x2
Because of (1.12), the dynamic pressure is governed also by
∂ 2p 2
2∂ p
= co (1.17)
∂t2 ∂x2
All unknowns are governed by the same equation due to linearity and the fact that all
coefficients are constants.
To complete the formulation, initial and bounday conditions must be added. These
matters will be discussed later in chapter 2.
As the process of linearization is frequently appplied in these notes, some comments
are in order here. To Þnd out the accuracy of linearization, it is useful to estimate
Þrst the scales of motion. Let A, T, L, U and P denote the scales of a0 , t, x, u and p0
respectively. It is natural to take L = co T . From (1.1) (1.5) and (1.6) we get the
relations among the scales of dynamical quantities
ao A U a2o AL
= , hence U =
T L ao T
EhA
P =
a2o
U 1P 1 1 EhA
= =
T ρL ρ L/T a2o
1.2. SOUND IN FLUIDS 5
It follows that
AL 1 1 EhA
=
ao T ρ L/T a2o
hence,
L2 Eh
2
= = c2o
T ρao
With these scales the ratio of a typical nonlinear term to a linear term is
u ∂u
∂x U 2 /L U A
∂u ∼ = =
∂t
U/T L/T ao
Hence the condition for linearization is that
A
¿1
ao
i.e., the amplitude of transverse oscillation is much smaller than the typical radius.
2 Sound in ßuids
The basic equations governing an inviscid and compressible ßuid are as follows. Mass
conservation:
∂ρ
+ ∇ · (ρu) = 0 (2.1)
∂t
Momentum conservation: Ã !
∂u
ρ + u · ∇u = −∇p (2.2)
∂t
We must add an equation of state
p = p(ρ, S) (2.3)
where S denotes the entropy. When no temperature gradient is imposed externally and
the gradient of the ßow is not too large, one can ignore thermal diffusion. The ßuid
motion is then adiabatic; entropy is constant. As a result p = p(ρ, So ) depends only on
the density. Eq. (2.1 ) can be written as
à ! à !
∂u ∂p
ρ + u · ∇u = − ∇ρ (2.4)
∂t ∂ρ S
We shall denote và !
u
u ∂p
C= t (2.5)
∂ρ S
1.2. SOUND IN FLUIDS 6
so that à !
∂u
ρ + u · ∇u = −C 2 ∇ρ (2.6)
∂t
It is easy to check that C has the dimension of velocity.
From thermodynamics we also have
à ! à !
∂p ∂p
=γ (2.7)
∂ρ S
∂ρ T
p = ρRT (2.8)
Liquids are much less compressible. One usually writes the equation of state as
à ! à !
∂ρ ∂ρ
dρ = dp + dT (2.10)
∂p T
∂T p
Denoting à !
1 ∂ρ
β=− (2.11)
ρ ∂T p
and
∂u 1
= − ∇p0 (2.15)
∂t ρo
Taking the curl of the second, we get
∂
∇×u=0 (2.16)
∂t
thus the velocity Þeld is irrotational if it is so initially. We can introduce a potential φ
by
u = ∇φ (2.17)
z
z=ζ ζ
O x
z=-h dy
dx
First, the law of mass conservation. The rate of volume increase in the column
∂ζ
dxdy
∂t
must be balanced by the net volume ßux into the column from all four vertical sides. In
shallow water, the horizontal length scale, characterized by the wavelength λ, is much
greater that the vertical length h. Water ßows mainly in the horizontal planes with
the velocity u(x, y, t), which is essentially constant in depth. Through the vertical sides
normal to the x axis, the difference between inßux through the left and outßux through
the right is
( )
∂
− [u (ζ + h) |x+dx − u (ζ + h) |x ] dy = − [u (ζ + h)] + O(dx) dxdy.
∂x
Similarly, through the vertical sides normal to the y axis, the difference between inßux
through the front and outßux through the back is
( )
∂
− [v (ζ + h) |y+dy − v (ζ + h) |y ] dx = − [v (ζ + h)] + O(dy) dydx.
∂y
Omitting terms of higher order in dx, dy, we invoke mass conservation to get
( )
∂ζ ∂ ∂
dxdy = − [u (ζ + h)] + [v (ζ + h)] + O(dx, dy) dxdy.
∂t ∂x ∂y
∂ζ
+ ∇ · [u(ζ + h)] = 0. (3.1)
∂t
1.3. SHALLOW WATER WAVES 9
This equation is nonlinear because of the quadratic product of the unknowns u and ζ.
Now the law of conservation of momentum. In shallow water the vertical momentum
balance is dominated by pressure gradient and gravity, which means that the distribution
of pressure is hydrostatic:
p = ρg (ζ − z) , (3.2)
where the atmospheric pressure on the free surface is ignored. Consider now momentum
balance in the x direction. The net pressure force on two vertical sides normal to the x
direction is
∂ Zζ ∂ Zζ
dxdy p dz = −dxdy ρg(ζ − z) dz
∂x −h ∂x −h
∂(ζ + h)
= −ρg(ζ + h) dxdy.
∂x
The hydrodynamic reaction from the sloping bottom to the ßuid is
∂h ∂h
−p dxdy = ρg(ζ + h) dxdy.
∂x ∂x
The change of ßuid momentum consists of two parts. One part is due to the time rate
of momentum change in the water column
( )
∂
[ρu(ζ + h)] dxdy,
∂t
and the other is due to the net ßux of momentum through four vertical sides:
∂ ∂
[ρu2 (ζ + h)]dxdy + [ρuv(ζ + h)]dydx.
∂x ∂y
Equating the total rate of momentum change to the net pressure force on the sides and
on the bottom, we get
∂ ∂ ∂
[ρu(ζ + h)] + [ρu2 (ζ + h)] + [ρuv(ζ + h)]
∂t ∂x ∂y
∂(ζ + h) ∂h
= −g(ζ + h) + g(ζ + h) .
∂x ∂x
The left-hand side can be simpliÞed to
à ! ( )
∂u ∂u ∂u ∂ζ ∂ ∂
+u +v (ζ + h) + u + [u(ζ + h)] + [v(ζ + h)]
∂t ∂x ∂y ∂t ∂x ∂y
∂ζ ∂u ∂u
= +u +v
∂t ∂x ∂y
1.3. SHALLOW WATER WAVES 10
∂u ∂u ∂u ∂ζ
+u +v = −g . (3.3)
∂t ∂x ∂y ∂x
∂v ∂v ∂v ∂ζ
+u +v = −g . (3.4)
∂t ∂x ∂y ∂y
∂u
+ u · ∇u = −g∇ζ. (3.5)
∂t
where ∇ is the horizontal (two-dimensional) gradient operator,
à !
∂ ∂
∇= ,
∂x ∂y
Equations (3.1) and (3.5) are coupled nonlinear partial differential equations for three
scalar unknowns u and ζ.
Now the boundary and initial conditions. On a shoreline S, there can be no normal
ßux, therefore,
hu · n = 0 on S, (3.6)
where n denotes the unit normal vector pointing horizontally into the shore. This
condition is applicable not only along a cliff shore where h is Þnite, but also on a
shoreline where h = 0, as long as the waves are gentle enough not to break. In the latter
case the whereabout of the shoreline is unknown a priori and must be found as a part
of the solution.
At the initial instant, one may assume that the displacement ζ(x, y, 0) and the ver-
∂
tical velocity of the entire free surface ∂t
ζ(x, y, 0) is known. These conditions complete
the formulation of the nonlinear shallow water wave problem.
1 ∂ 2ζ
= ∇2 ζ, (3.12)
c2 ∂t2
√
where c = gh = O(λ/T ) is the characteristic velocity of inÞnitesimal wave motion and
∂2 ∂2
∇2 = +
∂x2 ∂y 2
Equation (1.6.12) is the two-dimensional extension of the wave equation. If, furthermore,
all conditions are uniform in the y direction, ∂/∂y = 0, (3.12) reduces to the familiar
form
1 ∂ 2ζ ∂2ζ
= . (3.13)
c2 ∂t2 ∂x2
1 ∂ 2Φ
∇2 Φ = (4.1)
c2 ∂t2
q
where c = dp/dρ is the speed of sound. Consider the ratio
1 ∂2Φ
c2 ∂t2 ω 2 /k 2
∼
∇2 Φ c2
√
As will be shown later, the phase speed of the fastest surface gravity wave is ω/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/sec2 , so that the ratio is at
most
40000 1
= ¿1
14002 49
1.4. CAPILLARY-GRAVITY WAVES ON THE SEA SURFACE 13
∇2 Φ = 0 (4.2)
∇·u=0 (4.3)
P = po + p (4.5)
po = −ρgz (4.6)
satisfying
0 = −∇po + −ρgez (4.7)
It follows that
∂u
ρ = −∇p (4.8)
∂t
Clearly the velocity Þeld is irrotational if it is so initially, hence u = ∇Φ and (4.2)
follows from (4.3). Note that
∂Φ
p = −ρ (4.9)
∂t
Refering to Figure ??. let the free surface be z = ζ(x, y, t). Then for a gently sloping
free surface the vertical velocity of the ßuid on the free surface must be equal to the
vertical velocity of the surface itseld. i.e.,
∂ζ ∂Φ
= , z = 0. (4.10)
∂t ∂z
Having to do with the velocity only, this is called the kinematic boundary condition.
1.4. CAPILLARY-GRAVITY WAVES ON THE SEA SURFACE 14
Let us assume that the air above the sea surface is essentially stagnant, because of
its relative small density we ignore its presence and assume the air pressure to be zero.
If surfrace tension is ignored continutiy of pressure requires that
p = po + p = 0, z = ζ.
∂2Φ ∂Φ T ∂ 3 Φ
+g − = 0, z=0 (4.14)
∂t2 ∂z ρ ∂x2 ∂z
When viscosity is neglected, the normal ßuid velocity vanishes on the rigid seabed,
n · ∇Φ = 0 (4.15)
(hx , .hy , 1)
n= q (4.16)
1 + h2x + h2y
Hence
∂Φ ∂h ∂Φ ∂h ∂Φ
=− − , z = −h(x, y) (4.17)
∂z ∂x ∂x ∂y ∂y
Homework No.1
1. During an earthquake, water in a reservoir exerts hydrodynamic pressure on a dam
that may fail. Formulate the dam-reservoir interaction problem under the following
idealizations. The reservoir is inÞnitely long and has a uniform rectangular cross section.
Water is present only on one side of the dam (x > 0) and has the constant depth h.
Before t = 0, all is calm. After t = 0 the dam is forced to vibrate horizontally so that
uo (y, z, t) = prescribed, 0 < t < T,
u(0, y, z, t) = (4.1)
0, t > T.
The free surface is exposed to constant atmospheric pressure. The reservoir bottom is
rigid and does not vibrate vertically (!!!). Neglect gravity but consider compressibility
of water because of the high frequency (∼ O(100)Hz). Express all governing equa-
tions including the boundary conditons in terms of the velocity potential φ deÞned by
(u, v, w) = ∇φ.
2.1. GENERAL SOLUTION TO WAVE EQUATION 1
I-campus project
School-wide Program on Fluid Mechanics
Modules on Waves in ßuids
T. R. Akylas & C. C. Mei
CHAPTER TWO
ONE-DIMENSIONAL PROPAGATION
where F and g are arbitrary functions of their arguments. In the x, t (space,time) plane
F(x − ct) is constant along the straight line x − ct = constant. Thus to the observer
(x, t) who moves at the steady speed c along the positivwe x-axis, the function F is
stationary. Thus to an observer moving from left to right at the speed c, the signal
described initially by F (x) at t = 0 remains unchanged in form as t increases, i.e., F is
a wave propagating to the right at the speed c. Similarly G propagates to the left at the
2.2 BRANCHING OF ARTERIES 2
speed c. The lines x − ct =constant and x + ct = contant are called the characteristic
curves (lines) along which signals propagate. Note that another way of writing (1.2) is
2 Branching of arteries
References: Y C Fung : Biomechanics, Circulation. Springer1997
M.J. Lighthill : Waves in Fluids, Cambridge 1978.
Recall that (1.1) governs both the pressure and the velocity in the blood
∂ 2p 2
2∂ p
= c (2.1)
∂t2 ∂x2
∂2u 2
2∂ u
= c (2.2)
∂t2 ∂x2
The two unknowns are related by the momentum equation
∂u ∂p
ρ =− (2.3)
∂t ∂x
Since
∂p
= p0+ + p0− ,
∂x
and
∂u−
ρ = −ρcu0+ + ρcu0−
∂t
where primes indicated ordinary differentiation with repect to the argument. Equation
(2.3) can be satisÞed if
p+ = ρcu+ , p− = −ρcu− (2.6)
2.2 BRANCHING OF ARTERIES 3
Q± = u± A = ±Zp± (2.7)
where
ρc
Z= (2.8)
A
is the property of the tube and is call the impedance.
Now we examine the effects of branching; Refering to Þgure 1, the parent tube,
characterized by wave speed c and impedance Z, branches into two characterized by c1
and c2 and Z1 and Z2 . An incident wave approaching the junction will cause reßection
and transmitted waves in the branches are p1 (t − x/c1 ) and p2 (t − x/c2 ) in x > 0. At
the junction x = 0, continuity of pressure and ßuxes requires
and
pi − pr p1 p2
= + (2.11)
Z Z1 Z2
DeÞne the reßection coefficient R to be the amplitude ratio of reßected wave to incident
wave, then ³ ´
1 1 1
pr (t) Z
− Z
+ Z2
R= = ³ 1 ´ (2.12)
pi (t) 1
+ 1
+ 1
Z Z1 Z2
Note that both coefficients are constants depending only on the impedances. Hence the
transmitted waves propagate in the direction of increasing x and are similar in form
to the incident waves except smaller by the factor T . On the incidence side waves the
incident and reßected waves propagate in opposite directions.
2.3. WAVES DUE TO INITIAL DISTURBANCES 4
∂ζ ∂(uh)
+ =0 (3.1)
∂t ∂x
and
∂u ∂ζ
= −g (3.2)
∂t ∂x
where ζ(x, t) is the vertical displacement of the free surface and u(x, t) the horizontal
velocity. The atmospheric pressure over the entire free surface is uniform and constant.
By cross-differentiation, ζ is seen to be governed by
à !
∂2ζ ∂ ∂ζ
= g h (3.3)
∂t2 ∂x ∂x
In the limit of constant depth (h =constant), the above equation reduces to the classical
wave eqaution
∂ 2ζ 2 q
2∂ ζ
= c , where c = gh. (3.4)
∂t2 ∂x2
Consider now a sea of inÞnite extent, −∞ < x < ∞. Let the initial surface displace-
ment and velocity be prescribed along the entire surface
∂ζ
(x, t) = G(x), (3.6)
∂t
where F (x) and G(x) are non-zero only in the Þnite domain of x. At inÞnities x → ±∞,
ζ and ∂ζ/∂t are zero for any Þnite t. In (3.4 ) the highest time derivative is of the second
order and initial data are prescribed for ζ and ∂ζ/∂t. Initial conditions that specify all
derivatives of all orders less than the highest in the differential equation are called the
Cauchy initial conditions. These conditions are best displayed in the space-time diagram
as shown in Figure 2.
u tt =c 2u xx
x
u=f(x ) ut =g(x)
Range of (x,t)
1
influence 1
c c
1 1
c c
x0 x-ct x+ct x
Domain of
dependence
where K is an arbitrary constant. Now φ and ψ can be solved from (3.8) and (3.8 as
functions of x,
1 1 Zx
φ(x) = [f (x) + K] − g(x0 )dx0
2 2c x0
1 1 Zx
ψ(x) = [f (x) − K] + g(x0 )dx0 ,
2 2c x0
where K and x0 are some constants. Replacing the arguments of φ by x + ct and of ψ
by x − ct and substituting the results in u, we get
1 1 Z x−ct
ζ(x, t) = f (x − ct) − g dx0
2 2c x0
1 1 Z x+ct
+ f (x + ct) + g dx0
2 2c x0
1 1 Z x+ct
= [f (x − ct) + f (x + ct)] + g(x0 ) dx0 , (3.10)
2 2c x−ct
u, t
O x
and x + ct, and on the initial velocity only along the segment from x − ct to x + ct.
Nothing outside the triangle matters. Therefore, to the observer at x, t, the domain
of dependence is the base of the characteristic triangle formed by two characteristics
passing through x, t. On the other hand, the data at any point x on the initial line t = 0
must inßuence all observers in the wedge formed by two characteristics drawn from x, 0
into the region of t > 0; this characteristic wedge is called the range of inßuence.
Let us illustrate the physical effects of initial displacement and velocity separately.
Case (i): Initial displacement only: f (x) 6= 0 and g(x) = 0. The solution is
1 1
ζ(x, t) = f (x − ct) + f (x + ct)
2 2
and is shown for a simple f (x) in Figure 4 at successive time steps. Clearly, the initial
disturbance is split into two equal waves propagating in opposite directions at the speed
c. The outgoing waves preserve the initial proÞle, although their amplitudes are reduced
by half.
Case (ii): Initial velocity only: f (x) = 0, and g(x) 6= 0. Consider the simple example
where
u, t
E F G H
A B C D x
O
Referring to Figure 5, we divide the x ∼ t diagram into six regions by the characteristics
with B and C lying on the x axis at x = −b and +b, respectively. The solution in various
regions is:
ζ=0
in the wedge HCD. The spatial variation of u is plotted for several instants in Figure
5. Note that the wave fronts in both directions advance at the speed c. In contrast to
Case (i), disturbance persists for all time in the region between the two fronts.
2.4. REFLECTION FROM A CLIFF 9
1 1 Z x+ct
ζ= [f(x + ct) + f (x − ct)] + g(τ )dτ, x > ct. (4.2)
2 2c x−ct
But for x < ct, this result is no longer valid. To ensure that the boundary condition
is satisÞed we employ the idea of mirror reßection. Consider a Þctitious extension of the
sea to −∞ < x ≤ 0. If on the side x < 0 the initial data are imposed such that f (x)
and g(x) are even in x, then ζ(0, t) = 0 is assured by symmetry. We now have initial
conditions stated over the entire x axis
where
f(x) if x > 0
F (x) =
f(−x) if x < 0
g(x) if x > 0
G(x) =
g(−x) if x < 0.
These conditions are summarized in Figure 6. Hence the solution for 0 < x < ct is
2.5. FORCED WAVES IN AN INFINITE DOMAIN 10
Note that the point (ct − x, 0) on the x axis is the mirror reßection (with respect to the
cliff x = 0) of left tip (x − ct, 0) of the characteristic triangle . The effect of the initial
velocity in the region (0, ct − x) is doubled.
∂u ∂ζ ∂P
= −g −g (5.1)
∂t ∂x ∂x
The wave equation is now inhomogeneous
∂ 2ζ 2
2∂ u
= c + q(x, t) t > 0, |x| < ∞, (5.2)
∂t2 ∂x2
x+ct=x 0+ct 0
t
x-ct=x0-ct 0
( x0,t 0 )
x
( x 0-ct 0 ,0) O ( ct0-x 0 ,0) ( x 0+ct0 ,0)
Because of linearity, we can treat the effects of initial data separately. Let us therefore
focus attention only to the effects of persistent forcing and let the initial data be zero,
" #
∂ζ
ζ(x, 0) = 0, = 0, (5.3)
∂t t=0
Let us hide the parametric dependence on α for the time being. The general solution
to the the inhomogeneous second-order ordinary differential equation is
Z t
q̄(τ ) h i
ζ̄ = C1 ζ̄1 (t) + C2 ζ̄2 (t) + ζ̄1 (τ )ζ̄2 (t) − ζ̄2 (τ )ζ̄2 (t) dτ, (5.7)
0 W
Thus the observer is affected only by the forcing inside the characteristic triangle
deÞned by the two characteristics passing through (x, t).
For non-zero initial data ζ(x, 0) = f (x) and ζt (x, 0) = g(x), we get by linear super-
position the full solution of D’Alambert
1 1 Z x+ct
ζ(x, t) = [f (x + ct) + f(x − ct)] + dξg(ξ)
2 2c x−ct
1 Zt Z x+c(t−τ )
+ dτ dξ q(ξ, τ ), (5.10)
2c 0 x−c(t−τ )
Homework
2.6. STRONG SCATTERING BY A STEP 13
If a sinuuoidal wave train of frequency ω arrives from x ∼ −∞, how does the step
change the propagation ?
In each zone of contant depth (i = 1, 2, 3), the shallow water equations read:
∂ζi ∂ui
+ hi =0 (6.2)
∂t ∂x
∂ui ∂ζi
+g =0 (6.3)
∂t ∂x
A monochromatic wave of frequency ω can be written in the form
therefore,
∂Ui
−iωηi + hi =0 (6.5)
∂x
∂ηi
−iωUi + g =0 (6.6)
∂x
which can be combined to
d2 ηi ω
+ ki2 ηi = 0, where k= √ (6.7)
dx2 ghi
The most general solution is a linear combination of terms proportional to
Together with the time factor e−iωt , the Þrst term is a wave train propagating from
left to right, while the second from right to left. The free-surface displacement of the
incident wave therefore can be written as
where the amplitude is taken to be unity for brevity. At a junction, the pressure and
the ßux must be equal, hence we impose the following boundary consitions,
dη1 dη2
η1 = η2 , and h1 = h2 ,, x = −a; (6.9)
dx dx
dη2 dη3
η2 = η3 , and h2 = h1 ,, x = a. (6.10)
dx dx
Far from the step, sinusoidal disturbances caused by the presence of the step must be
outgoing waves. Physically, this so-called radiation condition implies that, to the left
of the step, there must be a reßected wavetrain travelling from right to left. To the
right of the step, there must be a transmitted wavetrain travelling from left to right.
Accordingly, the wave heights in each zone of constant depth are:
The reßection and transmission coefficients R and T as well as A and B are yet unknown.
Applying the matching conditions at the left junction, we get two relations
(1 − s2 ) sin2 2k2 a
|R|2 = (6.24)
4s2 + (1 − s2 )2 sin2 2k2 a
It is evident that |R|2 + |T |2 = 1, meaning that the total energy of the scattered waves
is equal to that of the incident wave.
Over the shelf the free surface is given by
h i
2s (1 + s)eik2 (x−a) + (1 − s)e−ik2 (x−a)
η2 = (6.25)
(1 + s)2 e−ik2 a − (1 − s)2 eik2 a
Recalling the time factor e−iωt , we see that the free surface over the shelf consists of two
wave trains advancing in oppopsite directions. Therefore along the shelf the two waves
can interfere each other constructively, with the crests of one coinciding with the crests
of the other at the same moment. At other places the interference is destructive, with
the crests of one wave train coinciding with the troughs of the other. The envelope of
energy on the shelf is given by
h i
4s2 cos2 k2 (x − a) + s2 sin2 k2 (x − a)
|η|2 = (6.26)
4s2 + (1 − s)2 sin2 2k2 a
At the downwave edge of the shelf, x = a, the envelope is
4s2
|η|2 = (6.27)
4s2 + (1 − s)2 sin2 2k2 a
2.7. REFRACTION 16
Note that the reßection and transmission coefficients are oscillatory in k2 a. In par-
ticular for 2k2 a = nπ, n = 1, 2, 3..., that is, 4a/λ = n, |R| = 0 and |T | = 1 ; the shelf
is transparent to the incident waves. It is the largest when 2k2 a = nπ, corresponidng
to the most constructive interference and the strongest transmission Mininum transmis-
sion and maximum reßection occur when 2k2 a = (n − 1/2)π, or 4a/λ = n − 1/2, when
the interference is the most destructive. The corresponding transimssion and reßection
coefficients are
4s2 (1 − s2 )2
min|T |2 = , max|R|2 = . (6.28)
(1 + s2 )2 (1 + s2 )2
See Þgure 8.
The features of interference can be explained physically. When a crest Þrst strikes
the left edge at x = a, part of the it is transmitted onto the shelf and part is reßected
towards x ∼ −∞. After reaching the right edge at x = a, the tranmitted crest has a
part reßected to the left and re-reßected by the edge x = −a to the right. When the
remaining crest arrives at the right edge the second time, its total travel distance is an
integral multple of the wave length λ2 , hence is in phase with all the crests entering
the shelf either before or after. Thus all the crests reinforce one another at the right
edge. This is constructive interference, leading to the strongest tranmission to the right
x ∼ ∞. On the other hand if 2k2 a = (n − 1/2)π or 4a/λ = n − 1/2, some crests will be
in opposite phase to some other crests, leading to the most destructive interference at
the right edge, and smallest transmission.
more slowly than the wave phase in x . Hence we try the solution
η = A(x)eiθ(x) (7.3)
dθ
k(x) = (7.4)
dx
and assume
dA
dx
= O(kL)−1 ¿ 1
kA
In fact we shall assume each derivative of h, A or k is µ times smaller than kh, kA or
k 2 . Futhurmore,
à ! " à ! à ! #
d dη ω2 dA d dA d(khA) ω 2 A iθ
h + η = ik ikh + h + h +i + e =0
dx dx g dx dx dx dx g
with A1 /A0 = O(µ), A2 /A0 = O(µ2 ), · · ·. From O(µ0 ) the dispersion relation follows:
ω
ω 2 = ghk 2 , or k = √ (7.6)
gh
Thus the local wave number and the local depth are related to frequency according
to the well known dispersion relation for constant depth. As the depth decreases, the
wavenumber increases. Hence the local phase velocity
ω q
c= = gh (7.7)
k
also decreases.
From O(µ) we get,
dA0 d(khA0 )
ikh +i =0
dx dx
or
d
(khA20 ) = 0 (7.8)
dx
which means
khA20 = C 2 = constant
or,
q q
ghA20 = constant = gh∞ A2∞ (7.9)
Since in shallow water the group velocity equals the phase velocity, the above result
means that the rate of energy ßux is the same for all x and is consistent with the
original assumption of unidirectional propagation. Furthermore, the local amplitude
increases with depth as
à !1/4
A0 (x) h∞
= (7.10)
A∞ h
This result is called Green’s law.
In summary, the leading order solution is
à !1/4 à !1/4 µ Z x ¶
h∞ iθ−iωt h∞ 0 0
ζ = A∞ e = A∞ exp i k(x )dx − iω (7.11)
h h
3.2 PROGRESSIVE WAVES OVER CONSTANT DEPTH 1
I-campus project
School-wide Program on Fluid Mechanics
Modules on Waves in ßuids
T. R. Akylas & C. C. Mei
CHAPTER THREE
DISPERSION OF SURFACE WATER WAVES
Except in very shallow water, one of the most outstanding physical properties of
sea surface waves is dispersion in that waves of different wavelengths propagate at
different speeds. If the waves are sufficiently steep, nonlinearity is also important. The
interplay of dispersion and nonlinearity gives rise to a host of new phenomena unfamiliar
in classical physics and makes surface water waves a perenial subject of fascination and
challenge. In this module, only dispersion of inÞnitesimal waves will be discussed.
Φ = f (z)eikx−iωt (1.1)
implying
Φ = B cosh k(z + h)eikx−iωt (1.5)
Upon integration,
Bk sinh kh ikx−iωt
ζ = Aeikx−iωt = e (1.8)
−iω
where A denotes the surface wave amkplitude, it follows that
−iωA
B=
k sinh kh
and
−iωA
Φ = cosh k(z + h)eikx−iωt
k sinh Ã
kh !
−igA T k 2 cosh k(z + h) ikx−iωt
= 1+ e (1.9)
ω gρ cosh kh
Clearly s
Tk
c≈ , if k/km À 1, or λ/λm ¿ 1 (1.18)
ρ
Thus for wavelengths much shorter than 1.7 cm, capillarity alone is important, These
are called the capillary waves. On the other hand
r
g
c≈ , if k/km ¿ 1, or λ/λm ¿ 1 (1.19)
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 beocmes large, there must be a minimum
for some intermediate k. From
dc2 g T
=− 2 + =0
dk k ρ
the minmum c occurs when
r
gρ
k= = km , or λ = λm (1.20)
T
3.2 PROGRESSIVE WAVES OVER CONSTANT DEPTH 4
c*=c/ c m
2
1.75
1.5
1.25
1
0.75
0.5
0.25 λ*= λ/ λ m
0
1 2 3 4 5
The smallest value of c is cm . For the intermediate range where both capillarity and
gravity are of comparble importance; the dispersion relation is plotted in Þgure (1).
Next we consider longer gravity waves where the depth effects are essential.
q
ω= gk tanh kh (1.21)
For 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
3.2 PROGRESSIVE WAVES OVER CONSTANT DEPTH 5
c/√ gh
1
0.75
0.5
0.25
0 kh
2 4 6 8 10
√
depth, gh, hence they are non-dispersive. The dispersion relation is plotted in Þgure
(2) for a wide range of wavelengths.
All dynamical quantities diminish exponentially to zero as kz → −∞. Thus the ßuid
motion is limited to the surface layer of depth O(λ). Gravity and capillary-gravity waves
are therefore surface waves.
For pure gravity waves in shallow water, T = 0 and kh ¿ 1, we get
gkA ikx−iωt
u = e (1.30)
ω
w = 0, (1.31)
∂Φ
p = −ρ = ρgAeikx−iωt = ρgζ (1.32)
∂t
Note that the horizontal velocity is uniform in depth while the vertical velocity is neg-
ligible. Thus the ßuid motion is essentially horizontal. The total pressure
P = po + p = ρg(ζ − z) (1.33)
is hydrostatic and increases linearly with depth from the free surface.
Letting
à !
a gkA T k 2 cosh k(zo + h)
= 2 1+ (1.41)
b ω cosh kh gρ sinh k(zo + h)
we get
x02 z 02
+ 2 =1 (1.42)
a2 b
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 diminshes
to zero at the seaebed, hence the ellipse collapses to a horizontal line segment. In deep
water, the major and minor axes are equal
à !
gkA T k 2 kzo
a=b= 2 1+ e , (1.43)
ω gρ
therefore the orbits are circles with the radius diminishing exponentially with depth.
Also we can rewrite the trajectory as
à !
0 gkA T k 2 cosh k(zo + h)
x = 1+ sin(ωt − kxo ) (1.44)
ω2 gρ cosh kh
à !
gkA T k 2 sinh k(zo + h) π
z0 = 1+ sin(ωt − kxo − ) (1.45)
ω2 gρ cosh kh 2
When ωt − kxo = 0, x0 = 0 and z 0 = b. A quarter period later, ωt − ko = π/2, x0 = a
and z 0 = 0. Hence as time passes, the particle traces the ellptical orbit in the clockwise
direction.
3.2 PROGRESSIVE WAVES OVER CONSTANT DEPTH 8
ρZ 0 ³ 2 ´
Ēk = dz u + w 2 (1.46)
2 −h
Then
à !2 à !2 Z 0
ρ gkA T k2 1 h i
Ēk = 1+ 2 dz cosh2 k(z + h) + sinh2 k(z + h)
4 ω gρ cosh kh −h
à !2 à !2 à !2 à !2
ρ gkA T k2 sinh 2kh ρ gkA T k2 sinh kh
= 1+ 2 = 1+
4 ω gρ 2k cosh kh 4 ω gρ k cosh kh
à !2 à !
ρgA2 T k2 gk tanh kh ρgA2 T k2
= 1+ = 1 + (1.51)
4 gρ ω2 4 gρ
Note that the total energy is equally divided between kinetic and potential energies; this
is called the equipartition of energy.
3.3. DISPERSION OF TRANSIENT DISTURBANCE 9
We leave it as an exercise to show that the power ßux (rate of energy ßux) across a
station x is
dĒ Z 0 Z 0
= pu dz − T ζx ζt = −ρ Φt Φx dz − T ζx ζt = Ēcg (1.54)
dt −h −h
where s
g T k3
c= + (1.61)
k ρ
Note that cg = c when k = km , and
> >
cg < c, if k < km (1.62)
In the limit of pure capillary waves of k À km , cg = 3c/2. For pure gravity waves
cg = c/2 as in (1.58).
3.3. DISPERSION OF TRANSIENT DISTURBANCE 10
∂Φ ∂ 2ζ
−ρgζ − ρ + T 2 = 0, z = 0. (2.2)
∂t ∂x
The kinematic condition requires
ζt = Φz , z = 0. (2.3)
∂2Φ ∂Φ T ∂ 3 Φ
+ g − = 0, z=0 (2.4)
∂t2 ∂z ρ ∂x2 ∂z
Φx , Φz → 0, z → −∞. (2.5)
Since conditions (2.3) and (2.2) involve Þrst-order time derivatives, we must prescribe
the initial data for Φ(x, 0, 0) and ζ(x, 0) on the free surface. Physically Φ(x, 0, 0) is
equivalent to an impulsive pressure applied on the free surface. Here we shall only
illustrate the effects of a prescribed initial displacement of the free surface,
Φ(k, 0, 0) = 0 (2.16)
A(k, 0) = 0,
à ! à !
T k2 T k2
At (k, 0) = −ζ(k, 0) g + = −ζ o (k) g +
ρ ρ
Hence à !
T k2 sin ωt
A = −ζ o (k) g +
ρ ω
The solution for the tranformed potential is
à !
T k2 sin ωt |k|z
Φ(k, z, t) = −ζ o (k) g + e (2.17)
ρ ω
The transform of the surface displacement is
Φt (k, 0, t)
ζ(x, t) = − 2 = ζ o (k) cos ωt (2.18)
g + Tρk
By Fourier inversion the solutions are given by (2.8) and (2.9).
To be concrete we shall take
Sb
ζo (x) =
π(x3 + b2
which is a hump of area S; the Fourier transform is
ζ o (k) = Se−|k|b
for real f and very large t. Let us Þrst give a quick derivation of the mathematical
result.
If t is large, then as k increases along the path of integration both the real and
imaginary parts of the exponential function
df (ko )
= f 0 (ko ) = 0, a < ko < b. (2.21)
dk
Then
1
f (k) = f (ko ) + (k − ko )2 f 00 (ko ) + · · ·
2
and
With an error of O(1/t), we also replace the limits of the last integral by ±∞; the
justiÞcation is omitted here. Now it is known that
Z ∞ r
2 π ±iπ/4
e±itk dk = e
−∞ t
3.3. DISPERSION OF TRANSIENT DISTURBANCE 14
It follows that
Z b " #1/2 µ ¶
itf (k) itf (ko )±iπ/4 2π 1
I(t) = F (k)e dk ≈ F (ko )e 00
+O , if ko ∈ (a, b),
a t|f (ko )| t
(2.22)
where the sign is + (or −) if f 00 (ko ) is positive (or negative). It can be shown that if
ther is no stationary point in the range(a,b), then the integral I(t) is small
µ ¶
1
I(t) = O , if ko ∈
/ (a, b). (2.23)
t
which is plotted in Figure 4. Note that cg (k) is large for both small and large k :
r
1 g
cg ∼ , for small k,
2 k
and s
1 Tk
cg ∼ , for large k,
2 ρ
Hence there is a minimum cg occuring at ko where ω 00 (ko ) = 0.
For any speed x/t > min cg , eq. (2.27) has two roots (stationary points). At the
smaller root k1 < ko , ω 00 (ko ) < 0; at the larger one k2 > ko , ω 00 (ko ) > 0 . Adding the
contributions from both we get the Þnal result for the right-going wave
s
1 −k1 b 2π
ζ+ (x, t) ∼ e 00
cos (k1 x − ω(k1 )t − π/4)
2π t|ω (k1 )|
s
1 −k2 b 2π
+ e 00
cos (k2 x − ω(k2 )t + π/4) (2.28)
2π t|ω (k2 )|
Physically an observer travelling at the speed x/t sees two trains of simple harmonic
waves with wavenumbers k1 and k2 , corresponding respectively to gravity (longer) and
capillary (shorter) waves. The local wavelengths are such that their group velocities
match the speed of the observer. The faster the observer, the shorter the capillary
waves and the longer the gravity waves. If a snapshot is taken for all x > 0, then the
longer gravity waves are at the very front, followed by shorter and shorter gravity waves.
However the shortest capillary waves lead the longer ones, see Þgure 5. Because e−kb
3.3. DISPERSION OF TRANSIENT DISTURBANCE 16
is is the greatest at k = 0, the longest waves are the biggest. The entire disturbance
attenuates in time as t−1/2 .
Note also that for x/t ≈ min cg , the second derivative f 00 (ko ) = −ω 00 (ko ) = 0. Hence
the asymptotic formula breaks down. A better approximation is needed, and is left as
an exercise.
Homework Show by expanding
1
ω(k) = ω(ko ) + (k − ko )ω 0 (ko ) + (k − ko )3 ω 000 (ko ) + · · · (2.29)
6
that for large x and t,
à !1/3
2
ζ+ (x, t) ≈ 00
e−ko b Ai (Z) cos(ω(ko ) − ko x) (2.30)
tω (ko )
where Ai(Z) is the Airy function with the argument
à !1/3
2
Z= 000
(cg (ko )t − x))
ω (ko )t
It can be deÞned by the integral
à !
1Z∞ α3
Ai(−Z) = cos −Zα + dα (2.31)
π 0 3
and is related to Bessel function of order ±1/3,
à √ ! q
Z 1/2
2 Z 3/2 2 (Z) 3/2
Ai(−Z) = J 1 Z + J− 1 Z (2.32)
3 3 3 3 3
dx
= ω 00 (k1 )dk1
t
Therefore the total energy between two observers moving at the local group velocity
remains the same for all time. In other words, waves are transported by the local group
velocity even in transient dispersion.
where ∙ µ ¶ ¸
Z ∞
1
A(x, t) = dk A(k) exp i∆kx − i ωo0 ∆k
+ ωo00 (∆k)2 t + · · · (3.3)
0 2
Although the integration is formally extends from 0 to ∞, the effective range is only
from ko − (∆k)m to ko + (∆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 ωo and
wave number ko , 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, (3.3) reduces to
Z ∞
A(x, t) = dk A(k) exp [i∆k(x − ωo0 t)] (3.4)
0
Clearly A = A(x − ωo0 t). Thus the envelope itself is a wave traveling at the speed ωo0 .
This speed is called the group velocity,
¯
dω ¯¯
cg (ko ) = ¯ (3.5)
dk ¯ko
Note that the characteristic length and time scales are (∆km )−1 and (ωo0 ∆km )−1 respec-
tively, therefore much longer than those of the component waves : ko−1 and ωo−1 . In other
words, (3.3) is adequate for the slow variation of Ae in the spatial range of ∆km x = O(1)
and the time range of ωo0 ∆km t = O(1).
As a speciÞc example we let the amplitude spectrum be a real constant within the
narrow band of ko − κ, ko + κ,
Z ko +κ
ζ=A eikx−iω(k)t dk, κ ¿ ko (3.6)
ko −κ
then
Z κ
ζ = ko Aeiko x−iωo t dξeiko ξ(x−cg t) + · · ·
−κ
2A sin κ(x − cg t) iko x−iωo t
= e = Aeiko x−iωo t (3.7)
x − cg t
3.3. DISPERSION OF TRANSIENT DISTURBANCE 19
Z ∞
∂A
= (i∆k)A(k)eiS dk
∂x 0
Z ∞³ ´
∂ 2A 2
= −(∆k) A(k)eiS dk
∂x2 0
where
1
S = ∆k x − ωo0 ∆k t − ωo00 (∆k)2 t (3.11)
2
is the phase function. It can be easily veriÞed that
∂A ∂A iω 00 ∂ 2 A
+ ωo0 = o (3.12)
∂t ∂x 2 ∂x2
By keeping the quadratic term in the expansion, (3.12) is now valid for a larger spatial
range of (∆k)2 x = O(1). In the coordinate system moving at the group velocity, ξ =
x − cg t, τ = t, we easily Þnd
∂A(ξ, τ ) ∂A ∂A ∂A(ξ, τ ) ∂A
= − cg , =
∂t ∂τ ∂ξ ∂x ∂x
∂A iωo00 ∂ 2 A
= (3.13)
∂τ 2 ∂ξ 2
By manipulations similar to those leading to (3.10), we get
à !
∂|A|2 iω 00 ∂ ∂A ∂A∗
= o A∗ −A (3.14)
∂τ 2 ∂ξ ∂ξ ∂ξ
Thus the local energy density is not conserved over a long distance of propagation.
Higher order effects of dispersion redistribute energy to other parts of the envelope.
For either a wave packet whose envelope has a Þnite length ( A(±∞) = 0), or for a
periodically modulated envelope (A(x) = A(x + L)), we can integrate (3.14) to give
∂ Z
|A|2 dξ = 0 (3.15)
∂τ
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.
WAVE RESISTANCE OF A TWO-DIMENSIONAL OBSTACLE 1
I-campus project
School-wide Program on Fluid Mechanics
Modules on Waves in ßuids
T. R. Akylas & C. C. Mei
CHAPTER FOUR
WAVES DUE TO MOVING DISTURBANCES
2 D SHIP WAVES
RADIATION DUE TO OSCILLATING PRESSURE
SHIP WAVES
[Ref]: Lecture notes on Surface Wave Hydrodynamics by Theodore T.Y. WU, Calif.
Inst.Tech.
Lighhill, Waves in Fluids
and
1
λ∗1 = (1.3)
λ∗2
Thus, as long as c∗ = U∗ > 1 two wave trains are present: the longer gravity wave
with length λ∗1 , and the shorter capillary wave with length λ∗2 . Since cg 1 < c = U and
cg 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 ßux in both wave trains, we
get
Rc = (c − cg 1 )Ē1 + (cg 2 − c)Ē2 (1.4)
Recalling that
cg 1 λ2∗ + 3
=
c 2 λ2∗ + 1
we Þnd, Ã !
cg 1 2 1 1/λ∗ 1 1/λ∗
1− =1− 1+ 2 = − = −
c 2 λ∗ + 1 2 λ∗ + 1/λ∗ 2 2c2
For the longer wave we replace cg by cg ∗1 cm and λ∗1 in the preceding equation, and use
(1.2), yielding
cg ∗ 1 ³ ´1/2
1− = 1 − c4∗ (1.5)
c∗
Similarly we can show that
cg ∗2 ³ ´1/2 cg
− 1 = 1 − c4∗ = 1 − ∗1 (1.6)
c∗ c∗
Since à ! à !
ρgA21 1 ρgA21 1 1
Ē1 = 1+ 2 = λ∗1 + = ρgA21 λ∗2 c2∗ , (1.7)
2 λ∗ 1 2 λ∗1 λ∗1
we get Þnally
1 ³ ´ 1 ³ ´
R = ρg λ∗2 A21 + λ∗1 A22 (c2∗ − 1)1/2 = ρg λ∗2 A21 + λ∗1 A22 (U∗2 − 1)1/2 (1.8)
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 inÞnite train of waves; the wave resistance vanishes.
When U∗ < 1, no waves are generated; thedisturbance 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 A1 , A2 which are affected by the size (relative to the
wavelengths), shape and depth of submergence.
RADIATION OF SURFACE WAVES BY AN OSCILLATING PRESSURE 3
When the speed is sufficiently high, pure gravity waves are generated behind the
body. Power balance then requires that
µ ¶ Ã !
cg ρgA2 1 2kh
R= 1− Ē = − (1.9)
U 2 2 sinh 2kh
The wavelength generated by the moving body is given implicitly by
à !1/2
U tanh kh
√ = (1.10)
gh kh
√ √
When U ≈ gh the waves generatesd are very long, kh ¿ 1, cg → c = gh, and the
√
wave resistance drops to zero. When U ¿ gh, the waves are very short, kh À 1,
ρgA2
R≈ (1.11)
4
For intermediate sppeds the dependence of wave resistance on speed is plotted in Þgure
(1).
φz = ζt , z = 0 (2.2)
pa
+ φt + gζ = 0 (2.3)
ρ
where pa is the prescribed air pressure. Eliminating the free surface displacement we
get
(pa )t
φtt + gφz = − , z = 0. (2.4)
ρ
Let us consider only sinusoidal time dependence:
pa = P (x)e−iωt (2.5)
and assume
φ(x, z, t) = Φ(x, z)e−iωt , ζ(x, t) = η(x)e−iωt (2.6)
Φz = −iωη, z = 0 (2.8)
and
ω2 iω
Φz − Φ = P (x), z = 0. (2.9)
g ρg
DeÞne the Fourier tranoform and its inverse by
Z ∞
1 Z∞
f¯(α) = dx f (x), f (x) = dα f¯(α), (2.10)
−∞ 2π −∞
We then get the transforms of (2.1) and (2.4)
subject to
ω2 iω
Φ̄z − Φ̄ = P̄ , z = 0. (2.12)
g ρg
The solution Þnite at z ∼ −∞ for all α is
Φ̄ = Ae|α|z
ω2 iω P̄
|α|A − A=
g ρg
hence
iω P̄
ρg
A=
|α| − ω 2 /g
or
1 Z∞
iωP̄
ρg
Φ= dαeiαx e|α|z
2π −∞ |α| − ω 2 /g
Z ∞ Z ∞
iω 1 iαx |α|z ! 1
= dαe e dx0 e−iαx P (x0 ) ,
ρg 2π −∞ −∞ |α| − ω 2 /g
iω Z ∞ 0 0 1
Z ∞
! 1
= dx P (x ) dα eiα(x−x ) e|α|z (2.13)
ρg −∞ 2π −∞ |α| − ω 2 /g
Let
ω2
k= (2.14)
g
we can rewrite (2.13) as
iω Z ∞ 0 1Z∞ cos(α(x − x0 ))
Φ= dx P (x0 ) dα eαz (2.15)
ρg −∞ π 0 α−k
The Þnal formal solution is
If we chose
P (x0 ) = Po δ(x0 ) (2.17)
then
iωPo 1 Z ∞ cos(αx)
Φ → G(x, z) = dα eαz (2.18)
ρg π 0 α−k
RADIATION OF SURFACE WAVES BY AN OSCILLATING PRESSURE 6
is clearly the response to a concentrated surface pressure and the response to a pressure
distribution (2.16) can be written as a superposition of concentrated loads over the free
surface,
Z ∞
φ= dx0 P (x0 )G(x − x0 , z). (2.19)
−∞
where
iωPo −iωt 1 Z ∞ cos(αx)
G(x, z, t) = e dα eαz (2.20)
ρg π 0 α−k
In these results, e.g., (2.20), the Fourier integral is so far undeÞned since the integrand
has a real pole at α = k which is on the path of integration. To make it mathematically
deÞned we can chose the principal value, deform the contour from below or from above
the pole as shown in Þgure (2). This indeÞniteness is due to the assumption of quasi
steady state where the inßuence of the initial condition is no longer traceable. We must
now impose the radiation condition that waves must be outgoing as x → ∞. This
condition can only be satisÞed if we deform the contour from below. Denoting this
contour by Γ, we now manipulate the integral to exhibit the behavior at inÞnity, and to
verify the choice of path. For simplicity we focus attention on G. Due to symmetry, it
suffices to consider x > 0. Rewriting,
iωPo −iωt 1
G(x, z, t) = e (I1 + I2 )
ρg 2π
" #
iωPo −iωt 1 Z αz eiαx e−iαx
= e dα e + (2.21)
ρg 2π Γ α−k α−k
Consider the Þrst integral in (2.21). In order that the Þrst integral converges for
large |α|, we close the contour by a largfe circular arc in the upper half plane, as shown
in Þgure (3), where =α > 0 along the arc. The term
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 Þgure (4).
RADIATION OF SURFACE WAVES BY AN OSCILLATING PRESSURE 7
The contribution by the circular arc C vanishes by Jordan’s lemma. The left hand side
is
LHS = 2πieikx ekz (2.23)
by Cauchy’s residue theorem. By the change of variable α = iβ, the right hand side
becomes
Z 0
e−βx eiβz
RHS = I1 + i dβ
∞ iβ − k
KELVIN’S SHIP WAVE 8
Hence
Z ∞
e−βx eiβz
I1 = 2πieikx ekz + i dβ (2.24)
0 iβ − k
Now consider I2
Z
e−iαx eαz
I2 = dα (2.25)
Γ α−k
and the contour integral along the contour closed in the lower half plane,
I Z ∞
e−iαx eαz e−iαx eαz
− dα = I2 + 0 + dα
α−k 0 α−k
Again no contribution comes from the circular arc C. Now the pole is outside the
contour hence LHS = 0. Let α = −iβ in the last integral we get
Z ∞
e−βx e−iβy
I2 = −i dβ (2.26)
0 −iβ − k
Adding the results (2.24) and (2.26).,
Z ∞ Ã !
ikx kz ie−βxeiβz ie−βxe−iβz
I1 + I2 = 2πie e + dβ −
0 iβ − k −iβ − k
Z ∞ −βx
e
= 2πieikx ekz + 2 dβ 2 (β cos βy + k sin βy) (2.27)
0 β + k2
The Þrst term gives an outgoging waves. For a concentrated load with amplitude
Po , the displacement amplitude is Po /ρg. The integral above represent local effects
important only near the applied pressure. If the concentrated load is at x = x0 , one
simply replaces x by x − x0 everywhere.
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 Þrst two coordinate systems. The Þrst r = (x, y, z) moves with ship at the
uniform horizontal velocity U. The second r0 = (x0 , y 0 , z) is Þxed 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,
r0 = r + Ut (3.1)
in the stationary coordinates. Therefore the apparent frequency in the moving coordi-
nates is
σ =ω−k·U (3.4)
The last result is essentially the famous Doppler’s effect. 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 ßuid parcel on the ship’s path radiates waves in all directions and at all frequencies.
Wave of frequency ω spreads out radially at the phase speed of c = g/ω 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
σ = 0, (3.5)
i.e.,
ω k
ω = k · U, or c = = ·U (3.6)
k k
KELVIN’S SHIP WAVE 10
M2 P2 .
We now reverse the process, as shown in Þgure 8. 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 S1 and S2 . We then draw from P two
lines parallel to M S1 and M S2 , the two points of intersection Q1 an Q2 on the x axis
are just the two points of dependence.
Let 6 P Q1 O = 6 M S1 O = θ1 and 6 P Q2 O = 6 M S2 O = θ2 . then
P Si P Si
tan(θi + β) = = = 2 tan θi i = 1, 2.
M Si P Qi /2
hence
tan θi + tan β
2 tan θi =
1 − tan θi tan β
which is a quadratic equation for θi , with two solutions:
q
tan θ1
1± 1 − 8 tan2 β
= (3.8)
tan θ2
4 tan β
These two angles deÞne the local stationary wave crests crossing P , and they must
be perpendicular to P Q1 and P Q2 . There are no solutions if 1 − 8 tan2 β < 0, which
corresponds to sin β > 1/3 or β > 19.5◦ , i.e., outside the wake. At the boundary of the
q
wake, β = 19, 5◦ and tan β = 1/8, the two angles are equal
√
−1 2
θ1 = θ2 = tan = 55◦ . (3.10)
2
KELVIN’S SHIP WAVE 13
By connecting these segments at all points in the wedge, one Þnds two systems of wave
crests, the diverging waves and the transverse waves, as shown in Þgure div-trans.
Knowing that waves are conÞned 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 ßux
hence
A ∼ r1/2 (3.12)
This estimate is valid throughout the wedge except near the outer boundaries, where
A ∼ r−1/3 (3.13)
I-campus project
School-wide Program on Fluid Mechanics
Modules on Waves in ßuids
T. R. Akylas & C. C. Mei
CHAPTER FIVE
REFLECTION, TRANSMISSION, AND DIFFRACTION
Scope:
Reßection of sound at an interface. ( Reference : Brekhovskikh and Godin §.2.2.)
Diffraction by a circular cylinder, theory and simulation.
Diffraction by a wedge
- Parabolic approximation
- Exact theory and Numerical simulation.
1 ∂ 2φ
= ∇2 φ (1.1)
c2 ∂t2
where c denotes the sound speed. Recall that the ßuid pressure p = −ρo ∂φ/∂t also
satisÞes the same equation.
We Þrst generalize the plane sinusoidal wave in three dimensional space
where n is the unit vector in the direction of k. Here the phase function is
θ(x, t) = k · x − ωt (1.3)
INTRODUCTION TO TWO DIMENSIONAL SCATERING 2
The equation of constant phase θ(x, t) = θo describes a moving surface. The wave
number vector k = kn is deÞned to be
k = kn = ∇θ (1.4)
hence is orthogonal to the surface of constant phase, and represens the direction of wave
propagation. The frequency is deÞned to be
∂θ
ω=− (1.5)
∂t
∇2 φ = ∇ · ∇φ = ik · ikφ = −k 2 φ
∂2φ
= −ω 2 φ
∂t2
Hence (1.1) is satisÞed if
ω = kc (1.6)
Scattering by an object has been the focus of research in physics, electrical, acousti-
cal and oceanographical engineering for a long time. Depending on the geometry, the
mathematics can be quite involved. Mountains of literatures on analytical and numerical
methods can be found. In this chapter we shall limit our study to two space dimensions.
For a plane sound wave of single frequency scattered by a cylinder whose axis is parallel
to the incident crests, the two-dimensional, time-dependent potential can be written as
h i
Φ(x, y, t) = < φ(x, y)e−iωt (1.7)
∂ 2 φ ∂ 2 φ2 ω
∇2 φ + k 2 φ = + + k 2 φ = 0, k= (1.8)
∂x2 ∂y 2 c
On the rigid and perfectly reglective boundary B the normal velocity vanishes,
∂φ
=0 (1.9)
∂n
INTRODUCTION TO TWO DIMENSIONAL SCATERING 3
Let the total wave be the sum of the incident and scattered waves
φ = φI + φS (1.10)
then the scattered waves must further satisfy the radiation condition at inÞnity, i.e., it
can only radiate energy outward from the scatterer.
Th preceding boundary value problem goeverns wave scattering in a variety of physi-
cal contexts. Elastic shear waves scattered by a cylinderical cavity waves is one example.
The scattering of surface water waves in a sea of constant depth is in principle thee di-
mensional, yet it can be reduced to the same two-dimensional boundary value problem
above, if the scatterer (a breakwater, a storage tank , etc.,) has vertical side walls
extending the entire water depth. Let us explain why.
Conder a vertical cylindrical structure of of arbitrary plan form in a sea of constant
depth h. A train of monochromatic waves is incident from inÞnity at the angle α with
respect the x axis. The still water surface is in the x, y plane.
In the water region deÞned by 0 ≥ z ≥ −h, the velocity potential Φ(r, θ, z, t) must
satisfy the Laplace equation,
∂ 2 Φ 1 ∂Φ 1 ∂2Φ ∂2Φ
+ + + 2 =0 (1.11)
∂r2 r ∂r r2 ∂θ 2 ∂z
and subject to the linearized free surface boundary conditions
∂Φ
= −gζ (1.12)
∂t
∂ζ ∂Φ
= (1.13)
∂t ∂z
which can be combined to give
∂2Φ ∂Φ
2
+g = 0, z = 0. (1.14)
∂t ∂z
Along the impermeable bottom and coasts, the no ßux boundary conditions are
∂Φ
=0 on z = −h (1.15)
∂z
∂Φ
=0 at θ = 0 and νπ (1.16)
∂θ
REFELCTION AND TRANSMISSION ACROSS AN INTERFACE 4
and π + α is the angle of incidence measured from the x axis. A0 is the incident wave
amplitude.
Because of the vertical side-walls, we assume
cosh k(z + h) −iωt
Φ(r, θ, z, t) = A0 φ(x, y, t) e (1.19)
cosh kh
where φ(x, y, t) is the horizontal pattern of the potential normlized for an incident wave
of unit amplitude, and is related to the amplitude of the free surface displacement η(x, y)
by
igη(x, y)
φ(x, y) = − (1.20)
ω
Substituting (1.19) into the Laplace equation and using both the kinematic and
dynamic boundary conditions on the free surface, the Laplace equation in (x, y, z) is
then reduced to the two dimensional Helmholtz equation in (x, y),
∂2φ ∂2φ
+ + k 2 φ = 0, (x, y) in the ßuid. (1.21)
∂x2 ∂y 2
The no normal ßux boundary condition on the rigid vertical wall B becomes
∂φ
n · ∇η = = 0, (x, y) ∈ B (1.22)
∂n
since the normal to the cylinder wall is horizontal. Therefore the three-dimensional
water-wave problem is mathematically equivalent to the two-dimension sound problem,
arive from z > 0 at the incident angle of θ with respect to the z axis,
implying that
ki = (kxi , kzi ) = k(sin θ, − cos θ) (2.2)
In the lower medium z < 0 the transmitted wave has the pressure
and
[w] = 0 z = 0, (2.7)
where the square brackets signify the jump across the interface:
p
Z =− (2.9)
w
∂w ∂p
ρ = −iωρw = − , (2.10)
∂t ∂z
REFELCTION AND TRANSMISSION ACROSS AN INTERFACE 6
p −iωρp
= − ∂p (2.11)
w ∂z
It follows from the two continuity requirements that the impedance must be continuous
[Z] = 0 z = 0 (2.12)
Note Þrst that to satisfy the conditions of continuity for all x it is necessary that the y
factors match, so that
k sin θ = k1 sin θ1 (2.13)
or
sin θ sin θ1
= (2.14)
c c1
Eq. (2.13) or (2.14) is the famous Snell’s law of refraction. If c1 < c, waves are incident
from the faster medium, the direction of the refracted (or transmitted) wave is closer to
the normal to the interface. Now (2.6) requires that
1+R=T (2.15)
The impedance of the incident wave, the reßected wave, and the transmitted waves
are respectively
ρc ρc ρ1 c1
Zi = , Zr = − , Z1 = (2.16)
cos θ cos θ cos θ1
which are all constants, and the total impedance on the incidence/reßection side is
ρc exp(−2ikz cos θ) + R
Z= (2.17)
cos θ exp(−2ikz cos θ) − R
which is in general a complex function of z. Next we impose (2.6) and get
ρc 1 + R
Z1 = (2.18)
cos θ 1 − R
hence
Z1 cos θ − ρc
R= (2.19)
Z1 cos θ + ρc
This formula is written in a general form where the impedance of the lower medium can
be anything . For the present example it is given by (2.16) and
ρ1 c1 cos θ − ρc cos θ1
R= (2.20)
ρ1 c1 cos θ + ρc cos θ1
REFELCTION AND TRANSMISSION ACROSS AN INTERFACE 7
Let
ρ1 c
m= , n= (2.21)
ρ c1
where the ratio of sound speeds n is called the index of refraction. We get after using
Snell’s law that
q
sin2 θ
m cos θ − n cos θ1 m cos θ − n 1 − n2
R= = q (2.22)
m cos θ + n cos θ1 m cos θ + n 1 − sin2 θ
n2
2m cos θ
T =1+R= q (2.23)
sin2 θ
m cos θ + n 1 − n2
sin θc = n (2.24)
beyond which (θ > θc ) the square roots above become imaginary. We must then take
s s
sin2 θ sin2 θ
cos θ1 = 1− =i −1 (2.25)
n2 n2
with |R| = 1, implying complete reßection. As a check the transmitted wave is now
given by ∙ µ ¶¸
q
2
pt = T exp k1 ix sin θ1 + z sin θ/n2 −1 (2.27)
so the amplitude attenuates exponentially in z as z → −∞. Thus the wave train cannot
penetrate much below the interface.
The dependence of R on various parameters is best displayed in the complex plane
R = <R + i=R.
Case 1: n > 1. Here R is always real.
REFELCTION AND TRANSMISSION ACROSS AN INTERFACE 8
iIm(R) iIm(R)
(1.a) (1.b)
iIm(R) iIm(R)
It can be shown (see Appendix A) that the plane wave can be expanded in Fourier-Bessel
series :
∞
X
ikr cos(θ−θo )
e = ²n in Jn (kr) cos n(θ − θo ) (3.6)
n=0
²0 = 0, ²n = 2, n = 1, 2, 3, . . . (3.7)
φS (r, θ) = R(r)Θ(θ)
we Þnd
r2 R00 + rR0 + (k 2 r2 − n2 )R = 0, and Θ00 + n2 Θ = 0
where n = 0, 1, 2, . . . are eigenvalues in order that Θ is periodic in θ with period 2π. For
each eigenvalue n the possible solutions are
where Hn(1) (kr), Hn(2) (kr) are Hankel functions of the Þrst and second kind, related to
the Bessel and Weber functions by
Hn(1) (kr) = Jn (kr) + iYn (kr), Hn(2) (kr) = Jn (kr) − iYn (kr) (3.8)
For large radius the asymptotic form of the Hankel functions behave as
s s
2 i(kr− π − nπ ) 2 −i(kr− π − nπ )
Hn(1) ∼ e 4 2 , Hn(2) ∼ e 4 2 (3.10)
πkr πkr
In conjunction with the time factor exp(−iωt), Hn(1) gives outgoing wavew while Hn(2)
give incoming waves. To satisfy the radiation condition, we must discard all terms
involving Hn(2) . From here on we shall abbreviate Hn(1) simply by Hn . The scattered
wave is now
∞
X
φS = A (An sin nθ + Bn cos nθ) Hn (kr) (3.11)
n=0
The expansion coefficients (An , Bn ) must be chosen to satisfy the boundary condition
on the cylinder surface. Without loss of generality we can take θo = 0. On the surface
of the cylindrical cavity r = a, we impose
∂φI ∂φS
+ = 0, r = a
∂r ∂r
It follows that An = 0 and
and " #
∞
X J 0 (ka)
Φ = Ae −iωt
en i n
Jn (kr) − n0 Hn (kr) cos nθ (3.13)
n=0 Hn (ka)
The numerical simulations can be seen for a wide range of ka on the web (give web
link).
Of practical interest is the angular variation of pressure on the surface of the cylinder
r = a. Figure 2 shows that for small ka the pressure is relatively uniform in all directions.
For increasingly large ka, waves become stronger on the reßection side (reaching 2 at
the back θ = π). On the shadow side the wave intensity is weak.
SCATTERING BY A CIRULAR CYLINDER 12
150 30
0.5
180 0
210 330
240 300
270
(b) ka= 1
90
2
120 60
1.5
150 1 30
0.5
180 0
210 330
240 300
270
SCATTERING BY A CIRULAR CYLINDER 13
1.5
150 1 30
0.5
180 0
210 330
240 300
270
(d) ka= 5
90
2
120 60
1.5
150 1 30
0.5
180 0
210 330
240 300
270
SCATTERING BY A CIRULAR CYLINDER 14
(e) ka= 10
90
2
120 60
1.5
150 1 30
0.5
180 0
210 330
240 300
270
(f) ka= 15
90
2
120 60
1.5
150 1 30
0.5
180 0
210 330
240 300
270
It is also interesting to examine certain limits. For very long long waves ka ¿ 1 the
expansions for Bessel functions for small argument may be used,
xn 2 2n (n − 1)!
Jn (x) ∼ , Yn (x) ∼ log x, Yn (x) ∼ (3.14)
2n n! π πxn
Then the scattered wave has the potential
which is one way of stating the radiation condition for two dimensional scattered waves.
The far Þeld pattern of |A(θ)|2 for various ka can be numerically computed as shown in
Þg (3.1).
SCATTERING BY A CIRULAR CYLINDER 16
0.2
150 30
0.1
180 0
210 330
240 300
270
(b) ka= 1
90
1
120 60
0.8
0.6
150 30
0.4
0.2
180 0
210 330
240 300
270
SCATTERING BY A CIRULAR CYLINDER 17
(c) ka= 2
90
2.5
120 60
2
1.5
150 30
1
0.5
180 0
210 330
240 300
270
(d) ka= 3
90
6
120 60
150 30
180 0
210 330
240 300
270
SCATTERING BY A CIRULAR CYLINDER 18
(e) ka= 5
90
20
120 60
15
150 10 30
180 0
210 330
240 300
270
(f) ka= 10
90
100
120 60
80
60
150 30
40
20
180 0
210 330
240 300
270
Let us apply Green’s formula to φ1 and φ2 over a closed area bounded by a closed
contour C,
ZZ ³ ´ Z Ã ! Z Ã !
2 2 ∂φ1 ∂φ2 ∂φ1 ∂φ1
φ2 ∇ φ1 − φ1 ∇ φ2 dA = φ2 − φ1 ds + ds φ2 − φ1 ds
S B ∂n ∂n C ∂n ∂n
where n refers to the unit normal vector pointing out of S. The surface integral vanishes
on account of the Helmholtz equation, while the line integral along the cavity surface
vanishes by virture of the boundary condition, hence
Z Ã !
∂φ1 ∂φ2
ds φ2 − φ1 ds = 0 (4.5)
C ∂n ∂n
This mathematical result implies the conservation of energy. Physically, across any circle
the net rate of energy ßux vanishes, i.e., the scattered power must be balanced by the
incident power.
Making use of (4.4) we get
s
Z 2π
2
0== rdθ eikr cos(θ−θo ) + Ao (θ)eikr−iπ/4
0 πkr
s
2 ∗
· −ik cos(θ − θo )eikr cos(θ−θo ) − ik A (θ)eikr−iπ/4
πkr o
Z 2π ½
2
== rdθ −ik cos(θ − θo ) + (−ik)|Ao |2
0
s
πkr
ikr[cos θ−θo )−1]+iπ/4 2 ∗
+e (−ik) A
πkr o
s
2
+ e−ikr[cos θ−θo )−1]−iπ/4 (−ik) cos(θ − θo ) Ao
πkr
The Þrst term in the integrand gives no contribution to the integral above because of
periodicity. Since =(if ) = =(if ∗ ), we get
2 Z 2π
0=− |Ao (θ)|2 dθ
π 0 s
Z 2π
2
+= rdθ Ao (−ik) [1 + cos(θ − θo )]eiπ/4 eikr(1−cos(θ−θo ))
0 πkr
2 Z 2π
=− |Ao (θ)|2 dθ
π
0 s
−iπ/4 2 Z 2π
−< e Ao (k)r dθ[1 + cos(θ − θo )]eikr(1−cos(θ−θo ))
πkr 0
For large kr the remaining integral can be found approximately by the method of sta-
tionary phase (see Appendix B), with the result
s
Z 2π
2π iπ/4
dθ[1 + cos(θ − θo )]eikr(1−cos(θ−θo )) ∼ e (4.8)
0 kr
THIN BARRIER AND PARABOLIC APPROXIMATION 21
We get Þnally
Z 2π
|A|2 dθ = −2<A(θo ) (4.9)
0
Thus the total scattered energy in all directions is related to the amplitude of the
scattered wave in the forward direction. In atomic physics, where this theorem was
originated (by Niels Bohr), measurement of the scattering amplitude in all directions is
not easy. This theorem suggests an econmical alternative.
Homework For the same scatterer, consider two scattering problems φ1 and φ2 .
Show that
A1 (θ2 ) = A2 (θ1 ) (4.10)
For general elastic waves, see Mei (1978) : Extensions of some identities in elastody-
namics with rigid inclusions. J . Acoust. Soc. Am. 64(5), 1514-1522.
incident and the reßected wave, and the shadow zone III. The boundaries of these zones
are the rays touching the barrier tip. According to this crude picture of geometrical
optics the solution is
Ao exp(ik cos θx + ik sin θy), I;
φ= Ao [exp(ik cos θx + ik sin θy) + exp(ik cos θx − ik sin θy)], II (5.1)
0, III
Clearly (5.1) is inadquate because the potential cannot be discontinuous across the
boundaries. A remedy to ensure smooth transition is needed.
Consider the shadow boundary Ox0 . Let us introduce a new cartesian coordinate
system so that x0 axis is along, while the y 0 axis is normal to, the shadow boundary.
The relations between (x, y) and (x0 , y 0 ) are
, but the amplitude is slowly modulated in both x0 and y 0 directions. Substituting (5.4
into the Helmholtz equation, we get
( )
ikx! ∂ 2A ∂A 2 ∂2A
e 02
+ 2ik 0
− k A + 02
+ k2 A = 0 (5.5)
∂x ∂x ∂y
Expecting that the characteristic scale Lx of A along x0 is much longer than a wavelength,
kLx À 1, we have
∂A ∂2A
À 2ik
∂x0 ∂x02
Hence we get as the Þrst approximation the Schródinger equation
∂A ∂ 2 A
2ik + ≈0 (5.6)
∂x0 ∂y 02
In this transition zone where the remaining terms are of comparable importance, hence
the length scales must be related by
k 1 √
0
∼ 02 , implying ky 0 ∼ kx0
x y
Thus the transition zone is the interior of a parabola.
Equation (5.6) is of the parabolic type. The boundary conditions are
A(x, ∞) = 0 (5.7)
Now the initial-boundary value for A has no intrinsic length scales except x0 , y 0 them-
selves. Therefore the condition kLx À 1 means kx0 À 1 i.e., far away from the tip. This
THIN BARRIER AND PARABOLIC APPROXIMATION 24
where
−ky 0
γ=√ (5.11)
πkx0
is the similarity variable. We Þnd upon subsitution that f satisÞes the ordinary differ-
ential equation
f 00 − iπγf 0 = 0 (5.12)
f → 0, γ → −∞; f → 1, γ → ∞. (5.13)
Rewriting (5.12) as
f 00
= iπγ
f0
we get
log f 0 = iπγ/2 + constant.
1.2
0.8
|φ|
0.6
0.4
0.2
0
−10 −8 −6 −4 −2 0 2 4 6 8 10
Distance in wavelengths (λ)
is reduced to an initial boundary value problem. One can use Crank-Nicholson scheme
to march in ”time”, i.e., x0 .
Homework Find by the parabolic approximation the transition solution along the
edge of the reßection zone.
more straightforward to get quantitative results from direct numerical calculation of the
series, as exempliÞed by Chen (1982). To facilitate the understanding of the physics,
these results are presented in animated form in ”link-simulations”. The exact theory
will then be compared with the parabolic approximation.
Refering to Figure 6 the entire ßuid region can be divided into three zones according
to the crude picture of geometrical optics. I: the zone of incident and reßected waves,
II : the zone of incident waves and III ; the shadow. To ensure smooth transition in
all zones there is also the diffracted (or scattered) waves. The total potential can be
expressed in a compact form by
where φo is deÞned by
φi + φr π − α > θ > 0, in I;
φo (r, θ) = φi π + α > θ > π − α, in II; (6.19)
0 θ0 > θ > π + α, in III.
and φs is the scattered (or diffracted) waves. Both the incident wave
are known, where α denotes the angle of incidence. The unknown scattered wave φs
must satisfy the radiation condition and behaves as an outgoing wave at inÞnity, i.e.,
√ ∂φs
lim r( − ikφs ) = 0 (6.22)
r→0 ∂r
or
A(θ)eikr
φs ∼ √ at r → ∞ (6.23)
kr
where we have used i = eiπ/2 . By similar analysis the second integral is found to be
à ! ∙ ¸1
1 n(π − α) ikr− iπ 2π 2
I2 (θ) ≈ cos e 4 (6.37)
2 ν kr
It follows that
à !Z
√ ∂ π−α nθ
lim r − ik e−ikr cos(θ+α) cos dθ
r→∞ ∂r 0 ν
à ! ∙
¸1
√ ∂ 1 n(π − α) 2π 2
ikr− iπ
= lim r − ik cos( )e 4
r→∞ ∂r 2 ν kr
=0 (6.38)
hence
à !Z
√ ∂ π+α nθ
lim r − ik e−ikr cos(θ+α) cos dθ
r→∞ ∂r 0 ν
à ! ∙
¸1
√ ∂ 1 n(π + α) ikr− iπ 2π 2
= lim r − ik cos( )e 4
r→∞ ∂r 2 ν kr
=0 (6.40)
In summary, only the Þrst integral associated with the incident wave furnishes a
nonvanishing contribution to the expansion coefficients, i.e.,
à !Z
√ ∂ νπ nθ √ nα −i(kr+ π4 )
lim r − ik φ0 cos dθ ∼ 2 2πk cos e (6.41)
r→∞ ∂r 0 ν ν
With this result we get by substituting (6.33) and (6.41) into (6.31), the coefficients an
are found
nα −i nπ
an = 2π cos e 2ν (6.42)
ν
By inverse transform, (6.25), we get the exact solution,
" ∞
#
2 X nπ nα nθ
φ(r, θ) = J0 (kr) + 2 e−i 2ν Jn/ν (kr) cos cos (6.43)
ν n=1 ν ν
EXACT THEORY OF WEDGE DIFFRACITON 31
By employing the partial-wave expansion theorm, (Abramowitz and Stegun 1964), the
preceding equaion becomes
" ∞
#
X
φ(r, θ) = 2 J0 (kr) + 2 (−i)n Jn (kr) cos nα cos nθ (6.46)
n=1
Now we set
t = ieiθ z = kr.
where ²n is the Jacobi symbol. The above result may be viewed as the Fourier expan-
sion of the plane wave with Bessel functions being the expansion coefficients. In wave
propagation theories, each term in the series represents a distinct angular variation and
is called a partial wave.
Using the orthogonality of cos nθ, we may evaluate the Fourier coefficient
2 Z π ikr cos θ
Jn(kr) = e cos nθdθ, (A.2)
²n in π 0
which is one of a host of integral representations of Bessel functions.
B Approximation of an integral
Consider the integral
Z 2π
dθ[1 + cos(θ − θo )]eikr(1−cos(θ−θo ))
0
When the limits are approximated by (−∞, ∞), the inegral can be evaluated to give
s
Z ∞
ikrθ2 /2 2π iπ/4
A(θo ) e dθ = e A(θo )
−∞ kr
Near the second stationary point the integral vanishes since 1+cos(θ −θo ) == 1−1 = 0.
Hence the result (4.8) follows.
For the second integral I2 , we take the phase to be f2 (θ) = k cos(θ + α). The
stationary phase point must be the root of
I2 is approximately
∙ ¸1 ∙ ¸1
1 nθ2 −ikr cos(θ2 +α)− iπ 2π 2 1 n(π − α) ikr− iπ 2π 2
I2 (θ) ≈ cos( )e 4 = cos( )e 4 (C.6)
2 ν kr 2 ν kr
EXACT THEORY OF WEDGE DIFFRACITON 34
Lastly for the third integral I3 , the phase is f3 (θ) = k cos(θ − α). The point of
stationary phase is found from
I3 is approximately
∙ ¸1
1 n(π + α) ikr− iπ 2π 2
I3 (θ) ≈ cos( )e 4 (C.9)
2 ν kr
I-campus project
School-wide Program on Fluid Mechanics
Modules on Waves in ßuids
T. R. Akylas & C. C. Mei
CHAPTER SIX
INTERNAL WAVES IN A STRATIFIED FLUID
[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 Effects in StratiÞed Flows Cambridge U. Press.
M. J. Lighthill 1978, Waves in Fluids , Cambridge University Press.
d2 ζ dρ
ρ 2
=g ζ
dt dz
or
d2 ζ
+ N 2ζ = 0 (1.1)
dt2
1
where à !1/2
g dρ
N= − (1.2)
ρ dz
is called the Brunt-Väisälä frequency. Ths elementary consideration shows that once
a ßuid is displaced from its equilibrium position, gravity and density gradient provides
restoring force to enable oscillations. In general ther must be horizontal nonunifomities,
hence waves are possible.
We start from the exact equations for an inviscid and incompressible ßuid with
variable density.
For an imincompuresibel ßuid the density remains constant as the ßuid moves,
ρt + q · ∇ρ = 0 (1.3)
where q = (u, w) is the velocity vector in the vertical plane of (x, z). Conservation of
mass requires that
∇·q=0 (1.4)
with
ρ À ρ0 , p À p0 (1.7)
2
ρu0t = −p0x (1.10)
hence
Z z
p(z) = ρ̄(z)dz. (1.13)
0
u0 = ψz , w 0 = −ψx (1.17)
3
approximation and amounts to taking ρ to be contant in (Eq:17.1) only. With it
(1.18) reduces to
Note that because of linearity, u0 and w 0 satisfy Eqn. (1.20) also, i.e.,
0 0
(wxx + wzz 0
)tt + N 2 wxx =0 (1.21)
etc.
(p + p0 )z=ζ = 0. (1.22)
Therefore,
−ρgζ + p0 = 0, z = 0, (1.23)
implying
−ρgζxxt + p0xxt = 0, z = 0. (1.24)
Kinematic condition :
ζt = w, z = 0. (1.25)
0
−ρgζxxt = −ρg wxx
hence
0 0
wztt − gwxx = 0, on z = 0. (1.26)
4
Since w0 = −ψx , ψ also satisÞes the same boundary conditon
(1.27)
ψztt − gψxx = 0, on z = 0.
On the seabed, z = −h(x) the normal velocity vanishes. For a horizontal bottom we
have
ψ(x, −h, t) = 0. (1.28)
F =0 z = −h. (1.31)
5
Figure 1: From Phillips, 1977
This is called the rigid lid approximation, which will be adopted in the following for
simplicity.
With the rigid-lid approximation, the solution for F is
à √ !
N 2 − ω2
F = A sin k(z + h) (1.34)
ω
where √
N 2 − ω2
kh = nπ, n = 1, 2, 3... (1.35)
ω
This is an eigen-value condition. For a Þxed wave number k, it gives the eigen-frequencies,
N
ωn = r ³ ´2 (1.36)
nπ
1+ kh
For a given wavenumber k, this dispersion relation gives the eigen-frequency ωn . For a
given frequency ω, it gives the eigen-wavenumbers kn ,
nπ ω
kn = √ (1.37)
h N − ω2
2
For a simple lake with vertical banks and length L, 0 < x < L, we must impose the
conditions :
u0 = 0, hence ψ = 0, x = 0, L (1.38)
The solution is
q
N 2 − ωnm
2
ψ = A sin km x exp(−iωnm t) sin km (z + h) . (1.39)
ωnm
6
with
km L = mπ, m = 1, 2, 3, ... (1.40)
ψ = ψ0 ei(αx+βz−ωt)
Then
α2
ω2 = N 2
α2 + β 2
or
α
ω = ±N (1.42)
k
k = α + β2
2 2
(1.43)
For a given frequency, there are two possible signs for α. Since the above relation is
also even in β, there are four possible inclinations for the wave crests and troughs with
respect to the vertical; the angle of inclination is
ω
|θ| = cos−1 (1.44)
N
For ω < N, |θ| < π/2. There is no vertically propagating internal wave. This unique
property of anisotropy has been veriÞed in dramatic experiments by Mowbray and
Stevenson. By oscillating a long cylinder at various frrequencies vertically in a stratiÞed
ßuid, equal phase lines are only found along four beams forming ”St Andrew’s Cross”,
see Þgure (??) for ω/N = 0.7, 0.9. It can be veriÞed that angles are |θ| = 45◦ for
ω/N = 0.7, and |θ| = 26◦ for ω/N = 0.9, in close accordance with the condition 1.44).
Comparison between measured and predicted angles is plotted in Figure (3) for a wide
range of ω/N To under the physics better we note Þrst that the phase velocity is
~ = ± ω (α, β)
C (1.45)
k2
7
Figure 2: St Andrew’s Cross in a stratiÞed ßuid
8
Figure 3: Comparison of measured and predicted angles of internal waves
~g · C
C ~ = 0. (1.48)
Since
³ ´
~ +C
C ~ g = ± N α2 + β 2 , 0 = ± N (k, 0) (1.49)
k3 k2
~ and C
the sum of C ~ g is a horizontal vector, as shown by any of the sketches in Figure
3. Note that when tkhe phase velocity as an upward component, the group velocity has
9
a downward component, and vice versa. Now let us consider energy transport. from
(1.10) we get
−p0x = ρψzt = ρωβψo ei(αx+βz−ωt)
β
p0 = iωρ ψo ei(αx+βz−ωt) (1.50)
k
~ = 1 ρ2 |ψ|2 β (β, α)
E (1.52)
2 α
10
records indeed conÞrm this prediction. Within each of the four beams which have widths
comparable to the cylinder diameter, only one or two wave lengths can be seen.
For another interesting feature, consider the reßection of an internal wave from a
slope.
ω
Recall that θ = ± cos−1 N
, i.e., for a Þxed frequency there are only two allowable
directions with respect to the horizon. Relative to the sloping bottom inclined at θo the
inclinations of the incident and reßected waves must be different, and are respectively
θ + θo and θ − θo , see Þgure.
Let ξ be along, and η be normal to the slope. Since the slope must be a streamline,
ψi + ψr must vanish along η = 0 and be proportional to ei(αξ−ωt) ; the total stream
function must be of the form
(i) (r)
ψi ei(kt ξ−ωt)
+ ψr ei(kt ξ−ωt)
∝ sin βηei(αξ−ωt) .
(i) (r)
kt = kt = α
Therefore
k (i) cos(θ + θo ) = k (r) cos(θ − θo ),
as sketched in Figure 5. The incident wave and the reßected wave have different wave-
lengths! If θ < θo , there is no reßection; refraction takes place instead.
11
Figure 5: Internal wave reßected by in inclined surface
12