ATKINSON,J. H. (1975). Ghoreclrrlique 25, No. 2, 357-374.

Anisotropic elastic deformations in laboratory tests on

undisturbed London Clay

J. H. ATKINSON, PhD*

Anisotropic elasticity may be considered as a suitable L’ClasticitB anisotrope peut &tre consid&e comme
mathematical model for the small strain deformation un modele mathematique valable pour le comporte-
behaviour of an overconsolidated clay. The predic- ment de la dkformation d’une argile surconsolid6e
sous faible contrainte. Les pr&isions du modble
tions of the model will be compared with the results
seront comparbs avec les resultats de quelques
of some laboratory compression tests on samples of essais (de laboratoire) de compression faits sur les
undisturbed London Clay from the site of the Cchantillons d’argile intacte pris 21Londres 51I’empla-
Barbican Arts Centre and it will be seen that the cement du chantier du Barbican Arts Centre, et il
agreement between the model predictions and the sera montr6 que l’accord entre les previsions du
test results is satisfactory. The effects of stress and modele et les rksultats de l’essai est satisfaisant. Les
effets de la contrainte et de son histoire, sur les para-
stress history on the elastic parameters for undis-
metres Clastiques pour l’argile londonienne intacte,
turbed London Clay will be investigated by compar-
vont etre CtudiCs en comparant les rt%ultats de tests
ing the results of laboratory tests on samples taken de laboratoire faits sur des Cchantillons provenant
from several locations within the London Clay de plusieurs emplacements dans le bassin argileux
basin. londonien.

The problem of prediction or interpretation of soil deformations at states of stress and strain
below failure is an essential part of soil mechanics design. The problem may be approached
either through a mathematical model for the soil or through some form of the stress path
method (Lambe, 1967). If the former course is adopted then clearly the correct choice of
mathematical model may be at least as important as the choice of numerical values for the soil
parameters used in the model. Mathematical models in current use are founded on the
theories of elasticity and plasticity and if at all useful are capable of predicting soil deforma-
tions and stresses in a region which has generalized boundary conditions using a few soil
parameters obtained in relatively simple tests.
A criterion for the choice of a suitable mathematical model, whether elastic, plastic, or any
other kind, may be stated quite simply: if, using a particular mathematical model, all the
stresses and strains in a boundary value problem can be predicted with acceptable accuracy for
the particular problem then the behaviour of the soil may be regarded as approximately that of
the ideal material for which the mathematical model provides the exact solution. A distinc-
tion between elastic and plastic soil behaviour was suggested by Atkinson (1973b) on the basis
of the dependence of the strain increments on, on the one hand, the stress increment and, on
the other hand, the current stress state. He proceeded to show that both the drained strain
path and the undrained stress path of an elastically deforming soil must be linear in a cylin-
drical compression test irrespective of the linearity or non-linearity of the stress-strain
response.
In this Paper a mathematical model for the deformation behaviour of undisturbed London
Clay will be investigated by obtaining soil parameters in drained cylindrical compression tests,

* Cambridge University Engineering Department.

Downloaded by [ UC San Diego Libraries] on [18/09/16]. Copyright © ICE Publishing, all rights reserved.
 358 J.H. ATKINSON

predicting the behaviour in undrained cylindrical compression tests and in drained and un-
drained plane strain compression tests and comparing these with the results of laboratory tests.
It will be shown that the deformations of the undisturbed London Clay used, at least up to
shear strains of y = 1% or so, can be predicted with acceptable accuracy through the theory of
elasticity provided elastic anisotropy is included in the model.
Although in a general anisotropic material there need be no special conditions of symmetry,
a model containing a single axis of symmetry may be adopted for an undisturbed soil; through-
out the Paper the term anisotropic, without further qualification, should be taken to mean a
soil containing an axis of symmetry placed vertically in the ground.

NOTATION
D (subscript) drained horizontal strain (laboratory
D dilatancy (drained test) test)
E deformation modulus intermediate strain (labora-
Em 4, E, Young’s moduli in X, y and tory test)
2 axes (terms of total volume strain (laboratory
stresses) test)
E’,, E’w E’, Young’s moduli in x, y and z vertical strain (in the ground)
axes (terms of effective horizontal strain (in the
stresses) ground)
G shear modulus in plane per- horizontal strain (in the
pendicular to yz (terms of ground)
total stresses) angle between u’% and (T’A
G’ shear modulus in plane per- (degrees)
pendicular to yz (terms of Vl Poisson’s ratio in plane yz
effective stresses) (terms of total stresses)
H (subscript) horizontal (8= V’l Poisson’s ratio in plane yz
90°) (terms of effective stresses)
M stress path gradient (un- v3 Poisson’s ratio for strains in
drained test) yz plane due to unit strain lX
N degree of total stress aniso- (terms of total stresses)
tropy V’3 Poisson’s ratio for strains in
n degree of effective stress yz plane due to unit strain
anisotropy cX (terms of effective
P (subscript) plane strain stresses)
,
PC +(u’~ + u’~+ u’~) in labora- Q'A vertical stress (laboratory
tory test during consolida- test)
tion U’H horizontal stress (laboratory
I
PC +‘.J 1 -I-2&) in the ground test)
, ,
PIU maximum past value of plc UP
intermediate stress (labora-
T (subscript) cylindrical (tri- tory test)
I
axial) test ux vertical stress (in the ground)
U (subscript) undrained UY horizontal stress (in the
V (subscript) vertical (8 = 0”) ground)
CA vertical strain (laboratory horizontal stress (in the
test) ground)

Downloaded by [ UC San Diego Libraries] on [18/09/16]. Copyright © ICE Publishing, all rights reserved.
 ANISOTROPIC ELASTIC DEFORMATIONS IN LABORATORY TESTS ON UNDISTURBED LONDON CLAY 359

I
_ .5 +/ 5 IO I5

J Ap’ : lb/in.*
c

L L
Fig. 1. Volume and axial strains against increment of consolidation pressure for isotropic consolidation

LABORATORY TESTS
The deformation behaviour of samples of blue London Clay has been investigated in labora-
tory cylindrical and plane strain compression tests. The samples were obtained from hand-
cut blocks taken from a horizon about 50 ft above the base of the London Clay at the site of
the Barbican Arts Centre in the City of London. The samples, sample preparation and
laboratory tests have been described in detail elsewhere (Atkinson, 1973a) and only the essen-
tial features will be considered here.
The samples were prepared for testing by trimming hand cut blocks ; before compression load-
ing each sample was isotropically loaded to 70 lb/in2 while pore-pressures were measured and
was then allowed to consolidate against a back pressure of 40 lb/in2 to an initial mean effective
stress state of ~‘~=30 lb/in2 close to the calculated mean effective stress state in the ground.
Volume strains and axial strains measured during isotropic consolidation are shown plotted
against dp’, (the increment of mean effective stress during consolidation) in Fig. 1. In the
cylindrical compression tests the increment of consolidation pressure dp’, was small indicating
that the final mean normal stress plc = 30 lb/in2 was indeed close to that in the ground; in the
plane strain tests dp’, was somewhat larger due to swelling during sample preparation when
small quantities of water were flushed through the platens to remove air trapped behind the
rubber membrane (Atkinson, 1973a).
Each test proper involved slow compression loading (da, > 0: da, = 0) with different bound-
ary and sample drainage conditions and with the axis of soil symmetry placed either vertically
or horizontally in the apparatus. The variables investigated were

(a) stress path: triaxial (s’~= u’= (T)

plane strain pI, = 0 (P)
(b) drainage : drained du = 0 (D)
undrained Ey-= 0
(c) orientation : vertical e=o :::
horizontal e = 90” (II)

Downloaded by [ UC San Diego Libraries] on [18/09/16]. Copyright © ICE Publishing, all rights reserved.
 360 1. H. ATKINSON

(the upper case letters are used as subscripts to denote a particular test). All possible combina-
tions of stress path, drainage and orientation lead to eight independent tests and these were
completed together with a special plane strain test to investigate the question of path depen-
dency.

.
--F
b

E, :% EA: %
W
Fii. 2. Normal and transformed stress-strain plots: (a) drained cylindrical tests; (b) drained plane
strain tests

Downloaded by [ UC San Diego Libraries] on [18/09/16]. Copyright © ICE Publishing, all rights reserved.
 ANISOTROPIC ELASTIC DEFORMATIONS IN LABORATORY TESTS ON UNDISTURBED LONDON CLAY 361
Each test has been numbered and, for plotting the results, has a particular symbol and
set of subscripts; a schedule of the tests is contained in Table 1. The special plane strain test
PS-7 was conducted on a vertical sample (8=0) and involved cycling the stress path between
those of the drained test PS-2 and the undrained test PS-6.

5000

*.
5
1

Es”

0 2.0 3.0
1
40 0 I.0
m 4
2.0 30 40
e,:%

(cl

Fig. 2. Normal and transfonaed stress-strain plots: (c) undrained cylind&$:tests; (d) uadrained
strain tests

Downloaded by [ UC San Diego Libraries] on [18/09/16]. Copyright © ICE Publishing, all rights reserved.
 362 J. H. ATKINSON

W (4
Fig. 3. Strain paths in drained tests: (a) drained cylindrical vertical test; (b) drained plane strain vertical test;
(c) drained cylindrical horizontal test; (d) drained plane strain horizontal test

The results of the tests are shown graphically1 in Figs 2-5. Figs 2(a)-2(d) give deviator
stress against axial strain plots for the eight independent tests together with their transforma-
tions to Et against E* (see Appendix 3). It may be noted in passing that the transformed stress-
strain plots turn out to be approximately linear for small strains showing that the parabolic
fit for the normal stress-strain relationship was reasonable; values for Ei (the gradient of the
normal stress-strain curve at ~~‘0) have been included in the figures. Fig. 3 contains the
strain paths for the drained tests plotted as cV against En and Fig. 4 contains the stress paths
for the undrained tests plotted in u’*, JH space. Fig. 5 contains the stress path, the strain
path and the normal stress-strain curve for the special test PS-7.

Table 1. List of cylindrical and plane strain tests

,
Cylindrical I Plane Strain

Symbol Type of test

Vertical 0 DPV
Drained
Horizontal A DPH

Vertical q UPV
Undrained
Horizontal V UPH

* Here, as is the cast throughout the Paper, length and volume reductions are reckoned positive.

Downloaded by [ UC San Diego Libraries] on [18/09/16]. Copyright © ICE Publishing, all rights reserved.
 ANISOTROPIC ELASTIC DEFORMATIONS IN LABORATORY TESTS ON UNDISTURBED LONDON CLAY 363
7(
‘Or

3(

IO

Ui: lb/h2 U; : lb/in?

6) (4
Fig. 4. Stress path in undrained tests: (a) undrained cylindrical tests; (b) undrained plane strain@ts

INTERPRETATION OF THE TEST RESULTS

Basic deformation behaviour
Before a mathematical model can be employed to interpret or predict the test data it is
necessary to decide whether the deformations are of an elastic nature or whether they are of a
plastic nature; it should be noted here that at any point on the loading path the deformations
are regarded as being wholly elastic or wholly plastic and the point, if there is one, at which the
deformations change from elastic to plastic may be regarded as a yield point.
It has been shown (Atkinson, 1973b) that if the constitutive relationships for an elastic
material have the form
dc,, = F(d~‘,J . . . . . . . . . (1)
then in cylindrical compression tests the strain path must be linear in a drained test and the
stress path must be linear in an undrained test irrespective of the linearity or non-linearity
of the normal stress-strain behaviour while for a plastic material with constitutive relation-
ships of the form
dci, = G(~‘kl) e s e . e . . . . . (2)
the corresponding strain and stress paths must be non-linear. These requirements of elastic
and plastic materials can be shown to apply to plane strain compression tests with da,=O.
The deformation behaviour of the London Clay may be separated into elastic and plastic
stages by examination of the linearity or non-linearity of the strain and stress paths and a

Downloaded by [ UC San Diego Libraries] on [18/09/16]. Copyright © ICE Publishing, all rights reserved.
 364 J. H. ATKINSON

yield point may be found at the junction of the linear and non-linear parts of the paths.
Examination of the strain paths in Fig. 3 and the stress paths in Fig. 4 show clearly that in all
the tests these paths were initially linear and consequently the deformations may be regarded
as wholly elastic up to the junction between the linear and non-linear segments of the paths.
Consequently, within this range the deformations of undisturbed London Clay may be repre-
sented by relationships like those in eqn (1) with the appropriate values for the elastic para-
meters.
The form of the function Fin eqn (1) and the anisotropic elastic parameters adopted in the
Paper are defined in Appendix 1. To describe the deformation behaviour of an anisotropic
elastic material completely, values for five independent elastic parameters will be required
but if, as is the case with the present tests, only vertical (0 = 0) and horizontal (0 = 90”) samples
are considered then, as shown in Appendix 2, values for only four of these will be needed and
the deformations of the samples can be represented by eqns (9) and (10). The constitutive
relationships for an ideal isotropic elastic material may be recovered from eqn (7) by putting
E’,=E’, n=l*O, v’~=v’~=v’ and noting that G’=E’/2(1+v’).

Questions of elastic isotropy or anisotropy

Having established a range of elastic deformation behaviour for undisturbed London Clay
in compression tests it is necessary to investigate whether the elasticity is sufficiently isotropic
to allow the use of the simple isotropic elastic model or whether the rather more complicated
anisotropic elastic model will be required to provide acceptable predictions.
If a material is isotropic (irrespective of whether or not it is elastic) its behaviour must be
independent of its orientation in the testing machine for a given test. Comparison of the
behaviour of vertical and horizontal samples in like tests from the data in Figs 2-4 shows
clearly that sample orientation, in most instances, has a significant effect on the observed be-
haviour. It may be noted that in undrained plane strain tests Ei for vertical and horizontal
samples are equal (EUpV= EUpH= 3900 lb/S) but this is independent of the degree of elastic
anisotropy; similarly in Fig. 3 D,,= DTH=O-38 is a consequence of the particular degree of
elastic anisotropy exhibited by the London Clay used in the investigation.
That the London Clay is significantly anisotropic may be further demonstrated by investi-
gating the stress path parameter M. Putting n = 1 and Y’~= v’~into the relationships for M in
Table 5 leads to M, = -2 for cylindrical tests and MP= - 1 for plane strain tests, which, it
may be noted, are independent of E’, Y’and the total stress path. Inspection of the paths
in Fig. 4 shows that those for the cylindrical tests fall either side of n/r,= -2 and those for
the plane strain tests fall either side of MP= - 1. Furthermore it can be shown quite simply

Table 2. Comparison of measured and predicted test parameters for elastic deformations
I
Test Type ! Measured Predicted
-
I ) D 1 M ED, lb/W
_.
E,, lb/in2 ’
-__
D M ED, lb/in2 E,, lb/W
T-l 0.38
0.38
;I: -1.00 3500 1 -1.00 2300
- 2.45 4200 - 3.00 3800
z2 1700 0.20 1700
PS-4 3300 0.38 3200
PS-6 - 0.48 3900 - 0.42 3900
PS-5 - 1.90 3900 -2.38 3900

Downloaded by [ UC San Diego Libraries] on [18/09/16]. Copyright © ICE Publishing, all rights reserved.
 ANISOTROPIC ELASTIC DEFORMATIONS IN LABORATORY TESTS ON UNDISTURBED WNDON CLAY 365

(4 F-1
Fig. 5. Stress and strain paths for test PS-7: (a) stress path; (b) strain path and stress-straia plot

that in isotropic compression of an isotropic elastic material

which is unique. Comparing slopes of the dc, against dulc and de, against da’, curves in
Fig. 1 leads to
-da, = 1.27 for vertical samples
d CA

dev
- = 3.82 for horizontal samples
d <A
Consequently it may be concluded that the undisturbed London Clay used in the tests was
significantly anisotropic in its elastic behaviour.

Elastic deformations
Values for the test parameters D, A4 and E, (Appendix 3) have been included in Figs 2-4
and are summarized in Table 2. The parameters D and M for elastic deformations are, by
definition, taken from the linear parts of the stress and strain paths and there is no difficulty
in measuring them; values of Et were obtained from the transformed stress-strain curves of
Fig. 2 and correspond to the gradients of the normal stress-strain curves at their origins.
Values for D, M and I?, may be predicted from the anisotropic elastic model from the relation-
ships contained in Table 5 once values for the elastic parameters E’,, n, v’~ and vf3 have been
found.

Downloaded by [ UC San Diego Libraries] on [18/09/16]. Copyright © ICE Publishing, all rights reserved.
 366 1. H. ATKINSON

Values of these four independent elastic parameters may be extracted from the results of
tests T-l (drained cylindrical vertical) and T-5 (drained cylindrical horizontal) from
DTV = 2~‘~ = 0.38

DTH = vll +;VI3= 0.38

E DTV -- E’, = 1600 lb/in2

E DTH = 5 = 3200 lb/in2

n
which leads quite simply to
E’, = 1600 lb/in2
n = 0.5
VI1= 0
VI3= 0.19
Values for all other test parameters have been calculated from the relationships in Table 5
and have been included in Table 2.
Comparison of the measured and predicted test parameters in Table 2 demonstrates that the
agreement between the two sets is good; only for Eupv in test T-2 is the agreement for some
reason less satisfactory. Clearly the comparison must be exact for tests T-l and T-5 since
the anisotropic elastic parameters were calculated from the results of these tests. In addition
the bulk modulus for isotropic compression was Ki = 1100 lb/in2 in the tests whilst the elastic
parameters lead to a value of Ki = 1290 lb/in2 which may be regarded as acceptable corre-
spondence between experiment and theory. This, and the agreement found in Table 2, means
that once the elastic parameters have been found from two relatively simple tests the elastic
deformations under a variety of boundary and drainage conditions can be predicted through
the theory of elasticity provided anisotropy is included in the model.

Path dependency
It is a property of an ideal elastic material that if the stress state is shifted between two points
in stress space the deformations are independent of the stress path taken. To investigate to
what extent the deformations of undisturbed London Clay are path independent a special test
PS-7 was carried out. The stress and strain paths and the stress-strain curve for PS-7 are
contained in Fig. 5 together with the paths for PS-2 and PS-6. Inspection of Fig. 5 shows that
up to the end of the drained loading stage (D) the path for PS-7 lay very close to one or other
of the paths for PS-2 or PS-6 depending on the particular drainage conditions at that time.
However when the path was reversed (D + E) the strain recovery was incomplete although
later the stress-strain curve approached that for the undrained test. From these results it may
be concluded that so long as the strains increase monotonically they may be regarded as path
independent (within the elastic range) but serious departures may occur if the directions of the
strain increment are reversed.
Test PS-7 is rather more important than just being a test for path independency since the
stages A -+ D are similar to an undrained-drained foundation loading path. The test indi-
cates that in a foundation loading problem in London Clay the long-term fully drained defor-
mations may be found from a drained analysis irrespective of the field undrained loading stage
always assuming that the correct values for the drained elastic parameters can be obtained
from a laboratory test.

Downloaded by [ UC San Diego Libraries] on [18/09/16]. Copyright © ICE Publishing, all rights reserved.
 ANISOTROPIC ELASTIC DEFORMATIONS IN LABORATORY TESTS ON UNDISTURBED LONDON CLAY
367

I5
t

Fig. 6. States of stress and

strain at yield at the elastic limit

Yield at the eIastic limit

For the cylindrical and plane strain compression tests the states of stress and strain at the
yield point when the strain and stress paths cease to be linear have been extracted from Figs 3
and 4. These have been plotted as r”% against p’/pfc in Fig. 6; although the points are some-
what scattered they cannot be separated according to type of test, drainage or sample orienta-
tion. Within the range of experimental data (i.e. O-7<p’/pfc < 1.4) the points fall close to a
line given by
y(%)=a-$ . . . . . . . . . .
with a = 1.0. Eqn (3) may suggest a yield criterion for undisturbed London Clay below which
the deformations of the clay may be approximated to those of an ideal anisotropic elastic
material.

COMPARISONS BETWEEN THE PRESENT DATA AND OTHER LABORATORY TEST RESULTS
It has been shown that the small strain deformations of undisturbed London Clay in labora-
tory tests can be predicted from the theory of elasticity once the relevant elastic parameters
have been found. While viewing these results however, it should be appreciated that only a
rather limited set of tests has been used; all involved compression loading from the same initial
isotropic stress state although in undrained tests the horizontal effective stress unloaded. In
addition it should be noted that the deformation moduli used (i.e. initial tangents) were
arbitrarily defined to provide a convenient way of describing a non-linear curve by a single
parameter and their numerical values need have no special practical significance. Neverthe-
less, it is instructive to compare the present results, and in particular the dependence of the
numerical values for the elastic parameters on initial stress state and stress history, with data
from laboratory tests on samples of undisturbed London Clay taken from other locations
within the London Basin. The question of predicting the in situ deformation behaviour of
London Clay from the results of laboratory tests will not be discussed here and will be left for
consideration elsewhere.
Most of the data were obtained from laboratory cylindrical compression tests originally
intended to investigate the strength of London Clay; these test results have been re-interpreted
in terms of the parameters used in the Paper. Particular use has been made of tests by Webb
(1966) on samples from Ashford Common, by Agarwal (1967) on samples from Wraysbury
and by Som (1968) on samples from High Ongar; but only in the case of Sam’s tests was the
measurement of small strains the primary purpose of the investigations.

Downloaded by [ UC San Diego Libraries] on [18/09/16]. Copyright © ICE Publishing, all rights reserved.
 368 J. H. ATKINSON

Stress path parameter M in laboratory tests

In Table 3 values of M for several different locations and tests have been assembled together
with values predicted for the elastic parameters n=05, v’~=O, vfg=O-2. The plane strain
tests described by Ove Arup and Partners (1972) were performed at Imperial College using
samples of approximately 3 in. cube between two pairs of rigid platens and because of the short
length in the Q,= 0 direction the results may have been influenced by platen friction. Webb’s
(1966) samples were consolidated isotropically to pfc = 30-l 50 lb/in2 while the other tests were
unconsolidated except for Atkinson’s (1973a) samples which were consolidated to the mean
normal stress in situ. The values for M quoted were obtained by direct measurement of the
linear parts of the stress paths taking a mean value of observations from several tests.
All the values of M in Table 3 are in reasonable agreement with the predictions of the
anisotropic elastic model; in no case does MP= - 1 or MT= -2 as would be the case had

Table 3. Undrained stress path parameters for elastic deformations in laboratory tests on undisturbed London
Clay
- -
Location Reference Loading MTV MPH
n=0.5 v’l=O /,=0.20 All -1.00 -3.00 - 0.42 -2.38
__
High Ongar Som (1968) Compression -1.07
Extension -1.24
Swiss Cottage Arup (1972) Compression - 0.90 - 0.65 - 1.50
Extension - 0.90
Barbican Arup (1972) Compression -1.00 - 0.65 - 2.25
Extension -1GO
Barbican Atkinson (1973a) Compression -l+m - 2.45 - 0.48 -1.90
Ashford
Common Webb (1966) Compression - 0.70 - 3.30
-

Table 4. Deformation moduli measured in laboratory tests on samples of undisturbed London Clay

Location Reference

Consolidated tests

High Ongar Som (1968)

/
Ashford Webb (1966) ;; 1;; (
Ashford Level C
Ashford Level E ;: 1::
~~V~sp,ury Lovenbury (1969)
Agarwal(1967) 120 75 75

Barbican Atkinson (1973a) :: 108 140 176 50 82 190 194

---I--
Unconsolidated tests _I-----
Ashford 1 Ward et al. (1965) I I 11251200 / 1 1
Ashford 1 Webb(1966< ’
High Ongar Som (1968)
Central London Skempton and Henkei (1957) 1
Barbican Arup (1972) 188 280
Swiss Cottage Arup (1972) 1:: 110
Wraysbury _ / Agarwal(l967) ~ 200
Varibus _ / W&d et a/. (1959) / 120 ; 200

Downloaded by [ UC San Diego Libraries] on [18/09/16]. Copyright © ICE Publishing, all rights reserved.
 ANISOTROPIC ELASTIC DEFORMATIONS IN LABORATORY TESTS ON UNDISTURBED LONDON CLAY 369

q
q Barbican (Atkinson. 19732) Fig. 7. Variation of dila-
0 Vertical samples tancy with consolidation
pressure in drained cylin-
l Horizontal sampler
drical compression tests:
(a) Ashford Common
(Webb, 1966); (b) High

i:__ \T,;;,\ ;;;I ) , ongar(Som’1968)

loo I.50 0 50 loo Is0

p c ‘: Iblin.2 p,‘: lb/in.*

(4 (b)

the elasticity been isotropic. Furthermore there is no evidence of loading path (compression
or extension) or consolidation pressure affecting the values of M.

Strain path parameter D in laboratory tests

Values of the strain path parameter DT obtained from consolidated drained cylindrical
compression tests on vertical and horizontal samples, by direct measurement of the initially
linear portions of the strain paths, have been plotted aganst consolidation pressure in Fig. 7
for samples from Ashford Common, High Ongar and the Barbican. Although the points are
rather scattered a trend of reducing DT with increasing consolidation pressure is apparent;
noting that for vertical samples DTv= 2~‘~ the trend seems to suggest a dependence of v’~ on
P’~ with the Poissons ratio reducing with increasing consolidation pressure. If values of v’~
are adjusted to a common consolidation pressure of 30 lb/in2 the Ashford Common and
Barbican values are in agreement with v‘3 approximately O-20 while for the High Ongar data
v’~ approximates to O-30; the difference between these values may be attributable, at least in
part, to differences in overconsolidation ratio. If, however, the variation of v’a with consoli-
dation pressure is ignored then the mean values of v’a are close to Wroth’s (1971) and Som’s
(1968) values of O-12 and O-16 for Ashford Common and High Ongar respectively.
Since, in Fig. 7, there is no clear distinction between D, for vertical and horizontal samples
DT. may be approximated to DTH and, from Table 5,

V’I =
( >
vrg2-i . . . . . . . . .

For the present, v’~ may be estimated from the empirical relationship in eqn (4); for it= 1-O
v’~= v’a which is correct and for n < l-0 v’~< v’~. However, Henkel(1971), without stating his
reasons, assumed a relationship v’~= $(v’~+v’~) which, noting that v’~=v’Jn leads to

dl = +/,(l+$ . . . . . . . . .

Eqn (5) is correct for n= l-0 when v’~ = v’~ but for n < 1-Opredicts v’~> v’~differing significantly
from eqn (4).
It is not immediately obvious whether ~‘3 should be greater or less than vlg for an undisturbed
overconsolidated clay; however it is pertinent to note that it can be shown (Appendix 5) that
for undrained loading in terms of total stress elastic parameters v1< vQfor N< 1.0 but this does

Downloaded by [ UC San Diego Libraries] on [18/09/16]. Copyright © ICE Publishing, all rights reserved.
 370 J. H. ATKINSON

not necessarily mean that Y’~should be less than Y’~for n < 1.0. Som (1968) quoted n=0*87,
Y’~= 0.37 and v’~ = O-16for which v’~ > v’~ but the present interpretation of his data is in agree-
ment with eqn (4) from which v’~< v’~. The difference in n, the degree of elastic anisotropy,
between that found by Som (1968) and that obtained in this Paper may be attributable to the
effect on II of overconsolidation ratio discussed in the next section. Som’s (1968) value of
v’~= 0.37 seems to be rather large for soil; however, it may be noted that Som (1968) calculates
v’~ from the results of isotropic consolidation tests and drained cylindrical compression tests
on vertical samples whereas the Author’s method of estimating v’~ from drained cylindrical
compression tests on vertical and horizontal samples is better conditioned. For the present
it would seem that better general agreement can be obtained with values of vfl smaller than v’~.

Deformation mod& in laboratory tests

For the present purpose of comparing deformation moduli of samples of undisturbed Lon-
don Clay from different sites and from different sources the modulus will be redefined as Es,,-
the secant from the origin to a point on the stress-strain curve at 50% of the peak deviator
stress. In drained tests the values of calculated tangent and measured secant moduli turn out
to be nearly equal while in undrained tests the secant moduli are somewhat higher than the
tangent moduli due to an initial very stiff response that is smoothed out by the method of
calculating Ei (Appendix 3).
The stiffness of a soil for a particular loading measured by a deformation modulus (however
this is defined) is generally regarded as being dependent on the mean normal stress at the start
of shear (p’,J and on the stress history. For example Ladd (1964) found a unique value for
E/p’, in cylindrical compression tests on a given soil provided normally consolidated and over-
consolidated samples were considered separately. Values of E/p’, for undisturbed London
Clay from various locations and for various kinds of test are contained in Table 4; these were
obtained by plotting Es0 against plc for several tests of the same kind and fitting a straight line
through the origin (Atkinson, 1973a). For consolidated tests plc is simply the consolidation
pressure whilst for unconsolidated tests, in the absence of direct measurements of pore-
pressure, plc has been calculated from the field stresses assuming K0 = 2.5. Although the data
are rather sparse there appears to be little good correlation in Table 4 between the results of
tests of the same kind on samples from different locations. For cylindrical compression tests
on vertical samples the drained and undrained moduli seem to fall in the ranges Efx/pfc=
45-75 and E&~‘~=75-125. For consolidated drained tests, with the exception of Som’s
(1968) results, the degree of anisotropy n lies around 0.5 while for unconsolidated undrained
tests the degree of total stress anisotropy N is about 0.6.
The effect of stress history on deformation moduli may be examined by following Wroth
(1971) and expressing stress history in terms of the overconsolidation ratio defined in terms of
mean normal stress (i.e. OCR=p’,/p’, where p’,,, is the maximum past mean normal stress).
Using data from Webb (1966) from drained cylindrical compression tests on vertical samples of
London Clay from Ashford Common Wroth (1971) found a linear variation of E’,/P’~ with
In (OCR). For London Clay in situ the correct value for the overconsolidation ratio is rather
difficult to estimate; the problem lies in determining a value for plrn for a particular location
even if pfc can be found for unconsolidated samples. For the present purpose p’,,, has been
calculated from an estimated depth of erosion of 1250 ft at Ashford Common (Bishop et al.,
1965) to 500 ft at Bradwell (Skempton, 1961) together with K,=O.64 for virgin consolidation
(Som, 1968).
In Fig. 8 values of E/p’, have been plotted against OCR on a natural scale for cylindrical
compression tests on samples from Ashford Common, High Ongar and from several sites in

Downloaded by [ UC San Diego Libraries] on [18/09/16]. Copyright © ICE Publishing, all rights reserved.
 ANISOTROPIC ELASTIC DEFORMATIONS IN LABORATORY TESTS ON UNDISTURBED LONDON CLAY

Horizontal samples l Barbican (Atkinson. 19734

203 0 0 A Swiss Cottage (Arup, 1972)

/ n High Ongrr (km, 1968)

Fig. 8. Variation of E/p’, with overcoasoli-

da&n ratio ia cylindrkil compression tests:
(a) consolidated drained tests, samples from
Ashford Common (Webb, 1%); (b) uncon-
solidated undraiied tests, vertical samples

(b)

central London. Although the data are rather scattered a distinct trend of decreasing E/pla
with decreasing OCR is apparent; no better fit would have been obtained if, following Wroth
(1971), OCR had been plotted on a log scale but it may be noted that the present method of
estimating the overconsolidation ratio differs from that used by Wroth (1971). The same
trend of decreasing E/p’, with decreasing OCR was found for consolidated drained tests on
samples from High Ongar and for consolidated undrained tests on samples from Ashford
Common and Wraysbury (Atkinson, 1973a). In Fig. 8(a) Webb’s (1966) vertical and hori-
zontal samples have been plotted separately and the data indicate strong evidence of decreas-
ing degree of elastic anisotropy with decreasing overconsolidation (i.e. with increasing iso-
tropic reconsolidation before shear the subsequent degree of elastic anisotropy may be reduced);
a similar trend was found in Webb’s consolidated undrained tests.

DISCUSSION
Laboratory tests on samples of undisturbed London Clay taken from the Barbican Arts
Centre site have demonstrated that the small strain deformation behaviour of the clay can
be satisfactorily interpreted through the theory of elasticity provided anisotropy is included in
the model. The part of the loading path for which the anisotropic elastic model represents
the soil behaviour can be identified with a linear strain path in a drained test or a linear stress
path in an undrained test. A further test showed that the elastic deformations were approxi-
mately path-independent provided the loading increased monotonically.
Comparison of the present tests with others on undisturbed London Clay suggests that the
elastic parameters may be dependent on the mean normal stress and on the stress history.
Taken together the tests suggest that the degree of elastic anisotropy in laboratory tests on
London Clay is of the order of O-5, the vertical stiffness in the ground being less than the
horizontal stiffness.
In all the tests described in the Paper the samples were loaded from an initial state of iso-
tropic stress; it remains to be seen whether samples behave in a similar way for other loading

Downloaded by [ UC San Diego Libraries] on [18/09/16]. Copyright © ICE Publishing, all rights reserved.
 372 J. H. ATKINSON

paths andfor other initial stress states close to those in the ground. The question of the defor-
mation behaviour of London Clay in the ground has not been considered.

ACKNOWLEDGEMENTS
The Author is indebted to Ove Arup and Partners for facilities in obtaining block samples of
London Clay from the Barbican site. The work forms part of a programme of research into
the deformation of undisturbed soils supported by the Science Research Council under the
direction of Professor A. W. Bishop and was completed while the Author was a Research
Student at Imperial College.

APPENDIX 1: ANISOTROPIC ELASTIC PARAMETERS

In the theory of elasticity increments of stress and increments of strain may be related by equations of the
form
dc,, = [&,I] do’kl . . . . . . . . . . . (6)
in which [&r] is an array of elastic parameters that may or may not be constants and i, j and k, I=x, y, z.
An anisotropic elastic soil has a single axis of symmetry x placed vertically in the ground and a plane of
symmetry yr placed horizontally; the elastic properties in any direction in the yz plane are identical but dif-
ferent from those in the x direction. For this kind of elastic anisotropy [F& must contain five independent
parameters (Jaeger, 1969); if, following Atkinson (1973a) these are chosen to be
E’X
n = E’JE’,
Y’1 = day = dy.
YI3 = dXE = dxy
G’ = G’,, = G’,,

then eqns (6) become

des 1 - Y’z --V’S do’,
dc, - Y’S ?I - nv’1 do’,
dcz - v’a - nv’1 n do’,
1 . . .
=- E’, drx,, (7)
&xv E’, IG

dy.w 2n(l+ Y’1) dr,,

dr:x IE’x
G__
dr 2%
_
in which y and z may be interchanged.

APPENDIX 2: LABORATORY AXES AND SAMPLE ROTATION

In a laboratory test the machine axes are denoted A, P, H where, in plane strain +=O and in axial sym-
metry o’~=(T’~. A sample may be placed in the machine with the x axis at some angle 0 to the A axis in a
plane normal to the P axis; then, the relationships between stress and strain increments may be written
de,, = lG%jl*I&r1 Ii%.1 do’,, . . . . . . . . . (8)
with p. q and r, s= A, P, H. In eqn (8) [T”] is the co-ordinate transformation matrix for stress and [T”]* is

1 -v’
-v’
&.s
the inverse of the co-ordinate transformation matrix for strain. If ddA dulp and do’= are increments of
principal stress and if t?= (N- l)(a/2) where N is an integer then eqn (8) becomes
For Nodd: 0=0 or 180”

I[ I
do’*
- Y’S n -n/l do’, . . . . . . . (9)
- vfg - ?tY’l n da’H

Downloaded by [ UC San Diego Libraries] on [18/09/16]. Copyright © ICE Publishing, all rights reserved.
 n-FIV’
-P’~~
ANISOTROPIC ELASTIC DEFORMATIONS IN LABORATORY TESTS ON UNDISTURBED LONDON CLAY 373
For N even : 9 = 90”
dot
- ?zY’l n -v’s da’p . . . . . . (10)
- v’a - Y’s 1 I[ da’RI
It may be noted that eqns (9) and (10) do not contain the independent shear modulus G’; this means that the
behaviour of vertical (0=0) and horizontal (0=90) samples can be expressed in terms of only four elastic
parameters.

APPENDIX 3: LABORATORY TEST PARAMETERS

From any laboratory test two parameters should be extracted; one should describe a stress-strain coupling
and the other should relate strain increments in drained tests or principal effective stress increments in un-
drained tests. (These are analogous to determining Young’s modulus and Poisson’s ratio from a single
uni-axial compression test on an isotropic elastic material.) For plane strain or cylindrical compression
tests it is convenient to define the parameters:
(for drained tests)
dalA - ddA
En = apparent drained modulus . . . . . .
dca

diIatancy (Rowe, 1962) . . . . . . (12)

(for undrained tests)

apparent undrained modulus . . . . . (13)

&f+ stress path gradient (Henkel, 1971) . . . . . (14)

a
Whether or not the values for these parameters turn out to be constants depends on such things as elastic
linearity, plastic hardening or indeed whether the soil deformations are elastic or plastic. However, both D
and M have an importance beyond being simple test parameters since it has been shown (Atkinson, 1973b)
that if D or M are constants-the stress or strain paths are linear-the soil satisfies the requirements of
elasticity in eqn (6).
For soils neither En nor Eu are likely to be constant since observed stress-strain relationships are usually
non-linear irrespective of whether or not the deformations are elastic. A non-linear stress-strain curve may
be approximated to a curve of the form
(a’*-o’JJ> = a(~*)-@(Q)2 . . . . . . . . . . (15)
and, by differentiating
Et = E,-b(cJ . . . . . . . . . . . (16)
where Et and E, are the gradients at Q and at Q*= 0 respectively. From a tabulated set of test results Et at a
point n on the curve may be approximated to
E(n) = o’(n+l)-0-l)
t e*(n+l)--*(n-l) . * . . . . . . . .

where u’=(u’~- 08). By plotting E(n) against C.&Z)the gradient of the stress-strain curve at the origin may
be obtained from tabulated laboratory test results.

APPENDIX 4: RELATIONSHIPS BETWEEN ELASTIC PARAMETERS AND LABORATORY TEST

PARAMETERS
By setting dulH = 0 for drained compression tests or de, = 0 for undrained tests together with either d+ = 0
for plane strain tests or da’p=du’H for cylindrical compression tests eqns (9) and (10) can be used to express
the test parameters E,,E,,D and M in terms of the anisotropic elastic parameters Etx,n,~‘I and v’+ After
considerable manipulation the expressions in Table 5 emerge for the special cases of 8=0 and 8= 90”
(Atkinson, 1973a); the familiar expressions for an isotropic elastic material may be obtained quite simply by
setting n = 1 and v’~= v’~= v’in Table 5.

Downloaded by [ UC San Diego Libraries] on [18/09/16]. Copyright © ICE Publishing, all rights reserved.
 374 J. H. ATKINSON

Table 5. Laboratory test parameters in terms of anisotropic elastic parameters

Loading / O&Mat&n / D 1 M 1 ED

Cylindrical
I
)
compression

v’a(1+ v’1) _ D&l - DPH) E’,&v 1-MPV

~ 0 E
1 - v’s”/n &H( 1 - DPY) v's(1+ Y’1) DPVDpy- Mpv
Plane strain ’ -
v’s 1 E’, DPH
90 EUPV
n(l - Y’l) Mpy V’,(l -I-V’1)

APPENDIX 5 : UNDRAINED BEHAVIOUR OF AN ANISOTROPIC ELASTIC SOIL

Summing eqns (9) and setting dc,=O for undrained behaviour
duA(l-2v,)+(do,+do,)(N-vs-Nvl) = 0 . . . . . . . (18)
where the elastic parameters are now those relating increments of strain to increments of total stress. For an
elastic material eqn (18) must hold for any loading path-i.e. for da, = 0 or for (dop + duH) = 0 in turn. Hence
v3 = !Z . . . . . . . . . . . . (19)
VI = l-1/(2N) . . . . . . . . . . . (20)
Now only three independent elastic parameters E,,Nand G are required to describe the material behaviour ;
for an isotropic elastic material only one parameter E, together with v=j, is needed.

REFERENCES
Agarwal, K. B. (1967). The influence of size and orientation of sample on the undrained strength of London
Clay. PhD thesis, University of London.
Atkinson, J. H. (1973a). The deformation of undisturbed London Clay. PhD thesis, University of London.
Atkinson, J. H. (1973b). A note on elasticity and plasticity in soils. Ghotechnique 23, No. 4, 565-571.
Bishop, A. W., Webb, D. L. & Lewin, P. I. (1965). Undisturbed samples of London Clay from the Ashford
Common Shaft: strength-effective stress relationships. Ge’otechnique 15, No. 1, 1-31.
Henkel, D. J. (1971). The relevance of laboratory measured parameters in field studies. Proc. Roscoe
Memorial Symp., Cambridge, 669-675.
Jaeger, J. C. (1969). Elasticity, fracture andflow. (2nd Ed.) London: Methuen.
Ladd, C. C. (1964). Stress-strain modulus of clay in undrained shear. Jnl Soil Mech. Fdn Diu. Am. Sot.
Engrs 90, SM5, 103-132.
Lambe, T. W. (1967). The stress path method. Jnl Soil Mech. Fdn Div. Am. Sot. Civ. Engrs 93, SM6.
309-332.
Lovenbury, H. T. (1969). Creep characteristics of London Clay. PhD thesis, University of London.
Ove Arup & Partners (1972). Private communications.
Rowe, P. W. (1962). The stress dilatancy relation for static equilibrium of an assembly of particles in con-
tact. Proc. Roy. Sot. A. 269, 50&527.
Skempton, A. W. (1961). Horizontal stresses in an overconsolidated Eocene clay. Proc. 5th Znt. Conf.
Soil Mech. and Fdn Engng, Paris 1, 351-357.
Skempton, A. W. & Henkel, D. J. (1957). Tests on London Clay from deep borings at Paddington, Victoria
and South Bank. Proc. 4th Znt. Conf. Soil Mech. and Fdn Engng l, lOO-105.
Som, N. N. (1968). The efict of stress path on the deformation and consolidation of London Clay. PhD
thesis, University of London.
Ward, W. H., Marsland, A. & Samuels, S. G. (1965). Properties of the London Clay at the Ashford Com-
mon shaft: in-situ and undrained strength tests. Giotechnique 15, No. 4, 321-344.
Ward, W. H., Samuels, S. G. & Butler, M. E. (1959). Further studies of the properties of London Clay.
Gdotechnique 9, No. 2, 33-58.
Webb, D. L. (1966). Mechanicalproperties of undisturbed samples of London Clay and Pierre Shale. PhD
thesis, University of London.
Wroth, C. P. (1971). Some aspects of the elastic behaviour of overconsolidated clay. Proc. Roscoe
Memorial Symp., Cambridge, 347-361.

Downloaded by [ UC San Diego Libraries] on [18/09/16]. Copyright © ICE Publishing, all rights reserved.