MIT OCW Wave Propagation Lecture Part 3

3.1.
Reection of sound by an interface 1

1.138J/2.062J/18.376J, WAVE PROPAGATION
Fall, 2006 MIT
Notes by C. C. Mei
CHAPTER THREE
TWO DIMENSIONAL WAVES
1 Reection and tranmission of sound at an inter-
face
Reference : L. M. Brekhovskikh and O. A. Godin: Acoustics of Layered Media I.
Springer..2.2.
The governing equation for sound in a honmogeneous uid is given by (7.31) and
(7.32) in Chapter One. In term of the the veloctiy potential dened by
u = (1.1)
it is
1
c
2
t
2
=
2
(1.2)
where c denotes the sound speed. Recall that the uid pressure
p = /t (1.3)
also satises the same equation.
1.1 Plane wave in Innite space
Let us rst consider a plane sinusoidal wave in three dimensional space
(x, t) =
o
e
i(kxt)
=
o
e
i(knxt)
(1.4)
Here the phase function is
(x, t) = k x t (1.5)
3.1.Reection of sound by an interface 2
The equation of constant phase (x, t) =
o
describes a moving surface. The wave
number vector k = kn is dened to be
k = kn = (1.6)
hence is orthogonal to the surface of constant phase, and represens the direction of wave
propagation. The frequency is dened to be
=
t
(1.7)
Is (1.4) a solution? Let us check (1.2).
=
_

x
,

y
,

z
_
= ik
2
= = ik ik = k
2
t
2
=
2
Hence (1.2) is satised if

= kc. (1.8)
Sound in an innite space is non-dispersive.
1.2 Two-dimensional reection from a plane interface
REfering to gure ??, let us cnsider two semi-innite uids separated by the plane
interface along z = 0. The lower uid is distinguished from the upper uid by the
subscript 1. The densities and sound speeds in the upper and lower uids are , c and
1
, c
1
respectively. Let a plane incident wave arive from z > 0 at the incident angle of
with respect to the z axis, the sound pressure and the velocity potential are
p
i
= P
0
exp[ik(xsin z cos )] (1.9)
The velocity potential is
i
=
iP
0
exp[ik(xsin z cos ] (1.10)

The indient wave number vector is
k
i
= (k
i
x
, k
i
z
) = k(sin , cos ) (1.11)
Figure 1: Plane wave incident towards the interface of two uids.
The motion is conned in the x, z plane.
On the same (incidence) side of the interface we have the reected wave
p
r
= Rexp[ik(x sin + z cos )] (1.12)
where R denotes the reection coecient. The wavenumber vector is
k
r
= (k
r
x
, k
r
z
) = k(sin , cos ) (1.13)
The total pressure and potential are
p = P
0
exp[ik(x sin z cos )] + Rexp[ik(x sin + z cos )] (1.14)
=
iP
0
exp[ik(x sin z cos )] + Rexp[ik(x sin + z cos )] (1.15)

In the lower medium z < 0 the transmitted wave has the pressure
p
1
= TP
0
exp[ik
1
(x sin
1
z cos
1
)] (1.16)
where T is the transmission coecient, and the potential
1
=
iP
0
T exp[ik
1
(x sin
1
z cos
1
)] (1.17)
Along the interface z = 0 we require the continutiy of pressure and normal velocity,
i.e.,
p = p
1
, z = 0 (1.18)
k
i
k
r
k
1
1
Figure by MIT OCW.
and
w = w
1
, z = 0, (1.19)
Applying (1.18), we get
P
0
_
e
ikxsin
+ Re
ikxsin
_
= TP
0
e
ik
1
xsin
1
, < x < .
Clearly we must have
k sin = k
1
sin
1
(1.20)
or,
sin
c
=
sin
1
c
1
(1.21)
This is just Snells law. With (1.20), we must have
1 + R = T (1.22)
Applying (1.19), we have
iP
0
_
k cos e
ik sin
+ Rk cos e
ik sin
=
iP
0
_
k
1
cos
1
Te
ik
1
sin
1
which implies
1 R =
k
1
cos
1
1
k cos
T (1.23)
Eqs (1.22) and (1.23) can be solved to give
T =
2
1
k cos
k
1
cos
1
+
1
k cos
(1.24)
R =

1
k cos k
1
cos
1
1
k cos + k
1
cos
1
(1.25)
Alternatively, we have
T =
2
1
c
1
cos
c cos
1
+
1
c
1
cos
(1.26)
R =

1
c
1
cos c cos
1
1
c
1
cos + c cos
1
(1.27)
Let
m =

1
, n =
c
c
1
(1.28)
where the ratio of sound speeds n is called the index of refraction. We get after using
Snells law that
R =
mcos ncos
1
mcos + ncos
1
=
mcos n
_
1
sin
2
n
2
mcos + n
_
1
sin
2
n
2
(1.29)
The transmission coecient is
T = 1 + R =
2mcos
mcos + n
_
1
sin
2
n
2
(1.30)
We now examine the physics.
1. If n = c/c
1
> 1, the wave enter from a faster to a slower medium, then >
1
and
sin /n < 1 always. R is real. In particular, for normal incidence =
1
= 0,
R =
mn
m + n
(1.31)
is real. If m > n > 1, i.e.,
1
/ > c/c
1
> 1, then R is positive. If n > m > 1 then
R is negative. As increases,
1
also increases. R decreases until at = /2, so
that
R =
n
n
= 1 (1.32)
Hence R lies on a segment of the real axis as shown in Figur 1.a. or Figure 1.b.
Note that there is no reection R = 0 at the special incidence angle =
B
, called
the Brewster angle:
sin
B
=
n
1 m
2
(1.33)
2. If however n < 1, i.e., the wave enters from a slower medium to a faster medium,
then
1
> . for suciently small , R is real. When increases to a critical value
, dened by
sin = n (1.34)
sin
1
= 1 so that
1
becomes /2, and
R =
mcos
mcos
= 1
Below this critical angle (0 < < ), R is real. In particular, when = 0, (1.31)
applies. See Figure 2.c for m > n and in 2.d. for m < n.
Figure 2: Complex reection coecient. From Brekhovskikh and Godin .2.2.
When > , the square roots above become imaginary. We must then take
cos
1
=
_
1
sin
2
n
2
= i
_
sin
2
n
2
1 (1.35)
The reection coecient is now complex
R =
mcos in
_
sin
2
n
2
1
mcos + in
_
sin
2
n
2
1
(1.36)
with [R[ = 1, implying complete reection. As a check the transmitted wave is
now given by
p
t
= T exp
_
k
1
_
ixsin
1
+ z
_
sin
2
/n
2
1
__
(1.37)
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 + iR. It is clear
from (1.36 ) that R < 0 so that R falls on the half circle in the lower half of the
complex plane as shown in Figures 2.c and 2.d.
3.2.Equations for Elastic Waves 7
2 Equations for elastic waves
Refs:
Gra: Wave Motion in Elastic Solids
Aki & Richards Quantitative Seismology, V. 1.
Achenbach. Wave Propagation in Elastic Solids
Let the displacement vector at a point x
j
and time t be denoted by u
i
(x
j
, t), then
Newtons law applied to an material element of unit volume reads
2
u
i
t
2
=

ij
x
j
(2.1)
where
ij
is the stress tensor. We have neglected body force such as gravity. For a
homogeneous and isotropic elastic solid, we have the following relation between stress
and strain
ij
= e
kk
ij
+ 2e
ij
(2.2)
where and are Lame constants and
e
ij
=
1
2
_
u
i
x
j
+
u
j
x
i
_
(2.3)
is the strain tensor. Eq. (2.2) can be inverted to give
e
ij
=
1 +
E

ij
kk
ij
(2.4)
where
E =
(3 + )
+
(2.5)
is Youngs modulus and
=

2( + )
. (2.6)
Poissons ratio.
Substituting (2.2) and (2.3) into (2.1) we get
ij
x
j
=
e
kk
x
j
ij
+

x
j
_
u
i
x
j
+
u
j
x
i
_
=
e
kk
x
i
+

2
u
i
x
j
x
j
+

2
u
j
x
i
x
j
= ( + )

2
u
j
x
i
x
j
+
2
u
i
3.2.Equations for Elastic Waves 8
In vector form (2.1) becomes
2
u
t
2
= ( + )( u) +
2
u (2.7)
Taking the divergence of (2.1) and denoting the dilatation by
e
kk
=
u
1
x
1
+
u
2
x
2
+
u
3
x
3
(2.8)
we get the equation governing the dilatation alone
t
2
= ( + ) +
2
= ( + 2)
2
(2.9)
or,
t
2
= c
2
L
2
(2.10)
where
c
L
=
+ 2
(2.11)
Thus the dilatation propagates as a wave at the speed c
L
. To be explained shortly, this
is a longitudinal waves, hence the subscript L. On the other hand, taking the curl of
(2.7) and denoting by the rotation vector:
= u (2.12)
we then get the governing equation for the rotation alone
2

t
2
= c
2
T
2
(2.13)
where
c
T
=
_
(2.14)
Thus the rotation propagates as a wave at the slower speed c
T
. The subscript T indicates
that this is a transverse wave, to be shown later.
The ratio of two wave speeds is
c
L
c
T
=
> 1. (2.15)
Since
=
1
2
1 (2.16)
3.3. Free waves in innite space 9
it follows that the speed ratio depends only on Poissons ratio
c
L
c
T
=
_
2 2
1 2
(2.17)
There is a general theorem due to Helmholtz that any vector can be expressed as
the sum of an irrotational vector and a solenoidal vector i.e.,
u = +H (2.18)
subject to the constraint that
H = 0 (2.19)
The scalar and the vector H are called the displacement potentials. Substituting this
into (2.7), we get

2
t
2
[ +H] =
2
[ +H] + ( + ) [ +H]
Since =
2
, and H = 0 we get
_
( + 2)
2

t
2
_
+
_
2
H
2
H
t
2
_
= 0 (2.20)
Clearly the above equation is satisied if
( + 2)
2

t
2
= 0 (2.21)
and
2
H
2
H
t
2
= 0 (2.22)
Although the governing equations are simplied, the two potentials are usually coupled
by boundary conditions, unless the physical domain is innite.
A typical seismic record is shown in Figure 2.
3 Free waves in innite space
The dilatational wave equation admits a plane sinusoidal wave solution:
(x, t) =
o
e
ik(nxc
L
t)
(3.1)
Figure 3: A typical seismic record. P: Longitudinal wave; S: Transverse wave; R :
Rayleigh surface wave. From Billingham & King. Wave Motion, Cambridge U Press.
1500
ERM
Waves in Elastic Solids
A typical seismogram during an earthquake, showing
the arrival of the P(primary), S(secondary) and R(Rayleigh)
waves. Note that the Rayleigh waves have the largest amplitude.
1000
500
P
1 2 3
T
i
m
e

(
s
)
0
S
R
Figure by MIT OCW.
Here the phase function is
(x, t) = k(n x c
L
t) (3.2)
which describes a moving surface. The wave number vector k = kn is dened to be
k = kn = (3.3)
hence is orthogonal to the surface of constant phase, and represents the direction of
wave propagation. The frequency is
= kc
T
=
t
(3.4)
A general solution is
= (n x c
L
t) (3.5)
Similarly the the following sinusoidal wave is a solution to the shear wave equation;
H = H
o
e
ik(nxc
T
t)
(3.6)
A general solution is
H = H(n x c
T
t) (3.7)
Note:
We can also write (3.5) and (3.9) as
= (t
n x
c
L
) (3.8)
and
H = H(t
n x
c
T
) (3.9)
where
s
L
=
n
c
L
, s
T
=
n
c
T
(3.10)
are called the slowness vectors of longitudinal and transverse waves respectively.
In a dilatational wave the displacement vector is parallel to the wave number vector:
u
L
= =
n (3.11)
3.4. Elastic waves in a plane 12
from (3.5), where
is the ordinary derivative of with repect to its argument. Hence

the dilatational wave is a longitudinal (compression) wave. On the other hand in a
rotational wave the displacement vector is perpendicular to the wave number vector,
u
T
= H = e
x
_
H
z
y

H
y
z
_
+e
y
_
H
x
z

H
z
x
_
+e
z
_
H
y
x

H
x
y
_
= e
x
_
H
z
n
y
H
y
n
z
_
+e
y
(H
x
n
z
H
z
n
x
) +e
z
_
H
y
n
x
H
x
n
y
_
= n H
(3.12)
from (3.7). Hence a rotational wave is a transverse (shear) wave.
4 Elastic waves in a plane
Refs. Gra, Achenbach,
Aki and Richards : Quantitative Seismology, v.1
Let us examine waves propagating in the vertical plane of x, y. All physical quantities
are assumed to be uniform in the direction of z, hence /z = 0, then
u
x
=

x
+
H
z
y
, u
y
=

y

H
z
x
, u
z
=
H
x
y
+
H
y
x
(4.13)
and
H
x
x
+
H
y
y
= 0 (4.14)
where
x
2
+

2
y
2
=
1
c
2
L
t
2
, (4.15)
2
H
p
x
2
+

2
H
p
y
2
=
1
c
2
T
2
H
p
t
2
, p = x, y, z (4.16)
Note that u
z
is also governed by (4.16).
Note that the in-plane displacements u
x
, u
y
depend only on and H
z
, and not on
H
x
, H
y
. Out-of-plane motion u
z
depends on H
x
, H
y
but not on H
z
. Hence the in-
plane displacement components u
x
, u
y
are independent of the out-of-plane component
u
z
. The in-plane displacements (u
x
, u
y
) are associated with dilatation and in-plane
shear, represented respectively by and H
z
, which will be refered to as the P wave and
the SV wave. The out-of-plane displacement u
z
is associated with H
x
and H
y
, and will
be refered to as the SH wave.
3.5. Reection of elastic waves from a plane boundary 13
From Hookes law the stress components can be written
xx
=
_
u
x
x
+
u
y
y
_
+ 2
u
x
x
= ( + 2)
_
u
x
x
+
u
y
y
_
2
u
y
y
= ( + 2)
_
x
2
+

2
y
2
_
2
_
y
2

2
H
z
yx
_
(4.17)
yy
=
_
u
x
x
+
u
y
y
_
+ 2
u
y
y
= ( + 2)
_
u
x
x
+
u
y
y
_
2
u
x
x
= ( + 2)
_
x
2
+

2
y
2
_
2
_
x
2
+

2
H
z
xy
_
(4.18)
zz
=

2( + )
(
xx
+
yy
) = (
xx
+
yy
) =
_
x
2
+

2
y
2
_
(4.19)
xy
=
_
u
y
x
+
u
x
y
_
=
_
2

2
xy

2
H
z
x
2
+

2
H
z
y
2
_
(4.20)
yz
=
u
z
y
=
_
2
H
x
y
2
+

2
H
y
yx
_
(4.21)
xz
=
u
z
x
=
_

2
H
x
xy
+

2
H
y
x
2
_
(4.22)
Dierent physical situations arise for dierent boundary conditions. We shall con-
sider rst the half plane problem bounded by the plane y = 0.
5 Reection of elastic waves from a plane boundary
Consider the half space y > 0, < x < . Several types of boundary conditions
can be prescribed on the plane boundary : (i) dynamic: the stress components only
(the traction condition); (ii) kinematic: the displacement components only, or (iii). a
combination of stress components and displacement components. Most dicult are
(iv) the mixed conditions in which stresses are given over part of the boundary and
displacements over the other.
We consider the simplest case where the plane y = 0 is completely free of external
stresses,
yy
=
xy
= 0, (5.23)
and
yz
= 0 (5.24)
It is clear that (5.23) aects the P and SV waves only, while (5.24) aects the SH
wave only. Therefore we have two uncoupled problems each of which can be treated
separately.
5.1 P and SV waves
Consider the case where only P and SV waves are present, then H
x
= H
y
= 0. Let all
waves have wavenumber vectors inclined in the positve x direction:
= f(y)e
ixit
, H
z
= h
z
(y)e
ixit
(5.25)
It follows from (4.15) and (4.16) that
d
2
f
dy
2
+
2
f = 0,
d
2
h
z
dy
2
+
2
h
z
= 0, (5.26)
with
=
2
c
2
L
2
=
_
k
2
L
2
, =
2
c
2
T
2
=
_
k
2
T

2
(5.27)
We rst take the square roots to be real; the general solution to (5.26) are sinusoids,
hence,
= A
P
e
i(xyt)
+ B
P
e
i(x+yt)
, H
z
= A
S
e
i(xyt)
+ B
S
e
i(x+yt)
(5.28)
On the right-hand sides the rst terms are the incident waves and the second are the
reected waves. If the incident amplitudes A
P
, A
S
and are given, what are the properties
of the reected waves B
P
, B
S
? The wave number components can be written in the polar
form:
(, ) = k
L
(sin
L
, cos
L
), (, ) = k
T
(sin
T
, cos
T
) (5.29)
where (k
L
, k
T
) are the wavenumbers, the (
L
,
T
) the directions of the P wave and SV
wave, respectively. In terms of these we rewrite (5.28)
= A
P
e
ik
L
(sin
L
xcos
L
yt)
+ B
P
e
ik
L
(sin
L
x+cos
L
yt)
(5.30)
H
z
= A
S
e
ik
T
(sin
T
xcos
L
yt)
+ B
S
e
ik
T
(sin
T
x+cos
T
yt)
(5.31)
In order to satisfy (5.23) (
yy
=
xy
= 0) on y = 0 for all x, we must insist:
k
L
sin
L
= k
T
sin
T
, ( = ) (5.32)
This is in the form of Snells law:
sin
L
c
L
=
sin
T
c
T
(5.33)
implying
sin
L
sin
T
=
c
L
c
T
=
+ 2
=
k
T
k
L
(5.34)
When (5.23) are applied on y = 0 the exponential factors cancel, and we get two
algebraic conditions for the two unknown amplitudes of the reected waves (B
P
, B
S
) :
k
2
L
(2 sin
2
2
)(A
P
+ B
P
) k
2
T
sin 2
T
(A
S
B
S
) = 0 (5.35)
k
2
L
sin 2
L
(A
P
B
P
) k
2
T
cos
T
(A
S
+ B
S
) = 0. (5.36)
Using (5.34), we get
2 sin
2
2
=
2
(2 sin
2
T
1) =
2
cos 2
T
The two equations can be solved and the solution expressed in matrix form:
_
_
_
B
P
B
S
_
_
_
=
_
_
S
PP
S
SP
S
PS
S
SS
_
_
_
_
_
A
P
A
S
_
_
_
(5.37)
where
S =
_
_
S
PP
S
SP
S
PS
S
SS
_
_
(5.38)
denotes the scattering matrrix. Thus S
PS
represents the reected S-wave due to incident
P wave of unit amplitude, etc. It is straightforward to verify that
S
PP
=
sin 2
L
sin 2
T

2
cos
2
2
T
sin 2
L
sin 2
T
+
2
cos
2
2
T
(5.39)
S
SP
=
2
2
sin 2
T
cos 2
T
sin 2
L
sin 2
T
+
2
cos
2
2
T
(5.40)
S
PS
=
2 sin 2
L
cos 2
T
sin 2
L
sin 2
T
+
2
cos
2
2
T
(5.41)
S
SS
=
sin 2
L
sin 2
T

2
cos
2
2
T
sin 2
L
sin 2
T
+
2
cos
2
2
T
(5.42)
In view of (5.33) and
=
c
L
c
T
=
_
2 2
1 2
(5.43)
The scattering matrix is a function of Poissons ratio and the angle of incidence.
(i) P- wave Incidence : In this case
L
is the incidence angle. Consider the special
case when the only incident wave is a P wave. Then A
P
,= 0 and A
S
= 0 and only
S
PP
and S
SP
are relevant. . Note rst that
L
>
T
in general . For normal incidence,
L
= 0, hence
T
= 0. We nd
S
PP
= 1, S
PS
= 0 (5.44)
there is no SV wave. The refelcted wave is a P wave. On the other hand if
sin 2
L
sin 2
T

2
cos
2
2
T
= 0 (5.45)
then S
PP
= 0, hence B
P
= 0 but B
S
,= 0; only SV wave is reected. This is the case
of mode conversion, whereby an incident P waves changes to a SV wave after reection.
The amplitude of the reected SV wave is
B
S
A
P
= S
PS
=
tan 2
T
2
(5.46)
(ii) SV wave Incidence : Let A
P
= 0 but A
S
,= 0. In this case
T
is the incidence
angle. Then only S
SP
and S
SS
are relevant. For normal incidence,
L
=
T
= 0,
S
SS
= 1, and S
SP
= 0; no P wave is reected. Mode conversion (B
P
,= 0, B
S
= 0) also
happens when (5.45) is satised. Since
L
>
T
, there is a critical incidence angle
T
beyond which the P wave cannot be reected back into the solid and propagates only
along the x axis. At the critical angle
sin
L
= 1, (
L
= /2), or sin
T
= 1/ (5.47)
by Snells law. Thus for = 1/3, = 2 and the critical incidence angle is
T
= 30
.
Beyond the critical angle of incidence, the P waves decay exponentially away from
the free surface. The amplitude of the SV wave is linear in y which is unphysical,
suggesting the limitation of unbounded space assumption.
Figure 4: Amplitude ratios for incident P waves for various Possions ratios. From Gra:
Waves in Elastic Solids. Symbols should be converted according to : A
1
A
P
, A
2

B
P
, B
1
A
S
, B
2
B
S
.
Figure 5: Reected wave amplitude ratios for incident SV waves for various Possions
ratios. From Gra: Waves in Elastic Solids. Symbols should be converted according to
: A
1
A
P
, A
2
B
P
, B
1
A
S
, B
2
B
S
.
3.6. Rayleigh surface waves 19
5.2 SH wave
Because of (5.2)
H
x
x
+
H
y
y
= 0
we can introduce a stream function so that
H
x
=
y
, H
y
=

x
(5.48)
where
2
=
1
c
2
T
t
2
(5.49)
Clearly the out-of-plane dispacement is
u
z
=
H
x
y
+
H
y
x
=

2
x
2
+

2
y
2
=
2
(5.50)
and
yz
=

y
2
=

c
2
T
t
2
(5.51)
The zero-stress boundary condition implies
y
= 0 (5.52)
Thus the problem for is analogous to one for sound waves reected by a solid plane.
Again for monochromatic incident waves, the solution is easily shown to be
=
_
Ae
iy
Ae
iy
_
e
ixit
(5.53)
where
2
+
2
= k
2
T
(5.54)
We remark that when the boundary is any cylindrical surface with axis parallel to
the z axis, the the stress-free condition reads
zn
= 0, on B. (5.55)
where n is the unit outward normal to B. Since in the pure SH wave problem
zn
=
u
z
n
=

n
2
=

c
2
T
t
2
Condition (5.55) implies
n
= 0, on B. (5.56)
Thus the analogy to acoustic scattering by a hard object is true irrespective of the
geomntry of the scatterer.
6 Rayleigh surface waves
Refs. Gra, Achenbach, Fung
In a homogeneous elastic half plane, in addition to P, SV and SH waves, another
wave which is trapped along the surface of a half plane can also be present. Because
most of the action is near the surface, this surface wave is of special importance to
seismic eects on the ground surface.
Let us start from the governing equations again
x
2
+

2
y
2
=
1
c
2
L
t
2
, (6.1)
2
H
z
x
2
+

2
H
z
y
2
=
1
c
2
T
2
H
z
t
2
(6.2)
We now seek waves propagating along the x direction
= '
_
f(y)e
ixit
_
, H
z
= '
_
h(y)e
ixit
_
(6.3)
Then f(y), h(y) must satisfy
d
2
f
dy
2
+
_
2
/c
2
L
2
_
f = 0,
d
2
h
dy
2
+
_
2
/c
2
T

2
_
h = 0, (6.4)
To have surface waves we insist that
=
_
2
/c
2
L
, =
_
2
/c
2
T
(6.5)
be real and postive. Keeping only the solutions which are bounded for y , we get
= Ae
y
e
i(xt)
, H
z
= Be
y
e
i(xt)
. (6.6)
The expressions for the displacements and stresses can be found straightforwardly.
u
x
=
_
iAe
y
Be
y
_
e
i(xt)
, (6.7)
u
y
=
_
Ae
y
+ iBe
y
_
e
i(xt)
, (6.8)
xx
=
__
2
2
2
_
Ae
y
2iBe
y
_
e
i(xt)
, (6.9)
yy
=
__
2
+
2
_
Ae
y
+ 2iBe
y
_
e
i(xt)
, (6.10)
xy
=
_
2iAe
y
+
_
2
+
2
_
Be
y
_
e
i(xt)
(6.11)
On the free surface the traction-free conditions
yy
=
xy
= 0 require that
_
2
+
2
_
A + 2iB = 0, (6.12)
2iA +
_
2
+
2
_
B = 0. (6.13)
For nontrivial solutions of A, B the coecient determinant must vanish,
_
2
+
2
_
2
4
2
= 0, (6.14)
or
_
2
2

2
c
2
T
_
2
4
2

2
c
2
L

2
c
2
T
= 0 (6.15)
which is the dispersion relation between frequency and wavenumber . From either
(6.12) or (6.13) we get the amplitude ratio:
A
B
=
2i
2
+
2
=

2
+
2
2i
, (6.16)
In terms of the wave velocity c = /, (6.15) becomes
_
2
c
2
c
2
T
_
2
= 4
_
1
c
2
c
2
L
_1
2
_
1
c
2
c
2
T
_1
2
. (6.17)
or, upon squaring both sides, nally
c
2
c
2
T
_
_
c
c
T
_
6
8
_
c
c
T
_
4
+
_
24
16
2
__
c
c
T
_
2
16
_
1
1
2
_
_
= 0. (6.18)
where
k =
c
L
c
T
=
+ 2
=
_
2 2
1 2
The rst solution c = = 0 is at best a static problem. In fact = = and A = iB,
so that u
x
= u
y
0 which is of no interest.
We need only consider the cubic equation for c
2
. Note that the roots of the cubic
equation depend only on Poissons ratio, through
2
= 2(1 )/(1 2). There can be
three real roots for c or , or one real root and two complex-conjugate roots. We rule
out the latter because the complex roots imply either temporal damping or instability;
neither of which is a propagating wave. When all three roots are real we must pick the
one so that both and are real. We shall denote the speed of Rayleigh wave by c
R
.
Figure 6: The velocity of Rayleigh surface waves c
R
. From Fung Foundations of Solid
Mechanics.
For c = 0, the factor in curley brackets is
. = 16
_
1
c
2
T
c
2
L
_
< 0
For c = c
T
the same factor is equal to unity and hence positive. There must be a solution
for c such that 0 < c < c
T
. Furthermore, we cannot have roots in the range c/c
T
> 1.
If so,
2
=
2
_
1
c
2
c
2
T
_
< 0
which is not a surface wave. Thus the surface wave, if it exists, is slower than the shear
wave.
Numerical studies for the entire range of Poissons ratio (0 < < 0.5) have shown
that there are one real and two complex conjugate roots if > 0.263 . . . and three real
roots if < 0.263 . . .. But there is only one real root that gives the surface wave velocity
c
R
. A graph of c
R
for all values of Poissons ratio, due to Knopo , is shown in Fig. 6.
A curve-tted expression for the Rayleigh wave velocity is
c
R
/c
T
= (0 87 + 1 12)/(1 + ). (6.19)
For rocks, = and =
1
4
, the roots are
(c/c
T
)
2
= 4, 2 + 2/
3, 2 2/
3. (6.20)
Figure by MIT OCW.
0.2
0 0.2
Poisson's Ratio
0.4
0.4
c
R
/c
L
c
R
/c
T
c
T
/c
L
0.6
0.8
0.90
0.92
0.94
0 0.2
Poisson's Ratio
0.4
3.7. Scattering of SH waves by a circular cavity 23
The only acceptable root for Rayleigh wave speed c
R
is
(c
R
/c
T
)
2
= (2 2/
3)
1
2
= 0 9194 (6.21)
or
c
R
= 0.9588c
T
. (6.22)
The particle displacement of a particle on the free surface is, from (6.7) and (6.8)
u
x
= iA
_

2
+
2
2
_
e
i(xt)
(6.23)
u
y
= A
_
+

2
+
2
2
_
e
i(xt)
(6.24)
Note that
a = A
_

2
+
2
2
_
= A
_
+
k
2
T
2
_
> 0
b = A
_
+

2
+
2
2
_
= A
_
( )
2
+ k
2
L
2
_
> 0
hence
u
x
= a sin(t x), u
y
= b cos(t x)
and
u
2
x
a
2
+
u
2
y
b
2
= 1 (6.25)
The particle trajectory is an ellipse. In complex form we have
u
x
a
+ i
u
y
b
= exp i (t x /2) (6.26)
Hence as t increases,a particle at (x, 0) traces the ellipse in the counter-clockwise direc-
tion. See gure (6).
Figure 7: Displacement of particles on the ground surface in Rayleigh surface wave From
Fung Foundations of Solid Mechanics.
7 Scattering of monochromatic SH waves by a cav-
ity
7.1 The boundary-value problem
We consider the scattering of two-dimensional SH waves of single frequency. The time-
dependent potential can be wirtten as
(x, y, t) = '
_
(x, y)e
it
(7.1)
where the potential is governed by the Helmholtz equation
2
+ k
2
=

2
x
2
+

2
2
y
2
+ k
2
= 0, k =

c
T
(7.2)
To be specic consider the scatterer to be a nite cavity of some general geometry. On
the stress-free boundary B the shear stress vanishes,
zn
=
2
c
2
T
'
_
n
e
it
_
= 0 (7.3)
hence
n
= 0, on B (7.4)
Let the incident wave be a plane wave
I
= Ae
ikx
(7.5)
Figure by MIT OCW.
Direction of wave propagation
Particle path
Instantaneous
wave surface Instantaneous
particle velocity
Displacement vectors
Let the angle of incidence with respect to the positive x axis be
o
. In polar coordinates
the incident wave vector is
k = k(cos
o
, sin
o
), x = r(cos , sin ) (7.6)
and the potential is
I
= Aexp [ikr(cos
o
cos + sin
o
sin )] = Ae
ikr cos(o)
(7.7)
Let the total wave be the sum of the incident and scattered waves
=
I
+
S
(7.8)
For the scattered wave the boundary condition on the cavity surface is
S
n
=
S
n
, on B. (7.9)
In addition, the scattered wave must satisfy the radiation condition at innity, i.e., it
must propagate outward at innty.
Note that this boundary value problem is identical to the one for sound plane sound
wave in air (or water) scattered by a perfectly rigid cylinder.
We treat below the special case of a circular cavity of readius a This is one of the
few geomentries that can be solved analytically.
7.2 The circular cavity
In polar coordinates the governing equation reads
1
r
r
_
r
r
_
+
1
r
2
2
+ k
2
= 0, r > a. (7.10)
Since
I
satises the preceding equation, so does
S
.
First, It is shown in Appendix A that the plane wave can be expanded in Fourier-
Bessel series :
e
ikr cos(o)
=
n=0
n
i
n
J
n
(kr) cos n(
o
) (7.11)
where
n
is the Jacobi symbol:
0
= 0,
n
= 2, n = 1, 2, 3, . . . (7.12)
Each term in the series (7.11) is called a partial wave.
By the method of separation of variables,
S
(r, ) = R(r)()
we nd
r
2
R
+ rR
+ (k
2
r
2
n
2
)R = 0, and
+ n
2
= 0
where n = 0, 1, 2, . . . are eigenvalues in order that is periodic in with period 2.
For each eigenvalue n the possible solutions are
n
= (sin n, cos n),
R
n
=
_
H
(1)
n
(kr), H
(2)
n
(kr)
_
,
where H
(1)
n
(kr), H
(2)
n
(kr) are Hankel functions of the rst and second kind, related to
the Bessel and Weber functions by
H
(1)
n
(kr) = J
n
(kr) + iY
n
(kr), H
(2)
n
(kr) = J
n
(kr) iY
n
(kr) (7.13)
The most general solution to the Helmholtz equation is
S
= A
n=0
(A
n
sin n + B
n
cos n)
_
C
n
H
(1)
n
(kr) + D
n
H
(2)
n
(kr)
, (7.14)
For large radius the asymptotic form of the Hankel functions are
H
(1)
n

_
2
kr
e
i(kr
n
2
)
, H
(2)
n

_
2
kr
e
i(kr
n
2
)
(7.15)
In conjunction with the time factor exp(it), H
(1)
n
gives an outgoing wave while H
(2)
n
gives an incoming wave. To satisfy the radiation condition, we must discard all terms
involving H
(2)
n
. From here on we shall abbreviate H
(1)
n
simply by H
n
. The scattered
wave is now
S
= A
n=0
(A
n
sin n + B
n
cos n) H
n
(kr) (7.16)
The expansion coecients (A
n
, B
n
) must be chosen to satisfy the boundary condition
on the cavity surface
1
Once they are determined, the wave is found everywhere. In
1
In one of the numerical solution techniques, one divides the physical region by a circle enclosing the
cavity. Between the cavity and the circle, nite elements are used. Outside the circle, (7.16) is used.
By constructing a suitable variational principle, nite element computation yields the nodal coecients
as well as the expansion coecients. See (Chen & Mei , 1974).
particular in the far eld, we can use the asymptotic formula to get
S
A
n=0
(A
n
sin n + B
n
cos n) e
in/2
_
2
kr
e
ikri/4
(7.17)
Let us dene the dimensionless directivity factor
/() =
n=0
(A
n
sin n + B
n
cos n) e
in/2
(7.18)
which indicates the angular variation of the far-eld amplitude, then
S
A/()
_
2
kr
e
ikri/4
(7.19)
This expression exhibits clearly the asymptotic behaviour of
S
as an outgoing wave.
By dierentiation, we readily see that
lim
kr
r
_
S
r

S
_
= 0 (7.20)
which is one way of stating the radiation condition for two dimensional SH waves.
Let us complete the solution.
Without loss of generality we can take
o
= 0. On the surface of the cylindrical
cavity r = a, we impose
I
r
+

S
r
= 0, r = a
It follows that A
n
= 0 and
n
i
n
AJ
n
(ka) + B
n
kH
n
(ka) = 0, n = 0, 1, 2, 3, . . . n
where primes denote dierentiation with respect to the argument. Hence
B
n
= A
n
i
n
J
n
(ka)
H
n
(ka)
The sum of incident and scattered waves is
= A
n=0
n
i
n
_
J
n
(kr)
J
n
(ka)
H
n
(ka)
H
n
(kr)
_
cos n (7.21)
and
= Ae
it
n=0
n
i
n
_
J
n
(kr)
J
n
(ka)
H
n
(ka)
H
n
(kr)
_
cos n (7.22)
Figure 8: Angular distribution of scattered energy in the far eld in cylindrical scattering
The limit of long waves can be approximatedly analyzed by using the expansions for
Bessel functions for small argument
J
0
(x) 1
x
2
4
+ O(x
4
), J
n
(x)
x
n
2
n
n!
, ; n = 1, 2, 3...
Y
0
(x)
2
log x, Y
n
(x)
2
n
(n 1)!
x
n
, n = 1, 2, 3... (7.23)
Then the scattered wave has the potential
S
A
H
0
(kr)
J
0
(ka)
H
0
(ka)
2iH
1
(kr)
J
1
(ka)
H
1
(ka)
cos + O(ka)
3
=

2
(ka)
2
_
i
2
H
0
(kr) H
1
(kr) cos
_
+ O(ka)
3
(7.24)
The term H
0
(kr) coresponds to a oscillating source which sends istropic waves in all
directions. The second term is a dipole sending scattered waves mostly in forward and
backward directions. For large kr, the angular variation is a lot more complex. The far
eld pattern for various ka is shown in g 4.
On the cavity surface surface, the displacement is proportional to
2
(a, ) or
2
(a, ). The angular variation is plotted for several ka in gure 5.
For numerical simulations, see the website
http://ocw.mit.edu/OcwWeb/Civil-and-Environmental-Engineering/1-138JFall-2004/Simulations/
Remark on energy ux: At any radius r the total rate of energy outux by the
Figure 9: Polar distribution of (a, ) on a circular cylinder.
3.8. The optical theorem 30
scattered wave is
r
_
2
0
d
rz
u
z
t
= r
_
2
0
d'
_
k
2
r
e
it
_
'[ik
2
e
it
]
=
k
4
r
2
_
2
0
d'
_
i
r
_
=
k
4
r
2

_
2
0
d
_
r
_
(7.25)
where overline indicates time averaging over a wave period 2/. The integral can be
evaluated by using the asymptotic expression.
Remark: In the analogous case of plane acoustics where the sound pressure and
radial uid velocity are respectively,
p =
o
t
, and u
r
=

r
(7.26)
the energy scattering rate is
r
_
2
0
dpu
r
=

o
r
2
'
_
C
d
_
i
r
_
=
o
r
2

_
C
d
_
r
_
(7.27)
8 The optical theorem
For the same scatterer and the same frequency , dierent angles of incidence
j
dene
dierent scattering problems
j
. In particular at innty, we have
j
A
j
_
e
ikr cos(
j
)
+/
j
()
_
2
kr
e
ikri/4
_
(8.1)
Let us apply Greens formula to
1
and
2
over a closed area bounded by a closed
contour C,
__
S
_
2
_
dA =
_
B
_
1
n

1
2
n
_
ds +
_
C
ds
_
1
n

1
1
n
_
ds
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
_
C
ds
_
1
n

1
2
n
_
ds = 0 (8.2)
3.8. The optical theorem 31
By similar reasioning, we get
_
C
ds
_
1
n

2
n
_
ds = 0 (8.3)
where
1
denotes the complex conjugate of
1
.
Let us choose
1
=
2
= in (8.3), and get
_
C
ds
_
n
_
ds = 2
__
C
ds
n
_
= 0 (8.4)
Physically, across any circle the net rate of energy ux vanishes, i.e., the scattered power
must be balanced by the incident power.
Making use of (8.1) we get
0 =
_
2
0
rd
_
e
ikr cos(o)
+
_
2
kr
/
o
()e
ikri/4
_
_
ik cos(
o
)e
ikr cos(o)
ik
_
2
kr
/
o
()e
ikr+i/4
_
=
_
2
0
rd
_
ik cos(
o
) +
2
kr
(ik)[/
o
[
2
+e
ikr[cos o)1]+i/4
(ik)
_
2
kr
/
o
+ e
ikr[cos o)1]i/4
(ik) cos(
o
)
_
2
kr
/
o
_
The rst term in the integrand gives no contribution to the integral above because of
periodicity. Since (if) = (if
), we get
0 =
2
_
2
0
[/
o
()[
2
d
+
_
2
0
rd
_
/
o
(ik)
_
2
kr
[1 + cos(
o
)]e
i/4
e
ikr(1cos(o))
_
=
2
_
2
0
[/
o
()[
2
d
'
_
e
i/4
_
r
_
2
kr
_
2
0
d/
o
()[1 + cos(
o
)]e
ikr(1cos(o))
__
For large kr the remaining integral can be found approximately by the method of sta-
tionary phase (see Appendix B), with the result
_
2
0
d/
o
()[1 + cos(
o
)]e
ikr(1cos(o))
/
o
(
o
)
_
2
kr
e
i/4
(8.5)
3.9. Diraction of SH waves by a long crack 32
We get nally
_
2
0
[/
o
[
2
d = 2'/
o
(
o
) (8.6)
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
1
and
2
.
Show that
/
1
(
2
+ ) = /
2
(
1
+ ) (8.7)
For general elastic waves, see Mei (1978) for similar and other identities in elastody-
namics with rigid inclusions. J . Acoust. Soc. Am. 64(5), 1514-1522.
9 Diraction of SH wave by a long crack - the parabolic
approximation
References
Morse & Ingard, Theoretical Acoustics Series expansions.
Born & Wolf, Principle of Optics Fourier Transform and the method of steepest de-
scent.
B. Noble. The Wiener-Hopf Technique.
If the obstacle is large, there is always a shadow behind where the incident wave
cannot penetrate deeply. The phenomenon of scattering by large obstacles is usually
referred to as diraction.
Diraction of plane incident SH waves by a long crack is identical to that of a hard
screen in acoustics. The exact solution was due to A. Sommerfeld. We shall apply
the boundary layer idea and give the approximate solution valid far away from the tip
kr 1 by the parabolic approximation, due to V. Fock.
Refering to gure () let us make a crude division of the entire eld into the illuminated
zone I , dominated by the incident wave alone, the reection zone II dominated the sum
Figure 10: Wave zones near a long crack
of the incident and the reected wave, and the shadow zone III where there is no wave.
The boundaries of these zones are the rays touching the crack tip. According to this
crude picture the solution is
=
_
_
A
o
exp(ik cos x + ik sin y), I
A
o
[exp(ik cos x + ik sin y) + exp(ik cos x ik sin y)], II
0, III
(9.1)
Clearly (9.1) is inadquate because the potential cannot be discontinuous across the
boundaries. A remedy to provide smooth transitions is needed.
Consider the shadow boundary Ox
. Let us introduce a new cartesian coordinate

system so that x
axis is along, while the y
axis is normal to, the shadow boundary.

The relations between (x, y) and (x
, y
) are
x
= xcos + y sin , y
= y cos xsin (9.2)

Thus the incident wave is simply
I
= A
o
e
ikx
(9.3)
Following the chain rule of dierentiation,
x
=

x
x
+

y
x
= cos

x
sin

y
y
=

x
y
+

y
y
= sin

x
+ cos

y
we can show straightforwardly that
x
2
+

2
y
2
=

2
x
2
+

2
y
2
so that the Helmholtz equation is unchanged in form in the x
, y
system.
We try to t a boundary layer along the x axis and expect the potential to be almost
like a plane wave
(x,
, y
) = A(x
, y
)e
ikx
(9.4)
, but the amplitude is slowly modulated in both x
and y
directions. Substituting (9.4

into the Helmholtz equation, we get
e
ikx
2
A
x
2
+ 2ik
A
x
k
2
A +

2
A
y
2
+ k
2
A
_
= 0 (9.5)
Expecting that the characteristic scale L
x
of A along x
is much longer than a wavelength,

kL
x
1, we have
A
x
kA, hence 2ik

A
x

2
A
x
2
We get as the rst approximation the Schr odinger equation
2
2ik
A
x
+

2
A
y
2
0 (9.7)
In this transition zone where the remaining terms are of comparable importance, hence
the length scales must be related by
k
x

1
y
2
, implying ky
kx
Thus the transition zone is the interior of a parabola.

Equation (9.7) is of the parabolic type. The boundary conditions are
A(x, ) = 0 (9.8)
2
In one-dimensional quantum mechanics the wave function in a potential-free eld is governed by
the Schr odinger equation
ih
t
+
1
2M
x
2
= 0 (9.6)
A(x, ) = A
o
(9.9)
The initial condition is
A(0, y
) =
_
_
_
0, y
> 0,
A
0
, y
< 0
(9.10)
he initial-boundary value for A has no intrinsic length scales except x
, y
themselves.
Therefore the condition kL
x
1 means kx
1 i.e., far away from the tip. This

problem is somwhat analogous to the problem of one-dimensional heat diusion across
a boundary. A convenient way of solution is the method of similarity.
Assume the solution
A = A
o
f() (9.11)
where
=
ky
kx
(9.12)
is the similarity variable. We nd upon subsitution that f satises the ordinary dier-
ential equation
f
if
= 0 (9.13)
subject to the boundary conditions that
f 0, ; f 1, . (9.14)
Rewriting (9.13) as
f
= i
we get
log f
= i/2 + constant.
One more integration gives
f = C
_

exp
_
iu
2
2
_
du
Since
_

0
exp
_
iu
2
2
_
du =
e
i/4
2
we get
C =
e
i/4
2
and
f =
A
A
o
=
e
i/4
2
_

exp
_
iu
2
2
_
du =
e
i/4
2
_
e
i/4
2
+
_

0
exp
_
iu
2
2
_
du
_
(9.15)
Dening the cosine and sine Fresnel integrals by
C() =
_

0
cos
_
v
2
2
_
dv, S() =
_

0
sin
_
v
2
2
_
dv (9.16)
we can then write
e
i/4
2
__
1
2
+ C()
_
+ i
_
1
2
+ S()
__
(9.17)
In the complex plane the plot of C() +iS() vs. is the famous Cornus spiral, shown
in gure (9).
The wave intensity is given by
[A[
2
A
2
o
=
1
2
_
_
1
2
+ C()
_
2
+
_
1
2
+ S()
_
2
_
(9.18)
Since C, S 1/2 as , the wave intensity diminshes to zero gradually into
the shadow. However, C, S 1/2 as in an oscillatory manner. The wave
intensity oscillates while approaching to unity asymptotically. In optics this shows up
as alternately light and dark diraction bands.
In more complex propagation problems, the parabolic approximation can simplify the
numerical task in that an elliptic boundary value problem involving an innite domain
is reduced to an initial boundary value problem. One can use Crank-Nicholson scheme
to march in time, i.e., x
.
Homework: Find by the parabolic approximation the transition solution along the
edge of the reection zone.
Figure 11: Cornus spiral, a plot of the Fresnel integrals in the comiplex plane of C() +
iS(). Abscissa: C(). Ordinate : iS().
Figure 12: Diraction of a normally incident plane sound wave on a rigid half screen.
Ordinate: f() = A/A
o
. Abscissa: = ky
kx
.
Figure by MIT OCW.
0
0.2
0.4
0.8
0.6
1.0
1.2
|E
z
|
-2 2 4 -4
0.2
0.2
x
0
0.2
0.2
0.4
0.6
0.4
0.6
1.5
2.5
2.0
0.5
0.5
2.0
Z
1.0
1.5
2.5
1.0
Z'
0.4 0.6 0.8
0.4 0.6 0.8
Figure by MIT OCW.

3.10. Exact theory of Wedge Diraction 38
10 Exact theory of Wedge Diraction
Refs. J. J. Stoker 1957 , Water waves. pp 120-125
Born & Wolf, 1950 Principles of Optics
Noble, 1963 The Wiener-Hopf Technique
The diraction of plane incident waves by a semi-innite barrier is a celebrated
problem in classical physics and was rst solved by A. Sommerfeld. Several analytical
treatments are available, including the mathematically very elegant technique of Wiener
& Hopf is available . In the last section we gave an approximate theory by parabolic
approximation. Here we present an exact theory for the more general case of a wedge,
by the more elementary method of series expansion. Though the result is in an innite
series and not in closed form, quantitative information can be calculated quite readily
by a computer.
The basic ideas were described for water waves by Stoker (1957) for a wedge. Ex-
tensive numerical results were reported by Dr. H.S. Chen (Army Corps of Engineers,
Tech Rept: CERC 878-16, 1987). Based on Stokers analysis more extensive numerical
computations have been carried out by Dr. G.D. Li. To facilitate the understanding
of the physics, these results are presented in animated form in our subject website (see
Simulations):
http://ocw.mit.edu/OcwWeb/Civil-and-Environmental-Engineering/1-138JFall-2004/
We shall describe the problem for water waves; the analysis and results can be
adapted for cracks, sound,... etc.
Refering to Figure 10, we consider a vertical wedge of arbitrary apex angle in a
sea of constant depth h. Let the tip of the wedge be the z axis orgin and the x axis
coincide with one wall. The still water surface is the x, y plane. In the cylindrical polar
coordinate system (r, , z) with (x = r cos , y = r sin ), z, the walls are given by = 0
and with 1 < < 2. A train of monochromatic waves is incident from innity at the
angle with respect the x axis.
In the water region dened by 0 < < and 0 z h, the velocity potential
(r, , z, t) must satisfy the Laplace equation,
r
2
+
1
r
r
+
1
r
2
2
+

2
z
2
= 0 (10.1)
Figure 13: Coordinate system and subregions.
and subject to the linearized free surface boundary conditions
t
= g (10.2)
t
=

z
(10.3)
whicdh can be combined to give
t
2
+ g
z
= 0, z = 0. (10.4)
Along the impermeable bottom and walls, the no ux boundary conditions are
z
= 0 on z = h (10.5)
= 0 at = 0 and (10.6)
The incident wave train is given by
i
=
igA
0
cosh k(z + h)
cosh kh
(r, )e
ikr cos()it
(10.7)
where k is the real wavenumber satisfying the dispersion relation
2
= gk tanh kh, (10.8)
and + is the angle of incidence measured from the x axis. A
0
is the incident wave
amplitude.
Because of the vertical side-walls, the three dimensional problem can be reduced to
a two dimensional one by letting
(r, , z, t) = A
0
cosh k(z + h)
cosh kh
(r, )e
it
(10.9)
where (r, , t) is the horizontal pattern of the velocity potential normlized for an inci-
dent wave of unit amplitude.
Substituting Equation (10.9) into the Laplace equation and using both the kinematic
and dynamic boundary conditions on the free surface, the Laplace equation is then
reduced to the Helmholtz equation
r
2
+
1
r
r
+
1
r
2
2
+ k
2
= 0, 0. (10.10)
with the following boundary conditions on the rigid walls of the wedge:
= 0 at = 0 and
0
(10.11)
The free surface displacement from the mean water level z=0 can be represented by
(r, , t) =
1
g
t
= A
0
(r, )e
it
(10.12)
Note that is dimensionless.
Refering to Figure 10, the entire water region can be divided into three zones accord-
ing to the crude picture of geometrical optics. I: the zone of incident and reected plane
waves, II : the zone of incident plane wave and III ; the shadow with no plane wave.
This crude picture is discontinuous at the border lines separating the zones. Siknce the
physical solution must be smooth every where we must nd the transitions. Let us use
the ideas of boundary layers.
Let the total potential be expressed in a compact form by
=
o
(r, ) +
s
(r, ), for all 0 < < (10.13)
where
o
consists of only the plane waves,
o
(r, ) =
_
i
+
r
> > 0, in I;
i
+ > > , in II;
0
0
> > + , in III.
(10.14)
Here
i
denotes the incident wave
i
= e
ikr cos()
(10.15)
where denotes the angle of incidence, and
r
the reected wave
r
= e
ikr cos(+)
(10.16)
The correction is the diracted wave
s
which must satisfy the radiation condition and
behaves as an outgoing wave at innity, i.e.,
lim
r0
r(
s
r
ik
s
) = 0 (10.17)
or
s

/()e
ikr
kr
at r (10.18)
10.1 Solution by Fourier series (or, nite Fourier Transform)
Let us solve the scattered wave formally by Fourier series
(r, ) =
1
0
(r) +
2
n=1
n
(r) cos
n
(10.19)
then, the Fourier coecients are:
n
(kr) =
_

0
(kr, ) cos
n
d (10.20)
From (10.10), each Fourier coecient satises
r
2
n
r
2
+ r
n
r
+
_
(kr)
2
_
n
_
2
_

n
= 0 (10.21)
The general solution nite at the origin is
n
(kr) = a
n
J
n/
(kr) (10.22)
where the coecients a
n
, n = 0, 1, 2, 3, ... are to be determined.
The Fourier coecient of (10.13) reads
n
(kr) = a
n
J
n/
(kr) =
_

0
s
cos
n
d +
_

0
o
cos
n
d (10.23)
or
s,n
= a
n
J
n/
(kr)

o,n
(10.24)
Applying the operator lim
r
r(/r ik) to both sides of (10.23), and using the

Sommerfeld radiation condition (10.17), we have
lim
r
r
_

r
ik
__
a
n
J
n/
(kr)
_

0
o
cos
n
d
_
= 0 (10.25)
We now perform some asymptotic analysis for large kr to evaluate a
n
.
The rst part can be treated explicitly for large kr, since
J
n/
(kr)
_
2
kr
cos
_
kr
n
2

4
_
(10.26)
It follows that
lim
r
r
_

r
ik
_
J
n/
(kr) =
_
2k
e
i(kr
n
2
+
4
)
(10.27)
For the second part, we substitute
o
from (10.15) and (10.16) to rewrite the integral as
_

0
o
cos
n
d =
1
..
_

0
e
ikr cos()
cos
n
d +
2
..
_

0
e
ikr cos(+)
cos
n
d
+
3
..
_
+
e
ikr cos()
cos
n
d (10.28)
Each of the integrals above can be evaluated for large kr by the method of stationary
phase (again). Details are given in the next subsection. only the results are cited below.
The rst integral is appoximately
I
1
() = cos
_
n
_
e
ikr+
i
4
_
2
kr
_1
2
+ O
_
1
kr
_
(10.29)
from which
lim
r
r
_

r
ik
__

0
e
ikr cos()
cos
n
d
= lim
r
r
_

r
ik
_
_
cos
_
n
_
e
ikr+
i
4
_
2
kr
_1
2
_
= 2
2k cos
_
n
_
e
ikr
i
4
(10.30)
where we have used i = e
i/2
. By similar analysis the second integral is found to be
I
2
()
1
2
cos
_
n( )
_
e
ikr
i
4
_
2
kr
_1
2
(10.31)
It follows that
lim
r
r
_

r
ik
__

0
e
ikr cos(+)
cos
n
d
= lim
r
r
_

r
ik
_
_
1
2
cos(
n( )
)e
ikr
i
4
_
2
kr
_1
2
_
= 0 (10.32)
Finally the third integral is approximately
I
3
()
1
2
cos(
n( + )
)e
ikr
i
4
_
2
kr
_1
2
(10.33)
hence
lim
r
r
_

r
ik
__
+
0
e
ikr cos(+)
cos
n
d
= lim
r
r
_

r
ik
_
_
1
2
cos(
n( + )
)e
ikr
i
4
_
2
kr
_1
2
_
= 0 (10.34)
In summary, only the rst integral associated with the incident wave furnishes a
nonvanishing contribution to the expansion coecients, i.e.,
lim
r
r
_

r
ik
__

0
o
cos
n
d 2
2k cos
n
e
i(kr+
4
)
(10.35)
With this result we get by substituting (10.27) and (10.35) into (10.25), the coecients
a
n
are found
a
n
= 2 cos
n
e
i
n
2
(10.36)
By the inverse transform, (10.19), we get the exact solution,
(r, ) =
2
_
J
0
(kr) + 2
n=1
e
i
n
2
J
n/
(kr) cos
n
cos
n
_
(10.37)
Numerical computations by MATLAB is straightforward. Aside from its own practical
interest, this solution is useful for checking strictly numerical methods for diraction
problem for other geometries.
10.2 Asymptoic approximation of Integrals
For the rst integral I
1
, we take the phase to be f
1
() = k cos( ). The points of
stationary phase must be found from
f
1
() = k sin( ) = 0, (10.1)
hence = , . Only the rst at
1
= lies in the range of integration (0, )
and is the stationary point. Since
f
1
(
1
) = k cos(
1
) = k < 0 (10.2)
the integral is approximately
I
1
() cos(
n
1
)e
ikr cos(
1
)+
i
4
_
2
kr
_1
2
= cos(
n
)e
ikr+
i
4
_
2
kr
_1
2
(10.3)
For the second integral I
2
, we take the phase to be f
2
() = k cos( + ). The
stationary phase point must be the root of
f
2
() = k sin( + ) = 0 (10.4)
or = , . The stationary point is at
2
= which is the the upper limit
of integration. Since
f
2
(
2
) = k cos(
2
+ ) = k > 0 (10.5)
I
2
is approximately
I
2
()
1
2
cos(
n
2
)e
ikr cos(
2
+)
i
4
_
2
kr
_1
2
=
1
2
cos(
n( )
)e
ikr
i
4
_
2
kr
_1
2
(10.6)
Lastly for the third integral I
3
, the phase is f
3
() = k cos( ). The point of
stationary phase is found from
f
3
() = k sin( ) = 0 (10.7)
or = , +. Only the point
3
= + is acceptable and coincides with the upper
limit of integration. Since
f
3
(
3
) = k cos(
3
) = k > 0, (10.8)
I
3
is approximately
I
3
()
1
2
cos(
n( + )
)e
ikr
i
4
_
2
kr
_1
2
(10.9)
10.3 Two limiting cases
(1) A thin barrier. Let the wedge angle be 0 by setting = 2. Equation (10.37) then
becomes
(r, ) = J
0
(kr) + 2
n=1
e
i
n
4
J
n/2
(kr) cos
n
2
cos
n
2
(10.10)
(see Stoker (1957)).
(2) An innite wall extending from x = to . Water occupying only the half
plane of y 0 and the wedge angle is 180 degrees. The diracted wave is absent from
the solution, and the total wave is only the sum of the incident and reected waves :
(r, ) = e
ikr cos()
+ e
ikr cos(+)
(10.11)
By employing the partial-wave expansion theorm, (Abramowitz and Stegun 1964), the
preceding equaion becomes
(r, ) = 2
_
J
0
(kr) + 2
n=1
(i)
n
J
n
(kr) cos ncos n
_
(10.12)
which agree with (10.37) for = 1.
For the animated version of the wave patterns for dierent wedge angles snd angle
of incidence, please visit the website.
Appendix A. Partial wave expansion 46
A Partial wave expansion
A useful result in wave theory is the expansion of the plane wave in a Fourier series
of the polar angle . In polar coordinates the spatial factor of a plane wave of unit
amplitude is
e
ikx
= e
ikr cos
.
Consider the following product of exponential functions
e
zt/2
e
z/2t
=
_

n=0
1
n!
_
zt
2
_
n
__

n=0
1
n!
_
z
2t
_
n
_
t
n
_
(z/2)
n
n!

(z/2)
n+2
1!(n + 1)!
+
(z/2)
n+4
2!(n + 2)!
+ + (1)
r
(z/2)
n+2r
r!(n + r)!
+
_
.
The coecient of t
n
is nothing but J
n
(z), hence
exp
_
z
2
_
t
1
t
__
=
t
n
J
n
(z).
Now we set
t = ie
i
z = kr.
The plane wave then becomes
e
ikx
=
N=
e
in(+/2
J
n
(z).
Using the fact that J
n
= (1)
n
J
n
, we nally get
e
ikx
= e
ikr cos
=
n=0
n
i
n
J
n
(kr) cos n, (A.1)
where
n
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 coecients. 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 n, we may evaluate the Fourier coecient
J
n
(kr) =
2
n
i
n
_

0
e
ikr cos
cos nd, (A.2)
which is one of a host of integral representations of Bessel functions.
Appendix B. Approximate evaluation of an integral 47
B Approximate evaluation of an integral
Consider the integral
_
2
0
d[1 + cos(
o
)]e
ikr(1cos(o))
For large kr the stationary phase points are found from
[1 cos(
o
)] = sin(
o
) = 0
or =
o
,
o
+ within the range [0, 2]. Near the rst stationary point the integrand
is dominated by
2/(
o
)e
ikt(o)
2
/2
.
When the limits are approximated by (, ), the inegral can be evaluated to give
/(
o
)
_

e
ikr
2
/2
d =
_
2
kr
e
i/4
/(
o
)
Near the second stationary point the integral vanishes since 1 +cos(
o
) = 1 1 = 0.
Hence the result (8.5) follows.

MIT OCW Wave Propagation Lecture Part 3

Загружено:

Сведения о документе

Оригинальное название

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

MIT OCW Wave Propagation Lecture Part 3

Загружено:

Авторское право:

Доступные форматы

3.1.

Reection of sound by an interface 1

Hence (1.2) is satised if

exp[ik(xsin z cos ] (1.10)

exp[ik(x sin z cos )] + Rexp[ik(x sin + z cos )] (1.15)

is the ordinary derivative of with repect to its argument. Hence

. Let us introduce a new cartesian coordinate

axis is along, while the y

axis is normal to, the shadow boundary.

= y cos xsin (9.2)

3.9. Diraction of SH waves by a long crack 34

we can show straightforwardly that

directions. Substituting (9.4

is much longer than a wavelength,

kA, hence 2ik

Thus the transition zone is the interior of a parabola.

1 i.e., far away from the tip. This

Figure by MIT OCW.

r(/r ik) to both sides of (10.23), and using the

Вам также может понравиться