RAYLEIGH WAVES I N A TWO-LAYER HETEROGENEOUS MEDIUM

Margery Newlaitds

(Communicated by Robert Stoneley)

(Received 1950 June 20)

Summary
A method is developed for the solution of the equations for Rayleigh waves
in a semi-infinite incompressible medium in which a crust of rigidity varying
linearly with depth lies'on top of a uniform elastic medium of great depth.
The method is extended to deal with compressible media and in both cases
is applied to a model Earth consisting of a crust of depth 37-5 km.in which
the rigidity increases linearly from 2.3 x 1011 to 4.53 x 101l dyneslcm.' over
the ultrabasic material of infinite depth in which the rigidity is constant and
equal to 6-47 x 10ll dynes/cm*. The variation of wave velocity with wave-
length is obtained numerically, the same features emerging as in the problem of
two homogeneous layers, i. e. a distinct minimum group velocity when the
disturbance is of wave-length approximately twice the depth of the upper layer
and, in the compressible case, an additional less pronounced minimum for
wave-length about six times the depth of the layer.
Two sets of readings taken at Glittingen are chosen for comparison.
With the second at least there is no real agreement,so that the effect of adjusting
the constants of our trial model arises.
The applicability of the theory to the problem of microseisms is considered
and it seems that it may be found useful in the study of those microseisms of
period a fraction of a second.
Finally, x(l),~ ( ~those
1 , solutions of a particular equation which it is found
convenient to use here, are related to the appropriate Whittaker confluent
hypergeometric function, already known to be a solution.

I. INCOMPRESSIBLE
MEDIA
Introduction
A study of Rayleigh waves in the Earth's crust must of necessity allow for the
changing properties of that crust. The treatment of it as a two-layer semi-infinite
medium would be only a first step towards the true description, but if we were
t o specify three or more layers the algebra and computation involved in obtaining
the equation for the wave velocity would be intractable. Above the ultrabasic
layer, which is of considerable depth, there are three other distinct layers, the
intermediate, the granitic and, at the surface, the sedimentary.
We therefore attempt to represent the variation by a distribution which is
continuous throughout a single finite layer lying on an infinitely deep layer in
which the properties are constant. Provided then that the disturbance penetrates
considerably into the lower medium, i.e. has a wavelength of at least twice
the depth of the upper layer, we may hope that our Earth model will be a good
approximation to the true stratified form in that it may reproduce the features
of the observed group-velocity-period relation.
II0 Margwy Newlands
The problem of Rayleigh waves in an incompressible medium in which the
density, p, is constant and the rigidity, p, varies linearly with depth was solved
formally by Stoneley (I)in terms of the Whittaker functions Whk, with suitably
chosen k and m, but in applications it would be necessary to evaluate the corre-
sponding series for relatively large values of the variable. Few tabulated values
were available at the time of that investigation and only a few special cases have
been computed since. Moreover, there was no corresponding formal solution
in the compressible case and not much progress was made with this problem.
A new approach therefore seemed desirable.
Equations of Motion.-Suppose that we have a layer of infinite horizontal
extent and thickness T, of which the upper surface is free and in which p is.
constant and p increases uniformly with depth, from po at the upper surface
to p2 at the lower.
We shall define a system of right-handed axes Ox, Oy,Oz,with origin in the
lower surface of the layer, measuring z vertically downwards and x, y horizontally.
Then at any point in the layer, p =p2+ plz, where p2, pl are constants.
Now consider a plane wave of the harmonic type, propagated in the x-direction
with wave velocity c and wave-length 2 z - l ~ .
Let u, w be the displacements in the x-, z-directions at the point (x, y, z ) at any
time and suppose that apart from a factor exp { i ~ ( x -ct)}, u and w are functions
of z only.
The equations of vibratory motion in two dimensions of an elastic solid are

where, since the medium is incompressible,

au
A~-+-=o.
aw
ax ax
Substituting

in (1.3), and writing -l7for the limit of M as h and A tend to infinity and zero
respectively (2), the equation (1.3)reduces to

Also, by virtue of the linear variation of p, (1.1) and (1.2)take the fairly simple
forms
a ( -l7+ 2p1w + pK"2$6)
-
ax
+ &a (pV2x + pK"2x) =0 ,

a (pV*x +p."c"x) =0
a ( - n+2p1w +p."C"$6) - -
-
a2 ax
 Rayleigh waves in a two-layer heterogeneous medium II€

satisfied by
17 =2 p 1 W + pK2C2+,
pv2x = -pK2c2x.
Without causing ambiguity, we may write
4 =+(z)exp {&(x - ct)},
x = x(z) exp {iK(x-ct)}.
Then (1.5) and (1.9) are equivalent to
I
(1.10)

(1.11)
and solutions of the first of these equations are of the form
+=
Pcosh~z+Qsinh~z. (1.12))
, *bc2/pi, where Po is the velocity of distortional waves.
Putting b = ~ p J p ~k =
in medium of density p, rigidity p2, i.e. Po=1/1-La/p and [ = 2 ( ~ 2 + b ) , the
equation (I .II) becomes
(1.13)
and solutions are W+*([), where the W-functions are the Whittaker confluent
hypergeometric functions (3).
Instead of working in terms of the Whittaker functions, we attempt to solve:
the equation (1.11) directly. We first p u t Z = I + p1z/p2,so that theequation(I.II)$
becomes

then look for a series solution in powers of (~p~/pJ?,

writing
X(Z) = X O W + (KP2/P1)2x1(z)+ * - - + (KP2/PJ2"Xn(Z)+ * * > (1.14)l
so that we must have
ix;+ (%/Pl)"x';+ * - .1-(KP2/P1)2(I -PC2/P2Z){X0+ (KP2/Pl)2xl+- * .I =o. (1.15)
Thus if
x; =o,
x'; = (1 - PC2/P2Z)X0,

x11+1= (1 -PC2/P2Z)Xn, (1.16)

and the resulting series (1.14) both converges and involves two arbitrary constants,
it is a valid solution and in fact the most general solution of the equation (1.11).
We may take
XXI) = Xn(1) =or (1.17)
XQ) =AzPz/Pl, (1.18)
X O W =A, + A @ - IIh/Pl (1.19)
=A, + 4 2 .
G a
 (1.21)

This is in effect the operator method (4) of solution of differential equations.

From a rough knowledge of the range of c, the wave velocity, we know that
the factor (I -pc2/p2Z) is certainly bounded for all Z in a range in which p never
vanishes. Denote the maximum value of I I - pc2/p2ZIby F. Then

= ~~
1 4 11
t(l-Z/~2t)[A1+A2(t-I)~2/~~dt~

Q {I Ai(Z- I F/2 + I I)’I p 2 F / h 3 !}

and
IxllI <{IAl(Z- ~ P/(zn)!+
) ~ ~I A ,l( Z - 1 ) ~ ” + ~ 1 p ~ F n / p ~ I)!)
(2n+
.and thus the series converges like the exponential series.
With the data which we shall use later, a rough consideration shows that
four or five terms of the series will give a very reasonable degree of accuracy,
particularly in view of the uncertainty of the elastic constants.
Using the equality

(ZnlogZ/n!)dZ=Zn+l logZ/(n+ I) !- (Zn+l- ~ ) / ( n I+)@+ I) ! (22I),

solutions xcl) and xc2) are readily obtained as follows :

X ( l ) = I + (Kp2/pl)2{(gz2+(E-I)z-(E-(g)-EZlogZ)
+ ( ~ p ~ / p ~ ) ~ (3
{ &-~g)zs+
+ (&- :e2)Z2+(c2- g)Z
+ (& - + 19)-z log Z(g€z*-( g E 2 2 -
€; €a+ &)}
+ ...+ (Kp2/p1)2n{u&)+ log 2. V$i)-1) + - .- - (1.22)

$”=(z- I)+(Kp2/p1)2{gp+ (-(gE-(g)z2+(gz+

(&- g)+€zlOgz)
( -&-+)Z4+(L-L
+(~p~/p~)~{+ J~+ 1 2 2 3
12 18+* )
+.(- &+ &2)Z2+ (&+ - &2)Z+ (-A+ -&2)
&€ &€

+ z log Z(g€z2- &2Z + & -&a))

+ ...+ (Kp&.b1)2n{ U$Z+1+ log z .V$?-1) + ..., (1.23)
where UCl) UJ2)and Vfcl), Vi2)are polynomials in Z of degree j and K respectively
’ ’.
and E is written for pc2/p2.
Application.-The variation in the rigidity of the Earth‘s ’crust may perhaps
b e represented roughly by a linear increase to a depth of about 37.5 km. and a
constant value for greater depths.
 Rayhigh waves in a two-layer heterogeneous medium 113
As a mean value we take
2-31x1011 dynes/cm.a in the sedimentary layer of the upper 4 km. and
Jeffrey8 (5) gives
3.00 x 101l dynes/cm.2 in the granitic layer of the next 15 km.,
3-99 x 10ll dynes/cm.2 in the intermediate layer of the next 18-19 km.,
6-47 x 10ll dynes/cm.2 in the ultrabasic layer.
We shall therefore take as a model a surface layer of depth 37-5km. on a uniform
basic layer of great depth and shall have
po (at the surface) =2.31 x 10ll dynes/cm.2,
p2 (at the interface) =4-53 x 1011 dynes/cm.2,
p (at intermediate points) =p2+ plx,
pa(in the basic layer) =6.47 x 1ol1 dynes/cm.a,
where p1 is a constant and 2 is measured vertically downwards from the interface.
Denote the density by
po (upper layer) = 2-72 gm./~m.~,
p' (basic layer) =3-40 g m . / ~ m . ~ ,
5
where p'/po =6 =a.
Let the depth of the layer be T = 37.5 km.
We assume that in the upper layer
+ = ( P c o s h ~ X +Qsinh~2)cOs~(X--Ct),
(1.24)
x =(A,x(l)(Z)+ A,X(~)(Z)) sinK(x-ct),
and in the lower (6)
4 =Rexp( -KZ)COSK(X-~~),
x =Sexp ( - K ~ z ) sinK(x-ct),
where
n2 = (I- p'c2/p3),

and we have chosen the negative exponentials in (1.25) because our disturbance
is assumed to be a surface wave. With the usual notation the displacements are

(1.26)
From (1.28)we obtain P, Q, A,, A, in the form

Writing
+E P = alR + blS,
~ E =azR
Q bzS,+
BEA, =a,R + b3S,
&Az =a4R+ b4S. I
the consistency of (1.29)as simultaneous equations in R and S implies that

We thus have an equation for E and hence the wave velocity c. Values of E were
calculated for a considerable range of K ; the results are quoted below with the
corresponding wave velocity and the derived group velocity C. The latter
 Rayleigh waves in a two-layer heterogeneous medium 115
has a distinct minimum of about 2.03 km./sec. for waves of period 27 sec. As
a check on the theory the limiting form of equation (1.33)was found as KT+o,
i. e. as the depth of the layer becomes a negligible fraction of the wave-length of
the disturbance. It is
( 2 - pfc2/p3)2=4(I -pfc2/pS)l'2,
which is the equation for the velocity of Rayleigh waves in a semi-infinite medium
composed of the material of the basic layer.
The relevant solution of the foregoing equation is
pfC2/p3= 0.917 . ..,
which is more conveniently expressed in the form
E =P0C2/P2 =o*917* - x (PoPslPfc12)
=1*043....
We might now attempt to compare our theoretical results and actual observa-
tions, but for a better comparison allowance must be made for compressibility.

2. EXTENSION
OF THE THEORY MEDIA
TO COMPRESSIBLE
When the medium is compressible, the variation of the elastic constant A
enters into the theory and the natural extension is to the case where h varies
linearly with depth. It greatly assists numerical manipulation, without in any
way altering the principle of the method of solution, if we assume that h varies
at the same relative rate as p. Such results as are available on the elasticity of
rocks do suggest that this is in fact very near to the truth.
We shall therefore assume
ho/Po =Ai/Pi = h z / P z = h d P 3
and seek to take account of the compressibility of the media.
Equations of Motion.-We again substitute

in the equations of motion (1.1) and (1.2). We have now no incompressibility

condition, but on the other hand we have not a function II occurring in the
equations (1.6))(1.7)and (1.8). Thus we find, once more by virtue of the
linear variation of A and p, that these equations reduce to the simple forms

which are satisfied by

(A + 2p)v2$+PK2C2$ - 2p1ax -0,
116 Margery Newlands
Extending the previous method to these simultaneous equations we a m m e
solutions which, apart from a factor exp {iw(x- c t ) } , may be written

+
p/(X 2p) = 8, a constant,
and
h+2p=v,
thus defining v, v2, v3 and vl.
Substituting (2.5) and (2.6) in (2.3) and (2.4)we have

${fo+Zf1+**.}+ P1
(" > " ( ~ - 1 ) { f 0 + ~ f 1P+
l. . . }

- 2 - P1
KPa( I4
)
g o + K h g l + . . . =o,

+ ...} + ("
P1
(& -
Psz
I) {go+ 3 g 1+
P1
...}
-2 (%){f0+
P1Z ""fl+...}
P1 =o.
Hence
d2f
d2fo
dT=o; & = 28ZZO'
d p - z f o , etc'
gk2-2
dZ2
It is seen that four distinct solutions are obtained with
fo=I, go=o; fo=z, g o = o ; fo=o, go=r; fo=o, go=x.
Denote these by

(2.11)

Then, since the simultaneous equations are equivalent to a fourth degree equation
for each of 4 and x, we have as general solutions

Successive terms fl, f2,

4 =4 4 1 + A 2 4 2 + A 3 4 3 + A 4 4 &
x = 4 x 1 + A 2 X 2 + A 3 X 3 + A4X4.
...,g,, g,, ...,are obtained as follows.
1 (2.12)

When f o f 0,and goE0,

fi'r = (1 -c'/Z)f2r-2 + (28/21g2r-1,
& r + l = ( ~ -~/zlg2r-1+(2e'/zlf2r,

far+l =g2r =o, I

 Rayleigh waves in a two-layer heterogeneous medium 117
where

E'

&=I.
E =Pc"lP2,

= €e,
8 =p/v, constant,
I
The symbol 8' is introduced for the purpose of symmetry, the casesfo =o, go= I , z
being immediately deducible from the cases go=0,fo= I , z by interchange of E
and c', 0 and 8') for then

Integration follows the same lines as in the incompressible problem, only integrals
of the type l Z n l o g Z d Z and JZndZ being involved, and convergence of the
series solutions is similarly established. We find that
4 1 = ~ +(Kp2/p1)y(+-5e)z2+
( E ' + ~ (++ e--E')+ziogz[2ez+ (4e-qi)
- ~ ) ~ +

+( ~ p 2 / p ~ ) 4 { ( & - * y p ) Z 4 + (- i + @ - v + + ' + + ' 8 ) Z 3

+ (4 - e+ 82- p+d(I + 2e))za+ ( - f + €'a+ p+ @2- &(I + 2e))z
+ (&+ ge+ ke2-3' + i E ' 2 - g ( 1 + 2e)) + ziogz[. ..I)
+ ...+ (.p2/p1)"{ u,, + z log ZV2,4,1} + ....
1 (2.16)
4 2 = (KP2/CLl){Z--11
+
+ ( K C L ~ / P ~ 8) ~+( (P ) Z 3+ ( - - 3 ~+' 3 W 2+ (4 - 3 W + ( - 8 + +E' - 46)
+ z log z[ - 2ez+ - 2e)l 1
(€'

+ * * * + (KP2/PdZn+l{U2n+1,2 + z 1% ZV,,-l, 2). (2.17)

Xl=(Kp2/pi){ -2Z-k 2+2ZlOgz}+ -$-p)z3
( K P ~ / ~ ~ ) ~ { (

+ ( - 2e+ % + (48- zE- z E ' ) z +($+e; - -

%')z2+ $8)

+ ziog z[(i+ge)z2+ (48- -q+ ( I + 28 - zE-241)

+ - - + (Kp2/p1)2m+1{ +z log .Yan, + ..- -

Wan+l,1 1) (2.18)
KX2 = (Kp2/p1)2{Z2 I - - 2 2 log z}
+ ( ~ p ~ / p ~ ) ~ { ( ; + (;+ -& - g)z3+....>
+$)Z4+

+ - - + (Kp2/p1)2"{Wzn, + z log 2.
2 y2,4 2) -k . .'- , (2.191
with similar series for d3, 44, x3, x4, where Urn,*Vm,$,etc. ( i = q 2, 3, 4) are
polynomials of order m in 2.
Application.-We shall now apply these results to the previous problem of
Rayleigh waves in a model Earth consisting of a crust in which p (and here, A)
varies linearly with depth, on a base of constant p and A. Assuming that for

and (7) for

X>O {4 = COSK(X

x=Se-k"zsin~(~-ct),
-Cf),
(2.21)
I18 Margery Newlands
where m 2 = ( I -p'c2/v3),
n2 =(I - p'c2/p3),
we shall have at z = 0,
1

(2.22)
x =A39
&- k d X = A4'
dz p2dZ
The displacements are

(2.23)

.and the tractions are

(2.24)

and

(2.27)

ci = 4 I

4 =&
-&EPZ/PO)Xi+ (P1/Po)di- 4s:
- &h/PO)+l+ (P1ho)Xi- x:, I (2.28)

the boundary conditions at z = - T , viz. $2 = =o, regarded as simultaneous

equations in R and S, require for consistency

(2.29)
 Rayleigh waves in a two-layer heterogeneous medium 119
an equation for E, and hence the wave velocity c. As a check, we determine the
limiting form of equation (2.29) as KT+Oand find it is
(I - 9p'c2/p3)2 = (I - p'c2/p3)42( I - p'c2/v3)lla
the equation for the velocity of Rayleigh waves in a medium of the. material of
the lower semi-finite layer. The relevant root of this is
p'c2/p3 =0.8387 ...,
giving the result
~ = p ~ C= ~ 0.8387..
/ h . X popa/p'p2
= 0958.. ..
For the ratio p/(A+ zp) we may take as mean value 0-35... (it is generally greater
for the sedimentary rocks and smaller for the lower ultrabasic).
Calculations, similar to but considerably more tedious than those of Section I,
give a distinct minimum group velocity as before, but in addition, a definite but
less pronounced minimum for small KT. This was also observed by Lee in calcula-
tions on the velocity of Rayleigh waves in a two-layer semi-infinite medium in
which there is no variation of elastic constants and reference is made to it by
JefTreys (8) in a paper in which he produces the same feature by an approximate
method. I n this case it is a very flat minimum (9),
-.
i. e. the group velocity is almost
constant for periods between 50 and go sec.
R6hrbach (Fig. 3)
ZmdE-W
Compressible Incompressible components

Period C C K*T C C Period Period C

a+ 3'99 3'99 0'010 4-17 4-17 +a 40.2 3-36

238 3-90 3'95 0.250 4'15 4'13 227 43'0 3'44
121 3-76 3-89 0.500 4-13 4'09 1I 4 46.0 3.52
82.4 3-60 3-81 0.750 4-11 4-05 76'4 49'7 3'59
71.2 3'59 3'78 0.875 4.10 4-02 65-6 55.2 3.68
62.6 3.58 3-76 I '000 4.09 3'99 57'6
55'9 3-61 3'74 1'125 4-08 3'91 51.4 R6hrbach (Fig. 4)
50'5 3-62 3'73 1'250 4-05 3-82 46'5 ZandE-W
46.1 3'53 3'72 1'375 4-03 3'71 42'5 components
42'5 3-40 3-70 1.500 4-00 3'58 39'2
39'5 3-18 3'67 1.625 3'96 3'39 36.6 26.7 2-91
37-2 2.99 3.62 1.750 3-91 3.16 34'4 27-1 2-93
35'1 2.83 3.58 1.875 3-85 2.89 32.6 27.1 2-97
33'5 2-48 3'52 2'000 3'78 2-55 31.1 28.0 2.95
32.1 2.05 3-45 2.125 3-70 2-30 30.0 30-2 2-98
31-2 1.70 3-36 2'250 3-62 2-21 28.9 33'1 3-04
30'4 1'79 3-26 2'375 3'54 2'12 28.0 34'5 3-02
29.5 2.03 3-20 2.500 3'47 2-03 27'2 36.2 3-06
28.5 2-15 3-15 2.625 3-40 2'12 26 -4 38.2 3-13
27.5 2-30 3-11 2.750 3'35 2-44 25 -6 42'4 3-19
3 '07 2'875 3'32 48.8 3.25
50.1 3'34
52-6 3'40
56.8 3-50
65'3 3'59
69.2 3-72
74'3 3.86
I20 Margery Newlands
For comparison of our model and the Earth’s general structure over Eurasia,
some readings made at Gottingen and reproduced in two graphs by Rohrbach (10)
were used. It was necessary to enlarge these graphs considerably in order even
to separate a number of the points and also convenient to replace the parameter A
(the so-called “wave-length”) by the period, so that the re-reading of data may
not be accurate but certainly sufficientlyso to show that there is no real agreement
with the theoretical curve.
One set of readings (small circles) for periods of about 50 sec. and wave-length
about five times the depth of the model layer may be said to lie approximately
on the curve but they are too few to justify any positive conclusions. On the’
other hand, the second set (small crosses) do not.
Group velocity
in km./sac. Incomprorsible

3.8

2.0

30 60 90 I20
Period in seconds
FIG.I .-Vmiotion of grot@ velocity with period within probable range of validity of
the linear layer theory.
0 from Rdhrbach (Fig. 3). x from R8hrbach (Fig. 4).

At best we might attempt to obtain better agreement by adjusting the

constants of our model. It is perhaps worth indicating to what extent this may
be done without further computation.
Our calculations depend on the factors
KT, PiT/pa( = - ~ 0 ) / ~ 2 ) , 4 = ~ 0 8 / ~ 2 ) , E“( =~ ‘ ~ ~ 1 ~ 3 1
and the ratios
Po: Pa :Ps-
If then all these factors remain constant but we now have a layer of thickness m T
and rigidities npo, npa, np3, densities unchanged po, p’, the shape of the group-
velocity-period curve will be the same but the appropriate period and group
velocity readings will be derived from the original by multiplication by the factors
mn-lIa and nila respectively. For, if KTis unaltered, we must replace corresponding
K readings by Kim, and if E is unaltered, replace c by n%. The group velocity
and so varies like c and we read n W ! for C. The period KC similarly
C =d ( ~ c ) / &
varies as (KC)-’ and so we read mn-lla x original period.
 Rayleigh waves in a two-layer heterogeneous medium 121

3. APPLICATION OF THE THEORY TO THE PROBLEM OF MICROSEISMS

It now seems that the theory may have its best application in the study of
microseisms of periods a fraction of a second. These appear to be generated by-
man-made disturbances such as traffic or road working and probably do not
penetrate more than a hundred or two hundred feet. Immediately below the
surface and through, perhaps, the first twenty feet of material, e. g. gravel, chalk,
there will be a considerable and most probably linear increase in the elastic
constants as the material consolidates. At greater depths the material will be
more or less uniform with only a very much less rapid increase in h and p known to
exist to considerable depths and taking place linearly with the initial increase of
mean stress but tending to zero as the stress increases further (XI). I n our present
problem, concerning layers of thickness perhaps two hundred feet, we may be.
able to neglect this and treat our medium as consisting of a layer of 10-20ft. in
which consolidation takes place, i. e. h and p increase linearly from &, p,,, say, at the
surface to &, pa at the face overlying a semi-idinite medium in which they are
constant and equal to &, pa. We might assume further that at the surface the
ratio of the volume of cavity to volume of solid material, although important in
its effect on the elasticity, is sufficiently small for us to treat the density as constant.
throughout the whole medium.
It is hoped that the linear layer theory may help to explain some of the features
of the seismograms which are obtained under such conditions.

4. THERELATION BETWEEN x ( l ) ( Z ) ,x(*)(Z)AND Wk,ua([)

In the first part of this paper we found two functions x(')(Z) and x(a)(Z)
satisfying the equation (1.13), viz.

where
z = I + p1z/p2.
Before deriving x(l), we noted that Wk.112 ( f ) also satisfies (1.13), where the
W-function is a Whittaker confluent hypergeometric function and
=3(h-P2/Pl)E,

f =~ ( h - ~ Z / ~ l ) Z .
This form of solution was used by Stoneley (12) and he gives the expression for
WkSm(z) (I=) (derived from Goldstein's series (13)for e-l'seWk.m(z),where 2m is.
an integer). From this we may deduce the expression for Wk,1,2(f), and this is.
easily developed into the form
Wk.l,2(0=(w),+logZ- (W>B, (4.1)

where (w), and ( W), are polynomials in 2.

Similarly we may write
122 Margery Newlands
Since (1.13) is a linear equation of the second order, it has only two independent
solutions. Therefore a relation of the form
Wk.l,2(o=cx(')(z) + W 2 ' ( Z ) (4.4)
must exist, where C and D are constants. We find that ($l))B, ($2))B and ( W ) ,
are all multiples of the same series, as indeed they must be by the theory of the
linear equations.
We now proceed to determine C and D by comparison of the constant terms
and the Z terms in (x(l)),, (W), (comparison of ZlogZ terms give the same
relation as comparison of the constant terms).
Comparing coefficients of Zoin (4.4), we have
+ (A- +
x/r(a) = C[I - ( E - + ) ( ~ p ~ / p ~ )3~ &2)(~p2/p1)4. .] +.
+D[- I + -~ ) ( K c L + -A+
~ /( P ~)~ -+p2)(~c12hd4+ - - -1, (4.5)
and of 2, we have
2 % / P l ~ ( a - 1){log(2KPZIP.I)+ +(a) - +(I) -+(2N - {KPZ/Plr(a))
- I ) ( K P ~ / P+~ (e2
= C{(e 1 ) ~ -1
)~ - 8 ) ( ~ 1 " 2 / ~+
+ D{I + H K C L ~+/ C (kL+~&+) ~- & 2 ) ( ~ ~ d +~ l- -)-1,
4 (4.6)
where Y(z) is the logarithmic derivative of r(x), so that Y(m+ I) = m-l+ Y ( m )
and - Y(I) is Euler's constant. The determinant formed from the coefficients of
the right-hand sides of (4.5) and (4.6) is identically equal to unity, so that writing
-
6 = [ 4 o g (2KP2/Pl) + - W)- Y(2))l
we deduce that
C = [1/1"(a11[{(1- H K P ~ / C+L(8.~-) ~Q)(KPz/P~)~
( K P ~ P+~ )
) ~- M K P ~ / P+~ )- ~-11
+ (& + & -% 2 ) ( ~ ~ 2 /+~{; 1 )-4 ( K P ~ / P+~(46
and
- ( K ~ c-L(e ~- )I ) ( K P ~ / +
D = [~/r(a)lC{ P (~E ~- t ) ( ~ ~ 2 / ~ 1 ) '
- -
- (e2 - B ) ( K P ~ / ~-1)+~i{. - ( K P ~ / P+~(c) -~ H K P ~ / P ~+) ~->I.
This relationship was verified by comparing coefficients of Z2,Z8, Z4.

5. APPENDIX
It is perhaps worth while to consider the possibility of extending the above
method of solution of the elastic wave equations when the variation of h and p
with depth is quite general.
We shall write

The substitution
A =A(x),
P P"(z)- I
 Raykigh waves in a two-layer heterogeneous medium 123
no longer appears to lead to a simple separation of the equations of motion into
two second-order equations in 4 and x, and so we proceed directly to assume

+ +
u = (uo Ku1 + ...) sinK(x -ct), (5.2)

W = (W0fKW1+K2W2+ ...) COSK(X- Ct). (5.3)

Substituting these series in the equations of motion, we obtain

d d
&( / 4 + 2 ) = +
(pWr+l) A -dwr+l
dz + [A + 2p -pcqur (r 2 0).

d
-([A+2p]w;)=o,
dz
d
-([A+2p]wu;)=
dz
- -(Auo)-p-
d
dz
duo
dz ' } (5.5)

From the first of (5.4)and (5.5) we derive four independent solutions with leading:
terms
(I) u0=A1, wo=o,
1

Again effect the integrations from o to z so that

When A(z) and p(z) may be written A,+", p,,zn or AoeQZ,poegZ,for example, the
integrations involved are fairly straightforward, but the practical value of such
variations is very questionable.
We must first obtain a much better idea of the variationofthe Earth's properties
physically before there can be any merit in attempting a detailed application to
numerical data.
124 Margery Newlands, Rayleigh waves in a two-layer heterogeneous medium
ACKNOWLEDGMENTS
I wish to express my thanks to Dr R. Stoneley who proposed the above
investigation and offered much helpful criticism during the writing of this paper,
a n d also to the Department of Scientific and Industrial Research for a grant.
Neumham College,
Cambridge:
~ 9 5 May.
0

References
(I) 1.Stoneley, M.N., Geophys. Suppl., 3,222-232, 1934. See also C. L. Pekeris, Physics,
6, 133, 1935. For a semi-infinite medium with the property p/p=constant, he was
able to determine the wave velocity c as a power series in the wave-length for the
case p= C cosha (a+/?2)which by suitable choice of a is equivalent top=po cosh*p2,
poeaflz,orp.o(I+/?z)2.
(2) Fonowing A. E. H. Love, Some Problems in Geodynamics, Cambridge, p. 92,191I.
(3) E. T. Whittaker and G. N. Watson, Modem Analysis, 2nd Ed., Cambridge, p. 331,1915.
(4) H. Jeffreys, “ Operational Methods in Mathematical Physics ”, Cambridge Tracts in
Math., No. 23, 1927.
f5) H. Jeffreys, “ Further Study of Near Earthquakes ”, M.N., Geophys. Suppl., 3,
131-156, 1933-
(6) These are standard solutions in a semi-infinite incompressible medium of constant p, p.
(7) These are standard solutions in a semi-infinite compressible medium of constant p, A, p.
(8) H. JefFreys, (‘The Surface Waves of Earthquakes ”, M.N., Geophys. Suppl., 3, 253,
1935.
(9) The effect of this on the seismogram is discussed by H. Jeffreys, M.N., Geophys. Suppl.,
I,286-289, 1925.
(10)W.Rohrbach, “ uber die Dispersion seismischer Oberflachenwellen ”, 2.far Geophysik,
8, 113-129, 1932.
(11)See, for example, F. Birch and R. Law, “ Measurements of Compressibility at High
Temperature and Pressures ”, Bull. Geol. SOC.A m . ,46, 1219-1250, 1935.
(12)L a . Cit. (I), p. 225.
(12a)LOC.Cit. (I), p. 227-read e-112 for e-z in (23) and (24).
(13)S. Gdldstein, Proc. Lond. Math. SOC.,Ser. 2, 34, 110, 1932.