PLANE STRAIN AND AXIALLY SYMMETRIC

CONSOLIDATION OF A CLAY LAYER ON A

SMOOTH IMPERVIOUS BASE
By R. E. GIBSON {King's College, London),
R. L. SCHIFFMAN {University of Illinois at Chicago Circle)
and S. L. PU {Formerly at Rensselaer Polytechnic Institute)

[Received 7 October 1969. Revise 3 April 1970]

Downloaded from http://qjmam.oxfordjournals.org/ at University of Sussex on May 18, 2012

SUMMARY
Loaded clay strata settle gradually due to the squeezing out of water from the pores
of the soil. It is important for the civil engineer to be able to assess the magnitude
and progress of this consolidation settlement. Analysis has hitherto been confined
to cases where the thickness of the stratum is very great compared with the dimen-
sions of the loaded area.
The problem of the porous elastic layer of infinite lateral extent, to the surface
of which a uniform pressure is applied over an infinite strip or over a circular area,
is treated here. Some results of interest have been evaluated and are presented in
graphical form.

1. Introduction
IN foundation engineering it is common to employ the one-dimensional
theory of Terzaghi (1, 2) to estimate the progress of consolidation and
settlement of loaded clay layers. The theory is based on the assumption
that both the displacement of this porous medium and the flow of water
through it occur only in a vertical direction, a situation which rarely arises
in practice.
Some workers (3-6) have sought to extend the theory to take account of
more general states of deformation and flow by assuming that the pressure
in the fluid filling the pores of the medium is governed by an equation of
the heat-conduction type. This unlinking of pore fluid flow from the prog-
ress of deformation of the medium requires that the components of total
(bulk) stress remain unaltered everywhere as consolidation proceeds (7).
Indeed, in order to provide a rational basis for the unlinked theory, the
plausibility of this invariance has been argued on physical grounds.
Such assumptions are unnecessary in the theory of three-dimensional
consolidation developed by Biot (8-12). Only the physical properties
of the porous compressible medium and of the fluid filling the pore space
require specification; these have been discussed in detail elsewhere (8,
13-15). The simplest form of the theory, which we adopt here, is encoun-
tered when the skeletal medium is perfectly elastic, the fluid (usually
[Quart. Joora. Mecb. and Applied Math., Vol. XXIII, Pt. 4, 1970]
5092.4 Mm
506 B. E. GIBSON, R. L. SCHIFFMAN AND S. L. PU

water) is incompressible and the relative velocity between the two phases
is governed by Darcy's law.
In order to assess the effects of the assumptions leading to theories of
the first kind (unlinked), comparisons were made between these theories
and the comparable situations evolving from Biot's theory (15). The
unlinked theory generally produced results which were in poor agreement
with those derived from the true three-dimensional theory.
Earlier work by the authors (14-19) and others (10, 20-23) on particular
problems of three-dimensional consolidation has been concerned exclu-
sively with a porous elastic medium occupying the half-plane or half-space.

Downloaded from http://qjmam.oxfordjournals.org/ at University of Sussex on May 18, 2012

Their solutions have application only to problems arising in engineering
practice where the thickness of the clay stratum is great compared with
the dimensions of the loaded area. When this is not so the use of numerical
results appropriate to the half-plane or half-space to estimate, for example,
surface displacements, may well lead to serious errors the magnitudes of
which cannot at present be assessed.
This problem merits attention and we shall use Biot's theory in this
paper to evaluate the displacements in a loaded clay layer (0 < z < h) of
infinite lateral extent resting on a rigid impervious medium. At the inter-
face (z = h) the contact is presumed to be perfectly smooth. We recognize
that a condition of complete adherence might reflect more closely the
restraint likely to be present in nature, but the algebra is then found to be
incomparably heavier and for this reason we have deferred examination
of this allied problem.
We make use of the method developed in an earlier paper (17), where
the equations governing problems of plane strain and axially symmetrical
consolidation were formulated in terms of two displacement functions
E and S, each of which is dependent on two space variables and time and
obeying the equations,
cV4E = V\dEjdt), V2S = 0. (1)
Solutions to plane strain and axially symmetrical problems are developed
in parallel by means of these displacement functions for those cases where
the surface loading is distributed uniformly over a limited area D of the
plane boundary (z = 0).

2. The governing equations

We restate briefly and with a minimum of comment, but with sufficient
completeness for our present purpose, the equations which govern the
displacement of an isotropic, homogeneous, fully saturated porous elastic
medium (9, 11) for conditions of plane strain and axial symmetry (17).
 CONSOLIDATION OF A CLAY LAYER ON A SMOOTH BASE 507

To conform with present usage in soil mechanics and with our earlier
work, we regard compressive stress as positive.

2.1 Plane strain. In plane strain the components u, w of displacement

in the x- and z- coordinate directions and the excess pore water pressure
a must satisfy the equations

Be 1 da
' dx Gdx

Downloaded from http://qjmam.oxfordjournals.org/ at University of Sussex on May 18, 2012

(2)
dz Gdz
Y72 <k
e v e = —,
Bt
where the dilatation e is given by

6 ==
( dudx dw\
Jzj
the operator V2 by V2 = 92/3a;2 + 92/9z2, and G is the shear modulus of
the porous skeleton. The coefficient of consolidation c is defined by
c = 2Grjk[yw, where h is the coefficient of permeability of the medium, y w is
the unit weight of water and the auxiliary elastic constant rj is given in
terms of the Poisson's ratio v of the porous skeleton by r\ = (1 —r)/(l —2v).
It has been shown (17) that problems of the type we are concerned with
can be reduced to the determination of two displacement functions E
and 8, related to the displacement components and the excess pore water
pressure by

_ JdE dS dE , dS o
W = \-Z S,
dx dx dz dz (3)

2G

and governed by the field equations (1).

The (primed) components of effective stress acting in the skeleton are
connected with the components of total stress in the bulk material (skele-
ton and fluid) by equations such as

a
O'xx = xx— a< a
'xz = a
xz- W

Since the strains of the skeleton are connected with the effective stresses
508 R. E. GIBSON, R. L. SCHIFFMAN AND S. L. PU
through Hooke's law, it follows that the total stresses are given by

2G a X
n •5 2 ' -5 )

dx dz (5)
2G dz2

2-
d2sZdz2+dz'
2G dx dz dx dz

2.2 Axial symmetry. In those problems which possess axial symmetry

Downloaded from http://qjmam.oxfordjournals.org/ at University of Sussex on May 18, 2012

the displacement components ur, w in the r- and z-directions must satisfy
1 in -i\^e 1 do -
r—2«r-(2»?-l)----- =0,
(6)
r dr G dr
together with (2)2 and (2)3, where

~dr2^rdr dz2'
The components of stress and displacement are again given by relations
of the type (3), (4) and (5) above, save that u is replaced by uT, the space
variable x is replaced by r and

(7)
2G r dr r dr dz
The displacement functions still obey equations (1) if the appropriate
form of V2 is used and it is this circumstance which permits problems of
plane strain and axial symmetry to be treated in parallel. This duality has
been retained in the following sections by a slight and obvious extension
to the existing notation.
3. Solutions of the governing equations
In order to solve equations (1) we introduce new functions E and S
related to E and S by the following repeated Fourier cosine (or Hankel) and
Laplace transforms, the form of which is suggested by the symmetry of
the problems we shall be concerned with presently:
c+too

=~ f
c—too 0
C-f-lCO CO
(8)
 CONSOLIDATION OF A CLAY LAYER ON A SMOOTH BASE 509

b b

111111111 "Xtt

rfS/*S//SSS/SS f///////,

Downloaded from http://qjmam.oxfordjournals.org/ at University of Sussex on May 18, 2012

FIG. 1

where the kernels are

K(x, £) = (2/7r)cos(a:f), K(r, f) =

and the Bromwich-Wagner contour in the p-plane is taken to the right.

of all the poles of the integrands.
By the use of these transforms it can readily be verified that the partial
differential equations (1) are reduced to the following ordinary differential
equations for E and 8:

(9)

(10)

where £8 = p+p/c. Use has been made here of the condition that
initially the dilatation e of the medium is everywhere zero. The solutions
of these equations are

= A1 si 4 cosh(£z), (11)
S = cosh(£z), (12)
where Au ... , A6, which are functions of £ andp, must now be determined
from the boundary conditions imposed on the plane surfaces z = 0, h.

4. The boundary conditions

Solutions are sought to the problems of axial symmetry and plane
strain shown in Fig. 1.
The layer 0 < z < h is loaded at t = 0 over a pressed area D of its
510 R. E. GIBSON, R. L. SCHIFFMAN AND S. L. PU

plane surface z = 0 by a uniform pressure/which is maintained thereafter.

The region D occupies either the strip \x\ < b or the circle r < b. The
entire surface remains free from shear stress and the excess pore water
pressure there is zero. These conditions on z = 0 can be stated thus:

azl = 0 over \x\, r > 0 for t < 0,

in \X\' T < , for t> 0, (13)

axl = 0, ari = 0, a = 0 everywhere.

Downloaded from http://qjmam.oxfordjournals.org/ at University of Sussex on May 18, 2012

The base (z = A) of the layer rests in contact with a smooth, rigid and
impervious medium, so that, over this plane,

axl = 0, <r« = 0, w = 0, dojdz = 0. (14)

It is from the boundary conditions (13)3 and (14), holding for t > 0,
that the coefficients Alt ... , Ae are to be determined.
From (3), (5) and (8), all these conditions may be expressed in terms
of E and 8 alone. For example, the condition a — 0 becomes
c + ioo oo

c—<oo 0

Since this relation is required to hold for all values of x or r and t it follows
that the integrand must vanish on z = 0. We obtain, therefore, using (11)
A k = (c£/rjp)A6. (16a)
All the remaining boundary conditions, save (13)2) are of this type, and
from them we find

Ax = ^ 8 (c£/ W )tanh(W, (16b)

A2 = A5{h cosech(gh)sech(£h)—(c£/»7p)tanh(£A)coth(ffe)}> (16c)
A3 = -Ab(c£lr)P)tanh(th), (16d)
^4 8 = -^ 5 tanh(|A). (16e)
The remaining constant ^4B must now be found from (13)2. This boundary
condition, in terms of E and S, is '

Tf[>*+^1 * f * W * (n/2G1 " ^ <17a>

outslde
c—<oo 0
1° A
 "CONSOLIDATION OF A CLAY LAYER ON A SMOOTH BASE 511
which reduces to

^l Af , f W = L ^ ' (17b)
o
dzJ*=o \r J \0 outside D,
by virtue of (13)x. A further inversion may be effected using Fourier's
(or Hankel's) theorem, and we find that
y

Downloaded from http://qjmam.oxfordjournals.org/ at University of Sussex on May 18, 2012

according to whether the problem is that of plane strain or axial symmetry
and it follows that
A __J_/r18in(6|)
B
~2Opffl(f)|6r 1 J l (6|)' (18a)

where
Q(f) = 1 + |A cosech(^)sech(|^)-(c^/ W )tanh(^)coth(^)+c| 2 /^. (18b)
In the more general oase where the pressure / depends upon r (or is an
even function of x) and is applied at some prescribed rate, the constant
A 6 can be found by the same procedure as that adopted above, namely,
by a double inversion of equation (17a). The operations involved are
purely formal, but their validity must be established a posteriori.
We note in the plane strain problem that if the pressure / is distributed
unsymmetrically, qua function of x, a complex Fourier transform in the
interval — oo < x < oo is required in place of the Fourier cosine trans-
form of (8). Finally, if the surface loading is distributed non-uniformly over
a circular area, the present method can be extended (19) by the introduc-
tion of a third displacement function.

5. The surface settlement

The displacement functions appropriate to these particular problems
may now be recovered from equations (8), but the resulting formulae
will be of little value for computational purposes unless some further
reduction can be effected. Our concern will be with the settlement of the
loaded surface z = 0 and the manner of its development with time, since
this is the information of greatest engineering interest.
The vertical displacement of any point of the medium is (cf. section 2)

dS _ dE
w= z-—8-—.
dz dz

On z = 0 it can easily be verified that the first and last terms vanish,
512 B. E. GIBSON, R. L. SCHIFFMAN AND S. L. PU

leaving
C+tOO CO

hi \ l^i*
where w0 is the surface settlement.
By introducing new variables,
X = £h, s= h£, (19)
the surface settlement may be expressed in the form

Downloaded from http://qjmam.oxfordjournals.org/ at University of Sussex on May 18, 2012

<? /x x
(2G/f)w0 = J r , X\D(X, «)A-1 tanh A/(A,<) <ZA, (20)
0 \ /

where

D(X, t) = (1+A cosech A sech A)"1 exp{-(A2c</&2)},

T(a;, A) = — sm(Xblh)cos(Xxlh),
•nX

T(r, X) = 6J1(A&/^)J0(Ar/^).

The integral along the Bromwich-Wagner contour in the s-plane is

?+<OO

- f
2TTI J
C(X, s)exp(c<s/A2) ds, (21a)

where,
C(X, s) = {s+L{X)-M(X)si tanh s*}"1, (21b)
L(X) = AV 1 ( 1 +^ cosech A sech A)"1—A2, (21c)
1 1
M(X) = A?;" coth A(l+A cosech A sech A)" . (21d)
The problem of locating the poles of C(X, s) is discussed in an Appendix
to this paper. It is shown there that the denominator of the integrand of
(21a) has an infinity of zeros along the negative real axis of s at s = —a2,,
where an are the roots of
a.2-L{X) = a-M(A)tan a, (22)
2
together with an isolated zero at s = A . The contour integral then reduces
to a circuit around these zeros, the residues at the poles along the negative
s-axis being

while at the isolated pole the residue is (2r] — l)/2r].

 CONSOLIDATION OF A CLAY LAYER ON A SMOOTH BASE 513

This enables us to express the settlement in the form,

w,

where
r>/ n V* exp{—(a£+A2)c</A2} ,„,•,
.r(a, /) = > - -, (^4b)
and
JP(aB, A) = — K l + t a n ^ + a ; 1 tan «.„). (24c)

Downloaded from http://qjmam.oxfordjournals.org/ at University of Sussex on May 18, 2012

M(X)
5.1 TAe immediate and final settlements. We note that in the limit as
t -*• 0 through positive values, the settlement wQ is finite and this is the
'immediate' settlement which occurs as the loading is applied. The most
simple expression for this quantity is not from (24a), but indirectly from
the final settlement relationship,

if",;*
wol , oo = 2i i 1 dX. (25)
\r / (2?7—\)G J A(l + A cosech Asech A)
o
We remark that (25) is a known expression for the settlement of the
surface of an elastic layer. This result might have been anticipated, for
in the limit t -> oo the pore water merely occupies the pore space at zero
pressure a — 0 while the elastic skeleton alone carries stress.
Now, at the instant of loading the porous saturated medium behaves
as an incompressible elastic solid of shear modulus G, since the presence of
pore water does not contribute to the rigidity of the skeleton. This
suggests that the 'immediate' settlement would be found from (25) in
the limit rj —>- oo:
(26)

This surprising result, that the ratio of the immediate to the final settle-
ment is the same for all points on the surface z = 0, has earlier been
established (16) for the semi-infinite medium. It would, clearly, not hold
for a layer whose base adhered fully to the underlying medium.

5.2 Limiting conditions. There are two limiting cases which prove to be
of some interest and we now examine these briefly.
The first results from allowing the thickness h of the layer to become
514 R. E. GIBSON, R. L. SCHIFFMAN AND S. L. PU

infinite. It is straightforward, but rather tedious, to show that the results

of an earlier paper (18, p. 223) can be recovered.
Secondly, if b, the half-width (radius) of D, approaches infinity the
layer becomes subject to a uniform pressure extending over its entire
surface. Introducing the variable /? = Xbjh into (20) and proceeding
to this limit leads to the same expression
y+i<o

= 1 f ex exp(cfa/A2)
r)fG 1m
2-ni J
J 2r)s~s$
2TJS~ tanh i
y—ioo

for the uniform settlement w0 in both our problems.

Downloaded from http://qjmam.oxfordjournals.org/ at University of Sussex on May 18, 2012

The integrand possesses an infinity of simple poles located along the
negative real s-axis at s = —a2, where an are the roots of
tan a = 2rja.,
together with an isolated pole at s = 0. I t then follows from (27) that
1 yv
+2 y "«"-/- i_ (28a)
'Zrj—l ^ 2r]—l—4rj2oCin
The final settlement is
w0(oo) = ^—, (28b)

while the 'immediate' settlement is,

= fh/4G. (28c)
These results merit some discussion. It has been shown that in the
limit b —*• oo the surface settlement is the same in both the plane strain
and axially symmetric problems. This is not entirely self-evident, for
the stress distribution beneath the load and the restraints at infinity are
quite different in the two cases, but the following considerations go some
way towards explaining the result.
A more complete analysis shows that the distribution of stress in the
medium is uniform at t = 0 + and t = oo, but not at other times. In
particular, for axial symmetry,
<r« = / , arr = aee = if, arB = 0 at t = 0+, (29a)
*«=/. on=agg = if(r]-l)l(2v-l), ar6 = 0 at t = oo, (29b)
and for plane strain,

<*«=/> <r** = 0. <rn = if. oXz = 0 at t = 0+ (29c)

l), ff« = 0 at I = oo.
(29d)
v CONSOLIDATION OF A CLAY LAYER ON A SMOOTH BASE 515
It is simple to show that, despite the different stress components, (29a)
v and (29c) both lead to (28c), while (28b) follows from (29b) and (29d).
It will also be noted that equation (28a) is quite different from that asso-
ciated with the conventional one-dimensional treatment of consolidation
,. (2). This difference stems from the nature of the restraints imposed in the
two cases. Terzaghi's theory is restricted to a condition of zero lateral
deformation, so that strain develops only in the direction in which the
y pore water flows. Our problem is not strictly one-dimensional in this
sense, for lateral straining does occur although the pore water flows only
in the z-direction. These lateral deformations can develop by virtue of

Downloaded from http://qjmam.oxfordjournals.org/ at University of Sussex on May 18, 2012

*"" the absence of shear stress on the base of the layer. It is likely that for a
layer with a rough base (u, uT = 0 on z = h), Terzaghi's equation would
be recovered in the limit b —*• oo.
Finally, we remark that beneath the loaded area \x\, r <b the case
h —>- 0 differs in no important respect from that considered above, save
y that equations (28)—and indeed the problem—become trivial at the limit.
6. Computation of the settlement
Expression (24a) for the surface settlement has a structure which seems
r to preclude any further simplification and we must resort to numerical
methods to reach results of engineering interest. Calculations have
therefore been carried out to determine the magnitude and time variation
Ix \
r of GwA - ,t\jbf and hence of the 'degree of consolidation'

(30)
\r I

This latter quantity provides a measure of the extent to which the time-
dependent part of the settlement (the consolidation settlement) has been
fully attained.
The general form of expressions (24) is common to problems in continuum
mechanics; they consist essentially of semi-infinite integrals with damped
oscillating integrands. This type of integral is readily evaluated, in
principle, by a variety of numerical techniques. However, serious prob-
lems may arise in seeking the establishment of a tolerable error, and a
balance between truncation and rounding errors. The computation of the
integrals for this problem is complicated by the relationship of the series
to the integral in equation (24). For each value of the variable of integra-
tion A an infinite series must be summed over the roots of the character-
istic equation (22).
516 R. E. GIBSON, R. L. SCHIFFMAN AND S. L. PU

The computations were carried out on a high speed digital computer

and the procedure was programmed for the general case covering a full
range of values of r\ and X. For small X the characteristic equation
becomes tan a = 2^a, while for large X (in the extreme case, r\ = 1)
tan a = a/A. Therefore the roots of (22) must lie in the interval

nn

for all values of X and r\ encountered. The particular roots can be narrowed
by a 'binary chop' followed by an iteration using Muller's method (24).

Downloaded from http://qjmam.oxfordjournals.org/ at University of Sussex on May 18, 2012

The infinite integrations were performed in two steps. The zeros of the
kernel F were first located. The integrals between successive zeros were
then found by a Gauss-Legendre technique, the complete integral being
evaluated by successive summation. The practical finite upper limit of
the semi-infinite integral was determined by limiting the relative error
between successive 'loop integrals' to a small value and by maintaining a
small absolute value for the last loop integration.
There are three sources of truncation errors in the calculations: (i)
in calculating the roots of (22); (ii) in summing the series; and (iii)
in the scheme of numerical integration used. The cumulative effect of
these errors was kept to within 10~4 on a relative basis by an ergodic
procedure.
The rounding error was noted in the course of the calculations by obser-
vations involving different computers. The computations were performed
both on an IBM 360/System computer with a 32 bit word size and a
CDC 6400 with a 60 bit word size. These represent approximately 8 and
12 decimal places, respectively. Parallel computations in extreme cases
indicate a relative rounding error of less than 10~4.

6.1 Numerical results. Expressions for the immediate and final settle-
ment of the surface of the layer have been given in section 2.1 but these

oo Eg:
• ^

0-2
-s
\ N
\ s s
T \ \ V 'I
o \ y \ \ s \ \
> \
\ \ \ \
A/6 = 0-2^ 0-5 l \ "via 00
0-8-
\ \ \ \ \s V
s
i-n. ^ ^ sv v

10- io- 3 10 10s

FIG. 2
 CONSOLIDATION OF A CLAY LAYER ON A SMOOTH BASE 517

00 —— rjb_
.
0-2 V
\
.0-4 N ?1
\ \ %
\
0 \
V
0
A *\ )
0-8 \
\
\ \ N*
10-4 lQ-3 ]0~2 10-1 l. O
^«.'•a a 101
5 ^ ^

1Q2

FIG. 3

Downloaded from http://qjmam.oxfordjournals.org/ at University of Sussex on May 18, 2012

are known results from the theory of elasticity which have been evaluated
numerically and they will not concern us further here.
Our aim has been to examine the influence of hjb and Poisson's ratio v
on the progress of settlement. The effect of layer thickness alone on the
progress of settlement at the centre of the loaded area is shown in Fig. 2
(strip) and Fig. 3 (circle) for v = 0, r\ = 1. The results have been plotted
in the form of curves of U(0, t) against ct/b2, where the modified coefficient \V
of consolidation, c = cjrj = 2Gk/yw, has been introduced to free the time
scale from Poisson's ratio. It is apparent that the layer thickness has an
important influence unless h/b > 10.
That Poisson's ratio also affects the consolidation rate is demonstrated
in Fig. 4 (6 = oo; strip and circle), Fig. 5 {hjb = 1; strip) and Fig. 6
(hjb = 1; circle).

0-2

0-4
\

0-6
»=0-49 0-45 0-3
S
\ V 01
0-8
\

10
io- 4 io- 2 IO- 1 10 10'
ctjh*
FIG. 4
518 R. E. GIBSON, R. L. SCHIFFMAN AND S. L. PU
0-2f> wm
I *• — (145
•*
=
(1 4
0-30 " —

(, :,
0-35
< \ \

b l - 0-40
ss ft
s y
\ s
s
0-45

Downloaded from http://qjmam.oxfordjournals.org/ at University of Sussex on May 18, 2012

J7(P=0

0-50
A/ft = HI \

0-55 •
10-" 10-' l-o 10'
et/h*
FIG. 5

Although extensive computations have been completed, the results

of which will be published later, the data given here is sufficient to show
that caution must be exercised in using simple Terzaghi theory, or solu-
tions applicable strictly only to very thick strata, to estimate the settle-
ment behaviour of loaded clay layers.

0-20

0-25

0-30

0-35-

0-40

0-45

0-50
10-4
 CONSOLIDATION OF A CLAY LAYER ON A SMOOTH BASE 519

REFERENCES
1. K. TEBZAGHI, Sber. Akad. Wiss. Wien, Abt. Ha 132 (1923) 125-138.
2. K. TEBZAGHI and O. K. FBOHLICH, Theorie der Setzung von Tonschichten (Deu-
ticke, Vienna, 1936).
3. L. RENDTJLIC, Bauingenieur 17, 51/53 (1936) 559-564.
4. S. MTOAYAMA and K. AKAI, Bull. Disaster Prevention Res. Inst., Kyoto Univ.,
8 (1954).
5. N. N. VEBIGIN, Rheology and Soil Mechanics (I.U.T.A.M. Symposium, Berlin,
1964), 231-234.
6. J. A. DE WET, Bull. Highw. Res. Bd. 342 (1962) 152-175.
7. R. E. GIBSON and P. LTJMB, Proc. Instn civ. Engrs 2 (1953) 182-198.

Downloaded from http://qjmam.oxfordjournals.org/ at University of Sussex on May 18, 2012

8. M. A. BIOT, AnnlsSoc. r. Sci. mid. nat. Brux. B 55 (1935) 110-113.
9. , J. appl. Phys. 12 (1941) 155-164.
10. , J. appl. Phys. 12 (1941) 426^-430.
11. , J. appl. Phys. 27 (1956) 459-467.
12. , J. Math. Mech. 12 (1963) 521-541.
13. , J. appl. Mech. 28 (1956) 91-96.
14. R. E. GIBSON and J. MCNAMEE, Q. Jl. Mech. appl. Math. 16 (1963) 115-127.
15. R. L. SCHTFFMAN, A.T-F. CHEN and J. C. JORDAN, J. Soil Mech. Fdns Div.
Am. Soc. civ. Engrs 95 (1969) 285-312.
16. R. E. GIBSON and J. MCNAMEE, Proc. <Uh Int. Conf. Soil Mech. 1 (1957) 297-299.
17. J. MCNAMEE and R. E. GIBSON, Q. Jl. Mech. appl. Math. 13 (1960) 98-111.
18. , Q. Jl. Mech. appl. Math. 13 (1960) 210-227.
19. R. L. SCHTFFMAN and A. A. FTXNGABOLI, Proc. 6th Int. Conf. Soil Mech. 2
(1965) 188-192.
20. M. A. BIOT and F. M. CLINGAN, J. appl. Phys. 12 (1941) 578-581.
21. G. DE JOSSEMN DE JONG, Proc. Uh Int. Conf. Soil Mech. 1 (1957) 320-323.
22. G. PABIA, J. Math. Phys. 36 (1958) 338-346.
23. G. V. HEINBICH and K. DESOYEB, Ing.-Arch. 30 (1961) 225-253.
24. D. E. MULLEB, Maihl Tabl. natn. Res. Coun., Wash. 10 (1956) 208-215.

APPENDIX
We seek to locate the poles of C(X, s) (equation (21b)) in the s-plane, which is
equivalent to locating the zeros of
Q(w) = {a>2 + JS(A)}cosh d> +K(k)6i sinh w
in the d5-plane, where d>2 = s and L and K( = —M) are given by equations (21c, d).
If <o = a+ir, then
|cosh (5|a = £(cosh 2a + cos 2T),
|sinh eo|2 = £(cosh 2a — cos 2T),
. .„ cosh 2<r + cos 2T cosh2cr —
coth a 2 cosh 2a — cos 2T > cosh 2a +1
so that |coth w\ > |tanh a\. Therefore, if \a\ > IT, then
|cosh m| > tanhu-. (Al)
2
Let H(a>) = Ku>l(d> +L). Clearly, there exists a real positive number R(X) such
that
|H(c5)| <^cosh 3t when \&\ > R. (A2)
 520 R. E. GIBSON, R. L. SCKEFFMAN AND S. L. PU

Take N to be the integer just greater than Rjir and divide the cu-plane into square
regions by the lines a = ±Nn, ±3NTT, ... ; T = ±NIT, SNTT, ... , the centre of the
square cunm being located at (2nNir, 2mN-n) where n, m = 0, ± 1 , ± 2 Note
that the points a> = ± ( — L)i lie within a>00.
Now let €>(di) = (dj2 + i ) cosh cu and ^(d)) = Ku> sinh to. I t follows from the
inequalities (Al) and (A2) that |®(cu)| > ^(tu)! at all points on the boundary of
u>nm and since neither *F(to) nor Yfcu) have poles, it follows from Rouche's theorem
that Q(a>) = O(d>) +x¥(a>) has the same number of zeros in wnm as <D(di).
We may therefore assert that

( roots in <J>OiO, \

Downloaded from http://qjmam.oxfordjournals.org/ at University of Sussex on May 18, 2012

2N roots in oJ0ii (i = ± 1 , ± 2 , ...),! (A3)
no roots elsewhere. )
We now show that all these roots are purely imaginary, with the possible exception
of two real roots in diOtO. Consider the points of intersection of the curves
y = -Kx/(xz -L), y = cot x. (A4)
If x = a is a point of intersection of these two curves then Q(±ia) = 0. We
distinguish between the two possibilities L > 1 and L < 0 and note that K < 0
since A > 0. I t can easily be verified that if L > 0 then the curves (A4) have
2N + 2 points of intersection if 0 < \x\ < NTT and 2N points of intersection within
2VTT< \X\ < 3N, 3Nn < \x\ < 5Nir Therefore, from (A3), it follows that
Q(a>) = 0 has only imaginary roots when L > 0.
When L < 0 the curves (A4) have 2iV points of intersection in each of the intervals
. 0 < \x\ < NIT, Nn < \x\ < 3NTT Then, from (A3), the roots of Q(&) = 0
are purely imaginary with the exception of two roots in co0,0 which we now show are
real.
Consider the curves
y = -Kx/(x2 + L), y = coth *
which intersect in only two points: these provide the remaining two real roots of
Q(u>) = Oin a) 0 0 .
There will, of course, be just one-half the number of poles of C in the s-plane as there
are zeros of Q in the co-plane.