You are on page 1of 33

Clayton, C. R. I. (2011). Geotechnique 61, No. 1, 537 [doi: 10.1680/geot.2011.61.1.

5]

Stiffness at small strain: research and practice


C . R . I . C L AYTO N 
Cet article presente le contexte de la 50e` me conference de
Rankine. Il se penche sur limportance croissante accordee, au cours des vingt dernie`res annees du 20e` me sie`cle,
a` la prediction des deplacements du sol en phase de
dimensionnement, a` la suite des lecons tirees dobservations sur le terrain. Le developpement historique de la
theorie de lelasticite est alors decrit, ainsi que les cadres
constitutifs dans lesquels il a ete propose dappliquer les
predictions geotechniques des deformations. Les facteurs
affectant la rigidite des sols et des roches tendres sont
evalues, et les resultats dune experience numerique,
evaluant limpact dun certain nombre de parame`tres de
rigidite sur les deplacements autour dune structure de
soute`nement sont decrits. On proce`de a` lexamen, et a`
une discussion critique, de certaines methodes adoptees
in situ et en laboratoire pour la determination de parame`tres de rigidite. La communication se termine avec la
proposition dune strategie pour la determination et
lintegration des donnees de rigidite, et des developpements necessaires pour loptimisation de letat actuel des
connaissances.

This paper provides the background to the 50th Rankine


Lecture. It considers the growth in emphasis of the prediction of ground displacements during design in the past
two decades of the 20th century, as a result of the lessons
learnt from field observations. The historical development
of the theory of elasticity is then described, as are the
constitutive frameworks within which it has been proposed that geotechnical predictions of deformation should
be carried out. Factors affecting the stiffness of soils and
weak rocks are reviewed, and the results of a numerical
experiment, assessing the impact of a number of stiffness
parameters on the displacements around a retaining structure, are described. Some field and laboratory methods of
obtaining stiffness parameters are considered and critically discussed, and the paper concludes with a suggested
strategy for the measurement and integration of stiffness
data, and the developments necessary to improve the
existing state of the art.
KEYWORDS: anisotropy; deformation; elasticity; geophysics;
ground movements; in situ testing; laboratory equipment;
laboratory tests; stiffness

INTRODUCTION
The rapid development of computing power and of numerical modelling software over the past 40 years has made
sophisticated analysis of geotechnical problems accessible to
most engineering practices. Typically, computer packages
now offer a wide range of constitutive models, which the
design engineer needs to choose among, and then obtain
parameters for. For structures designed to be far from failure, for example supporting urban excavations, strains in the
ground are small. A sound knowledge of stiffness parameters
at small strain is essential, if realistic predictions of the
ground movements that may affect adjacent buildings or
underlying infrastructure are to be made.
This paper discusses the geotechnical background to the
measurement of stiffness parameters, briefly reviewing the
lessons learnt from field observation and back-analysis of
foundation and deep excavation behaviour. It describes the
historical development of elastic theory, and the constitutive
frameworks within which it can be applied to soil and weak
rock behaviour. It reviews what is now known about the
complex stiffness behaviour of soil and weak rocks in the
context of what is, arguably, the simplest of constitutive
models. A numerical experiment, to assess the importance of
different parameters for the displacement of a particular structure, a singly propped retaining wall, is described. It is shown
that for this particular problem most parameters significantly
affect predicted displacements. Methods of determining the
required stiffness parameters are then explored, and the usefulness of seismic field testing, dynamic laboratory testing and
advanced triaxial testing is examined. Finally, strategies for
integrating the data are discussed, and conclusions are drawn.

GEOTECHNICAL BACKGROUND
James Bell (1989) has described the 19th century as the
Age of Design by Disaster. According to him, surprisingly
few engineers working in this period carried out analyses of
their design concepts before beginning construction. Given
the significant construction problems faced by civil engineers
at the beginning of the 20th century, early soil mechanics
understandably focused on preventing failure.
But by the late 1970s the emphasis had changed. For
many practising engineers soil mechanics was becoming a
mature science, because most failure mechanisms were
understood, and with good practice could be identified and
avoided with some certainty. The start of global urbanisation
changed all that, as the pressure to redevelop inner city
infrastructure produced more and more challenges, many of
which now related to ground movements and their effects on
adjacent structures and buried infrastructure. At the same
time the need to build nuclear and other key facilities
increased, requiring analysis for the effects of large, albeit
sometimes infrequent, seismic events. The rise of numerical
modelling in the 1960s, and the huge increase in computing
power since then, has given us increasingly sophisticated
analytical tools for use in practice (e.g. Zienkiewicz et al.,
1968; Simpson, 1981; Britto & Gunn, 1987; Potts, 2003).
The determination of the parameters needed for such analyses has, perhaps understandably, lagged behind the
development of numerical modelling.
Burland (1989) gives a good account of how the interaction of field observations and numerical modelling of the
deformations associated with foundations and excavations in
the London area led, in the UK, to the development of more
appropriate stiffness models for the ground. Back-analysis of
construction in London showed that field stiffnesses were
much greater than those obtained from routine laboratory
tests, for example in the oedometer or triaxial apparatus
(Cole & Burland, 1972; St John, 1975; Clayton et al., 1991),

 School of Civil Engineering and the Environment, University of


Southampton, UK

5
Delivered by ICEVirtualLibrary.com to:
IP: 152.78.201.162
On: Mon, 14 Feb 2011 15:36:53

CLAYTON

that stiffness increased significantly with depth (Marsland &


Eason, 1973; Burland & Kalra, 1986), that stiffness was
anisotropic (Cole & Burland, 1972), and greater at small
strains than at large (Burland & Hancock, 1977; Simpson et
al., 1979). These concepts seem to be true for weak rocks
such as the chalk, as well as for stiff clays (Ward et al.,
1968; Burland & Lord, 1970; Kee, 1970; Burland et al.,
1973; Hobbs, 1975; Matthews & Clayton, 2004).
Atkinson & Sallfors (1991) noted that, as far as they
could determine, no previous state-of-the-art or general
reports to major international conferences in the previous
decade had considered specifically the determination of soil
deformation parameters. The 1991 International Society for
Soil Mechanics and Foundation Engineering conference in
Florence was a turning point, at which the importance of
soil stiffness to both theoreticians and practising engineers
became accepted. Now, almost 20 years on, this paper
revisits the issue of stiffness determination, and in particular
is concerned with small-strain stiffness, which was for many
the defining feature of the research in the period leading up
to the Florence conference.

CONSTITUTIVE FRAMEWORKS FOR STIFFNESS


The stiffness of a body (or structure) is defined as the
resistance of that body to deformation under applied force.
It is derived from:
(a) the shape of the body
(b) boundary conditions, such as fixities and load positions
(c) the stiffness properties of the constituent materials
(Youngs moduli, etc.).
Thus deformation depends upon stiffness, which in turn
depends on the stiffness properties that are the subject of
this paper. In geotechnical engineering practice stiffness is
normally defined within the context of the mathematical
theory of elasticity, although this is not strictly necessary.
The development of the theory of elasticity is described
below.

Historical development
The recognition of linear load/deformation behaviour is
widely attributed to Hooke (1676), who wrote at the end of
his A description of helioscopes that
To fill the vacancy of the ensuing page, I have here added
a decimate of the centesme of the Inventions I intend to
publish, though possibly not in the same order, but as I can
get opportunity and leasure; most of which, I hope, will be
as useful to Mankind, as they are yet unknown and new.
The third of these Inventions was on
The true Theory of Elasticity or Springiness, and a
particular Explication thereof in several Subjects in which
it is to be found: And the way of computing the velocity of
Bodies moved by them. ceiiinosssttuu.
In his treatise De Potentia Restitutiva, or of Spring, Hooke
(1678) explained his anagram as Ut tensio sic vis, which is
roughly translated as extension is proportional to force. As
we would see it today, this is a description of linearity.
Hooke also recognised elastic behaviour, that is, the behaviour of a material that returns to its original shape after
loading is removed. In the same work he states that
. . . it is very evident that the Rule or Law of Nature in
every springing body is, that the force or power thereof to
restore it self to its natural position is always proportionate
to the Distance or space it is removed therefrom.

In reality, according to Bell (1989), Hookes measurements


on long iron wires were too insensitive to show linearity. As
early as 1687 James Bernoulli produced data for the gut
string of a lute that suggested a parabolic relationship between load and deformation at small strains (although Leibnitz assumed his data were hyperbolic). Over 100 years later,
in about 1810, two independent sets of experiments, by
Duleau and by Dupin, led to conflicting conclusions. Duleau
(1820), testing forged iron for a bridge over the Dordogne
river, found linear behaviour at small strain. Dupin (1815),
testing wooden beams for ships, found a non-linear response.
By the end of the 19th century, and following the findings
of the Royal Commission on Application of Iron to Railway
Structures (1849), which recommended that Hookes law
should be replaced by experimentally based, well-documented non-linearity, several non-linear laws (parabolic, by
Eaton Hodgkinson, a member of the Royal Commission;
hyperbolic, by Homersham Cox; and non-linear exponential,
by Carl Bach) had been proposed. Nineteenth-century data
for cast iron, showing Coxs hyperbolic law, are given in
Fig. 1. The results are similar in form to those obtained
from triaxial testing on intact chalk (Heymann et al., 2005).
Writing in his classic work on The experimental foundations
of solid mechanics, Bell (1989) has noted that
The dilemma of Leibniz in the 17th century over the
apparently conflicting experiments of Hooke and James
Bernoulli has been resolved in favor of the latter. The
experiments of 280 years have demonstrated amply for
every solid substance examined with sufficient care, that
the strain resulting from small applied stress is not a linear
function thereof.
However, the impact of this realisation has generally been
small, for as Viggiani (2000) notes, the achievements of
linear elasticity theory are well known to all of us; modern
engineering is still largely based on it.
Parameters
Although Hooke recognised the concept of the stiffness of
a body, the idea of an elastic property was not developed
until 1727, by Leonhard Euler, and not measured until 1782,
by Giordano Riccati. The concept of stresses in solids had
been introduced by Coulomb in 1773, in his classic paper,
which also dealt with the pressure of soil on retaining
structures. Young later published the idea of his modulus
in his book of 1807, although his definition does not align
with what we would now term Youngs modulus (Todhunter, 1886; Timoshenko, 1953; Cooper, 1978). Although
Young recognised the distinction between extension and
Normalised secant Youngs modulus, Esec /E0

12
Stiffness continues to rise
10
E
08

Reference modulus, E0

(1 )

06
04
02
0
00001

Experimental data for cast iron


Coxs Hyperbolic law
0001

001

01

Axial strain: %

Fig. 1. Normalised stiffness data for cast iron (Royal Commission on Application of Iron to Railway Structures, 1849; Cox,
1856)

Delivered by ICEVirtualLibrary.com to:


IP: 152.78.201.162
On: Mon, 14 Feb 2011 15:36:53

STIFFNESS AT SMALL STRAIN: RESEARCH AND PRACTICE


shear stiffnesses, he did not suggest a different measure for
these modes of deformation.
The concept of a Poissons ratio dates back to 1814
(Cooper, 1978). The PoissonCauchy theory of 1848 produced the uni-constant theory of elasticity, a consequence of
which was the prediction that for homogeneous isotropic
solids Poissons ratio should be exactly 1/4 for all materials.
In contrast Greens (1828) work, although largely unknown
at the time, had identified an elastic system with 21 elastic
constants, reduced to two in the case of isotropy. Later
experimental work by Wertheim, Kupffer, Neumann and
Kirchhoff (Timoshenko, 1953) subsequently lent support to
Greens findings. By the end of the 19th century the framework of elasticity was fully developed.

Application in geotechnical engineering


Probably the most commonly assumed behaviour in practical geomechanics is that of isotropic linear elasticity.
Characterisation of an isotropic elastic solid requires the
determination of only two material parameters (from four
possible measurements, i.e. Youngs modulus E and Poissons
ratio , or shear modulus G and bulk modulus K) for
calculations of strain or deformation, and therefore an
assumption of isotropic elasticity has the merit of simplicity.
However, as noted by Bishop & Hight (1977), there are
many reasons to believe that the ground will generally be
anisotropic, or at least transversely isotropic.
As found by Green, the characterisation of an anisotropic
elastic solid requires the determination of 21 independent
elastic constants. Given the complexity of subsurface geometry, and the spatial variability of soil and rock, this is
beyond the reach of practical soil mechanics. But in many
cases it may be sufficient to assume transverse isotropy, or
cross-anisotropy as it is also known, for which there are
seven measurable parameters, and a further two (for example, dip and dip direction) necessary to define the orientation
of the plane of isotropy in the general case where it is not
horizontal. For a transversely isotropic material where the
plane of isotropy is horizontal, the seven elastic parameters
are
Ev Youngs modulus for loading in the vertical direction
Eh Youngs modulus for loading in the horizontal direction
vh Poissons ratio relating to the horizontal strain caused
by an imposed vertical strain
hv Poissons ratio relating to the vertical strain caused by
an imposed horizontal strain
hh Poissons ratio relating to the horizontal strain caused
by an imposed horizontal strain in the normal direction
Gv shear modulus in the vertical plane
Gh shear modulus in the horizontal plane
where the subscripts v and h refer to the vertical and
horizontal directions.

Skeleton and pore fluid interaction. The discussion so far has


ignored the fact that most geomaterials are not solid, but are

particulate or voided, and have at least two and often three


phases:
(a) a skeleton, or frame, for example an assembly of
particles in contact with, and often cemented to, each
other
(b) pore fluid, which will normally be water for a saturated
material, and water and air for an unsaturated material.
The skeletal stiffness of uncemented soil is a function of
effective stress, and is often low in comparison with the
stiffness of water, which may then be considered incompressible (Bishop & Hight, 1977). For a saturated soil, therefore,
there are two cases and two sets of stiffness parameters that
may be required:
(a) the undrained, short-term, or end of construction
case, where shear strains have occurred but excess pore
pressures remain, and volumetric strain is assumed to
have been prevented because of the low permeability of
the soil relative to the rate of loading/unloading, and
the incompressibility of the pore fluid relative to the
soil skeleton
(b) the drained, long term or effective stress case,
where both volumetric and shear strains have occurred,
and any excess pore pressures set up during loading
have fully dissipated (Bishop & Bjerrum, 1960).
The stiffness of an isotropic soil material can be defined
in terms of a number of different sets of parameters, the
most commonly used in soil mechanics being shown in
Table 1. For the isotropic drained case the engineer can
choose to measure either the effective Youngs modulus and
the effective Poissons ratio, or the shear modulus and the
drained bulk modulus (K9 dp9/dV , where p9 is mean
effective stress and V is the volumetric strain). Parameter
set 1 (Table 1) can be readily measured (assuming isotropy)
in the triaxial test. The computational convenience of parameter set 2 lies in the fact that G remains the same in the
undrained and drained cases, since it involves change in
shape without change in volume, and the contribution to
stiffness of the shear modulus of water can be assumed to
be negligible at low rates of strain.
In the isotropic case, the relationships between drained
and undrained Youngs modulus and Poissons ratio, shear
modulus and bulk modulus can be obtained from
E9
Eu

21 9 21  u
E9
K9
31  29
Eu
Ku
31  2 u

Undrained

Eu

Drained

Parameter set 1
Undrained Youngs
modulus, Eu
Effective Youngs
modulus, E9

(1)
(2)
(3)

If the pore fluid is assumed incompressible (but see Bishop


& Hight, 1977), then the undrained bulk modulus is infinite,
and from equation (3)  u 0.5. Hence
1:5E9
1 9

(4)

Table 1. Isotropic drained and undrained parameter sets


Case

Undrained Poissons
ratio,  u
Effective Poissons
ratio, 9

Parameter set 2
Shear
modulus, G
Shear
modulus, G

Delivered by ICEVirtualLibrary.com to:


IP: 152.78.201.162
On: Mon, 14 Feb 2011 15:36:53

Undrained bulk
modulus, Ku
Drained bulk
modulus, K9

CLAYTON

Options for anisotropic parameter determination. In the case


of a transversely isotropic elastic solid it is only necessary
to measure five of the seven parameters identified above,
since Love (1892) (see Pickering, 1970) proved that
hv vh

(5)
Eh
Ev
and
Gh

Eh
21 hh

(6)

In the case of a drained material the engineer may choose to


work with one of a number of parameter sets (Pickering,
1970; Lings et al., 2000; Lings, 2001): for example,
E9v , E9h , 9vh , 9hh and Gv , or
E9v , E9h , 9vh , Gh and Gv , or
E9v , E9h , 9hv , 9hh and Gv , or
E9v , 9hv , 9hh , Gh and Gv :
The choice is arbitrary from the point of view of computation, since conversion between different parameter sets can
readily be achieved using equations (5) and (6). However, as
will be seen later, some parameters are more readily measured than others, and a combination of field and/or laboratory
techniques will be required to obtain a full five-parameter
set. For example, Youngs modulus E9v and Poissons ratio 9vh
are readily obtained from a drained triaxial compression test
with local axial and radial strain measurement. The determination of Gh and E9h is more challenging.
Limits. Thermodynamic considerations require that the strain
energy of an elastic material should always be positive
(Pickering, 1970). It follows that for an isotropic elastic
material Youngs modulus E should be greater than zero, and
Poissons ratio should fall between 1.0 and +0.5 (Pickering,
1970; Gibson, 1974). For a drained transversely isotropic
elastic material, E9v , E9h , Gv and Gh must all be greater than
zero,
1 < 9hh < 1

(7)

and
E9v
1  9hh  29vh 2 > 0
E9h

(8)

As in the isotropic stiffness case, Youngs moduli and


Poissons ratios are different in the drained (long-term) and
undrained (short-term, or end of construction) conditions,
while shear moduli remain the same (Table 1). In the
undrained case fewer parameters are required, because
(9)
uvh 0:5
Euh
2Euv

(a) Parameters at very small (ideally reference) strain levels


(e.g. E0 , 0 and G0 ). These depend upon, for example,
(i)
void ratio
(ii) grain characteristics such as particle size and
shape
(iii) current effective stresses
(iv) structure (here used in the sense of Kavvadas &
Anagnostopoulos, 1998)
(v) stress history
(vi) fabric (in the sense of Rowe, 1972) and particle
arrangement
(vii) discontinuities
(viii) rate of loading
(b) Stiffness parameters are altered by increasing strain and
changing stress levels, during loading or unloading.
Factors controlling stiffness under operational conditions include
(i)
strain level
(ii) loading path and changes in effective stress
(iii) changes in loading path
(iv) recent stress history

Typical strain ranges:


Retaining walls

Chowdhury & King (1971)


uhv

FACTORS AFFECTING MEASURED STIFFNESS


Any measurement of stiffness, whether made in the field
or in the laboratory, needs to be critically reviewed in the
context of those factors that will control the stiffness of the
ground around a structure. If conditions are not the same,
then the measured stiffness will be different, and may be of
limited value or require modification when making predictions of displacements. Therefore, in the following paragraphs, key factors affecting stiffness are reviewed.
The effect of strain level has already been noted in the
section on Historical background above. Experimental
physicists have established beyond doubt that (even for
materials much more competent than soil and weak rock),
there is no linear stressstrain behaviour. Superficially at
least, soils and weak rocks appear to behave in a similar
way to other materials, and it has been observed that for a
wide range of stiffness (e.g. Clayton & Heymann, 2001)
behaviour is sufficiently constant below a strain level of
about 0.001% for this to be taken, for practical purposes, as
the strain range within which to measure the very-smallstrain reference modulus values (E0 or G0 ) (Fig. 1).
Stiffness parameters may therefore, for practical purposes,
be considered constant at very small strains, but can be
expected to reduce as strains increase above this level. This
was the approach of Atkinson & Sallfors (1991). Because
the strain levels around well-designed geotechnical structures
such as retaining walls, foundations and tunnels are generally small (Fig. 2), measurements are required in order to
determine two sets of parameters:

(10)

Stiffness, G

Foundations
Tunnels

Gibson (1974)
uhh 1 

Euh
2Euv

(11)
00001

Gibson (1974).
Thus the parameter set can be reduced, as noted by
Atkinson (1975), to Euv , Euh and Gv .

0001

001

01

10

Shear strain, s: %

Fig. 2. Typical stiffness variation and strain ranges for different


structures (redrawn from Mair, 1993)

Delivered by ICEVirtualLibrary.com to:


IP: 152.78.201.162
On: Mon, 14 Feb 2011 15:36:53

STIFFNESS AT SMALL STRAIN: RESEARCH AND PRACTICE


(v)
(vi)

destructuring
changes in loading rate.

(See also Hight & Higgins, 1995; Jardine, 1995.)


Stiffness at very small strain
Tests on reconstituted materials have shown the important
influence that state (void ratio, current effective stresses)
and stress history can have on stiffness. The shear modulus
of a granular material at very small strain levels is affected
fundamentally by three factors:
(a) the void ratio of the specimen
(b) interparticle contact stiffness, which will depend upon
particle mineralogy, angularity and roughness, and
effective stress
(c) deformation and flexing within individual particles,
which will depend on particle mineralogy and shape.
If interparticle stiffness is removed, for example by cementing, then for a given particle shape and mineralogy the
combined effect of void ratio and particle flexing can be
seen. Fig. 3 shows shear modulus (Gv0 ) measured in the
resonant column apparatus, for cemented Leighton Buzzard
fraction E sand (rotund, uniform, D50  0.1 mm). Two types
of cement have been used: methane hydrate (Clayton et al.,
2010) and epoxy resin (Fleris, personal communication). The
location of the cement in both cases is primarily at the grain
contacts. In the case of methane hydrate this was achieved
using the excess gas method, where damp sand was
compacted into a mould to form an unsaturated specimen,
and the water then combined (under suitable thermobaric
conditions) with methane gas to form disseminated hydrate.
The epoxy-bonded specimens were formed by tumbling the
sand grains in a pre-measured quantity of epoxy resin, and
then either compacting them, or rubbing them through a
grillage into a mould. This allowed very high void ratios to
be obtained, which for both types of specimen were calculated taking into account the volume of the sand grains and
of the cement.
The results show, for this sand, a unique relationship for
upper-bound shear modulus against void ratio, independent
of the effective stress applied during testing. Lower values
of stiffness were obtained for hydrate volumes of less than
about 5% of the void space, below which stiffness is
sensitive to effective stress, suggesting that cementing of the
interparticle contacts is incomplete (Clayton et al., 2005).
The epoxy resin data are for 2%, 4% and 6% Araldite by

weight of sand, equivalent at 4%, for example, to between


4.5% and 13% of the void space between the sand grains.
For fully cemented sand, over the range tested, void ratio
has an approximately 20-fold effect on stiffness.
Hardin (1961) suggested that the shear wave velocity (and
therefore the shear modulus) of sands was influenced not
only by void ratio but also by mean effective stress. Experimental work by Hardin & Richart (1963), Hardin & Drnevich (1972) and Iwasaki & Tatsuoka (1977), carried out in
the resonant column apparatus, subsequently, as might be
expected, supported this view. For this reason, the results of
resonant column testing on pluviated or compacted sands
have classically been shown normalised by a function of
effective stress, in addition to being plotted against void
ratio.
Figure 4 shows the results of a recent survey of reported
data from resonant column testing (Bui, 2009) for both
sands and clays. For dimensional consistency the effective
stress applied during testing needs to be normalised, in this
case by atmospheric pressure, patm . The introduction of
Hertzian contact theory into expressions for the shear modulus of a pack of identical elastic spheres suggests that G0
should approximately be a function of effective stress to the
power of 1/3 (Duffy & Mindlin, 1957; Goddard, 1990) In
experiments, the exponents of individual pluviated sands
have been found to vary between approximately 0.4 and 0.6,
and (as in Fig. 4) a value of 0.5 has been observed by many
researchers and used to normalise their stiffness data (Hardin
& Black, 1966, 1968; McDowell and Bolton, 2001).
A number of equations have been derived to capture the
trends of such data. Based upon Buis survey of existing
data, a reasonable expression is
3

Gv0 Cp 1 E p9= patm 0 5 (MPa)

(12)

where Gv0 is the very-small-strain shear modulus in the


vertical plane (MPa), Cp is a constant (in MPa), e is the
void ratio of the specimen under test, p9 is the (isotropic)
effective stress applied to the specimen, and patm is atmospheric pressure (in the same units as p9).
Trend lines for the data normalised by effective stress, for
Cp 300 MPa and Cp 600 MPa, are shown in Fig. 4.
These trends are shown without normalisation, and for
Cp 450 MPa, in Fig. 3. From equation (12) and Fig. 3 it
can be seen that a tenfold increase in effective stress
produces only a threefold increase in stiffness, which is
200

Leighton Buzzard E plus Araldite


Leighton Buzzard E plus hydrate
Typical rotund sand, without cement,
at various stress levels, : kPa

Gv0: MPa

4000

3000

Structured Leighton Buzzard sand


no inter-grain compliance

2000
: kPa
1600
1000 800
400
200
100
0
0

Gv0/(/patm)05: MPa

5000
Gv0 Cp(1 e)3(/patm)05 (MPa)

100

Cp 600

0
05

10
15
Void ratio, e

20

Cp 300
0

25

Fig. 3. Effect of void ratio on cemented Leighton Buzzard


fraction E sand

2
Void ratio

Fig. 4. Stiffness Gv0 normalised by effective stress, as a function


of void ratio. From resonant column tests on pluviated sands
and reconstituted clays (Bui, 2009)

Delivered by ICEVirtualLibrary.com to:


IP: 152.78.201.162
On: Mon, 14 Feb 2011 15:36:53

CLAYTON

small in comparison with the increase that can be produced


by quite modest amounts of cement.
For clays, overconsolidation produces significant decreases
in void ratio, even at quite modest stress levels. Perhaps for
this reason, very-small-strain stiffness data for clays (in this
case from bender element testing) have been normalised by
a void ratio function before being plotted as a function of
stress (e.g. Pennington et al., 1997). For sands, relatively
large stresses are needed in the short term to induce the
grain crushing necessary to produce major changes in void
ratio, although in the longer term and over the geological
timescale this is not the case, as shown by the evidence
from aged natural materials, such as locked sands (Dusseault & Morgenstern, 1979). Fig. 5 shows resonant column
test results for two undisturbed specimens of natural materials: a sandy facies of the Eocene London clay, from a site
to the west of London, and a Lower Cretaceous (Folkestone
Beds) locked sand from a site to the south of London
(Cresswell & Powrie, 2004). Superimposed on these data are
predictions made using equation (12), with a value of Cp of
300 adopted for the sandy clay, and 1200 for the sand. The
difference in value presumably reflects that fact that for
sands the development of flats between particles will have
taken much of the compressibility out of the grain contacts,
whereas for the clay, particle flexure is an important mechanism. These data show how preliminary estimates of shear
modulus can be made on the basis of mean effective stress,
void ratio, stress history, particle grading and particle shape.
As noted above, there are other factors affecting stiffness,
so it is not surprising that equation (12) cannot completely
normalise the data. For example, data presented by Clayton
et al. (2006) and by Xu et al. (2007) show the influence of
particle arrangement on stiffness under horizontal cyclic
triaxial loading.

strain, and stiffness degradation data, are required for predictions of ground movements.
Figure 6(a) shows, as an example, the degradation of
normalised vertical Youngs modulus with increasing strain,
for triaxial compression data taken from Heymann (1998).
The test results given in the figure show a remarkable
consistency, which is increased once the higher values, for
tests involving reversed and repeated loadings, are excluded.
Given that E0 for these materials varied from approximately
24 MPa for the Bothkennar clay to 240 MPa for the London
Clay, and to 4800 MPa for the intact chalk, it is notable that
there is so little scatter around E0:01 /E0:001  0.8, and E0:1 /
E0:001  0.4. Jardines linearity index (L E0:1 /E0:01 ; Jardine
et al., 1984) is approximately 0.5.
If identical specimens are tested, or the same specimen is
tested several times without significant destructuring, then
undrained triaxial tests will produce the same very-smallstrain Youngs modulus, E0 , regardless of the approach path,
and whether tested in triaxial compression or extension.
Loading path direction does, however, have some effect at
slightly higher strains. Fig. 6(b) shows, as might be expected, that when soil is loaded towards the nearest failure

Normalised Youngs modulus, E uv/E uv0001

10

10
Loading towards isotropic stress
08

06

Shear modulus, Gv0: MPa

800
700
600
Locked sand (measured)
Locked sand (predicted Cp 1200)
Eocene sandy clay (measured)
Eocene sandy clay (predicted Cp 300)

500
400
300
200

02

100

200
300
400
500
600
Mean effective stress, p: kPa

700

600
Loading in triaxial
compression

400

200

Compression
s
Extension

0
0

01

800

0
0001

100

001
Axial strain: %
(a)

Normalised Youngs modulus, E uv/p0

900

04

0
0001

Change of stiffness with increasing strain


Jardine et al. (1986) and Mair (1993) have shown that the
typical strain levels around geotechnical structures such as
retaining walls, spread foundations, piles and tunnels fall in
the range where soil stiffness changes most dramatically
with strain, and that for many structures they are in the
range 0.010.1% (Fig. 2). Thus both stiffness at very small
1000

Multiple loadings

Loading in
triaxial extension
001
01
Local axial strain: %
(b)

800

Fig. 5. Shear modulus G0v from resonant column tests on two


natural undisturbed materials. Eocene sandy clay results on a
number of specimens reconsolidated to their approximate in situ
stress levels. Lower Cretaceous locked sand results for a single
block sample tested at a range of isotropic effective stress levels

Fig. 6. (a) Degradation of vertical Youngs modulus with


increasing axial strain. Triaxial compression data from intact
chalk, destructured chalk, undisturbed London Clay, and
undisturbed Bothkennar Clay (Heymann, 1998). (b) Degradation of vertical Youngs modulus with strain, for the same
specimen tested with the same initial effective stress, under
triaxial compression and extension (Clayton & Heymann, 2001)

Delivered by ICEVirtualLibrary.com to:


IP: 152.78.201.162
On: Mon, 14 Feb 2011 15:36:53

STIFFNESS AT SMALL STRAIN: RESEARCH AND PRACTICE


envelope, the initial rate of stiffness degradation with strain
is higher than when loading takes place away from the
nearest failure envelope. The influence of approach loading
path (the recent stress history of Atkinson et al., 1990),
while apparently unimportant at very small strain levels (less
than about 0.05% in tests by Clayton & Heymann, 2001),
can be significant at higher strains (Gasparre et al., 2007).

Anisotropy
Anisotropy can be recognised at a number of scales. At
the very small (laboratory) scale, anisotropic effects have
been variously described as inherent and induced. Inherent anisotropy results from grain characteristics (principally
form; Abbireddy et al., 2009; Clayton et al., 2009a) and
the depositional process. Casagrande & Carillo (1944) described this type of anisotropy as a physical characteristic
inherent in the material and entirely independent of the
applied stresses and strains. Fig. 7 shows a computed
tomography (CT) scan of pluviated platy sand-sized material. The orientation of the particles normal to the direction of
gravity is clear, and suggests that stiffness will be higher in
the horizontal than in the vertical direction.
Induced anisotropy is caused by stress or strain changes
following deposition, particularly those resulting from the
post-depositional application (as is normal) of different effective stresses in the horizontal and vertical directions (for
examples in relation to the London Clay, the reader is
referred to Burland et al., 1979). Changes in principal stress
directions can cause disruption of strong force chains within
granular materials (Thornton & Zhang, 2010), and changes
in memory as a result of particle rotation.
As a result of the in situ stress regime, most materials are
likely to exhibit anisotropic stiffness. Youngs modulus measured in a laboratory specimen is controlled primarily by the
effective stress in the direction of loading (Hardin & Bland-

(a)

(b)

Fig. 7. CT scan showing preferred particle orientation of 1 mm


pluviated glass glitter (Abbireddy, 2008): (a) horizontal section
(view from top); (b) vertical section (view from side)

11

ford, 1989; Yamashita & Suzuki, 1999), although because of


the Poisson effect it will also be somewhat influenced by the
effective stresses in the normal directions. Shear modulus is
controlled by the effective stresses acting in the plane of
distortion (Roesler, 1979; Yu & Richart, 1984; Stokoe et al.,
1995; Bellotti et al., 1996). This means that in a transversely
isotropic material, horizontal shear modulus Gh is a function
of horizontal effective stress alone (Butcher & Powell,
1997), whereas vertical shear modulus Gv is a function of
both vertical and horizontal effective stress.
Anisotropy also needs to be assessed at larger scales. The
stiffness of softer materials may be increased by the inclusion, for example, of more sandy or cemented layers (e.g.
claystones within the London Clay, and hydrate sheets
within deep ocean sediments; Fig. 8). The stiffness of weak
rocks is significantly reduced by fracturing, jointing and (in
the case of the Chalk) dissolution associated with stress
relief, and weathering (Lord et al., 2002, Matthews &
Clayton, 2004). In the unusual example shown in Fig. 9 the
dominant joint set, probably associated with a plane of
stiffness isotropy, is sub-vertical. More normally in the chalk
it is sub-horizontal, associated with the shallow dip of
bedding.
At the largest scale, relevant for example to the volume of
soil loaded by a large foundation, or unloaded during deep
basement or tunnel excavation, many soils and rocks show
evidence of heterogeneity in the form of bedding and of
layering of different materials within that bedding. Fig. 10
shows piezocone results in gold tailings; the rhythmic
deposition of finer and coarser materials results from variations in the position of the central pool, which (as a result
of the management of the dam) moves around the depositional area with time. Similar rhythmic deposition can be
seen in many natural deposits, for example varved clays
deposited in glacial lakes, and in the Cenomanian Chalk,
where the layering is driven by global climate changes
resulting from Milankovitch cycles (Hart, 1987).
As might be expected from the discussion above, anisotropy of stiffness has been widely recognised in natural
materials, both in seismic geophysical testing (Butcher &
Powell, 1997) and in laboratory measurements (Ward et al.,
1959; Atkinson, 1975; Graham & Houlsby, 1983). However,
few studies have been carried out in sufficient depth to
determine the full set of anisotropic stiffness parameters,
two notable exceptions being reported by Lings et al. (2000)
for the Gault Clay, and by Gasparre et al. (2007) for the
London Clay. Values of effective Youngs moduli and of
shear moduli for the London Clay at Heathrow Terminal 5
are shown in Fig. 11. Bearing in mind the likely variation of
the ratio of effective horizontal to vertical stress (K0 ) to be
expected over the 30 m profile shown in Fig. 11 (Burland et
al., 1979), these observations suggest that anisotropy of
very-small-strain stiffness seen here is likely to be dominated by factors other than effective stress ratio.
The effect of loading, and ultimately of destructuring, on
the anisotropy of stiffness remains a matter of some debate.
Jovicic & Coop (1998) suggest that very large plastic strains
are necessary to affect the inherent anisotropy of weak rocks
and stiff clays, while test data for isotropic effective stress
loading of undisturbed Bothkennar Clay (Clayton et al., 1992)
and chalk (Clayton & Heymann, 2001) suggest that even small
strains may be sufficient to change the degree of stiffness
anisotropy as a result of comprehensive destructuring.
Cyclic loading and rate effects
It has long been held that the observed stiffness of soil is
strongly dependent on the rate at which it is tested. As a
result, the stiffness values obtained from field seismic or

Delivered by ICEVirtualLibrary.com to:


IP: 152.78.201.162
On: Mon, 14 Feb 2011 15:36:53

12

CLAYTON

Hydrate veins

(a)

(b)

Fig. 8. Sub-vertical orientation of methane hydrate veins in very soft deep ocean
sediment: (a) CT scan of methane hydrate veins in a very soft deep ocean
sediment core; (b) lower hemisphere projections from three core sections,
showing preferred orientations of hydrate veins

qc: MPa
4
6

10

20

qc

21

22

Depth: m

u
Slimes

23

Sands

24

25
0

Fig. 9. Structured chalk, showing preferred orientations of


discontinuities (image courtesy of Professor R. N. Mortimore,
University of Brighton)

200

400
600
u: kPa

800

1000

Fig. 10. CPT profiles of pore pressure and cone resistance from
gold tailing (Obuasi, Ghana) showing interlayered slimes (fines)
and sands (data courtesy of Professor E. Rust, University of
Pretoria)

Delivered by ICEVirtualLibrary.com to:


IP: 152.78.201.162
On: Mon, 14 Feb 2011 15:36:53

STIFFNESS AT SMALL STRAIN: RESEARCH AND PRACTICE


0

Effective Youngs moduli, E: MPa


200
400

13

Shear moduli, G: MPa


100
200

10

Depth: m

Depth: m

10

20

20

30

30
Ev

Gv

Eh

Gh

Triaxial tests

Bender element tests

Hollow cylinder apparatus

Static torsional test


Resonant column apparatus

Fig. 11. Profiles of small-strain effective Youngs moduli and shear moduli for the London Clay at
Heathrow Terminal 5 (modified from Gasparre et al., 2007)

600

laboratory dynamic tests were thought for many years to be


large overestimates of the static stiffnesses required for most
engineering predictions of ground deformations. Two factors
are now known to have been behind the development of this
view.

Rate effects are now considered to be relatively unimportant


at very small strain levels. For example, for the tests on stiff
clays and mudstones reported by Tatsuoka & Shibuya
(1992), stiffness was found to be almost independent of
strain rate for strains ,0.001%. At higher strains, three
significant effects have been observed: see for example
Isenhower & Stokoe (1981), Tatsuoka & Shibuya (1992) and
Lo Presti et al. (1997). First, the extent of the elastic
plateau increases with strain rate, so that the results of
resonant column tests, cyclic and monotonic loading tests
cannot be expected to be the same at small (as distinct from
very small) strains. Second, shear stiffness becomes more
sensitive to rate of loading at intermediate strains, say
between 0.01% and 0.1% strain (see also Sorensen et al.,
2007; Fig. 12), and finally the stressstrain response under
cyclic loading can be expected to be stiffer than under
monotonic loading. Because of this, Lo Presti et al. (1997)
conclude that the very high cyclic strain rates imposed by
resonant column testing make it not very suitable for the
measurement of static monotonic stiffness degradation. The

Deviator stress, q: kPa

(a) The effects of sampling disturbance in many cases led


to reductions in the stiffness measured in laboratory
tests, through destructuring during sampling and as a
result of associated decreases in effective stress.
(b) Tests carried out on reconstituted and destructured
materials did indeed demonstrate significant rate effects,
but the material tested was not representative of natural
material.

Intact London Clay

400
.
a 08%/h
.
a 02%/h
.
a 005%/h
200

2
3
Shear strain, s: %

Fig. 12. Effects of changes of strain rate during shear on


deviatoric stress. Intact London Clay (Sorensen et al., 2007)

results are likely to provide a lower limit when compared


with other measurements.
SIGNIFICANCE OF PARAMETERS FOR PREDICTED
PERFORMANCE: A NUMERICAL EXPERIMENT
As noted above under Geotechnical background, backanalysis of monitored construction, and particularly of deep
excavations in the London Clay, has indicated the complexity of soil behaviour, and has suggested that, particularly if
displacement patterns are to be predicted, soil needs to be

Delivered by ICEVirtualLibrary.com to:


IP: 152.78.201.162
On: Mon, 14 Feb 2011 15:36:53

CLAYTON

14

treated as a non-linear transversely isotropic material. A


significant number of stiffness parameters are required for
such a material, as described in the section on Constitutive
frameworks for stiffness, and the question then arises as to
whether all of these need be determined with the same
accuracy. For example, given that they relate to strain levels
much smaller than are expected in the zone of influence, are
the values of E0 and G0 really significant when attempting
to predict deformations close to such structures?
The answer to such a question must be that the importance of different parameters depends on the ground, on the
structure, and on the aspect of performance to be predicted.
For example, it might intuitively be expected that horizontal
Youngs modulus would be particularly significant in controlling the horizontal displacement of retaining structures,
while vertical stiffness may be more significant when predicting the settlement of spread foundations. Therefore in
practice some kind of sensitivity analysis will be required in
many cases, in order to identify which parameters dominate
the particular problem under consideration.
As a demonstration of the significance of different stiffness parameters, a numerical experiment has been carried
out, to estimate the ground deformations around a singly
propped retaining wall. The underpinning methodology is
described below. The analyses were carried out using twodimensional FLAC version 5.0 (Itasca, 2005). Stiffness
degradation was implemented using a FISH function.
Problem geometry
Figure 13 shows the geometry of the selected problem.
Dimensions are similar to those of a dual-carriageway highway underpass. The excavated depth is 8 m, and the full
width of the excavation is 30 m (i.e. the distance from the
wall face to the centreline of the excavation is 15 m). The
0.6 m thick retaining wall is 16.5 m long, and is supported
by props at 1 m below ground level, with loads equivalent to
an 8.75 m centrecentre spacing. A preliminary parametric
study was undertaken to explore the effects of mesh size
and boundary locations (Iqbal, personal communication). For
the analyses reported here, computational time was reduced
by placing the vertical boundary 80 m back from the face of
the retaining wall, with the basal boundary 40 m below
ground level. A single soil type was used in each analysis.
The wall was wished in place (Gunn et al., 1992), and a
uniform value of K0 1 was therefore used to calculate
starting in situ stress levels (Gunn & Clayton, 1992). Excavation was modelled in 1 m stages, with the prop being
installed after the first excavation step.

Soil models
Analyses were run with two sets of variables:
(a) Uniform stiffness, or stiffness increasing with depth.
(b) A range of constitutive models:
Case 1 Linear elastic soil, with Eu 100 MPa.
Case 2 Case 1, but with MohrCoulomb plastic yield
at su 100 kPa.
Case 3 Linear elastic soil, with stiffness increasing
with depth.
Case 4 As in Case 3, but with stiffness decreasing
with strain. A base case was used to explore
the effects of some variables (e.g. very-smallstrain stiffness, and rate of stiffness degradation) on the displacements predicted by this
model.
Case 5 As in Case 4 base case, but with various
degrees of transverse isotropy (Euh . Euv , etc.).
Analysis Case 3 was based on the short-term parameters
deduced by Hooper (1973) from movement around the Hyde
Park Cavalry Barracks excavation.
As the predictions were for the undrained (short-term)
case, only one parameter (Eu ; recall that  u  0.5, and G
and K are dependent on Eu ) was required for the isotropic
Cases 1, 2, 3 and 4. In Case 3 Eu varied with depth, as
shown by the dashed line in Fig. 14.
In Case 4 stiffness increased with depth but reduced with
increasing strain. The base case adopted a reference stiffness, Eu0 , arbitrarily taken as four times the values backanalysed from Hyde Park Cavalry Barracks (hpcb) (Hooper,
1973). Stiffness degradation was modelled by assuming
constant values of tangent stiffness above, below and between fixed octahedral strain limits shown in Table 2.
Figure 14 shows the variations of stiffness with depth at
different strain levels, and Fig. 15 compares the stepped
input tangent stiffness values with the secant Youngs modulus degradation curve computed from them, for soil at
10 m depth.
The transversely isotropic cases explored in Case 5
required three independent parameters (Euv , Euh and Gv ). The
ratio Euv =Euh was varied from analysis to analysis, and Gv
was obtained from

Undrained secant Youngs modulus, E usec: MPa


100
200
300
400

500

0
Hyde Park Cavalry Barracks
Uniform stiffness
5

BA
8m

Props at 1 m depth

Depth: m

15 m

10

85 m
15

Base case (Case 4)


for non-linear analyses
E0 4Ehpcb 0002%
0006%
0002%
002%
0006%

20
CL
A

Fig. 13. Selected retaining wall geometry for numerical analysis

02% 006%

006% 002%

Fig. 14. Comparison of undrained secant Youngs moduli values


at different strain levels, as a function of depth, for non-linear
base case, E0
4Ehpcb

Delivered by ICEVirtualLibrary.com to:


IP: 152.78.201.162
On: Mon, 14 Feb 2011 15:36:53

STIFFNESS AT SMALL STRAIN: RESEARCH AND PRACTICE


Table 2. Base case reduction in tangent stiffness values

2
0

Stiffness ratio, Eu =Euhpcb

Octahedral strain level: %


,0.002
0.0020.006
0.0060.02
0.020.06
0.060.2
0.20.6
.0.6

4
2.5
1.5
0.7
0.35
0.15
0.05

Depth below ground level: m

At 10 m below ground level

10

12

14

Case 2
Case 1

10
Base of wall
20

Case 4

Cases 1 and 2

Case 3

30

E0 4. Eu hpcb
1
2
3
4

200

Resulting secant
Youngs modulus
100

Uniform linear elastic, Eu 100 MPa


Uniform linear elastic, MohrCoulomb
Linear, stiffness increasing with depth
Non-linear elastic, stiffness
increasing with depth

Fig. 16. Horizontal displacements on plane AA (back of wall)


for four soil models
Input tangent
Youngs modulus

0
00001

0001

001

01

10

Strain: %

Fig. 15. Secant moduli resulting from input tangent moduli, for
base case at 10 m below ground level

Gv

Euh Euv
2Euh Euv

(13)

Lekhnitskii (1981)
As noted by both Simpson (1992) and Atkinson (2000),
the mobilised strength at any strain level is equal to the area
under the tangent stiffnessstrain curve, up to that strain
level. Therefore, even though the soil is modelled as elastic,
there are restrictions on the values of stiffness that can be
used as input. For example, increasing the rate of stiffness
degradation with strain will reduce the available strength at
a given strain level, and in an undrained retaining wall or
spread foundation, analysis may prevent stability within
reasonable deformation limits. For the base case described
above, the mobilised undrained shear strength at 1% strain is
of the order of 60 kPa at 10 m depth, which is a relatively
low value for the London Clay (Marsland, 1972; Hight,
1986).
Impact of model and parametric variations on predicted
displacements
In order to simplify the discussion, the following key
outputs are compared below for different soil models and
parametric values:
horizontal wall displacements
vertical displacements at original ground level
vertical displacements at excavation level
bending moments and prop loads.

Figure 16 shows horizontal displacements on the plane of


the back of the wall (shown as AA in Fig. 13). The use of
a constant stiffness profile with depth (Case 1) leads to
unrealistic predictions of the pattern of displacement of the
wall, when compared with field observations in the London

Clay. Introduction of MohrCoulomb yielding (Case 2) has


little effect. Increasing stiffness with depth (Case 3) has a
major impact on the shape of wall deflections, predicting
maximum horizontal movements at about excavation level
(as observed in practice, e.g. by Burland & Hancock, 1977).
The predicted shape is further enhanced by the introduction
of higher stiffnesses at small strains (Case 4). It is clear that,
for a problem of this type, determination of the stiffness
profile must be a priority.
Figure 17 shows the vertical displacements at original
ground level, behind the wall. Again, there is little difference
between Case 1 and Case 2. Cases 1, 2 and 3 show heave,
and tilt away from the wall, at between approximately 10 m
and 20 m. In contrast, Case 4 shows settlement between 5 m
and 25 m behind the wall, associated with tilt towards the
excavation. This mirrors the case record at New Palace Yard,
where on the basis of a linear-elastic analysis Big Ben was
predicted to tilt away from the excavation for the new House
Back of wall
20

15
Cases 1 and 2

Heave: mm

Undrained secant Young's modulus, E usec: MPa

Base of excavation

300

(a)
(b)
(c)
(d )

15

Horizontal displacement: mm

10
Excavation
5

Case 3
0

Case 4
5

60

40

20
Distance behind wall: m

20

Fig. 17. Heave of soil at original ground level (BB in Fig. 12)
for four different soil models

Delivered by ICEVirtualLibrary.com to:


IP: 152.78.201.162
On: Mon, 14 Feb 2011 15:36:53

Case 1

Depth: m

Case 2

Case 3

Case 4

10

Undrained secant Youngs modulus, E usec: MPa

CLAYTON

16

500
E0 2 E0
in Case 4

400
300

Case 4 base case


Linear range
extended to 002%

200
100
E0 reduced to Case
4 at 0006% strain
0
00001

0001

001

01

10

Strain (%)
(a)
20

20

40

60

80

100

120

140

160

Bending moment: kNm

Excavation

Uniform linear elastic, Eu 100 MPa


Uniform linear elastic, MohrCoulomb
Linear, stiffness increasing with depth
Non-linear elastic, stiffness
increasing with depth

Heave: mm

1
2
3
4

Back of wall

10

180

Fig. 18. Predicted bending moments in the wall, for four


different soil models

E0 reduced to Case
4 at 0006% strain

4
Linear range
extended to 002%

Base case
Case 4

0
2

(a) reference stiffness moduli (E0 and G0 )


(b) rate of stiffness degradation.
Figure 19(a) shows four variations of stiffness at very small
strains. The effects on surface settlement can be seen in Fig.
19(b). These variants were produced by changing the magnitude and range of the very-small-strain tangent modulus, Eu0 .
Both the shape and the magnitudes of settlements behind the
wall are affected. Fig. 20 shows different rates of stiffness
degradation. The shaded area is taken from the experimental
results previously shown in Fig. 6(a). The left-hand curve
shows Case 4 base case values, and the other two curves
show additional lower rates of stiffness degradation assumed
for additional analyses. At any given intermediate strain
level the expected stiffness varies very significantly, depending upon the line adopted. For example, at 0.02% strain the
stiffness increases by 50%, and then doubles, as one moves
from the base case through to the reduced rates of stiffness
degradation shown by the other two lines in Fig. 20. For this
problem, there are very significant associated reductions in
the predicted deformations of the wall, the ground surface,

4
60

40

20
0
Distance in front of wall: m
(b)

20

Fig. 19. Effect of changes in very-small-strain stiffness on


vertical movement behind the wall: (a) variations in Eu at
10 m depth; (b) predictions of surface settlement behind the
wall
Undrained secant Youngs modulus, E usec: MPa

of Commons car park, but was observed in reality to tilt


towards it (Burland & Hancock, 1977; Simpson et al., 1979)
Figure 18 shows predicted bending moments in the wall,
for Cases 1 to 4. Again, there is little difference between the
predictions for Case 1 and Case 2. As would be expected
from the deformed shapes in Fig. 16, the introduction of a
stiffness increase with depth leads to a significant increase
(more than 50%) in maximum bending moment, and this is
further enhanced by the increase of stiffness at small strains.
The use of the simple models leads to lower predictions of
bending moment. These changes do not affect prop load,
however, which for this example were found to remain fairly
constant (10%), being largely a product of initial horizontal effective stress and wall geometry.
In addition to the variations in constitutive models described above, a number of parametric variations have been
used in conjunction with Case 4 (isotropic stiffness increasing with depth and decreasing with strain) in order to
explore the sensitivity of key outputs to uncertainties in

E0 2 E 0
in Case 4

300
E0 4 Eu hpcb

At 10 m below ground level


Shaded area from Fig. 6(a)

200

100

0
00001

Reduced rates
of stiffness
degradation
Base case
Case 4

0001

001
01
Strain: %

10

Fig. 20. Undrained secant moduli against strain, showing base


case 4 and two reduced rates of stiffness degradation, compared
with observed values in the triaxial test (Fig. 6(a))

and (Fig. 21) the vertical movements at base of excavation


level.
Finally, Fig. 22 shows the effect of undrained modular
ratio (Euh =Euv ) on the predicted maximum horizontal wall
movement. Maximum wall movement was normalised by the
value from the isotropic (Case 4, base case) analysis.
Horizontal Youngs modulus has a large effect, and a modular ratio of 2.5, approximately the value expected in the
London Clay Formation (Fig. 11), halves the predicted
magnitude of wall movement.

Delivered by ICEVirtualLibrary.com to:


IP: 152.78.201.162
On: Mon, 14 Feb 2011 15:36:53

STIFFNESS AT SMALL STRAIN: RESEARCH AND PRACTICE

MEASURING STIFFNESS PARAMETERS


The selection of methods for measuring stiffness at any
given site needs to be made in the context of a number of
factors:

Linear elastic, stiffness


increasing with depth

30

Base case Case 4


Non-linear elastic,
stiffness increasing with
depth

(a) the variability of the ground


(b) the relative merits of field and laboratory measurement
techniques
(c) prior experience of the use of the technique in the
given ground conditions
(d ) the availability of equipment and personnel in the
country or region where the work is to be carried out
(e) the need for redundancy of data.

Heave: mm

20

Reduced rates of
stiffness degradation
(see Fig. 20)

10

80

60

40
20
Distance in front of wall: m

20

Fig. 21. Effects of rate of stiffness degradation on predicted


vertical movements at excavation level

Ratio of maximum wall displacements

12
10
08
06
uvh 049

04
Gh

02

E uh
2(1uhh)

uhh 1 E uh /2E uv
Gv

E uhE uv
2E uh E uv

0
1

15
20
Modular ratio: Eh /Ev

17

25

Fig. 22. Case 5: effect of stiffness anisotropy on maximum wall


displacement

In summary, for the particular problem that has been


analysed here, an assumption of uniform stiffness with depth
leads to unrealistic wall deflections and low predictions of
bending moments. Increasing stiffness with depth gives
better estimates, when compared with field observations in
the London Clay. It is clear that the determination of a
reliable stiffness profile must be a priority in any investigation. Although a number of different combinations of soil
stiffness model may seem from the figures to give broadly
similar estimates of ground movements, high initial stiffness,
coupled with stiffness degradation with increasing strain, is
needed to mimic the pattern of observed ground surface
movements for structures that take the soil to intermediate
strain levels, for example at the House of Commons car
park. Predicted displacement patterns are sensitive to most
parameters, including very-small-strain stiffness, rate of stiffness degradation, and anisotropy.
These observations may not be generally true, however. In
stiffer materials, strain levels may be much smaller, and
closer to the elastic plateau. In softer materials, significant
destructuring may take place, and an elastic approach to
deformation modelling may not be appropriate.

This section first discusses these issues, before passing on to


give examples of a range of techniques that have been used
by the author.
The heterogeneity of the ground is important, because
even in the most intensely investigated site it is unlikely that
more than one part in one million of the volume of ground
affected by construction will be sampled, seen (for example
in trial pits or as core), or mechanically explored (e.g. using
penetrometers) (Broms, 1980). If, as is frequently the case,
there is a high degree of vertical variability but relatively
little lateral variability (e.g. as a result of stratification or
weathering), then, having established lateral correlations between different layers (for example by profiling, by index
testing, or by classification testing) it may be practical to
determine the stiffness of the different layers. But if lateral
continuity cannot be established, then the priority must be to
carry out profiling, perhaps deducing stiffness from simple
and approximate correlations (e.g. between CPT or SPT and
Youngs modulus). In such situations the advanced and
generally more reliable methods of stiffness measurement
described in the paper are unlikely to be of practical use.
The relative merits of field and laboratory testing have
been well rehearsed over the years (e.g. Dyer et al., 1986;
Clayton et al., 1995b). In terms of stiffness determinations
(as will be discussed further below), field seismic testing
techniques can be significantly affected by background noise.
But because they can be very effective in determining
subsoil geometry and heterogeneity, are carried out at the in
situ stress level, and can test large volumes of soil (so
including the effects of smaller-scale heterogeneities, such as
fractures, and large particle sizes), they remain attractive for
major projects, such as deep excavations, tall structures and
seismically sensitive projects (e.g. nuclear power plants). In
situ test methods can, in most cases, avoid the worst effects
of borehole and sampling disturbance, although installation
and bedding effects can still be significant when relatively
small volumes of soil are tested close to the wall of an
exploration hole (for example during pressuremeter testing).
Laboratory tests can also suffer from background noise
(of various types), and can be impractical, because long
testing times can delay the design process. In addition, all
laboratory test specimens will have been disturbed to some
extent by drilling and sampling. Sample disturbance can
make the results of laboratory tests unrepresentative, through
three mechanisms:
(a) removal of total stresses (so called perfect sampling;
Skempton & Sowa, 1963), which includes the removal
of any shear stresses that exist in the ground
(b) changes in effective stress, as a result of tube sampling
strains (Clayton et al., 1998), or air entry and swelling
(c) destructuring (Clayton et al., 1992; Hight & Jardine,
1993).
On the positive side, as will be seen, laboratory tests
generally have controlled boundary conditions, and for this
reason can be used to obtain a wider range of parameters

Delivered by ICEVirtualLibrary.com to:


IP: 152.78.201.162
On: Mon, 14 Feb 2011 15:36:53

CLAYTON

18

than in situ tests. In addition they are (hopefully) carried out


in a better-regulated (laboratory) environment than are field
tests.
The availability of test equipment, experienced test personnel and written standards and method statements is
clearly very important in all testing, but is critical for many
stiffness tests, which can be complex. Experience of testing
in similar soil conditions is essential. In this respect laboratory testing has the advantage that samples can be flown to
key laboratories, while field seismic testing must sometimes
rely on (reasonably) local skills.
Finally, when good determinations of stiffness are essential, for example because the range of measured values
produce significantly different designs, there is a need for
data redundancy. Poorly conducted tests, or tests affected by
background noise (for example) can then be identified and
ignored. Combinations of field and laboratory tests, tests for
different stiffness parameters (Eu and G, for example) and
tests at different strain levels are helpful in this respect.
Marsland (1986) has stated that
the choice of test methods and procedures is one of the
most important decisions to be made during the planning
and progress of a site investigation. . . . In assessing the
suitability of a particular test it is necessary to balance the
design requirements, the combined accuracy of a test and
associated correlations, and possible differences between
test and full-scale behaviour.
A great many techniques exist from which stiffness parameters can be derived, ranging from the simple SPT to the
sophisticated self-boring pressuremeter. This paper considers
a limited selection of more unusual techniques, based on the
authors experience and belief that they will have value in
many situations. In particular, two classes of test are
reviewed:
(a) field geophysics
(i)
continuous surface wave testing
(ii) down-hole geophysics
(iii) cross-hole geophysics
(b) laboratory methods
(i)
bender element testing
(ii) resonant column testing
(iii) advanced triaxial testing.

wave testing on the London Clay at Brent, which Burland


(1989) compared with undrained Youngs modulus values at
0.01% axial strain made using local-strain instrumentation
on specimens of the London Clay Formation from Canons
Park, North London, noting that the dynamic values of
undrained Youngs modulus were only about 30% greater
than the values of Eu(0:01) .
Over a period, Hoar & Stokoe (1978), Abbiss (1981), Chu
et al. (1984), Sully & Campanella (1995), Bellotti et al.
(1996), Hight et al. (1997) and others have demonstrated the
potential for measuring stiffness anisotropy. But despite the
practical potential for seismic field tests to provide valuable
stiffness data, seismic techniques remain relatively unknown
in general geotechnical engineering practice. The following
sections describe the technical background, and some test
methodologies, and give examples of their application.

Background. The seismic field geophysical techniques used


in geotechnical engineering make use of two types of seismic
wave:
(a) body waves, which travel through the body of a solid,
unaffected by its surface, with a velocity and ray path
controlled only by the density and stiffness, and their
variation
(b) surface waves, which in general propagate along the
interfaces between materials with different densities
and/or stiffnesses, or along the ground surface.
There are two types of body waves: primary (P), first
arriving, compressional waves; and secondary (S), or shear
waves. P waves induce volumetric strain (Fig. 23(a)), and
therefore travel at a speed related to the undrained volumetric stiffness of the ground, since the dominant frequencies (20400 Hz, according to Woods, 1994) do not allow
drainage. In saturated near-surface soils, values of compressional wave velocity are typically found to be of the order
of 1500 m/s, the calculated undrained bulk modulus being
similar to that of water rather than that of the volumetric
Direction of wave travel

Field geophysics
Up until the 1980s it seems to have been widely assumed
that stiffnesses measured in dynamic (laboratory and field
seismic) tests might be about one order of magnitude higher
than those needed for analysis of ground movements, and
were therefore only of practical significance for dynamic
problems, such as the effects of machinery vibration, or
earthquake loading on construction (Ballard & MacLean,
1975; ASCE, 1976). During the late 1970s and the 1980s,
and partly as a result of the realisation by geotechnical
researchers that statically measured small-strain stiffness was
much higher than previously thought, it became apparent
that field seismic testing might be used to determine stiffness values for more routine, static, geotechnical design.
Abbiss (1979) used first arrival times in a seismic refraction survey, coupled with an interpretation based on Dobrins
(1960) equation for seismic velocity increasing linearly with
depth, to determine the Youngs modulus values of the
fractured Chalk Mundford, and found encouraging agreement
with stiffness values obtained from both down-hole
(865 mm) plate tests, and values back-figured from observed
ground movements beneath an 18.6 m diameter tank loading
test. He later reported (Abbiss, 1981) stiffness values derived
from continuous surface wave and seismic refraction shear

Compression wave
(a)

Shear wave

(b)

Shear wave

(c)

Fig. 23. Compressional and shear wave travel: (a) volumetric


distortion. Vp depends upon the volumetric compressibility of
both soil skeleton and pore water. (b) Shear distortion in the
vertical plane; Gv0 rV 2s hv . Rayleigh waves travel at similar,
but slightly slower, speeds. (c) Shear distortion in the horizontal
plane. Gh0 rV 2s hh

Delivered by ICEVirtualLibrary.com to:


IP: 152.78.201.162
On: Mon, 14 Feb 2011 15:36:53

STIFFNESS AT SMALL STRAIN: RESEARCH AND PRACTICE


skeletal stiffness of the soil. This makes the measurement of
P-wave velocity unattractive in most geotechnical surveys.
Shear waves (Figs 23(b) and 23(c)) induce change in
shape without change in volume, and (provided the correct
bulk density can be estimated or measured) the stiffnesses
determined from them are then independent of whether or
not the ground is saturated. Shear waves travel at a velocity
that is a function of soil density and shear stiffness in the
plane of distortion, arriving after the compressional waves.
Thus, if a seismic source rich in both P and S waves is used,
the S-wave first arrivals may be obscured by the P waves,
and the travel time overestimated. Provided that they can be
detected, shear-wave arrivals can be used to determine the
shear modulus, G. From this, Youngs modulus and bulk
modulus can be calculated, if Poissons ratio is known or can
be estimated, and the ground stiffness is assumed to be
isotropic. In anisotropic ground, shear wave velocities (from
different modes of distortion; compare Figs 23(b) and 23(c))
can in principle be used to determine both Gv and Gh .
Most of the energy input by a source at the ground
surface will travel away from the point of input as a
Rayleigh wave. The Rayleigh wave is a species of surface
wave (the other being the Love wave) that results from the
interaction of compressional and shear waves at the ground
surface, propagating away from a surface energy source with
an elliptical motion in the vertical plane. In given ground
conditions the Rayleigh wave will travel a little slower than
that of a vertically polarised shear wave. It is a function of
bulk density Gv and Poissons ratio, and, all other things
being equal, for Poissons ratios of 0.25 and 0.5 the shear
wave velocities will be greater than the Rayleigh wave
velocities by 9% and 5% respectively. Rayleigh waves are
dispersive; when (as is usual) stiffness varies with depth,
their velocity (VR ) varies with wavelength (), because longer wavelength energy engages with deeper, stiffer ground.
Figure 24 illustrates the layouts and principles of three
established field geophysics techniques that will be discussed
below. Fig. 24(a) shows continuous surface wave (CSW)
testing. A vibrator, which may be mechanical, servo-hydraulic or electro-magnetic, applies a single-frequency sinusoidal
force at the ground surface. Rayleigh waves travel away
from the vibrator, and are detected by co-linear geophones
at a range of distances from the source. By varying the input
frequency a profile of phase velocity against wavelength is
obtained, from which a stiffnessdepth profile can be computed.
Figure 24(b) shows down-hole seismic testing. This uses a
surface source (a sledgehammer striking a weighted metal
beam, for example) to input shear wave energy to the
ground. For practical reasons, and to avoid significant energy
travelling down the borehole and its casing, the energy
source is offset from the top of the hole, such that the travel
distance (typically calculated on the basis of a straight ray)
is greater than the depth. In a noisy environment, data from
a number of blows can be stacked (i.e. added to each
other) to improve the signal-to-noise ratio of the received
signal. The arrival of seismic energy is detected at depth
either by geophones clamped within a plastic-cased borehole
(to avoid borehole collapse while allowing transfer of energy
from the ground to the geophones), or by geophones within
a seismic CPT. In either case, it is desirable to have two sets
of three orthogonally orientated geophones in each detector
array, separated vertically by about 1 m. This allows the
travel time to be determined from waveforms detected at
both sensors from the same hammer blow.
Figure 24(c) shows the principle of cross-hole seismic
testing. Three co-linear boreholes, lined with grouted plastic
(ABS) casing and at a 57 m separation, are generally used.
A borehole verticality survey is required in order to calcu-

19
Geophones

Gv

(a)

Three-component geophones

Gv

Seismic CPT

(b)

Down-hole hammer

G v, G h

Three-component geophones

(c)

Fig. 24. Three established field seismic testing techniques:


(a) continuous surface wave; (b) down-hole; (c) cross-hole

late the actual distance between the boreholes at each test


depth (typically 1 m intervals), since some deviation from
vertical will have occurred during drilling and casing installation. A down-hole shear wave energy source and two sets
of three-component geophones are lowered to the bottom of
the holes and progressively raised, and clamped to the borehole walls to generate shear waves and take data, typically
at 1 m intervals. The use of two sets of receivers avoids the
issue of trigger accuracy, but increases the cost of this type
of test. The inter-borehole distance is divided by the travel
time at each depth, determined either on a first break or
peak-to-peak basis, to calculate the shear wave velocity.
Most commonly, the energy source is clamped in the borehole and struck vertically, to produce a vertically polarised
horizontally travelling shear wave, from which Gv can be
calculated. Horizontally polarised, horizontally travelling
shear wave sources have also been used (Hoar & Stokoe,
1978; Woods & Henke, 1979; Sully & Campanella, 1995;
Butcher & Powell, 1997), from which Gh can, in favourable
conditions, be determined.

Delivered by ICEVirtualLibrary.com to:


IP: 152.78.201.162
On: Mon, 14 Feb 2011 15:36:53

CLAYTON

(a) The test data, once processed, produces estimates of


only a single parameter, Gv , at very small strain.
(b) On a noisy site it may be very difficult to record
signals with the necessary coherence to avoid the
generation of a scattered stiffnessdepth profile.
(c) The depth of investigation is limited. Experience with
lightweight vibrators suggests that it will be about 5
8 m in a stiff clay, rising to 1020 m in weak rock.
(d ) Interpretation of CSW data relies generally on very
simplistic interpretations (for example, wavelength/3),
although more sophisticated inversion (Haskell
Thomson method referred to by Lai & Rix, 1998) or
dynamic finite-element modelling methods (Clayton et
al., 1995a) can be used.
(e) The ground may vibrate in a number of modes, which
in routine testing may not be recognised; and near-field
effects may be significant.
( f ) The Poissons ratio uncertainty leads to a possible error
in predicted stiffness of about 10%.
(g) In complex ground, interpretation may be made
uncertain by aliasing.
(h) A complex (irregular) ground surface can significantly
affect Rayleigh wave propagation, leading to difficulty
in interpreting data.
Despite these limitations there are many situations where the
advantages of CSW make it an invaluable tool.
(a) It is a relatively low-cost technique.
(b) It is non-intrusive, which contributes to its low cost, but
is also an advantage when working on contaminated
land.
(c) The test requires relatively little space, for shallow
depths.
(d ) It can be used to determine the stiffness profiles of
near-surface materials, which are important (for example) in linear and low-rise projects (highways, pipelines,
housing).
(e) It can provide stiffness profiles in highly weathered and
fractured ground, and where coarse particles (e.g.
boulders) prevent most other methods being used.
Figure 25 shows a recent example (Heymann et al.,
2008), where CSW testing was carried out for the South
African Gautrain project, and the results compared with
stiffnesses back-analysed from the ground movements beneath a 20 m 3 20 m 3 10 m high load provided by concrete
kentledge. The material tested was composed of chert gravels and boulders in a matrix of hillwash sand to approximately 2 m, underlain by dolomite residuum comprising
wad and chert in highly variable proportions down to
bedrock. It is almost impossible to obtain values of stiffness
in such materials, except through expensive and timeconsuming area load tests. Stiffnesses back-figured from the
data from two extensometers (A and B) located under the
kentledge are shown, for three depth ranges, on the righthand side of Fig. 25. The reduction of stiffness with increasing strain can clearly be seen, and the stiffnesses of deeper
materials tend to be greater. Stiffnesses derived from CSW

1200

Back-analysed
field data

CSW
data
B
A

02 m

800

26 m

/26

612 m

Continuous surface wave testing. An introductory text on the


practicalities of geotechnical surface wave testing is provided
by Matthews et al. (1996). Although spectral analysis of
surface waves (SASW) (Nazarian & Stokoe, 1984; Stokoe &
Nazarian, 1985) is the more economical in terms of
equipment and test time, and can allow greater depths to
be explored, the CSW method, which uses a mono-frequency
source, has been found to provide better data in a
geotechnical setting, because unwanted background noise is
more easily recognised, avoided or filtered.
CSW has several disadvantages.

Youngs modulus, Ev: MPa

20

Depth
02 m
26 m
612 m

B
A
A

400

Lai & Rix (1998)


0

001
Vertical strain: %

01

Fig. 25. Comparison of stiffnesses derived from CSW with those


back-analysed from ground movements beneath a loaded area.
Test site 55: extensometers A and B (redrawn from Heymann et
al., 2008)

testing, interpreted in two ways (Gazetas, 1982; Butcher &


Powell, 1996; Lai & Rix, 1998), are shown on the left-hand
side of the graph. As might be expected in such difficult
ground conditions, there is considerable variation in measured stiffness, but the values obtained from CSW testing
appear to give a reasonably conservative estimate when
compared with those from full-scale measurements at smaller strain levels. Matthews et al. (2000) have similarly shown
good agreement between CSW stiffness measurements and
those obtained from 1.8m diameter plate loading tests on
weathered and fractured chalk.

Down-hole geophysics. The potential problems of measuring


stiffnesses in an urban setting and on a live construction site
are illustrated by a case history given by Hope et al. (1998).
At the time of the seismic surveys reported in this paper a
highway was under construction in an old railway cutting,
which created complex ground surface geometry in the area
of testing. The available space within which to carry out the
surveys was very limited, and lay immediately alongside the
site. Ground conditions consisted of 34 m of made ground
and glacial till, overlying weathered mudstones and sandstones. Six seismic methods were applied in an attempt to get
stiffness data, in order to enhance a dataset previously
developed using dilatometers and pressuremeters. Problems
of ground-borne vibration were created by traffic on nearby
roads, by construction plant, and by a nearby electricity
substation.
Of the six methods initially proposed (parallel cross-hole
with vertically polarised shear waves, CSW, downhole seismic profiling, SASW, shear wave refraction, and uphole
seismic profiling), only two (CSW and downhole seismic
profiling) could achieve a good enough signal-to-noise ratio
to produce credible results. The data from these are shown
in Fig. 26, along with the dataset from dilatometer and
pressuremeter testing along the length of the road. A number
of lessons can be learnt.
(a) The limited effective depth of CSW testing (about 5
7 m) can be seen.
(b) The difficulties of down-hole testing near to ground
surface are obvious, as the scatter of this dataset within
3 m of ground level shows.
(c) Despite all the problems, data were produced, but this
was only because flexible and varied arrangements
could be used to obtain them. A rigid contract,

Delivered by ICEVirtualLibrary.com to:


IP: 152.78.201.162
On: Mon, 14 Feb 2011 15:36:53

STIFFNESS AT SMALL STRAIN: RESEARCH AND PRACTICE


0

100

200

300

21

Shear modulus, Gv0: MPa


100

Gv0: MPa

200

Depth: m

Depth: m

10

10

20

Down-hole profiling
15

Continuous surface wave


London

Dilatometer
Weak rock pressuremeter
Self-boring pressuremeter

Site locations
30

20

Fig. 26. Comparison of CSW and down-hole stiffness measurements for a noisy weak-rock site with complex surface geometry
(from Hope et al., 1998)

A1 North Circular (Gordon, 1997)


Chattenden (Hope, 1993)
Surrey Research Park (Gordon, 1997)

Cross-hole geophysics. During the 1990s the author and


colleagues from the University of Surrey carried out a
number of seismic surveys in the south-east of England. The
data from three cross-hole surveys in the London Clay are
supplemented by those from a later survey reported by Hight
et al. (2007) in Fig. 27. Also shown in Fig. 27 is an inset
map, showing the locations of the four sites. Some sites (e.g.
the Surrey Research Park at Guildford, to the south-west of
London) would normally be considered relatively quiet, but
the presence of several railway lines and a trunk road, all
within a couple of kilometres from the site, meant that care
had to be taken when recording data, and much had to be
rejected on the basis of observed background noise. The A1
North Circular Road site was urban, located on a trunk road,
and data were therefore taken at the quietest time, in the early
hours of the morning.
Despite these difficulties, and the fact that the various
sites are located tens of kilometres from each other, there is
a remarkable consistency between the datasets for vertical
shear modulus (Gv ), with only a few measurements falling
outside the shaded area (at the Heathrow site, the shallowest
are probably due to a layer of gravel at ground surface). In
the early 1990s we observed that the stiffness parameters we
were obtaining from geophysics were similar to those that
we had been using in numerical modelling, based on backanalysis of excavations in the London Clay. Fig. 28 therefore
takes the data from these surveys and compares them with
data from back-analysis, and with estimates of Youngs
modulus (E) based on the results from routine laboratory
tests. As might be expected from the early work of Ward et
al. (1959), the stiffnesses back-analysed from measurements
of foundation and retaining wall movements are much
greater (by about an order of magnitude) than the stiffnesses

Heathrow T5 (Hight et al., 2007)

Fig. 27. Vertical shear moduli (Gv ) against depth, from four
cross-hole seismic surveys in the London Clay around London

obtained from routine laboratory testing (in this case the


oedometer testing). Even the enhanced values, using Butlers
(1975) proposed correlation with undrained shear strength,
are four or five times too low. Thus, even though the verysmall-strain stiffness values obtained from seismic geophysics overpredict the back-figured results, they are relatively
close to them, and at the very least provide a benchmark
against which to assess the stiffnesses provided by other
methods of measurement.
Two difficulties potentially arise with the interpretation of
Undrained Youngs modulus, Eu or E uv: MPa
0

100

200

300

400

500

600

Constrained modulus from oedometer


Eu 220.su (triaxial) (Butler, 1975)

Depth below ground level: m

preventing repeat visits to site and restricting the


method to be used, would probably have failed.
(d ) The value of the seismic dataset is illustrated by the
number of data points obtained, the tight grouping of
the seismic dataset (as compared with the pressuremeter
and dilatometer datasets), and the agreement between
two different methods of obtaining stiffness from
seismic data.

10

20

30
Back-analysis of case records

Cross-hole geophysics

Fig. 28. Youngs moduli against depth for the London Clay,
from cross-hole geophysics (assuming isotropy and v
0.5; see
shaded area in Fig. 27), back-analysis of case records, and
routine laboratory testing at Grand Buildings, Trafalgar Square
(modified from Clayton et al., 1991)

Delivered by ICEVirtualLibrary.com to:


IP: 152.78.201.162
On: Mon, 14 Feb 2011 15:36:53

CLAYTON

22
0

100

Shear modulus, G0: MPa


200
300

400

500

0
Gravel

Down-hole Gv
Cross-hole Gv
Cross-hole Gh

Shear wave velocity, Vs: m/s


1000

2000

10

20

20

30
30

Depth: m

Depth below GL: m

10

40

40

50

Fig. 29. Cross-hole data from London Clay Formation, using


Bison (Vsv ) and BRE (Vsh ) shear wave hammers (Butcher,
personal communication)

50

60

cross-hole seismic data in terms of stiffness. First, there is


the issue of noise, which even though a reasonable dataset
may be obtained can still affect results. Fig. 29 shows the
results of down-hole and cross-hole testing carried out in the
London Clay using three sources. A surface source and a
Bison vertically polarised down-hole shear wave hammer
(see Clayton et al., 1995b) were used to determine values of
Vsv and hence Gv from down-hole and cross-hole surveys.
The BRE horizontally polarised shear wave hammer was
used to determine Vsh (and hence Gh ). Two sets of data were
collected: that is, there were two down-hole surveys and four
cross-hole surveys, each in different boreholes. Three points
can be made from Fig. 29.
(a) Below the gravel and above about 30 m depth the Gv
data are tightly grouped, and the down-hole and
vertically polarised cross-hole surveys yield similar
values of shear modulus.
(b) In general, values of horizontal shear modulus (Gh ) are
higher than those of Gv , indicating significant stiffness
anisotropy. The dashed line in Fig. 29 is not intended to
represent the trend of Gh with depth; it has been drawn
at a stiffness of twice the solid line, which has been
used to represent the trend of Gv data with depth. Fig.
11 has shown the horizontal shear modulus in the
London Clay at Heathrow Terminal 5, assessed from
laboratory tests, to be of the order of twice the vertical
shear modulus.
(c) The increased scatter in Gh , compared with the scatter
in Gv values, probably results from the lower energy
input available from horizontally polarised shear wave
hammers. As will be seen later, noise tends to lead to
longer estimated travel times, and therefore lower
interpreted stiffnesses. Thus some values of Gh may
be underestimates.
In the example above, down-hole and cross-hole surveys
gave similar estimates of Gv , in the London Clay Formation.
Although this is theoretically true for a transversely isotropic
elastic medium, it not always observed in practice. A comparison of down-hole and cross-hole determined seismic
velocities is shown in Fig. 30 (Pinches & Thompson, 1990).
It can be seen that the down-hole values are almost always
lower than those determined from cross-hole testing. This is
because the measured down-hole travel time results from
averaging of the velocity of layered strata, which in this case
have significantly different stiffnesses. In contrast, in a
cross-hole survey the first arrival time results from energy
travelling through the stiffest layers, provided that these are

Limestone band

Siltstone and gypsum

Mudstone

Down-hole
measurements

Siltstone/limestone
Fissile mudstone

Cross-hole
measurements

Fig. 30. Comparison of down-hole and cross-hole seismic


velocities in layered ground (Pinches & Thompson, 1990)

of sufficient thickness to act as wave guides for the energy.


In a cross-hole survey in layered ground, measured stiffness
is likely to represent the stiffest layers, rather than the
average stiffness.
Laboratory testing methods
Laboratory testing plays a vital role in determining the
stiffness of geomaterials, but as already noted can suffer
from various disadvantages.
(a) It must be possible to sample and prepare specimens of
a representative volume of soil. This will not be
feasible if, for example, the stiffness of the ground is
controlled by widely spaced discontinuities, or if it
contains very coarse material, such as cobbles and
boulders.
(b) The specimens must, as far as practical, be undisturbed.
Even the most undisturbed samples will have undergone some change in both deviatoric and mean
effective stress, which need to be compensated for in
some way.
(c) Advanced laboratory testing may take many weeks or
months, and requires sophisticated apparatus used by
technical staff trained and experienced in its use.
While much advanced testing is carried out under quasistatic loading, the potential use of laboratory dynamic testing, such as the resonant column apparatus (RCA), bender
elements, and cyclic triaxial testing to measure static stiffness, has also been recognised for some time. Some promising techniques are discussed below.

Delivered by ICEVirtualLibrary.com to:


IP: 152.78.201.162
On: Mon, 14 Feb 2011 15:36:53

STIFFNESS AT SMALL STRAIN: RESEARCH AND PRACTICE


Bender element testing. Bender elements, which appear to
have been introduced to the UK by the Bangor marine
geophysics group (Schultheiss, 1981), were first used in the
USA by Shirley in the late 1970s (Shirley, 1978; Shirley &
Hampton, 1978). The bender element test has become
increasingly popular over the past decade or so, thanks to
its perceived simplicity. This is something of an illusion,
however, as the increasing number of publications relating to
new variants, parallel testing and standardisation perhaps
suggest.
The piezoelectric crystals initially used to generate shear
waves in rock specimens were very stiff compared with soil,
which resulted in a large mismatch in characteristic impedance between the specimen and the source (Schultheiss,
1982; Thomann & Hryciw, 1990). This problem was reduced
through the use of more flexible piezo-ceramic bender
elements, sometimes referred to as piezo-ceramic bimorphs.
The bender element consists of two thin piezo-ceramic plates
bonded rigidly together, with a conductor between them and
on their outer surfaces (Fig. 31(a)). Application of a voltage
causes the plates to extend or contract, and the bimorph to
try to bend, generating seismic waves in the soil in which it
is embedded. Distorting a bender element generates charge,
allowing a similar device (if wired appropriately) to detect
incoming waves. Determination of the travel distance and
travel times allow the calculation of wave velocity and, from
density and orientation, the relevant shear modulus (Gv or
Gh ).
In principle all that is required for a bender element
determination of stiffness is a set of (normally two) bi-

Flexure with applied voltage

23

morphs, a signal generator and a (storage) oscilloscope.


Because they are compact, bender elements have been
installed in oedometers (Schultheiss, 1981; Dyvik & Olsen,
1989; Thomann & Hryciw, 1990), in direct simple shear
apparatus (Dyvik & Olsen, 1989), triaxial specimens
(Schultheiss, 1981; Bates, 1989), inside the resonant column
apparatus (Bennell et al., 1984; Dyvik & Madshus, 1985;
Ferreira et al., 2006), and indeed on unconfined samples
immediately after recovery from boreholes (Hight, 1998;
Hight et al., 2003; Landon et al., 2007). Various configurations have been used (Schultheiss, 1981; Bates, 1989; Viggiani & Atkinson, 1995; Pennington et al., 1997; Clayton et
al., 2004), the more common of which are shown in Fig.
31(b).
Considerable accuracy (for example 5% of G0 and 2%
of Vs; e.g. Hight et al., 1997; Pennington et al., 1997) has
been claimed for stiffness measurements made using bender
elements, but while it is true that their installation into
standard geotechnical testing apparatus is relatively straightforward, and that with modern data logging and computer
power, high-speed acquisition and processing is not a problem, it has become clear that interpretation is not easy
(Ferreira & da Fonseca, 2005), or necessarily repeatable.
Figure 32 shows the input traces and received data for a
single bender element test carried out on a specimen of
natural clay, using the conventional top cap and base pedestal triaxial mounting shown in Fig. 31(b), with a GDS
Instruments bender element drive and data acquisition system. In order to be able to assess near-field effects the
period of the single input pulse was varied, equivalent to
input frequencies of 10 kHz (the maximum the system could
provide) down to 2.5 kHz. Several features can be seen in
Fig. 32.
(a) Although the traces are relatively free from noise, the
difficulty of picking a first break and therefore the
travel time is clear. In this exercise, first breaks were
picked at the first significant rise (i.e. movement in the
same direction as the input) in the trace, as approximately indicated in the figure.
(b) Noise levels appear to increase at lower frequencies.
(c) Despite the significant (fourfold) change in the period
of the input signal, the period of the received signals
does not alter greatly.
(d ) The periods of the input and received signals are
closest at the highest frequency (10 kHz).

(a)

Transmitters
Transmitted signals

100 kHz
75 kHz

First
breaks

100 kHz

Amplitude

Receivers

50 kHz 25 kHz

75 kHz
50 kHz
25 kHz

Gv0

Received signals stiff clay

Gv0, Gh0
(b)

Fig. 31. Bender element configurations: (a) sketch of bender


element (Shirley, 1978; Schultheiss, 1983; Dyvik & Madshus,
1985); (b) layouts within a triaxial test (Schultheiss, 1981; Bates,
1989; Viggiani & Atkinson, 1995; Pennington et al., 1997)

01

02

03

04

05

Time: ms

Fig. 32. Example input and received waveforms at different


frequencies in a bender element test on undisturbed stiff clay

Delivered by ICEVirtualLibrary.com to:


IP: 152.78.201.162
On: Mon, 14 Feb 2011 15:36:53

CLAYTON

(a) For this case, perhaps as a result of noise and/or nearfield effects (Clayton et al., 2004, and see below), the
scatter in estimated shear wave velocities is considerably higher at low input frequencies than at high input
frequencies.
(b) For the small data sample shown (a larger number of
participants might be expected to give a greater range),
the variation in shear modulus (Gv ) calculated from
first-break travel times is 5.4% at 10 kHz, rising to
15.4% at 2.5 kHz.
(c) The data show that it is easier to pick consistent peakto-peak travel times than first break travel times. There
is considerable consistency in the shear wave velocities
estimated using the peak-to-peak method.
(d ) However, the results obtained from using peak-to-peak
travel times are not consistent with those from firstbreak determinations until the received period is
approximately the same as the transmitted period.
These data suggest that previous estimates of the accuracy
of bender element determinations of velocity and stiffness
have been optimistic. They also suggest that, while carrying
out commercial bender element tests, there is merit in
(a) systematically using a range of input frequencies for
every test
(b) having more than one person interpret the traces, in the
same environment (e.g. in a spreadsheet)
(c) estimating shear-wave velocity and therefore stiffness
on the basis of both first-break and peak-to-peak travel
times.
The important issue of noise is further explored in the
data shown in Fig. 34. Five noisy bender element traces
taken on the same specimen are divided into those with a
(relatively) good signal-to-noise ratio (above) and those with
a poor signal-to-noise ratio (below). The ratio (S/N)a shown
in the figure is not the conventional value (signal power
divided by noise power), but is the ratio of peak-to-peak
noise prior to the first break, divided by the maximum peakto-peak amplitude in the received wave train. This value is
more easily estimated in the laboratory during testing. The
estimated position of the first breaks is shown by the arrows.
The arrival of the seismic wave is detected significantly later
when there is more noise. A survey of about 200 traces from

0001

Amplitude: V

The received data in Fig. 32 were examined independently


by four engineers, with instructions to estimate the first
break and peak-to-peak travel times. The shear wave velocities calculated from these results are shown in Fig. 33.
The following can be seen.

(S/N)a 122, 132, 137

First
breaks
0

0001

Time

0001
First
breaks?

Amplitude: V

24

(S/N)a 33, 40

0001

Time

Fig. 34. Effect of signal-to-noise ratio on picked arrival time:


noisy bender element data

tests on stiff clays in the London area, carried out by the


author, has indicated that bender element tests with a (S/N)a
ratio of less than 10 should be considered unacceptable.
Leong et al. (2005) suggest a conventional receiver S/N ratio
of at least 4dB.
Reporting on the results of round robin testing carried out
under the auspices of ISSMGE Technical Committee TC29,
Yamashita et al. (2009) give comparisons of shear modulus
(Gv ) values obtained in different laboratories across the
world on loose and dense pluviated rotund uniformly graded
Toyoura sand, and compare these with the results of resonant
column tests by Iwasaki & Tatsuoka (1977). Fig. 35 shows
an extract from their results, and indicates a large scatter in
estimated shear modulus. Helpfully, however, their results
suggest that the simplest method of bender element testing,
using first-break or peak-to-peak travel times derived from
the propagation of a single sine pulse input waveform,
provides reasonably consistent results.
The wide range of issues that has been identified in the
manufacture and use of bender elements is summarised in
Table 3. At this stage, from a practitioner viewpoint, it
would seem sensible to use simple test and interpretation
techniques, and regard the bender element test as semiempirical, requiring validation in each new material and test
arrangement, through comparison with results from other

320
Four independent peak-to-peak picks
400
Resonant column tests
(Iwasaki & Tatsuoka, 1977)

First break - 1
First break - 2
First break - 3
First break - 4

280

Shear modulus, Gv0: MPa

Shear wave velocity: m/s

300

260
240
220
200
0

4
6
8
Source frequency: kHz

10

12

Fig. 33. Effect of subjectivity in picking first breaks and first


peaks, using the data presented in Fig. 32

200 kPa
Circles: saturated
Triangles: dry

200

100

50
First breaks
Peak-to-peak
25
06

07

Frequency domain and


cross-correlation
Void ratio

08

09

Fig. 35. TC29 international parallel bender element tests on


Toyoura sand (redrawn from Yamashita et al., 2009)

Delivered by ICEVirtualLibrary.com to:


IP: 152.78.201.162
On: Mon, 14 Feb 2011 15:36:53

STIFFNESS AT SMALL STRAIN: RESEARCH AND PRACTICE

25

Table 3. Variants in bender element testing


Issue

Brief description

Effect

Suggested standard

Most laboratories adopt 10 mm width, 0.5


1.0 mm thickness, and a protrusion of ,5 mm
(Yamashita et al., 2009). Shorter protrusion (c.
1 mm) is favoured in the UK when testing stiff
natural materials, where otherwise a slot needs to
be cut and the bender filled around during
installation. Fixing epoxy is set back inside end
caps, and the space filled with silicone rubber
compound, to allow flexing of the bender
element.
Experience suggests that the wiring of benders as
Using different (series and parallel)
Bender
Transmitter and receiver
bimorphs as transmitter and receiver (Dyvik S-wave transmitters, P-wave transmitters and Sconstruction
bender configurations,
wave receivers gives advantages. Self& Madshus, 1985) improves shear wave
extender elements, and selfgeneration and reception. Extender elements monitoring bender elements may be useful, as
monitoring configurations.
they allow the detection of time lags and phase
(Lings & Greening, 2001) allow P-wave
changes between driving voltage and actual
generation. Use of a self-monitoring
movement of the transmitter, resulting from
configuration, as per Schultheiss (1982),
characteristic impedance mismatch.
allows the movement of the bimorph (as
distinct from its driving voltage) to be
determined.
Mounting across or along specimens allows The addition of side-mounted bender elements is
Bender elements may be
Location of
relatively easy to achieve using grommets and Omounted in the rigid base and shorter travel distances (giving reduced
bender
rings similar to those for mid-plane pore water
attenuation and higher S/N ratio), and
top caps of cells (e.g.
elements
pressure measurement (Sodha, 1974).
Schultheiss, 1981), or through determination of the velocity of waves
polarised in different directions. Very-smallthe triaxial membrane
strain anisotropy of stiffness can then be
(Pennington et al., 1997;
inferred from shear wave velocities.
Clayton et al., 2004).
Travel distance Small tip-to-tip travel distance Closely spaced benders display more scatter Calculated shear wave velocities may be affected
affects travel time resolution. in calculated shear wave velocity (Yamashita if large bender element penetrations are used in
conjunction with small specimens (e.g. vertically
et al., 2009). Data from widely spaced
Large distance increases
benders are more noisy. The estimated travel in oedometers or DSS specimens). Higher
attenuation.
transmitter frequencies will be needed to keep the
time may be affected.
wavelength down (see Input frequency, below).
Square wave pulses contain a broad spectrum Practice suggests single sine pulses produce
Input wave
Square (Schultheiss, 1981,
acceptable traces, giving repeatable first break or
form
1982, 1983), continuous sine of frequencies. Low frequencies place the
peak-to-peak travel times, and permitting a more
or pulsed sine waves (de Alba receiver in the near field.
restricted and controllable input frequency, and
& Baldwin, 1991).
therefore wavelength: see below (Thomann &
Hryciw, 1990).
It is suggested that extender elements (Lings &
Input wave
P or S wave.
In soft saturated soil P waves travel at c.
Greening, 2001) should be routinely used to warn
mode
14501550 m/s (the P-wave velocity of
of misinterpretation.
water), much faster than S waves. In
unsaturated soils P waves travel only approx.
50% faster, and can obscure or be confused
with S waves.
The receiver should be at least 23 wavelengths
Input frequency Low source frequencies
Long wavelengths place the receiver in the
from the transmitter (Sanchez-Salinero et al.,
produce long wavelengths.
near field, affecting the received waveform
1986; de Alba & Baldwin, 1991; Leong et al.,
and picked travel time.
2005). It is suggested that, for routine triaxial
testing, results are returned for a range of
frequencies from about 2.5 kHz to 12.5 kHz).
Picked arrival times become uncertain.
In order that accuracy is not degraded by
Data
Low voltage and temporal
resolution. the sampling time interval should be
acquisition
resolution reduce quality of
less than 1/100th of the travel time between
captured traces.
transmitter and receiver. Voltage resolution
should be better than 1/100th of the amplitude of
the received signal (Yamashita et al., 2009).
A study of some relatively noisy UK data
Signal-to-noise Noise affects low-amplitude
Picking of first arrival or peak times is
suggests a minimum amplitude signal/noise ratio
ratio
received signals.
subjective. A low signal-to-noise ratio
(S/N)a of 10. Leong et al. (2005) suggest a
increases scatter and tends to increase the
estimated travel time.
receiver S/N ratio of at least 4 dB.
First-break and first-peak-to-first-peak travel
Different processing methods lead to
First break, peak-to-peak,
Method of
time detection are favoured by Yamashita et al.
different travel times. With mixed
cross-correlation, or phase/
determining
(2009), who found significant differences in
frequencies pulse broadening may occur,
frequency relationship.
travel time
owing to attenuation of the higher-frequency some cases where cross-correlation and phase/
component, leading to increases in measured frequency methods were used (see also Ferreira
& da Fonseca, 2005). It is suggested that results
travel time when using peak-to-peak
from both the first break and the peak-to-peak
detection. Receiver first breaks may be
methods be routinely reported.
hidden by noise, leading to increases in
measured travel time when using this
method.
Bender
dimensions

Width and protrusion into the


specimen

Travel distance is generally accepted as


bender tip-to-tip.
Reduced travel distance causes reduced
accuracy.
Installation of highly protruding bender
elements is more difficult in specimens of
stiff/hard natural sediment.

Delivered by ICEVirtualLibrary.com to:


IP: 152.78.201.162
On: Mon, 14 Feb 2011 15:36:53

26

CLAYTON

types of test (e.g. resonant column and field geophysical


tests, as by Dyvik & Madshus, 1985; Davis & Bennell,
1986, Bennell & Taylor Smith, 1991), other laboratories
determinations on the same material, and in-laboratory comparisons with previous tests on similar materials. The tests
value lies in its simplicity, its relatively low cost and its
potential for determining anisotropy of shear modulus.

Pressure
vessel

LVDT

Counter
weight

Accelerometer
Magnet

Coil

(a) Set up specimen within the apparatus. An estimate of


the expected very-small-strain shear modulus will be
useful in judging apparatus compliance, which may
need correction, and specimen slippage effects, which
may be avoidable by cementing the specimen to the
platens or, for weaker materials, by using vanes
protruding into the specimen (Drnevich, 1978; Clayton
et al. 2009b).
(b) Re-establish the in situ effective stress(es) on the
specimen, by applying suitable cell and (elevated) back-

Coil
Top cap

Cross
arm

Specimen

Resonant column testing. The resonant column apparatus,


which has been successfully used for more than 40 years in
the field of soil dynamics, provides a method of determining
the shear modulus (G) and Youngs modulus (E or Eflex ) of
soils and weak rocks at very small strain levels, and of
obtaining estimates of the rate of stiffness degradation with
increasing strain. Different apparatus configurations allow
vibration of a soil specimen in torsion (Hardin & Music,
1963; Stokoe et al., 1980; Menq & Stokoe, 2003), in flexure
(Cascante et al., 1998) and axially (Drnevich, 1972).
Torsional testing appears not to suffer from bedding effects
(although compliance is an issue, as will be discussed below),
and is therefore to be preferred for stiffness determinations.
Figure 36(a) shows a simple schematic diagram of a
Stokoe resonant column apparatus. An electromagnet drive
head, to which four magnets are attached, is bolted to the
specimen top cap. Torsion can be applied by running current
through the four coils in which the magnets sit, which are
held in place by a substantial support frame. Flexure can be
applied by running current through two diametrically opposed coils. At the start of a test a relatively low sinusoidal
drive voltage is applied, and a frequency sweep is carried
out. As the frequency is increased, the amplitude of the
vibrations, measured by an accelerometer mounted on the
drive head, increases up to a peak, and then decays. This is
shown by the lowest curve in Fig. 37. The peak amplitude,
which occurs at low levels of damping at the resonant
frequency, is recorded. Given the mass polar moment of
inertia of the drive head and top platen, the specimen mass
and its dimensions, and assuming linear elasticity, the shear
modulus (Gv ) of the soil can be calculated.
The process is then repeated with a higher applied voltage. Measured amplitude, and therefore strain, increases. At
first, at the lowest strain levels, the peak frequency is
unaffected by the increasing voltage, but as shear strain
increases shear stiffness decreases, and the peak frequency
of the system drops (Fig. 37). The shear modulus at very
small strain (G0 ), and a curve of shear modulus against
shear strain (stiffness degradation), can be obtained from the
results.
Resonant column testing is the subject of ASTM standard
D 4015-07 (ASTM, 2007), which provides generic guidance
for the calibration and operation of a range of resonant
column devices to determine both stiffness and damping, but
does not attempt to provide engineering guidelines on the
appropriate use of the apparatus, nor of problems that may
be encountered. This can be found elsewhere in the literature
(e.g. Bennell et al., 1984; Bennell & Taylor Smith, 1991). A
suitable test procedure might be described as follows.

Pedestal

Cell base

(a)

(b)

Fig. 36. (a) Schematic drawing and (b) photograph of Stokoe


resonant column apparatus

pressures, and allow drainage while monitoring volume


change and specimen height.
(c) Immediately after re-establishment of in situ stresses,
measure the very-small-strain shear modulus (Gv ) (and
normally damping) of the specimen, and monitor it at
regular intervals on a logarithmic scale (for example at
approximately 1, 2, 4, 8, 16, 32 min, 1, 2, 4, 8, 16,
32 h, etc. after the start of this stage). Low-amplitude

Delivered by ICEVirtualLibrary.com to:


IP: 152.78.201.162
On: Mon, 14 Feb 2011 15:36:53

STIFFNESS AT SMALL STRAIN: RESEARCH AND PRACTICE

01

001
125

130

135
Drive frequency: Hz

140

145

Fig. 37. Principle of operation of a resonant column apparatus;


results of a Stokoe resonant column test on dense Leighton
Buzzard sand

(d )

(e)

(f)
(g)

27

(2009b) report the results of extensive numerical, analytical


and experimental modelling, driven by the need to test stiffer
materials (Bennell & Taylor Smith, 1991). They provide
recommendations relating to

Increasing drive voltage

torsional frequency sweeps (each lasting about one


minute) should be carried out under drained conditions
until primary consolidation is complete and the creep
rate is well defined (Anderson & Stokoe, 1978),
perhaps for 10002000 min (12 days). The reference
shear strain level used for these measurements should
be as low as practical (e.g. taking into account
limitations imposed by instrument sensitivity and
background vibration) and should be non-destructive,
with a shear strain amplitude , 0.0001%.
Determine the relationship between shear modulus Gv ,
damping and shear strain. With the drainage valves
open, carry out an initial shear modulus (and damping)
measurement at the reference strain level. Then close
the drainage valves.
Test at a higher (say three times) strain level. Open the
drainage valves, allow a rest period of about 1 min,
then repeat the stiffness (and damping) measurements at
the reference strain level, under drained conditions.
Repeat (e) with progressively higher strain levels, for
example approximately 0.00001%, 0.00003%, 0.0001%
0.0003%, 0.001%, 0.003%, 0.01%, 0.03%.
Cease testing when the shear modulus determined at
the reference strain level does not return to its initial
value, since this indicates some level of specimen
destructuring as a result of the torsional strains that
have been imposed.

Tests can be carried out under isotropic effective stress


conditions in the Stokoe apparatus, or if required under
anisotropic effective stress conditions, in the Hardin apparatus. The advantages of the Stokoe apparatus are that it is a
relatively simple piece of equipment, and it can apply high
levels of torque, thus exploring a greater torsional strain
range.
The results for Eocene sandy clay shown in Fig. 5, for a
number of specimens tested under various degrees of stress
anisotropy, suggest that, apart from its impact on mean
effective stress, deviatoric stress has little effect on verysmall-strain shear modulus Gv , and that for practical
purposes this may be determined under isotropic stress
conditions. This view seems to be supported by bender
element data obtained by Gasparre et al. (2007) on London
Clay, during reconsolidation to anisotropic stress states, and
by the results of Yamashita & Suzuki (1999) from torsional
tests on pluviated Toyoura sand.
For the Stokoe resonant column device Clayton et al.

(a) the repeated use of (appropriately designed) calibration


bars before and after testing, to check for loosening of
apparatus components
(b) the use of calibration bars with different stiffnesses to
explore and correct for apparatus compliance, when
testing stiffer materials
(c) limits on frictional fixity between stiff (e.g. cemented)
specimens and the specimen platens (see also Drnevich,
1978)
(d ) the base of the apparatus, which needs to have a high
mass polar moment of inertia relative to that of the
drive head.
Figure 38 shows the results of some resonant column
determinations of very-small-strain stiffness on gypsiferous
mudrocks from Dubai. Note that the shear modulus values
are plotted on a logarithmic scale; there is (roughly) a one
order of magnitude difference between the stiffness determined from the pressuremeter initial loading and that from
the cross-hole seismic testing. Because of sample disturbance and apparatus compliance the resonant column determinations of Gv fall between these values. Note the very
high stiffness of the ground, and that the ratio of corrected
to uncorrected shear modulus for the apparatus used in these
tests is greater than 3 when the measured shear modulus
rises to 2 GPa. For this particular apparatus design it would
appear that tests on material with a shear modulus greater
than about 500 MPa would be significantly affected by
apparatus compliance. In tests on methane hydrate bearing
Shear modulus, Gv: GPa
1

01
0

Depth: m

Output amplitude: V

10

10

100

200
Initial loading pressuremeter shear modulus
Cross-hole shear modulus
Resonant column data, uncorrected
Resonant column data, corrected for apparatus compliance

Fig. 38. Comparison of resonant column very-small-strain stiffness measurements with those from pressuremeter and crosshole seismic testing; gypsiferous mudrocks with saline pore
water, Nakheel Tower, Dubai

Delivered by ICEVirtualLibrary.com to:


IP: 152.78.201.162
On: Mon, 14 Feb 2011 15:36:53

28

CLAYTON

sediments, for which shear modulus can rise to about 5 GPa


(Clayton et al., 2005), a Stokoe resonant column apparatus
with an enhanced cross-arm stiffness has been used.
Triaxial testing. A much improved appreciation of the effects
of disturbance and potential destructuring during drilling,
sampling, specimen preparation and the early stages of
laboratory testing (Hight, 1986; Baligh et al., 1987; Chandler
et al., 1992; Clayton et al., 1992; Hight & Jardine, 1993;
Clayton et al., 1998; Clayton & Siddique, 1999), and the
development and use of higher-quality drilling techniques and
improved samplers (Lefebvre & Poulin, 1979; La Rochelle et
al., 1981; Scarrow & Gosling, 1986; Harrison, 1991; Tanaka
& Tanaka, 1999; Tan et al., 2002), have led to a significant
increase in the availability of high-quality soil stiffness data.
The development of triaxial testing has been facilitated by
the growth in high-quality triaxial instrumentation over the
past 3040 years. Fig. 39 shows some examples.
(a) The introduction of high-stiffness water pressure sensors
(Fig. 39(a)) that could be mounted on de-airing blocks
close to the cell base allowed laboratories not only to
move away from the mercury null indicators described
by Bishop & Henkel (1962) but also to improve
response times dramatically (Whitman et al., 1961).
(b) Pore pressure measurements were then further improved
by the introduction of mid-plane pressure-measuring
systems. Fig. 39(b) shows the components for a midplane probe (Hight, 1982) using a submersible miniature pore pressure transducer. Experience suggests that
this device is more applicable in the testing of softer
sediments, where the initial suctions are relatively low,
and will not lead to cavitation behind the ceramic
during specimen set-up (Bishop & Henkel, 1962,
Appendix 6). For stiffer sediments (such as, for
example, the London Clay) the flushable probe shown
in Fig. 39(c) (Sodha, 1974) is preferred, since the gas
bubbles that form behind the high-air-entry ceramic as
a result of cavitation and dissolved-air ex-solution
during specimen set-up can be removed to re-establish
the stiffness of the pore pressure measuring system.
(c) The introduction of submersible load cells (Fig. 39(d)),
which should be mounted inside the triaxial cell
between the specimen top cap and the ram, has
removed the need to use rotating bushes, while
avoiding the unpredictable effects of ram friction. This
has become more important with the introduction of O
rings within the ram bushings in many cells.
(d ) The use of local strain measurement devices, such as
linear variable differential transformers (LVDTs) (Fig.
39(e)) or Hall effect sensors (Fig. 39(f)). These are
typically mounted on the mid-third of the specimen
height (and in Fig. 39(e) on a radial caliper).
(e) The use of electronic sensors and computer based data
acquisition has allowed near-continuous records of
force, pressure and displacement to be made, which is
particularly important in the early stages of triaxial
small-strain stiffness determination.
( f ) Finally, the introduction of bender elements (Fig. 39(g))
has allowed independent determination of G0 (see
earlier) during stiffness testing for, for example,
Youngs modulus.
Broadly, there are two classes of triaxial test currently
carried out commercially in the UK: routine tests and advanced tests.
Routine tests have been standardised in BS 1377 part 8
(BSI, 1990), and in ASTM standards (for example ASTM
D4767; ASTM, 2004). Standards for the more complex

cyclic triaxial test exist in the USA (ASTM D3999; ASTM,


2003) and Japan (Toki et al., 1995), but not in the UK. The
text below (see also Fig. 40) suggests how, practically, some
of the disadvantages of the standard triaxial test configuration may be overcome in commercial testing.
Load cell and other apparatus compliance. The stiffness of
the load cell or proving ring used to measure deviatoric load,
or the loading frame itself, is not generally an issue when
testing materials with low stiffnesses, but as has been seen in
the context of the resonant column apparatus, apparatus
compliance can be very significant at the higher stiffnesses
often associated with structured soils and weak rocks when
tested at small strains. For the triaxial apparatus some
researchers initially suggested reducing these effects through
separate measurement of compliance (e.g. Atkinson & Evans,
1985), but it seems now to be accepted that in practice it is
more straightforward, given the need also to remove bedding
effects (see below), to routinely use local strain measurement.
Ram misalignment. A large number of top cap arrangements
exist (e.g. Atkinson & Evans, 1985; Kuwano et al., 2001). The
arrangement that is used is known to affect the measured
stressstrain behaviour of soils (e.g. Jardine et al., 1985; Baldi
et al., 1988). However well trimmed and aligned a specimen
may be at the time of set-up in the triaxial apparatus, experience
suggests that misalignment of the specimen and top cap will
occur, for example during initial application of cell pressure, as
a result of fissure closure and bedding. The use of a rounded
ram end engaging with a dimpled top cap, specified in the
British Standard, then inevitably leads to sideways movement
and slippage at the start of a test, which can affect even local
measurements of strain. Fig. 40 suggests a practical method of
overcoming this, when carrying out the simplest of advanced
triaxial tests, that is, loading only in triaxial compression. The
dimple in the top cap is removed, and replaced with a flush
stainless steel insert.
Air trapped around top cap and in top cap ducts. It is
essential to ensure, during set-up, that air is removed not only
from between the membrane, filter drains (if used) and
specimen, but also from around the specimen top cap.
Experience suggests that this is made more difficult if lowair-entry porous stones and drainage ducts and leads are used
at this location. For saturated soils, failure to remove air from
blanked-off ducts, or from small voids between the edges of
the porous stones, the specimen and the pedestal or top cap,
causes the system to behave as if the specimen were
unsaturated, given a slower pore pressure response, and a
potential underestimate of pore pressure change. The
arrangement in Fig. 40 removes much of the opportunity to
trap air. Little is lost by removing top cap drainage provided
appropriate side drains can be used, as the theoretical studies
of Bishop & Gibson (1963) showed. The amount of air
trapped at the specimen base can be significantly reduced
through the use of a base pedestal with a flush-mounted highair-entry ceramic (Bishop & Henkel, 1962).
Bedding between soil, porous stones and platens. The issues
of compliance, discussed above, and bedding became
generally recognised in the late 1970s and early 1980s (e.g.
Brown & Snaith, 1974; Costa-Filho & Vaughan, 1980;
Daramola, 1980). In response various devices were developed
that could be mounted on the sides of triaxial specimens to
measure displacement locally (e.g. Jardine et al., 1984;

Delivered by ICEVirtualLibrary.com to:


IP: 152.78.201.162
On: Mon, 14 Feb 2011 15:36:53

STIFFNESS AT SMALL STRAIN: RESEARCH AND PRACTICE

(a)

(b)

29

(c)

(d)

(e)

(g)
(f)

Fig. 39. Examples of triaxial instrumentation: (a) external pore pressure transducer; (b) pore pressure transducer mid-plane probe;
(c) flushable mid-plane pore pressure probe; (d) internal load cell; (e) LVDT local strain measurement; (f) Hall effect local strain
measurement; (g) side-mounted bender element

Clayton & Khatrush, 1986; Goto & Tatsuoka, 1986; Ackerley


et al., 1987). In addition, standard instrumentation such as
submersible proximity sensors (Hird & Yung, 1989) and
LVDTs came into use. Most research laboratories now prefer
the high resolution (about 1 microstrain) that can be achieved
using LVDTs (Cuccovillo & Coop, 1997), but the stiff
electrical leads associated with their waterproofed cabling
cause concern regarding conformance. The ease of use of
Hall effect gauges (Clayton & Khatrush, 1986; Fig. 39(e))
makes them attractive to many commercial laboratories,
despite their lower resolution.
De-saturation of porous stones. Most fine-grained heavily
overconsolidated and deep samples arrive in the laboratory

with significant suctions. When a specimen is placed on


traditional coarse low-air-entry porous stones during set-up, it
immediately starts to imbibe water. Water can be quickly
released by the stones because of their high permeability, and
their low air-entry value means that, as water moves into the
specimen, it can easily be replaced by air. The use of flushmounted high-air-entry (say 250 kPa) ceramics, fixed into the
base (and top cap, if required) with epoxy resin slows this
process, as well as reducing the opportunity for air to be
trapped during set-up.
Pore pressure equalisation issues. In order to interpret
triaxial test data it is, of course, necessary that the pore
pressures in the middle of the specimen are reasonably

Delivered by ICEVirtualLibrary.com to:


IP: 152.78.201.162
On: Mon, 14 Feb 2011 15:36:53

CLAYTON

30
Stainless steel
insert

No top cap
drainage or
stone

Local axial strain


gauges

Flushable high air


entry mid-plane
p.w.p. probe

t (1 3)/2

Triaxial
compression

Bender element
(one of a pair)
1

Flush-mounted
high air entry
flushable porous
stone

s (1 3)/2
4, 5
3

Triaxial
extension

Radial strain
caliper

6
(a)

(a) establishment of pore pressures in the measuring system


(b) consolidation or swelling to estimated in situ mean
effective stress
(c) application of approximate estimated in situ deviatoric
stress, allowing for
(d ) a stress path to model recent stress history, if required,
and
(e) a rest period, to reduce creep to an acceptable level
( f ) undrained monotonic shearing to failure, in either
triaxial compression or extension.
In the UK the standard test procedure (BSI, 1990; 1999)
used for routine effective stress triaxial strength testing
involves three stages: saturation, consolidation and monotonic shearing to failure in triaxial compression. The saturation referred to in the standard, although not clearly worded,
is that of the system (the specimen, plus voids between
specimen and membrane and porous stones, and in base
pedestal and top cap ducts), rather than of the specimen
itself. Two methods are described: back-pressure saturation,
and saturation at constant water content, the latter using
application of cell pressure alone.
With notable exceptions, for example compacted fill and
coarse soils above the water table, many materials will have

Mimic recent
history?

Rest period to
reduce creep

Apply
deviatoric
stress

Establish
ve pwp

Isotropic
consolidation

uniform, and known. Routine triaxial tests measure pore


pressure by connecting a pressure transducer, mounted in a
de-airing block, to the cell base. The pore pressure at the
base of the specimen will not necessarily be the same as at
the mid-height of the specimen, for example during
consolidation or shearing stages. In conventional testing,
undrained loading to failure must be run sufficiently slowly to
ensure that sufficient equalisation of pore pressure occurs
between the mid-height of the specimen and the base by the
time that failure is reached. Drained shear stages, which for
reasons of economics are generally used only for coarser,
more free-draining materials, must be run sufficiently slowly
to ensure that a high degree of consolidation occurs, which in
practice means that the mid-plane and base pore pressures
should be close to each other. To overcome these difficulties,
advanced triaxial testing is normally carried out using a midplane probe (Sodha, 1974; Hight, 1982). A flushable probe is
preferred for stiffer materials, where the initial suction in the
specimen is likely to exceed 100 kPa, and cavitation will
occur. A miniature pore pressure transducer has the
advantage of smaller size.
Figure 41 shows the testing strategy for the simplest of
advanced triaxial tests. The test is divided into six stages:

Deviatoric stress

Fig. 40. Advanced triaxial test configuration

Pore pressure Cell pressure

Shear to
failure

Time

1 month ?
(b)

Fig. 41. Basic testing strategy for an advanced triaxial test

been saturated in the ground, before sampling. While backpressure saturation (which through the application of a very
low effective stress allows the specimen to swell) is adopted
in routine triaxial testing in the UK and elsewhere, it has
been found that when using advanced triaxial apparatus the
application of large steps of cell pressure, aimed at reaching
the in situ mean effective stress as rapidly as possible before
measuring a B value, works well. This is shown schematically in Fig. 41 (stage 01). Because of sampling disturbance effects, the mean effective stress at the end of this
stage would not be expected to equal the mean effective
stress in the ground. Estimates of K0 based on these data
(Skempton & Sowa, 1963) should be treated with caution.
Stages 12 and 23 (Fig. 41) aim to bring the specimen
back to its in situ effective stress regime, albeit with a high
(generally . 300 kPa) back-pressure, to ensure effective pore
pressure measurement. The most economic technique is to
apply a single increment of consolidation (or swelling) under
isotropic conditions (stage 12), and follow this with a
deviatoric stress ramp (or series of small steps) (stage 23),
during which the excess mid-plane pore pressure should be
monitored. If the excess pore pressure (the difference between the measured mid-plane and base pore pressures)
exceeds a specified proportion of the major principal stress
(say 5%), then loading (or unloading if going into triaxial
extension, as shown in Fig. 41(a)) should be slowed, to
ensure that the actual effective stress path does not deviate
excessively from its planned route.
The necessity of modelling recent history (stage 34) has

Delivered by ICEVirtualLibrary.com to:


IP: 152.78.201.162
On: Mon, 14 Feb 2011 15:36:53

STIFFNESS AT SMALL STRAIN: RESEARCH AND PRACTICE

STRATEGIES FOR MEASUREMENT AND


INTEGRATION OF DATA
Table 4 brings together the methods that have been
discussed above in the context of the depths for which they

can be expected to yield good data (for field seismic tests)


and the parameters that can be determined.
As the table shows, there are a relatively large number of
ways in which the very-small-strain shear modulus in the
vertical plane (Gv0 ) can be determined. In addition to the
methods shown in the table, if isotropy is assumed then an
estimate of G0 can also be obtained from advanced undrained triaxial testing (Poissons ratio being 0.5 in the
undrained case). Some of these methods (e.g. cross-hole,
down-hole and bender element testing) are of particular
value in that they can be used to detect and estimate
stiffness anisotropy.
Table 4 also highlights the need to carry out triaxial
testing, despite its complexity and time-consuming nature.
As can be seen, this is needed for measurement of both
undrained and drained Youngs moduli, and their degradation
with strain.
Figure 42 shows a comparison of very-small-strain shear
modulus (Gv ) data obtained from field geophysics, resonant
column and advanced triaxial testing of a site underlain by
Eocene sandy clay. Down to 3040 m there is good agreement between Gv values determined using down-hole and

100

Shear modulus, Gv: MPa


200
300

400

10

20

Depth: m

been debated for some time (Atkinson et al., 1990; Clayton


& Heymann, 2001), but it now appears that small recent
strain excursions do not affect measured stiffness, whereas
larger ones do (Gasparre et al., 2007; Hight et al., 2007). In
practice, even when they are thought to have occurred,
estimating the magnitude of recent stress history (e.g. due to
the deposition of Terrace Gravels over London Clay (Burland et al., 1979) or fill) may be difficult.
Inaccurate measurements of stiffness will be made if
significant creep deformations are allowed to occur during
loading, since these will add to or subtract from the deformations being produced during current loading. In the tests
reported by Clayton & Heymann (2001) rest periods of up
to 2 weeks were used, leading to creep rates before loading
of less than 0.01%, equivalent to less than 2% of the locally
measured shear rate at the start of each shear stage. Routine
practice for advanced testing in the UK (based on practice
at Imperial College; Jardine, 1995) has been to adopt a rest
period (stage 45) that extends until the creep rate becomes
less than 1% of the subsequent external (machine) rate of
loading. It should be noted that this could be considerably
less conservative than the rates used by Clayton & Heymann
(2001), depending upon bedding effects.
In the final stage (56) of this example test, deviatoric
stress is applied under undrained and approximately constant
rate of strain conditions, either increasing the deviatoric
stress to produce failure in triaxial compression, or reducing
it to produce failure in triaxial extension. In contrast with
routine testing, a more-or-less standard machine rate of
strain of around 5% per day is typically applied. Given that
the specimen has mid-plane pore pressure measurement,
equalisation between the base and mid-plane is not a factor
in deciding on the machine rate of strain.
The test described above has typically been used to
determine the very-small-strain vertical undrained Youngs
modulus (Euv ), its degradation with strain and the effective
stress failure envelopes, loading in both triaxial compression
and extension. Following Ward et al. (1959) and Atkinson
(1975), it is suggested that horizontally cut specimens could
provide estimates of the horizontal undrained Youngs modulus (Euh ). The apparatus can also be used (in slower tests)
to determine the drained Youngs modulus and Poissons
ratio (E9v and 9vh ), and, when loading horizontally (Lings et
al., 2000), E9h =(1  9hh ). When used with care, high-quality
instrumentation, and excellent temperature control (Gasparre
& Coop, 2006), triaxial testing can also determine the verysmall-strain stiffness for other (drained) stress paths.

31

30

40

50

60

70
Cross-hole Gv0

Resonant column Gv0

Down-hole Gv0

CAU txl Eu (001%)/3

Fig. 42. Comparison of stiffness data obtained from field


geophysics, resonant column and advanced triaxial testing;
undisturbed samples of Eocene sandy clay

Table 4. Methods of obtaining the various stiffness parameters

Profile max. depth


Initial stiffness
Short-term operational
stiffness
Long-term operational
stiffness

Continuous surface
wave test

Down-hole shear
wave survey

Cross-hole shear
wave survey

Bender element Resonant


testing
column tests

810 m
Gv0

2040 m
Gv0

.100 m
Gv0 , Gh0 ?

Gv0 , Gh0

Gv , Gh y
Gv0

Euv , Euh y
Euv , Euh y

Gv0 , Gh0

Gv , Gh

E9v , E9h , 9vh

 Depending upon verticality of holes, change in stiffness with depth, etc.


y
Horizontal specimens required for Ehu , E9h and Gh , but in situ deviatoric stress cannot be re-established.

Delivered by ICEVirtualLibrary.com to:


IP: 152.78.201.162
On: Mon, 14 Feb 2011 15:36:53

Advanced triaxial
testing

32

CLAYTON

cross-hole seismic testing. Very-small-strain stiffness resonant column (Gv0 ) results give a lower stiffness profile, even
though the specimens were returned to their estimated in situ
effective stresses presumably because of sample disturbance. Undrained triaxial stiffnesses, derived assuming isotropy (i.e. Gv Euv =3) from Hall effect local strain
measurements at 0.01% axial strain, are considerably below
the stiffnesses from resonant column testing, as might be
expected at the higher strain levels.
Estimation of the degree of stiffness anisotropy requires
good specification and successful use of a horizontally
polarised source during cross-hole seismic testing, or of
horizontally polarised bender element tests. The results of
bender element testing and horizontally polarised cross-hole
shear wave seismic testing are not shown in Fig. 42, because
at this site they were judged to be unreliable, and thus the
degree of stiffness anisotropy (if any) was not known. It is
suggested that the chances of recognising undrained and
shear stiffness anisotropy may be improved in future by the
use of horizontally cut specimens in the resonant column
and triaxial apparatuses.
Determination of the changes of stiffness parameters with
strain remains challenging. Changes in stiffness (Euv or Gv )
can be measured under static conditions in advanced triaxial
testing, or under dynamic conditions in the cyclic triaxial
test, and in the resonant column apparatus. However, as a
result of cycling and strain rates, the different methods can
be expected to give different results, with estimates based on
resonant column testing being unconservative. Determination
of stiffness degradation with increasing strain should take
place (at least) in both compression and extension, since
degradation rates will differ.
Very few determinations of the full set of drained (longterm) anisotropic stiffness parameters have been reported for
undisturbed natural soils, to date. Given the testing timescales, and the fact that data from several different test types
need to be combined, this is likely to be feasible only for
major projects.
CONCLUSIONS
Non-linear elasticity has proved to be an effective and
convenient basis on which, for many types of ground, to
determine geotechnical displacements. However, most soils
and weak rocks can be expected to display stiffness anisotropy, requiring the determination of at least three parameters
for the computation of displacements under short-term (undrained) conditions, and five in the long term, under drained
conditions.
Despite the fact that a linear range of stressstrain behaviour does not exist, for practical purposes it is convenient to
determine stiffness parameters at very small strains (say
,0.001%) and use these as a reference. Very-small-strain
stiffness can be used to establish the stiffness profile, which
has a profound influence on displacement patterns around
new and existing infrastructure. And for those projects where
strains will remain low, the operational stiffness will not be
greatly different from the very-small-strain value.
To assess the sensitivity of predicted displacements and
bending moments to different stiffness parameters, a propped
cantilever wall has been modelled using linear and nonlinear elasticity, and it has been shown that, for this particular problem, stiffness at very small strain, the change of
stiffness parameters with increasing strain, and the degree of
stiffness anisotropy had significant effects on computed
displacements and wall bending moments.
Methods of determining very-small-strain stiffness, in the
field using seismic geophysical methods, and in the laboratory using bender elements, the resonant column test and the

triaxial apparatus, have been described. They have been


shown to have considerable promise in determining small
strain stiffness moduli, although they are not without problems. The advantages and disadvantages of each technique
have been reviewed, and the important potential influences
of sampling disturbance recalled.
Different methods of determining very-small-strain stiffness parameters cannot be expected to yield the same
results, because of the volume affected by testing, layering,
sampling disturbance, the effects of test detail, different
methods of interpretation, and so on. The various in situ
seismic testing methods were shown to have different viable
depth ranges, and to be restricted to determinations of shear
modulus.
Laboratory testing, although complex, time consuming,
and affected by sampling disturbance, can provide a greater
range of stiffness data than field testing, and is unavoidable
if stiffness degradation with strain is to be determined. As a
result of rate effects and cyclic loading, the rates of stiffness
degradation with strain determined using advanced triaxial
testing, resonant column testing and cyclic triaxial testing
will not be the same.
If isotropy of stiffness is assumed, the necessary stiffness
parameters can be determined for both the undrained and
drained cases, using a limited range of loading paths in an
advanced triaxial test. If anisotropy is assumed, then in the
undrained case it may be feasible to estimate the effects of
increasing strain using a combination of resonant column,
cyclic triaxial and monotonic advanced triaxial testing, with
horizontally and vertically cut specimens. Determination of
the full suite of transversely isotropic drained stiffness parameters, and especially Poissons ratio values, remains a
challenge.
A great many (sometimes only partly understood) details
control the precise results of stiffness tests. For future
routine work, higher-quality sampling and more complete
best practice and method specifications will be needed.
Engineering judgement is currently necessary in order to
weigh the different measurements in the context of a given
ground model and engineering application. Comparison of
results from different field and laboratory tests, and against
previously published results in similar ground conditions,
should be used to check the integrity of stiffness data.

ACKNOWLEDGEMENTS
The author gratefully recognises the important contributions made by colleagues, friends and geotechnical companies. Individuals contributing to the work, with data,
discussion, criticism and support included Jim Bennell, Tony
Butcher, Manolis Fleris, Chris Haberfield, Gerhard Heymann,
David Hight, Qaiser Iqbal, Marcus Matthews, Rory Mortimore, Louise Otter, Justin Phillips, John Powell, Pat Power,
Jeffrey Priest, Mike Rattley, Emily Rees, Chris Russell, Eben
Rust, Hardev Sidhu, Ken Stokoe, Jerry Sutton, Roger
Thompson and Xu Ming.
Data, facilities, expertise and comment have been provided by AWE plc, Buro Happold, Coffey, Fugro GeoConsulting Ltd, Golder Associates and SGC Ltd.
Finally, this paper was reviewed and criticised by Gerhard
Heymann, Marcus Matthews, Jeffrey Priest, Xu Ming and
Antonis Zervos. Their comments were invaluable: any remaining errors are the authors.

NOTATION
Cp a material constant
E Youngs modulus
E9 Youngs modulus in terms of effective stress

Delivered by ICEVirtualLibrary.com to:


IP: 152.78.201.162
On: Mon, 14 Feb 2011 15:36:53

STIFFNESS AT SMALL STRAIN: RESEARCH AND PRACTICE


Eh
Ev
Eu
E0
e
G
Gh
Gv
G0
K9
Ku
p9
p9atm
V
u
hh
hv
vh

Youngs modulus for loading in horizontal direction


Youngs modulus for loading in vertical direction
undrained Youngs modulus
Youngs modulus at very small strain
void ratio
shear modulus
shear modulus for distortion in horizontal plane
shear modulus for distortion in vertical plane
shear modulus at very small strain
bulk modulus in terms of effective stress ( dp9/dV )
undrained bulk modulus
isotropic effective stress
atmospheric pressure
volumetric strain
undrained Poissons ratio
Poissons ratio relating to horizontal strain caused by imposed
horizontal strain in normal direction
Poissons ratio relating to vertical strain caused by imposed
horizontal strain
Poissons ratio relating to horizontal strain caused by imposed
vertical strain

REFERENCES
Abbireddy, C. O. R. (2008). Particle form and its impact on
packing and shear behaviour of particulate materials. PhD
thesis, University of Southampton
Abbireddy, C. O. R., Clayton, C. R. I. & Huvenne, V. A. I. (2009).
A method of estimating the form of fine particulates. Geotechnique 59, No. 6, 503511, doi: 10.1680/geot.2008.P.009.
Abbiss, C. P. (1979). A comparison of the stiffness of the Chalk at
Mundford from a seismic survey and a large scale tank test.
Geotechnique 29, No. 4, 461468, doi: 10.1680/geot.1979.29.
4.461.
Abbiss, C. P. (1981). Shear wave measurements of the elasticity of
the ground. Geotechnique 31, No. 1, 91104, doi: 10.1680/
geot.1981.31.1.91.
Ackerley, S. K., Hellings, J. E. & Jardine, R. J. (1987). Discussion
on A new device for measuring local axial strains on triaxial
specimens by Clayton and Khatrush (1986). Geotechnique 37,
No. 3, 413417, doi: 10.1680/geot.1987.37.3.413.
Anderson, D. G. & Stokoe, K. H. II (1978). Shear modulus: a time
dependent soil property. In Dynamic geotechnical testing, ASTM
STP 654, pp. 6690. West Conshohocken, PA: ASTM International.
ASCE (1976). Subsurface investigation for design and construction
of foundations of buildings. American Society of Civil Engineers, Manuals and Reports on Engineering Practice, No. 56, 61
pp. New York: ASCE.
ASTM (2003). Standard test methods for the determination of the
modulus and damping properties of soils using the cyclic
triaxial apparatus, ASTM D3999-91. West Conshohocken, PA:
ASTM International.
ASTM (2004). Standard test method for consolidated undrained
triaxial compression test for cohesive soils, ASTM D4767. West
Conshohocken, PA: ASTM International.
ASTM (2007). Standard test methods for modulus and damping of
soils by resonant-column method, ASTM D4015-07. West Conshohocken, PA: ASTM International.
Atkinson, J. H. (1975). Anisotropic elastic deformations in laboratory tests on undisturbed London Clay. Geotechnique 25, No. 2,
357374, doi: 10.1680/geot.1975.25.2.357.
Atkinson, J. H. (2000). Non-linear soil stiffness in routine design.
The 40th Rankine Lecture. Geotechnique 50, No. 5, 487508,
doi: 10.1680/geot.2000.50.5.487.
Atkinson, J. H. & Evans, J. S. (1985). Discussion on The measurement of soil stiffness in the triaxial apparatus by Jardine et al.
(1984). Geotechnique 35, No. 3, 378380, doi: 10.1680/
geot.1985.35.3.378.
Atkinson, J. H. & Sallfors, G. (1991). Experimental determination
of stressstraintime characteristics in laboratory and in situ
tests. General report to Session 1. Proc. 10th Eur. Conf. Soil
Mech. Found. Engng, Florence 3, 915956.
Atkinson, J. H., Richardson, D. & Stallebrass, S. E. (1990). Effect
of recent stress history on the stiffness of overconsolidated soil.
Geotechnique 40, No. 4, 531540, doi: 10.1680/geot.1990.40.
4.531.

33

Baldi, G., Hight, D. W. & Thomas, G. E. (1988). A re-evaluation of


conventional triaxial test methods. Proceedings of ASTM symposium on advanced triaxial testing of soil and rock, ASTM STP
977, pp. 219263. West Conshohocken, PA: ASTM International.
Baligh, M. M., Azzouz, A. S. & Chin, C.-T. (1987). Disturbance
due to ideal tube sampling. Proc. ASCE, J. Geotech. Engng 113,
No. 7, 739757.
Ballard, R. F. & MacLean, F. G. (1975). Seismic field methods for
in situ moduli. Proceedings of ASCE conference on in situ
measurement of soil properties, Vol. 1, pp. 121150.
Bates, C. R. (1989). Dynamic soil property measurements during
triaxial testing. Geotechnique 39, No. 4, 721726, doi: 10.1680/
geot.1989.39.4.721.
Bell, J. F. (1989). Experimental solid mechanics in the nineteenth
century. The 1989 William Murray Lecture. Exp. Mech. 29, No.
2, 57165.
Bellotti, R., Jamiolkowski, M., Lo Presti, D. C. F. & ONeill, D. A.
(1996). Anisotropy of small strain stiffness in Ticino sand.
Geotechnique 46, No. 1, 115131, doi: 10.1680/geot.1996.
46.1.115.
Bennell, J. D. & Taylor Smith, D. (1991). A review of laboratory
shear wave techniques and attenuation measurements with particular reference to the resonant column. In Shear waves in
marine sediments (eds J. M. Hovem, M. D. Richardson and R.
D. Stoll), pp. 8393. Dordrecht: Kluwer.
Bennell, J. D., Taylor Smith, D. & Davis, A. M. (1984). Resonant
column testing of marine sediments. Proc. Oceanography Int.,
Brighton, 1.9/1OI1.9/11.
Bishop, A. W. & Bjerrum, L. (1960). The relevance of the triaxial
test to the solution of stability problems. Proceedings of the
ASCE research conference on the shear strength of cohesive
soils, Boulder, CO, pp. 437501.
Bishop, A. W. & Gibson, R. E. (1963). The influence of the
provisions for boundary drainage on strength and consolidation
characteristics of soil measured in the triaxial apparatus. In
Laboratory shear testing of soils, ASTM STP 361, pp. 435451.
Philadelphia, PA: American Society for Testing and Materials.
Bishop, A. W. & Henkel, D. J. (1962). The measurement of soil
properties in the triaxial test, 2nd edn. London: Edward Arnold.
Bishop, A. W. & Hight, D. W. (1977). The value of Poissons ratio
in saturated soils and rocks stressed under undrained conditions.
Geotechnique 27, No. 3, 369384, doi: 10.1680/geot.1977.27.
3.369.
Britto, A. M. & Gunn, M. J. (1987). Critical state soil mechanics
via finite elements. New York: Ellis Horwood.
Broms, B. B. (1980). Soil sampling in Europe: state-of-the-art.
Proc. ASCE, J. Geotech. Div. 106, No. GT1, 6598.
Brown, S. F. & Snaith, M. S. (1974). The measurement of recoverable and irrecoverable deformations in the repeated load triaxial
test. Geotechnique 24, No. 2, 255259, doi: 10.1680/geot.
1974.24.2.255.
BSI (1999). Code of practice for site investigations, BS 5930
incorporating Amendment No. 1 (2007). Milton Keynes: British
Standards Institution.
BSI (1990). Methods of test for soils for civil engineering purposes.
Part 8: Shear strength tests (effective stress), BS 1377-8:1990
incorporating Amendment No. 1 (1995). Milton Keynes: British
Standards Institution.
Bui, M. (2009). Influence of some particle characteristics on the
small strain response of granular materials. PhD thesis, University of Southampton
Burland, J. B. (1989). Small is beautiful: the stiffness of soils at
small strains. Ninth Laurits Bjerrum Lecture. Can. Geotech. J.
26, No. 4, 499516.
Burland, J. B. & Hancock, R. J. R. (1977). Underground car park at
the House of Commons: geotechnical aspects. The Struct. Engr
55, No. 2, 87100.
Burland, J. B. & Kalra, J. C. (1986). Queen Elizabeth II Conference
Centre: geotechnical aspects. Proc. Instn Civ. Engrs, Part 1 80,
December, 14791503.
Burland, J. B. & Lord, J. A. (1970). The load deformation behaviour
of Middle Chalk at Mundford, Norfolk: a comparison between
fullscale performance and in situ and laboratory measurements.
Proceedings of the conference on in situ investigations in soils
and rocks, British Geotechnical Society, London, pp. 315.

Delivered by ICEVirtualLibrary.com to:


IP: 152.78.201.162
On: Mon, 14 Feb 2011 15:36:53

34

CLAYTON

Burland, J. B., Sills, G. C. & Gibson, R. E. (1973). A field and


theoretical study of the influence of non-homogeneity on settlement. Proc. 8th Int. Conf: Soil Mech. Found. Engng, Moscow
1.3, 3946.
Burland, J. B., Simpson, B. & St John, H. D. (1979). Movements
around excavations in London Clay. Proc. Eur. Conf. Soil Mech.
Found. Engng, Brighton 1, 1329.
Butcher, A. P. & Powell, J. J. M. (1996). Practical considerations
for field geophysical techniques used to assess ground stiffness.
Proceedings of the international conference on advances in site
investigation practice, London, pp. 701714.
Butcher, A. P. & Powell, J. M. (1997). Determining the modulus of
the ground from in situ geophysical testing. Proc. 14th Int.
Conf. Soil Mech. Found. Engng, Hamburg, 449452.
Butler, F. G. (1975). Heavily over-consolidated clays. Review paper:
Session III. Proceedings of the BGS conference on settlement of
structures, Cambridge, pp. 531578.
Casagrande, A. & Carillo, N. (1944). Shear failure of anisotropic
materials. Proc. Boston Soc. Civ. Engrs 31, 7487.
Cascante, G., Santamarina, C. & Yassir, N. (1998). Flexural excitation in a standard torsional resonant column. Can. Geotech. J.
35, No. 3, 478490.
Chandler, R. J., Harwood, A. H. & Skinner, P. J. (1992). Sample
disturbance in London clay. Geotechnique 42, No. 4, 577585,
doi: 10.1680/geot.1992.42.4.577.
Chowdhury, R. N. & King, G. J. W. (1971). Discussion on
Anisotropic elastic parameters for soil by Pickering (1970).
Geotechnique 21, No. 2, 181183, doi: 10.1680/geot.1971.
21.2.181.
Chu, H. Y. F., Lee, S. H. H. & Stokoe, K. H. II (1984). Effects of
structural and stress anisotropy on velocity of low-amplitude
compression waves propagating along principal stress directions
in dry sand, Report GR84-6. University of Texas at Austin.
Clayton, C. R. I. & Heymann, G. (2001). Stiffness of geomaterials
at very small strains. Geotechnique 51, No. 3, 245255, doi:
10.1680/geot.2001.51.3.245.
Clayton, C. R. I. & Khatrush, S. A. (1986). A new device for
measuring local axial strains on triaxial specimens. Geotechnique 36, No. 4, 593597, doi: 10.1680/geot.1986.36.4.593.
Clayton, C. R. I. & Siddique, A. (1999). Tube sampling disturbance: forgotten truths and new perspectives. Proc. Inst. Civ.
Engnrs, Geotech. Engng 137, No. 3, 127135.
Clayton, C. R. I., Edwards, A. & Webb, M. J. (1991). Displacements within the London Clay during construction. Proc. 10th
Eur. Conf. Soil Mech. Found. Engng, Florence 2, 791796.
Clayton, C. R. I., Hight, D. W. & Hopper, R. J. (1992). Progressive
destructuration of the Bothkennar Clay: implications for sampling and reconsolidation procedures. Geotechnique 42, No. 2,
219239, doi: 10.1680/geot.1992.42.2.219.
Clayton, C. R. I., Matthews, M. C., Gunn, M. J., Foged, N. &
Gordon, M. A. (1995a). Reinterpretation of surface wave test for
the resund crossing. Proc. 11th Eur. Conf. Soil Mech. Found.
Engng, Copenhagen 5, 141147.
Clayton, C. R. I., Matthews, M. C. & Simons, N. E. (1995b). Site
investigation. Oxford: Blackwell Science (available to download
at www.geotechnique.info).
Clayton, C. R. I., Siddique, A. & Hopper, R. J. (1998). Effects of
sampler design on tube sampling disturbance: numerical and
analytical investigations. Geotechnique 48, No. 6, 847867, doi:
10.1680/geot.1998.48.6.847.
Clayton, C. R. I., Theron, M. & Best, A. I. (2004). The measurement of vertical shear-wave velocity using side-mounted bender
elements in the triaxial apparatus. Geotechnique 54, No. 7,
495498, doi: 10.1680/geot.2004.54.7.495.
Clayton, C. R. I., Priest, J. A. & Best, A. I. (2005). The effects of
disseminated methane hydrate on the dynamic stiffness and
damping of a sand. Geotechnique 55, No. 6, 423434, doi:
10.1680/geot.2005.55.6.423.
Clayton, C. R. I., Xu, M. & Bloodworth, A. (2006). A laboratory
study of the development of earth pressure behind integral
abutments. Geotechnique 56, No. 8, 561571, doi: 10.1680/
geot.2006.56.8.561.
Clayton, C. R. I., Abbireddy, C. O. R. & Schiebel, R. (2009a). A
method of estimating the form of coarse particulates. Geotechnique 59, No. 6, 493501, doi: 10.1680/geot.2007.00195.
Clayton, C. R. I., Priest, J. A., Bui, M., Zervos, A. & Kim, S. G.

(2009b). The Stokoe resonant column apparatus: effects of


stiffness, mass and specimen fixity. Geotechnique 59, No. 5,
429437, doi: 10.1680/geot.2007.00096.
Clayton, C. R. I., Priest, J. A. & Rees, E. V. L. (2010). The effects
of hydrate cement on the stiffness of some sands. Geotechnique
60, No. 6, 435455, doi: 10.1680/geot.2010.60.6.435.
Cole, K. W. & Burland, J. B. (1972). Observations of retaining wall
movements associated with a large excavation. Proc. 5th Eur.
Conf. Soil Mech. Found. Engng, Madrid 1, 445453.
Cooper, M. L. (1978). Early contributions to the problems of
elasticity. Phys. Educ. 13, No. 6, 384387.
Costa-Filho, L. M. & Vaughan, P. R. (1980). Discussion on A
computer model for the analysis of ground movements in
London clay by Simpson et al. (1979). Geotechnique 30, No. 3,
336339, doi: 10.1680/geot.1980.30.3.336.
Cox, H. (1856). The deflection of imperfectly elastic beams and the
hyperbolic law of elasticity. Trans. Camb. Phil. Soc. 9, No. II,
177190.
Cresswell, A. & Powrie, W. (2004). Triaxial tests on an unbonded
locked sand. Geotechnique 54, No. 2, 107115, doi: 10.1680/
geot.2004.54.2.107.
Cuccovillo, T. & Coop, M. R. (1997). The measurement of local
axial strains in triaxial tests using LVDTs. Geotechnique 47, No.
1, 167171, doi: 10.1680/geot.1997.47.1.167.
Daramola, O. (1980). On estimating K0 for overconsolidated granular soils. Geotechnique 30, No. 3, 310313, doi: 10.1680/
geot.1980.30.3.310.
Davis, A. M. & Bennell, J. D. (1986). Dynamic properties of
marine sediments. In Ocean seismo-acoustics (eds T. Akal and
J. M. Berkson), pp. 501510. New York: Plenum Press.
de Alba, P. & Baldwin, K. (1991). Use of bender elements in
soil dynamics experiments. In Recent advances in instrumentation, data acquisition and testing in soil dynamics (eds S.
K. Bhatia and G. W. Blaney), ASCE National Convention
Special Geotechnical Publication, 12, pp. 86101. New York:
ASCE.
Dobrin, M. D. (1960). Introduction to geophysical prospecting, 2nd
edn. London: McGraw-Hill.
Drnevich, V. P. (1972). Undrained cyclic shear of saturated sands.
J. Soil Mech. Found. Div. ASCE 98, No. SM8, 807825.
Drnevich, V. P. (1978). Resonant column testing: problems and
solutions. In Dynamic geotechnical testing (ed. M. L. Silver),
ASTM STP 654, pp. 384398. Philadelphia, PA: American
Society for Testing and Materials.
Duffy, J. & Mindlin, R. D. (1957). Stressstrain relations and
vibrations of a granular medium. Trans. ASME, J. Appl. Mech.
24, 585593.
Duleau, A. (1820). Essai Theorique et Experimental sur la Resistance du Fer Forge. Paris: Huzard-Courcier.
Dupin, P. (1815). Experiences sur la flexibilite, la force et lelasticite des bois. Journal de lEcole Royale Polytechnique 10, 137
211.
Dusseault, M. B. & Morgenstern, N. R. (1979). Locked sands. Q. J.
Engng Geol. 12, No. 2, 117131.
Dyer, M., Jamiolkowski, M. & Lancellotta, R. (1986). Experimental
soil engineering and models for geomechanics. Proc. 2nd Int.
Symp. on Numer. Models Geomech., Ghent, 873906.
Dyvik, R. & Madshus, C. (1985). Lab measurements of Gmax using
bender elements. Proc. ASCE Annual Convention on Advances
in the Art of Testing Soils under Cyclic Conditions, Detroit, MI,
186196.
Dyvik, R. & Olsen, T. S. (1989). Gmax measured in oedometer and
DSS tests using bender elements. Proc. 12th Int. Conf. Soil
Mech. Found. Engng, Rio de Janeiro 1, 3942.
Ferreira, C. & da Fonseca, A. V. (2005). International parallel tests
on bender elements at the University of Porto, Portugal, internal
report. Porto: Faculdade de Engenharia da Universidade do Porto
(FEUP).
Ferreira, C., da Fonseca, A. V. & Santos, J. A. (2006). Comparison
of simultaneous bender elements and resonant column tests on
Porto residual soil. Proceedings of the geotechnical symposium
on soil stressstrain behavior: measurement, modeling and
analysis, Rome, pp. 523535.
Gasparre, A. & Coop, M. R. (2006). Techniques for performing
small strain probes in the triaxial apparatus. Geotechnique 56,
No. 7, 491495, doi: 10.1680/geot.2006.56.7.491.

Delivered by ICEVirtualLibrary.com to:


IP: 152.78.201.162
On: Mon, 14 Feb 2011 15:36:53

STIFFNESS AT SMALL STRAIN: RESEARCH AND PRACTICE


Gasparre, A., Nishimura, S., Minh, N. A., Coop, M. R. & Jardine,
R. J. (2007). The stiffness of natural London Clay. Geotechnique
57, No. 1, 3347, doi: 10.1680/geot.2007.57.1.33.
Gazetas, G. (1982). Vibrational characteristics of soil deposits
with variable velocity. J. Numer. Analyt. Methods Geomech. 6,
120.
Gibson, R. E. (1974). The analytical model in soil mechanics. 14th
Rankine Lecture. Geotechnique 24, No. 2, 115140, doi:
10.1680/geot.1974.24.2.115.
Goddard, J. D. (1990). Nonlinear elasticity and pressure-dependent
wave speeds in granular media. Proc. R. Soc. Lond. A 430, No.
1878, 105131.
Gordon, M. A. (1997). Applications of field seismic geophysics to
the measurement of geotechnical stiffness parameters. PhD
thesis, University of Surrey.
Goto, S. & Tatsuoka, F. (1986). Influence of the testing condition
on strength and deformation characteristics of granular material.
Proc. 21st Ann. Japanese Symp. Soil Mech. Found. Engng 1,
237240 (in Japanese).
Graham, J. & Houlsby, G. T. (1983). Anisotropic elasticity of a
natural clay. Geotechnique 33, No. 2, 165180, doi: 10.1680/
geot.1983.33.2.165.
Green, G. (1828). An essay on the application of mathematical
analysis to the theories of electricity and magnetism. Published
by the author.
Gunn, M. J. & Clayton, C. R. I. (1992). Installation effects and
their importance in the design of earth-retaining structures.
Geotechnique 42, No. 1, 137141, doi: 10.1680/geot.1992.42.
1.137.
Gunn, M. J., Satkunananthan, A. & Clayton, C. R. I. (1992). Finite
element modelling of installation effects. Proceedings of the
ICE conference on retaining structures, Cambridge, pp. 4655.
Hardin, B. O. (1961). Study of elastic wave propagation and
damping of granular materials. PhD thesis, University of Florida
at Gainesville, FL.
Hardin, B. O. & Black, W. L. (1966). Sand stiffness under various
triaxial stresses. J. Soil Mech. Found. Div. ASCE 92, No. SM2,
2742.
Hardin, B. O. & Black, W. L. (1968). Vibration modulus of
normally consolidated clay. J. Soil Mech. Found. Div. ASCE 94,
No. SM2, 353369.
Hardin, B. O. & Blandford, G. E. (1989). Elasticity of articulate
materials. J. Geotech. Engng ASCE 115, No. 6, 788805.
Hardin, B. O. & Drnevich, V. P. (1972). Shear modulus and
damping in soil: measurement and parameter effects. J. Soil
Mech. Found. Div. ASCE 98, No. 7, 603624.
Hardin, B. O. & Music, J. (1963). Apparatus for the vibration of
soil specimens during the triaxial test. In Symposium on instrumentation and apparatus for soils and rocks. ASTM STP 392,
pp. 5574. Philadelphia, PA: American Society for Testing and
Materials.
Hardin, B. O. & Richart, E. F. J. (1963). Elastic wave velocities in
granular soils. J. Soil Mech. Found. Div. ASCE 89, No. SM1,
3365.
Harrison, I. R. (1991). A pushed thinwall tube sampling system for
stiff clays. Ground Engng, April, 3034.
Hart, M. B. (1987). Orbitally induced cycles in the Chalk facies of
the United Kingdom. Cretaceous Res. 8, 335318.
Heymann, G. (1998). The stiffness of soils and weak rocks at very
small strains. PhD thesis, University of Surrey.
Heymann, G., Clayton, C. R. I. & Reed, G. T. (2005). Triaxial ultra
small strain measurements using laser interferometry. ASTM
Geotech. Test. J. 28, No. 6, 544552.
Heymann, G., Jacobsz, S. W., Vorster, T. E. B. & Storry, R. B.
(2008). Comparison of ground stiffness from seismic surface
wave and large scale load tests. Proc. 3rd Int. Conf. on Site
Characterization (ISC3), Taipei, 843848.
Hight, D. W. (1982). A simple piezometer probe for the routine
measurement of pore pressure in triaxial tests on saturated soils.
Geotechnique 32, No. 4, 396401, doi: 10.1680/geot.1982.
32.4.396.
Hight, D. W. (1986). Laboratory testing: assessing BS5930. In Site
investigation practice: Assessing BS5930 (ed. A. B. Hawkins),
Engineering Geology Special Publication No. 2, pp. 4351.
Geological Society of London.
Hight, D. W. (1998). Soil characterization: the importance of

35

structure and anisotropy. 38th Rankine Lecture. London: British


Geotechnical Society (personal communication).
Hight, D. W. & Higgins, K. G. (1995). An approach to the
prediction of ground movements in engineering practice: background and application. Proceedings of the international symposium on pre-failure deformation characteristics of geomaterials,
Sapporo, Vol. 2, pp. 909945.
Hight, D. W. & Jardine, R. J. (1993). Small strain stiffness and
strength characteristics of hard London Tertiary clays. Proc. Int.
Symp. on Hard Soils Soft Rocks, Athens, 533552.
Hight, D. W., Bennell, J. D., Chana, B., Davis, P. D., Jardine, R. J.
& Porovic, E. (1997). Wave velocity and stiffness measurements
of the Crag and Lower London Tertiaries at Sizewell.
Geotechnique 47, No. 3, 451474, doi: 10.1680/geot.1997.
47.3.451.
Hight, D. W., McMillan, F., Powell, J. J. M., Jardine, R. J. &
Allenou, C. P. (2003). Some characteristics of London clay.
Proceedings of the international workshop on characterization
and engineering properties of natural soils, Singapore, Vol. 2,
pp. 851907.
Hight, D. W., Gasparre, A., Nishimura, S., Minh, N. A., Jardine,
R. J. & Coop, M. R. (2007). Characteristics of the London Clay
from the Terminal 5 site at Heathrow Airport. Geotechnique 57,
No. 1, 318, doi: 10.1680/geot.2007.57.1.3.
Hird, C. C. & Yung, P. C. Y. (1989). The use of proximity
transducers for local strain measurements in triaxial tests. ASTM
Geotech. Test. J. 12, No. 4, 292296.
Hoar, R. J. & Stokoe, K. H. II (1978). Generation and measurement
of shear waves in situ. In Dynamic geotechnical testing, ASTM
STP 654, pp. 329. Philadelphia, PA: American Society for
Testing and Materials.
Hobbs, N. B. (1975). Factors affecting the prediction of settlement
of structures on rocks with particular reference to the Chalk and
Trias. Proceedings of the BGS conference on settlement of
structures, pp. 579610. London: Pentech Press.
Hooke, R. (1676). A description of helioscopes and some other
instruments made by Robert Hooke. London: John Martyn.
Hooke, R. (1678). Lectures de potentia restitutiva, or of spring.
London: John Martyn.
Hooper, J. A. (1973). Observations on the behaviour of a pile raft
foundation on London Clay. Proc. Inst. Civ. Engrs, Part 2 55,
No. 4, 855877.
Hope, V. S. (1993). Applications of seismic transmission tomography in civil engineering. PhD thesis, University of Surrey.
Hope, V. S., Clayton, C. R. I. & Sutton, J. A. (1998). The use of
seismic geophysics in the characterization of a weak rock site.
Proceedings of the symposium on geotechnical site characterization (eds P. K. Robertson and P. W. Mayne), pp. 479484.
Rotterdam: Balkema.
Isenhower, W. M. & Stokoe, K. H. II (1981). Strain-rate dependent
shear modulus of San Francisco Bay Mud. Proceedings of the
international conference on recent advances in earthquake engineering and soil dynamics, St Louis, MO, Vol. 2, pp. 597
602.
Itasca (2005). FLAC: fast lagrangian analysis of continua, version
5. Minneapolis, MB: Itasca Consulting Group, Inc.
Iwasaki, T. & Tatsuoka, F. (1977). Effects of grain size and grading
on dynamic shear moduli of sands. Soils Found. 17, No. 3,
1935.
Jardine, R. J. (1995). One perspective of the pre-failure deformation
characteristics of some geomaterials. Proceedings of the international symposium on pre-failure deformation characteristics of
geomaterials (eds S. Shibuya, T. Mitachi and S. Miura), Vol. 2,
pp. 855885. Rotterdam: Balkema.
Jardine, R. J., Symes, M. J. & Burland, J. B. (1984). The measurement of soil stiffness in the triaxial apparatus. Geotechnique 34,
No. 3, 323340, doi: 10.1680/geot.1984.34.3.323.
Jardine, R. J., Symes, M. J. & Burland, J. B. (1985). Authors reply
to discussion on The measurement of soil stiffness in the
triaxial apparatus. Geotechnique 35, No. 3, 380382, doi:
10.1680/geot.1985.35.3.380.
Jardine, R. J., Potts, D. M., Fourie, A. B. & Burland, J. B. (1986).
Studies of the influence of nonlinear stressstrain characteristics
in soilstructure interaction. Geotechnique 36, No. 3, 377397,
doi: 10.1680/geot.1986.36.3.377.
Jovicic, V. & Coop, M. R. (1998). The measurement of stiffness

Delivered by ICEVirtualLibrary.com to:


IP: 152.78.201.162
On: Mon, 14 Feb 2011 15:36:53

36

CLAYTON

anisotropy in clays with bender elements in the triaxial apparatus. Geotech. Test. J. 21, No. 1, 310.
Kavvadas, M. & Anagnostopoulos, A. G. (1998). A framework for
the mechanical behaviour of structured soils. Proc. 2nd Int.
Symp. on Geotechnics of Hard Soils Soft Rocks, Napoli 2,
603614.
Kee, R. (1970). The behaviour and design of foundations in Chalk.
MPhil thesis, Department of Civil Engineering and Construction,
Hatfield Polytechnic.
Kuwano, J., Katagiri, M., Kita, K., Nakano, M. & Kuwano, R.
(2001). A review of Japanese standards for laboratory shear
tests. In Advanced laboratory stressstrain testing of geomaterials (eds F. Tatsuoka, S. Shibuya and R. Kuwano), pp. 53
64. Lisse: Swets & Zeitlinger.
La Rochelle, P., Sarrailh, J., Tavenas, F., Roy, M. & Leroueil, S.
(1981). Causes of sampling disturbance and design of a new
sampler for sensitive soils. Can. Geotech. J. 18, No. 1, 5266.
Lai, C. G. & Rix, G. J. (1998). Simultaneous inversion of Rayleigh
phase velocity and attenuation for near-surface site characterization, Report No. GIT-CEE/GEO-98-2. School of Civil and
Environmental Engineering, Georgia Institute of Technology.
Landon, M. M., DeGroot, D. J. & Sheahan, T. C. (2007). Nondestructive sample quality assessment of a soft clay using shear
wave velocity. ASCE J. Geotech. Geoenv. Engng 133, No. 4,
424432.
Lefebvre, G. & Poulin, C. (1979). A new method of sampling in
sensitive clay. Can. Geotech. J. 16, No. 1, 226233.
Lekhnitskii, S. G. (1981). Theory of elasticity of an anisotropic
body. Moscow: Mir Publishers.
Leong, E. C., Yeo, S. H. & Rahardjo, H. (2005). Measuring shear
wave velocity using bender elements. Geotech. Test. J. 28, No.
5, 488498.
Lings, M. L. (2001). Drained and undrained anisotropic elastic
stiffness parameters. Geotechnique 51, No. 6, 555565, doi:
10.1680/geot.2001.51.6.555.
Lings, M. L., Pennington, D. S. & Nash, D. F. T. (2000). Anisotropic stiffness parameters and their measurement in a stiff
natural clay. Geotechnique 50, No. 2, 109125, doi: 10.1680/
geot.2000.50.2.109.
Lings, M. L. & Greening, P. D. (2001). A novel bender/extender
element for soil testing. Geotechnique 51, No. 8, 713717, doi:
10.1680/geot.2001.51.8.713.
Lo Presti, D. C. F., Jamiolkowski, M., Pallara, O., Cavallaro, A. &
Pedroni, S. (1997). Shear modulus and damping of soils.
Geotechnique 47, No. 3, 603617, doi: 10.1680/geot.1997.47.
3.603.
Lord, J. A., Clayton, C. R. I. & Mortimore, R. N. (2002). Engineering in chalk, CIRIA Report C574. London: Construction Industry Research and Information Association.
Love, A. E. H. (1892). A treatise on the mathematical theory of
elasticity, 2 vols. Cambridge: Cambridge University Press.
Mair, R. J. (1993). Developments in geotechnical engineering research: applications to tunnels and deep excavations. Unwin
Memorial Lecture 1992. Proc. Instn Civ. Engrs, Civ. Engng 3,
No. 1, 2741.
Marsland, A. (1972). Clays subjected to in situ plate tests. Ground
Engng 5, No. 6, 2431.
Marsland, A. (1986). The choice of test methods in site investigation. In Site investigation practice: assessing BS5930 (ed. A. B.
Hawkins), Engineering Geology Special Publication No. 2, pp.
289298. London: Geological Society.
Marsland, A. & Eason, B. J. (1973). Measurement of displacements
in the ground below loaded plate in deep bore holes. Proceedings of the symposium on field instrumentation, London, Vol. 1,
pp. 304317.
Matthews, M. C. & Clayton, C. R. I. (2004). Large diameter plate
tests on weathered in situ chalk. Q. J. Engng Geol. Hydrogeol.
37, No. 1, 6172.
Matthews, M. C., Hope, V. S. & Clayton, C. R. I. (1996). The use
of surface waves in the determination of ground stiffness
profiles. Proc. Inst. Civ. Engrs Geotech. Engng 119, No. 2,
8495.
Matthews, M. C., Clayton, C. R. I. & Own, Y. (2000). The use of
field geophysical techniques to determine geotechnical stiffness
parameters. Proc. Inst. Civ. Engrs Geotech. Engng 143, No. 1,
3142.

McDowell, G. R. & Bolton, M. D. (2001). Micro mechanics of


elastic soil. Soils Found. 41, No. 6, 147152.
Menq, F.-Y. & Stokoe, K. H. II (2003). Linear dynamic properties of
sandy and gravelly soils from large-scale resonant tests. In
Deformation characteristics of geomaterials (eds H. Di Benedetto,
T. Doanh, H. Geoffroy and C. Sauzeat), pp. 6371. Lisse: Swets
& Zeitlinger.
Nazarian, S. & Stokoe, K. H. II (1984). In situ shear wave
velocities from spectral analysis of surface waves. Proc. 8th
World Conf. Earthquake Engng, San Francisco 3, 3138.
Pennington, D. S., Nash, D. F. T. & Lings, M. L. (1997). Anisotropy
of G0 shear stiffness in Gault Clay. Geotechnique 47, No. 3,
391398, doi: 10.1680/geot.1997.47.3.391.
Pickering, D. J. (1970). Anisotropic elastic parameters for soil.
Geotechnique 20, No. 3, 271276, doi: 10.1680/geot.1970.
20.3.271.
Pinches, G. M. & Thompson, R. P. (1990). Crosshole and downhole
seismic surveys in the UK Trias and Lias. In Field testing in
engineering geology (eds F. G. Bell, M. G. Culshaw, J. C.
Cripps and J. R. Coffey), Engineering Geology Special Publication No. 6, pp. 299308. London: Geological Society.
Potts, D. M. (2003). Numerical analysis: a virtual dream or practical
reality? 42nd Rankine Lecture. Geotechnique 53, No. 6, 535
573, doi: 10.1680/geot.2003.53.6.535.
Roesler, S. K. (1979). Anisotropic shear modulus due to stress
anisotropy. J. Geotech. Engng ASCE 105, No. 7, 871880.
Rowe, P. W. (1972). The relevance of soil fabric to site investigation
practice. 12th Rankine Lecture. Geotechnique 22, No. 2, 195
300, doi: 10.1680/geot.1972.22.2.195.
Royal Commission on Application of Iron to Railway Structures
(1849). Royal Commissioners appointed to inquire into the
application of iron to railway structures. 1847: Report and
appendices, Report xviii, 435pp, 2 vols. Whitehall, 26 July.
London: HMSO.
St John, H. D. (1975). Field and theoretical studies of the behaviour of ground around deep excavations in London Clay. PhD
thesis, University of Cambridge.
Sanchez-Salinero, I., Roesset, J. M. & Stokoe, K. H. (1986).
Analytical studies of body wave propagation and attenuation,
Report GR 86-15. Civil Engineering Department, University of
Texas at Austin, TX.
Scarrow, J. A. & Gosling, R. C. (1986). An example of rotary core
drilling in soils. In Site investigation practice: Assessing BS
5930 (ed. A. B. Hawkins), Engineering Geology Special Publication No. 2, pp. 357364. London: Geological Society.
Schultheiss, P. J. (1981). Simultaneous measurement of P and S
wave velocities during conventional laboratory soil testing procedures. Mar. Geotechnol. 4, No. 4, 343367.
Schultheiss, P. J. (1982). The influence of packing structure on
seismic wave velocities in sediments. PhD thesis, University
College of Swansea, Wales.
Schultheiss, P. J. (1983). The influence of packing structure on
seismic wave velocities in sediments. University College of
North Wales Marine Geological Report No. 3/1, 186 pp. Bangor: University College of North Wales.
Shirley, D. J. (1978). An improved shear wave transducer. J. Acoust.
Soc. Am. 63, No. 5, 16431645.
Shirley, D. J. & Hampton, L. D. (1978). Shear wave measurements
in laboratory sediments. J. Acoust. Soc. Am. 63, No. 2, 607
613.
Simpson, B. (1981). Finite elements in design: with particular
reference to deep basements in London Clay. In Finite elements
in geotechnical engineering (eds D. J. Naylor and G. N. Pande),
pp. 213242. Swansea: Pineridge Press.
Simpson, B. (1992). Retaining structures: displacement and design.
32nd Rankine Lecture. Geotechnique 42, No. 4, 541576, doi:
10.1680/geot.1992.42.4.541.
Simpson, B., ORiordan, N. J. & Croft, D. D. (1979). A computer
model for the analysis of ground movements in London Clay.
Geotechnique 29, No. 2, 149175, doi: 10.1680/geot.1979.
29.2.149.
Skempton, A. W. & Sowa, V. A. (1963). The behaviour of saturated
clays during sampling and testing. Geotechnique 13, No. 4,
269290, doi: 10.1680/geot.1963.13.4.269.
Sodha, V. (1974). The stability of embankment dam fills of plastic
clays. MPhil thesis, Imperial College, University of London.

Delivered by ICEVirtualLibrary.com to:


IP: 152.78.201.162
On: Mon, 14 Feb 2011 15:36:53

STIFFNESS AT SMALL STRAIN: RESEARCH AND PRACTICE


Sorensen, K. K., Baudet, B. A. & Simpson, B. (2007). Influence of
structure on the time-dependent behaviour of a stiff sedimentary
clay. Geotechnique 57, No. 1, 113124, doi: 10.1680/geot.2007.
57.1.113.
Stokoe, K. H. II & Nazarian, S. (1985). Use of Rayleigh waves in
liquefaction studies. In Measurement and use of shear wave
velocity for evaluating dynamic soil properties (ed. R. D.
Woods), pp. 117. New York: ASCE.
Stokoe, K. H. II, Isenhower, W. M. & Hsu, J. R. (1980). Dynamic
properties of offshore silty samples. Proc. 12th Ann. Offshore
Technology Conf., Houston, TX, paper OTC 3771.
Stokoe, K. H. II, Hwang, S. K., Lee, N. K. J. & Andrus, R. D.
(1995). Effect of various parameters on the stiffness and damping of soils at small to medium strains. Keynote Lecture. Proc.
1st Int. Symp. on Pre-failure Deformation of Geomaterials,
Hokkaido 2, 785816.
Sully, J. P. & Campanella, R. G. (1995). Evaluation of in situ
anisotropy from crosshole and downhole shear wave velocity
measurements. Geotechnique 45, No. 2, 267282, doi: 10.1680/
geot.1995.45.2.267.
Tan, T.-S., Lee, F.-H., Chong, P.-T. & Tanaka, H. (2002). Effect of
sampling disturbance on properties of Singapore Clay. Proc.
ASCE, J. Geotech. Geoenviron. Engng 128, No. 11, 898906.
Tanaka, H. & Tanaka, M. (1999). Key factors governing sample
quality. In Characterization of soft marine clays (eds T. Tsuchida
and A. Nakase), pp. 5782. Rotterdam: Balkema.
Tatsuoka, F. & Shibuya, S. (1992). Deformation characteristics of
soils and rocks from field and laboratory tests. Keynote Lecture.
Proc. 9th Asian Conf. Soil Mech. Found. Engng, Bangkok 2,
101170.
Thomann, T. G. & Hryciw, R. D. (1990). Laboratory measurements
of small strain shear modulus under K0 conditions. Geotech.
Test. J. ASTM 13, No. 2, 97105.
Thornton, C. & Zhang, L. (2010). On the evolution of stress and
microstructure during general 3D deviatoric straining of granular
media. Geotechnique 60, No. 5, 333341, doi: 10.1680/
geot.2010.60.5.333.
Timoshenko, S. P. (1953). History of the strength of materials: With
a brief account of the history of elasticity and theory of
structures. London: McGraw-Hill.
Todhunter, I. (1886). A history of the theory of elasticity and of the
strength of materials, from Galilei to the present time (ed. K.
Pearson). Cambridge: Cambridge University Press.
Toki, S., Shibuya, S. & Yamashita, S. (1995). Standardization of
laboratory test methods to determine the cyclic deformation
properties of geomaterials in Japan. Proceedings of the international symposium on pre-failure deformation characteristics of
geomaterials (eds S. Shibuya, T. Mitachi and S. Miura), Vol. 2,
pp. 741784. Rotterdam: Balkema..
Viggiani, C. (2000). Does engineering need science? In Constitutive
modelling of granular materials (ed. D. Kolymbas), pp. 2535.
Berlin: Springer-Verlag.
Viggiani, G. & Atkinson, J. H. (1995). Interpretation of bender
element tests. Geotechnique 45, No. 1, 149154, doi: 10.1680/
geot.1995.45.1.149.
Ward, W. H., Samuels, S. G. & Butler, M. E. (1959). Further
studies of the properties of London Clay. Geotechnique 9, No.
2, 3358, doi: 10.1680/geot.1959.9.2.33.
Ward, W. H., Burland, J. B. & Gallois, R. W. (1968). Geotechnical
assessment of a site at Mundford, Norfolk, for a large proton
accelerator. Geotechnique 18, No. 4, 399431, doi: 10.1680/
geot.1968.18.4.399.
Whitman, R. V., Richardson, A. M. & Healy, K. A. (1961). Timelags in pore pressure measurements. Proc. 5th. Int. Conf. Soil
Mech. Found. Engng, Paris 1, 407411.
Woods, R. D. (1994). Borehole methods in shallow seismic exploration. In Geophysical characterization of sites, ISSMFE Technical
Committee #10 (ed. R. D. Woods), pp. 91100. Rotterdam:
Balkema.
Woods, R. D. & Henke, R. (1979). Seismic techniques in the
laboratory. In Geophysical techniques in the laboratory, Proc.
ASCE Specialty Session, Atlanta, GA, pp. 293322.
Xu, M., Clayton, C. R. I. & Bloodworth, A. (2007). The earth
pressure behind full-height frame integral abutments supporting
granular backfill. Can. Geotech. J. 44, No. 3, 284298.

37

Yamashita, S. & Suzuki, T. (1999). Youngs and shear moduli under


different principal stress directions of sand. Proc. 2nd Int. Symp.
on Pre-failure Deformation Characteristics of Geomaterials,
Torino 1, 149158.
Yamashita, S., Kawaguchi, T., Nakata, Y., Mikami, T., Fujiwara, T.
& Shibuya, S. (2009). Interpretation of international parallel test
on the measurement of Gmax using bender elements. Soils
Found. 49, No. 4, 631650.
Yu, P. & Richart, F. E. (1984). Stress ratio effects on shear modulus
of dry sand. J. Geotech. Engng ASCE 110, 3, 331345.
Zienkiewicz, O. C., Valliappan, S. & King, I. P. (1968). Stress
analysis of rock as a no tension material. Geotechnique 18, No.
1, 5666, doi: 10.1680/geot.1968.18.1.56.

VOTE OF THANKS
Emeritus Professor R. BUTTERFIELD, University of
Southampton.
Good evening. I am very pleased to have been asked to
propose a vote of thanks to Professor Chris Clayton for his
lecture on Stiffness at small strain: research and practice
for many reasons not least of which is that I have known
him and admired his great breadth of interests for many
years.
Above all I found his lecture refreshing, focused, clear
and highly relevant even if its embedded message was that
we still have some way to go before we can claim to be able
to predict the service state displacements of soil structures
in general. This is in part because, in addition to their basic
anisotropy and non-linearity, soil materials usually undergo
non-recoverable, inelastic plastic displacements that often
increase over time: they creep.
Professor Clayton has, I think, provided a very significant
contribution totally in the spirit of answering Rankines
(1858) question: In practical science what are we to do?
Historically, Roscoe in his 1970 Rankine Lecture mentioned being similarly motivated. Apparently, in 1951 Sir
John Baker asked him to provide foundations for a steelframed building that would collapse simultaneously with the
frame, and also their predicted displacements at working
load. (Baker initiated plastic-hinge design methods for steel
structures.) Roscoe was embarrassed to find that he couldnt
do so adequately using current soil mechanics knowledge,
and proclaimed: The soil mechanician should not be interested only in failure; he should be concerned with being
able to predict the movements of a foundation when subject
to given working loads. (Apparently soil mechanics was a
very masculine activity as recently as 1970!) His endeavours
led, of course, to the development of critical state soil
mechanics.
Then again, Gibson, in his 1974 Rankine Lecture, presented analytical solutions for elastic half-spaces in which
the shear modulus increased linearly with depth, and he
would certainly be gratified to find that this is now a wellsupported and useful model.
I must add here that I find it extremely worrying that
neither Roscoes nor Gibsons work would have been supported under the latest government guidelines, under which,
even in universities, demonstrable short-term economic benefit is to be the key determinant for research funding.
I am sure we all agree that Professor Clayton has
succeeded admirably in demonstrating how the moduli required in a range of non-linear elastic soil models can be
determined from a variety of both in situ and laboratory
tests, their strengths and their weaknesses and, in particular,
the practical circumstances in which they are most relevant.
Please join me in expressing your appreciation of his
quite excellent lecture.

Delivered by ICEVirtualLibrary.com to:


IP: 152.78.201.162
On: Mon, 14 Feb 2011 15:36:53

You might also like