Acta Geotechnica (2013) 8:293–309

DOI 10.1007/s11440-012-0201-8

RESEARCH PAPER

On the undrained strain-induced anisotropy of loose sand

T. Doanh • Ph. Dubujet • X. Protière

Received: 25 May 2012 / Accepted: 20 November 2012 / Published online: 22 December 2012
Ó Springer-Verlag Berlin Heidelberg 2012

Abstract An experimental study was carried out to 1 Introduction

investigate the effects of previous deviatoric strain histories
on the undrained behaviour of loose and saturated Toyoura Two kinds of anisotropy exist in the geotechnical literature.
sand and compared with known results of Hostun RF sand. The first kind concerns the inherent anisotropy Winherent
From an initial isotropic stress state, recent deviatoric created by the long-term process of one-dimensional sedi-
strain histories in the compression side of the triaxial plane mentation of natural soils or by the short-term axial com-
were generated by a standard drained presheared cycle up paction of soil specimen in laboratory. The second kind
to a specified mobilized stress ratio. Mainly, the fully liq- refers to the change of the internal structure due to some past
uefied, contractive, unstable and softening behaviour of loading, thus the development of the induced anisotropy
loose sand was progressively transformed into the non- Winduced , either by stress or by strain loading. For simplicity,
liquefied, dilative, fully stable and hardening behaviour of these two forms of anisotropy are often considered to be
dense-like sand, while remaining within a narrow range of independent, and the effects of anisotropy arbitrarily
loose density. The paper validates and extends the current decomposed into W ¼ Winherent þ Winduced . Ultimately, the
understanding of strain-induced anisotropy of loose sand. inherent anisotropy can be seen as the initial state of induced
New experimental data support the directional dependency anisotropy created by unknown past loading.
of the instability cone on the stress increment direction, It is well known that anisotropy is an important com-
suggest the bifurcation characteristics of loose sand and ponent of the mechanical behaviour of soils, and it can even
evidence the important role of past deviatoric strain be a critical component in a civil engineering project in
histories. many aspects, especially in the realm of small strains.
Natural soils, due to their horizontal layer formation, tend to
Keywords Anisotropy  Elasticity  Laboratory tests  be highly anisotropic, inhomogeneous and overconsolidat-
Liquefaction  Sands ed, thus notoriously difficult to apprehend for both experi-
mental and theoretical soil scientists, Muir Wood [53].
In the early days of soil mechanics, the inherent
anisotropy of natural soils is often ignored in laboratory
studies and neglected in many engineering projects. In
practice, the usual isotropic hypothesis simplifies greatly
T. Doanh (&) the soil behaviour and facilitates conveniently the geo-
Ecole Nationale des Travaux Publics de l’Etat, Laboratoire de
technical design, without drilling down into unnecessarily
Tribologie et Dynamique des Systmes, UMR 5513, 2 Rue M.
Audin, 69518 Vaulx en Velin Cedex, France details. Since isotropic, homogeneous and normally con-
e-mail: [email protected] solidated samples can be created in laboratory, granular
soils offer an alternative candidate to natural soils and
Ph. Dubujet  X. Protière
constitute a first useful laboratory step toward a better
Ecole Nationale d’Ingénieurs de Saint-Etienne, Laboratoire de
Tribologie et Dynamique des Systmes, UMR 5513, 58 Rue J. understanding of the latter. Furthermore, the study of iso-
Parot, 42023 Saint-Etienne Cedex 2, France tropic soil eliminates de facto Winherent and favours greatly

123
294 Acta Geotechnica (2013) 8:293–309

the experimental and theoretical studies devoted to The deviatoric strain histories are created by simple linear
Winduced . stress paths from an initial isotropic stress state in the
All equipments of soil mechanics were successively classical triaxial plane and performed on loose and moist-
used to investigate the effects of induced anisotropy on tamped sand samples having nearly the same initial void
initially isotropic sands. Conventional triaxial machines ratio prior to undrained shearing. Sect. 2 recalls some
were used first by Poorooshasb et al. [44, 45], Tatsuoka essential results of previous studies and describes the
et al. [48], Ishihara et al. [32–34] and Vaid et al. [50] experimental programme. Sect. 3 reports new experimental
among others, on dense sand using large drained triaxial findings on two fine sands with some global observations.
preshearing, and contributed successfully to identify the The validation and extension of the current knowledge of
yield surface in the development of non-associated elas- induced anisotropy are studied in Sect. 4 with some new
toplasticity framework. Later on, more complex machines specific features.
such as true triaxial, Arthur et al. [4]; simple shear, Arthur
et al. [2, 3]; hollow cylinder, Hight et al. [27]; and plane
strain machines, Tatsuoka et al. [49], for citing a few, were 2 Experimental programme
solicited next to explore the uncharted features of induced
anisotropy, principal stress rotation, and intermediate 2.1 Previous observations
principal stress. These pioneer experimental studies dis-
covered the key role of induced anisotropy created by Toyoura and Hostun sands are among the most studied
recent stress loading on the mechanical behaviour of soils. sands in the international geotechnical literature. Espe-
In a collective effort to unlock the underlying mecha- cially, the undrained behaviour of loose and normally
nisms of devastating earthquakes, Lanier et al. [40], Doanh consolidated Toyoura sand was investigated by Verdugo
et al. [18], Gajo et al. [24], Vaid [50] and Yamada et al. and Ishihara [52], Ishihara [31] and Yamada et al. [54]; and
[54], for example, noticed the same important effect of Hostun RF sand by Canou et al. [7], Konrad [35], Di Prisco
drained preloading histories on the undrained shear et al. [13], Gajo et al. [24] and Finge et al. [21], among
strength of isotropic loose sands and emphasized the role of many others. Induced anisotropy developed during mono-
the recent strain history. The strain developed in the past tonic or cyclic preshearing, either drained or undrained,
stress history has been identified experimentally as one of changes dramatically the deformation characteristics of
the most suitable candidates controlling the induced sand, irrespective of initial density. The inevitable changes
anisotropy. These experimental findings were introduced in of void ratio alone from the drained preloading cannot fully
constitutive modelling in terms of history strain-dependent account for these behaviours.
mechanisms, and induced anisotropy was qualitatively When restricted to the compression side of the classical
described [14, 19]. However, how to capture the quantita- triaxial plane, the essential features of the undrained
tive effects of induced anisotropy is still a hot topic in the induced anisotropy of loose Hostun RF sand of a previous
geotechnical literature. work [16] are recalled in Fig. 1. A set of 10 moist-tamped
Recently, our previous works [15, 16] evidence the
important role of recent deviatoric strain history on the
300
behaviour of loose Hostun RF sand, from any initial iso- Undrained failure
CC-C 10
tropic or anisotropic stress state, and offer some practical 9
250 10
indications for a quantitative description. The deviatoric 9
B 8 8
Deviatoric stress (kPa)

strain history constitutes a special and useful case of strain- 200 7

7
6
induced anisotropy. The main finding was the progressive Phase Transformation Line A 6
CID
transformation of the usual contractive and unstable 150
5
behaviour of loose sand into a dilative and stable behaviour 5

of dense-like sand by previous deviatoric strain history, 100 4

4
while remaining in a loose state. The granular mechanics,
50 2
3
even when restricted to the classical triaxial plane, are more 1 3
2
subtle and complex than it appears to most soil scientists. 1
0 4
The purpose of this paper is to validate and extend the 3
2 CC-E
known effects of deviatoric strain history on the undrained 1
-50
0 50 100 150 200
behaviour of loose sands. It represents a step forward in the
Effective mean pressure (kPa)
efforts to understand the strain-induced anisotropy and
introduces new experimental findings of an extensive Fig. 1 Undrained induced anisotropy of presheared and loose Hostun
programme on very loose and saturated Toyoura sand. RF sand

123
Acta Geotechnica (2013) 8:293–309 295

loose sand samples were preloaded along the classical (O5) when the final undrained loading and the past drained
drained stress path (CID, dashed line) up to a mobilized preloading are in opposite direction is not considered in
deviatoric stress level qmob (cross symbols) and unloaded this paper.
back down to the same initial isotropic stress state to These salient features are essentially governed by recent
eliminate the well-known effect of mean pressure. Full deviatoric strain created by drained linear stress path.
lines represent the subsequent monotonic undrained load- Recently, large undrained preloadings on Toyoura sand in
ing in compression, in the same (respectively opposite) Yamada et al. [54] give similar results.
direction of previous drained preloading in the upper Once identified the effects of past deviatoric strain his-
(respectively lower) part. Basically, the usual contractive tory, the next logical step is to verify and possibly extend
and unstable behaviour of loose sand is transformed pro- these effects to other clean sands. This paper proposes to
gressively into a dilative and stable behaviour of dense-like explore the mechanisms of undrained induced anisotropy
sand, depending on the magnitude of qmob. It is worth on Toyoura sand, using the identified mechanisms obtained
noting that all samples have nearly the same state of very on Hostun RF loose sand in the previous work [16], as
loose density at the beginning of final undrained shearing, background for evaluating new experimental results. It will
which is a requirement for minimizing the another well- focus particularly on two open questions linking to the two
known effect of density. Shearing from the same initial missing features in Fig. 1 : when subjected to small drained
isotropic stress state and the same void ratio permits to take preshearing, can loose sand experience a true softening
into account only the effects of previous deviatoric strain behaviour associated with the drop of deviatoric stress in
history. Dashed curves are the usual partial static lique- the stress–strain relationship, together with a decreasing
faction of a virgin isotropic loose sample in triaxial com- mobilized frictional angle and a full liquefied state below
pression as well as in extension. test C1? Can recent deviatoric strain history of large
Five open circles in Fig. 2 give the striking effects of drained preshearing create a fully stable behaviour with
past deviatoric strain histories on the undrained behaviour continuous dilatancy of dense-like sand, above test C10,
of very loose and saturated Hostun RF clean sand and while remaining in the same state of loose density?
support the preponderant influence of strain-induced
anisotropy : (O1) the directional dependency of the initial 2.2 New experimental programme
effective stress gradient on the direction of preloading
stress path, (O2) the common and nonlinear effective stress The reference Toyoura sand used in this study is a siliceous
response nearly up to the deviatoric tress peak, regardless clean sand, and the comparative Hostun RF sand and quartz
of the length of preloading stress path, (O3) the progressive clean sand. These sands have the same USCS classification,
appearance of the dilatancy domain associated with the although Toyoura sand is much finer. The physical prop-
phase transformation state and (O4) the progressive evo- erties of these uniform fine sands are shown in Table 1.
lution of the undrained behaviour toward that of dense-like Sand samples were prepared using a modified moist
sand. The dramatic change of the undrained behaviour tamping and under compaction method, proposed by Bjer-
rum et al. [6] and Ladd [36]. This method was commonly
chosen to create sand samples with large initial void ratios
4 D E at fabrication stage and to obtain a fully contractive
behaviour in triaxial compression shearing in laboratory.
3 Moist-tamped samples were prepared by placing predeter-
Failure line B
mined quantities of moist sand, mixed with 2 % of distilled
Deviatoric stress (kPa)

C
water in weight, and gently compacted in five layers of
Phase Transformation Line
prescribed thickness using a flat-bottom circular stainless
steel tamper of 20 mm in diameter. The targeted relative
2
density Dr was under 15 % before undrained shearing.
Short samples, 70 mm in height and 70 mm in diameter,
were prepared with enlarged and lubricated end plates to
A 1

C' Isotropic stress state Table 1 Hostun RF and Toyoura sands

B' 5
D'
Sand D50 (mm) Cu Cc emin emax
Effective mean pressure (kPa)
Hostun RF 0.38 1.90 0.97 0.624 0.961
Fig. 2 Schematic illustration of the effects of undrained induced Toyoura 0.19 1.33 1.02 0.590 0.960
anisotropy on cycle presheared and loose sand

123
296 Acta Geotechnica (2013) 8:293–309

produce homogeneous deformations at large strains. Latex

membrane of 0.3 mm of thickness was used. CO2 method failure

[37] with deaired distilled water and back pressure of up to PT

200 kPa were applied to fully saturate sand samples to

peak
obtain Skempton’s coefficient B higher than 0.95. p
peak

Deviatoric stress
The experimental works were conducted using a modi- q B mob
peak
fied classical triaxial system. All tests were sheared axially q
mob
in a strain-controlled mode with a deformation rate of
p
about 0.21 %/min, up to 20 % of axial strain ea in com- mob

pression. Additional information concerning the sample

preparation, the saturation procedure, the initial void ratio CID
determination, the repeatability of the undrained test and
the verification of the antifrictional system can be found
were checked in [17, 18, 20, 29]. A
The void ratio of each sample was evaluated from the Effective mean pressure
water content obtained at the end of the undrained test. All
soil grains and water were carefully retrieved in an dural- Fig. 3 Preloading drained linear stress path with constant direction
umin tray and weighted with a tolerance of 0.1 g. The void
ratio ec before undrained shearing was determined after the induced anisotropy of fully saturated sand. The direc-
oven-drying. As indicated by Verdugo and Ishihara [52], tion of all preloading stress path was fixed, a = 71°, from
the void ratios estimated from water content are more the positive mean effective pressure axis. These preloading
accurate and reliable than those determined from the usual deviatoric strain histories were the same as those used for
procedure of measuring sample dimensions. Hostun RF sand in the previous study [16].
Table 2 summarizes all tests of this study with their
2.3 Preloading stress path initial conditions and their results. Symbol ‘‘CC’’ indicates
the drained standard triaxial compression cycle preshear,
In Fig. 3, from an initial isotropic stress state at point A, a ‘‘C’’ the final undrained shearing mode in compression.
drained classical triaxial cycle ABA, up to a desired value e0–100 is the void ratio at the end of the isotropic consoli-
of mobilized frictional level gmob at point B, was per- dation stage at 100 kPa and ec at the end of the presheared
formed before the final undrained shearing in compression. cycle before undrained shearing. axial vol
mob and mob are the axial
The investigation points B were carefully distributed along and volumetric strain at the beginning of undrained
the whole standard drained stress path CID of virgin nor- shearing. gmob is the mobilized frictional level of the
mally consolidated sample in order to study the effects of drained presheared cycle, and adr contains the range of

Table 2 Undrained tests on loose and presheared Toyoura sand

Tests e0–100 ec Dr_c% qmobkPa gmob %Italic qpeakkPa gpeak peak % gTP axial
mob % vol
mob %
adr degree

Unstable behaviour, B1
T-CC-C1 0.947 0.947 3.63 27.4 0.25 17.8 42.0 0.59 0.21 – 0.063 0.037 215°6–243°4
T-CC-C2 0.955 0.946 3.83 93.9 0.72 50.6 79.6 0.79 0.14 – 0.568 0.482 215°7–247°3
T-CC-C3 0.957 0.941 5.24 128.1 0.90 63.5 100.6 0.95 0.21 – 1.285 0.874 215°3–250°0
T-CC-C4 0.957 0.936 6.44 139.4 0.95 67.4 108.8 1.01 0.28 – 1.791 1.060 215°9–248°7
Temporarily unstable behaviour, B2
T-CC-C5 0.958 0.934 7.04 142.9 0.97 68.6 109.9 1.03 0.36 – 2.379 1.215 215°5–253°2
T-CC-C6 0.954 0.930 8.04 153.0 1.02 71.7 120.4 1.07 0.40 – 2.372 1.224 217°0–247°7
T-CC-C7 0.940 0.915 12.26 173.5 1.10 78.0 136.4 1.15 0.40 1.37 2.543 1.378 216°7–247°7
T-CC-C8 0.941 0.905 14.87 196.8 1.19 84.3 164.2 1.25 0.86 1.37 4.989 2.059 215°4–244°7
T-CC-C9 0.948 0.910 13.46 207.5 1.23 86.9 183.3 1.28 1.32 1.35 6.443 2.192 217°6–240°8
T-CC-C10 0.950 0.905 14.90 224.8 1.29 90.9 216.4 1.36 2.10 1.38 8.440 2.591 219°1–246°7
Stable behaviour, B3
T-CC-C11 0.924 0.870 24.22 240.8 1.34 94.6 – – – – 9.698 2.818 –
T-CC-C12 0.920 0.890 18.77 256.4 1.39 98.0 – – – – 16.307 3.209 –

123
Acta Geotechnica (2013) 8:293–309 297

stress increment direction having negative second-order pressure–axial strain plane. The measured excess pore
work d2W. The variability of e0–100 seems to be randomly pressure DU was normalized by the initial radial stress
distributed. The zero strain state is chosen at the beginning before the final monotonic undrained shearing. The last
of undrained loading. bottom-left figure indicates the evolution of the mobilized
stress ratio g ¼ q=p0 versus normalized DU. All important
features of loose sand discussed in the paper are graphi-
3 Undrained behaviour of presheared loose sand cally identified and related using these four planes [15]. To
facilitate the visual comparison, virgin isotropic behaviour
The experimental undrained behaviour of loose Toyoura was superimposed as dashed line in all figures. Anisotropic
sand can be interpreted synthetically by grouping together scale is used to emphasize the differences in the response
four figures, as shown in Fig. 4. The first top-left figure of virgin and presheared behaviours. The results of the
shows the effective stress paths in the q  p0 plane in terms drained preloading can be found at Protière [46].
of the deviatoric stress q ¼ r0a  r0r and the effective mean

pressure p0 ¼ r0a þ 2r0r 3, where r0a and r0r are the effec- 3.1 Unstable behaviour with total liquefaction, B1
tive axial and radial stresses. The second top-right figure
shows the stress–strain relationship in the deviatoric–axial The test on virgin sample confirmed the undrained unstable
strain plane, and the third bottom-right figure, the evolution behaviour of Toyoura moist-tamped loose sand, as reported
of the excess pore pressure in the normalized excess pore by Verdugo [51]. In the q  p0 plane, the Lade’s instability

(a) 150 (b) 120

4

3
q
Drained presheared level peak
100
= 1.41
failure
q
peak
Deviatoric stress (kPa)

Deviatoric stress (kPa)

4
100 2 80
3

60 max
2
= 0.58
IL
max
50 40
CID
1
1
20
Virgin sample
Virgin sample 1 2 3 4
0 0
0 50 100 150 0 1 2 3 4 5 6 7 8
Effective mean pressure (kPa) Axial strain (%)

(c) 1 (d) 1 Virgin sample

Liquefaction 1 2 3 4
max

0.8 0.8
Normalized pore pressure

Normalized pore pressure

max
0.6 0.6

0.4 0.4
4

3
Virgin sample q
peak
0.2 2 0.2
q
1 peak

O
0 0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 0 1 2 3 4 5 6 7 8
Stress ratio Axial strain (%)

Fig. 4 Results on compression presheared cycle, unstable behaviour of Toyoura sand. a Effective stress paths. b Stress–strain relationships.
c Mobilized level d Pore pressure generation

123
298 Acta Geotechnica (2013) 8:293–309

line IL [39], defined for virgin isotropically and normally continuous increase in the stress ratio g up to gmax. Only a
consolidated sand (marked as dashed line), has a slope in brief softening behaviour of loose sand is observed, from
compression gLI = 0.58 (broken line), or a mobilized gmax (open circle) to the catastrophic collapse of instan-
frictional angle at peak /peak of about 15°3 in Fig. 4a. taneous liquefaction at large strains.
As expected from the previous work on Hostun RF sand, The Hill’s instability criterion of second-work order [28],
recalled in Fig. 1, the effective stress paths of the first series defined as d2 W ¼ dr : de with dr and de the incremental
on Toyoura loose sand in Fig. 4a offer similar evolution, stress and strain tensor, is reduced to d2 W ¼ dra  dea þ
showing a unique initial gradient of about 10.65 ± 1.01 and 2drr  der ¼ dq  dea in the case of undrained triaxial.
inclined toward the positive p0 axis, a common linear Therefore, d2W and dq have the same sign for undrained
response from the beginning of the undrained shearing up to compression. Consequently, loose sand of this series is
the deviatoric stress peak qpeak (open square), a progressive potentially unstable since d2W has negative values from
transition around the rounded peak, a descending part deviatoric stress peak qpeak onwards. It can be characterized
toward the origin of the axis and an unexpected global as unstable behaviour. This aspect will be developed further
collapse not shown in this figure. Full black circles on the in Sect. 4.3.
total stress path (CID) indicate the drained mobilized levels This unstable behaviour with full liquefaction and sud-
gmob prior to undrained shearing. Also expected from the den failure was not observed in loose Hostun sand having
instability concept, test C1, with a level gmob below the the same range of relative density [16].
instability line, does not modify this instability line; there-
fore, no additional tests with low gmob are required. 3.2 Temporarily unstable behaviour, B2
For tests C1 to C4, the peaks of the stress–strain rela-
tionship in Fig. 4b occur at small axial strain, below 0.3 %, The second series of six tests with higher mobilized level,
followed by a sharp drop of deviatoric stress. However, the well above gLI, explores and confirms the temporarily
constant residual value corresponding to the steady state of unstable behaviour in the sense of Hill, since d2W has
deformation (SSD) at large strains was never attained in this temporarily negative values from qpeak to null value of the
series. Instead, we observed a global collapse of the sample quasi-steady state of deformation (QSSD [8]). Then, a
within 2 seconds, which is the constant data sampling fre- tendency to dilatancy appears after a peak of excess pore
quency of our system. After an initial monotonic and con- pressure generation DUmax and stability returns with posi-
tinuous increase, the excess pore pressure DU reaches tive values of d2W. This series experiences three successive
suddenly the initial radial stress and the samples liquefy phases : stability before qpeak, temporarily instability from
immediately, as shown in Fig. 4d only for virgin sample for qpeak to qmin of QSSD and finally complete stability at large
clarity with dashed line relying the last two data points. It is strains. The length of the temporarily unstable phase, in
impossible to collect any data with our digital acquisition terms of axial strain in Fig. 5b, is greatly reduced as gmob
system during this catastrophic and instantaneous collapse, increases.
unless using a fast acquisition system. The effective stress paths in the q  p0 plane of Fig. 5a
Figure 4c offers new insights into the unstable behav- show a common nonlinear response up to the progressively
iour of loose sand. The initial common and linear part of sharp bend around qpeak, followed by different descending
the effective stress paths is translated into a common and parts to the point of phase transformation (PT [42, 48])
linear part in this figure. The subsequent evolution of all before the final climb-up along the undrained failure. Note
tests, after a progressive transition, is almost linear with that qpeak moves upwards in the direction of preloading
different gradient and parallel in a regular pattern, despite with increasing gmob.
some experimental scatter, up to gmax (horizontal arrow), The pore pressure generation in Fig. 5d indicates the
the maximum of the stress ratio during undrained loading. existence of a local extremum at small strain, of about
This graph reveals the absence of the dilatancy phenome- 0.21 %, vertical up arrow, well before the PT of DUmax at
non of this series, since no maximum of DU is observed large strain. This local extremum occurs within the com-
and no vertical gradient in the effective stress plane is mon nonlinear response.
noticed. Furthermore, it shows a possibility of a true soft- Figure 5c confirms the often overlooked hardening
ening behaviour of loose sand associated with a drop of characteristic of temporarily unstable behaviour due to a
deviatoric stress in the stress–strain relationship, together continuous increase in the stress ratio g [15].
with a decreasing mobilized frictional angle from gmax.
The large drop of deviatoric stress in the stress–strain 3.3 Stable behaviour, B3
relationship, usually misinterpreted as ‘‘strain softening’’
behaviour of loose sand in the geotechnical literature The last series of only two tests with highest mobilized
[7, 39], has in fact a hardening behaviour because of a level performed in this study, up to 98 % of undrained

123
Acta Geotechnica (2013) 8:293–309 299

(a) 250 (b) 250 10

Potentially unstable state
10 9
Drained presheared level
10 9 q q
200 = 1.41 8 200 max min
failure
9
= 1.37
Deviatoric stress (kPa)

Deviatoric stress (kPa)

7
TP 8
8
6
150 150 U
5 max
7

6 7
5
100 100
= 0.58
IL

6
50 50
CID
5

Virgin sample Virgin sample

0 0
0 50 100 150 200 0 5 10 15 20
Effective mean pressure (kPa) Axial strain (%)

(c) 1 (d) 1
Liquefaction U 5
max Virgin sample
5
6
0.8 6
0.8
Normalized pore pressure

U
Normalized pore pressure
max

7
0.6 0.6
7

8
0.4 0.4
8
Virgin sample
9 9
0.2 0.2
10 10
= 1.37 small
TP
O = 1.41 U
failure
0 0
0 0.5 1 1.5 0 5 10 15 20
Stress ratio Axial strain (%)

Fig. 5 Results on compression presheared cycle, intermediate behaviour of Toyoura sand. a Effective stress paths. b Stress–strain relationships.
c Mobilized level d Pore pressure generation

failure, offers a new stable behaviour of dense-like sand, the line of phase transformation cannot be defined for this
never seen before for loose and presheared Hostun sand. series although it seems that test T-CC-C12 follows a large
This series remains stable, shown by always positive d2W, portion of the PTL identified in previous Sect. 3.2.
due to the remarkable disappearance of qpeak in the stress–
strain relationship of loose sand in Fig. 6b. It is highly
probable that a large plateau of deviatoric stress can be 4 Discussions
detected with gmob & 0.91-0.94 between test T-CC-C10
and T-CC-C11. This plateau indicates the frontier between We now focus on the validation and the extension of some
the temporarily unstable and stable behaviour. experimental striking features of strain-induced anisotropy
The common nonlinear response in the q  p0 plane can in the compression side shown schematically in Fig. 2 with
be fairly large, up to gmob in Fig. 6a. Large mobilized level four open circles. The essential idea is that a more broader
triggers a continuous decrease in pore pressure from the picture of strain-induced anisotropy of granular material
local extremum at small strain in Fig. 6d, indicating a emerges with new results on Toyoura sand.
dilative behaviour even within the common nonlinear
response. This only point of Umax cannot be identified to 4.1 Undrained strain-induced anisotropy revisited
the usual PT, resulting in a very low corresponding
mobilized angle, and also due to the lack of vertical gra- To facilitate the visual comparison, the global behaviour
dient at pmin in the effective stress path [30]. Technically, of two sands is given in different representations in

123
300 Acta Geotechnica (2013) 8:293–309

(a) 400 (b) 400

12
Drained presheared level

12
11
300 300
Deviatoric stress (kPa)

Deviatoric stress (kPa)

11
12
11
= 1.41
failure
200 200
= 1.37
TP
CID

100 = 0.58 100

IL

Virgin sample
Virgin sample
0 0
0 50 100 150 200 250 0 1 2 3 4 5 6 7 8
Effective mean pressure (kPa) Axial strain (%)

(c) 1 (d) 1
Liquefaction = 1.37
TP
0.8 = 1.41 0.8
failure Virgin sample
0.6 0.6
Normalized pore pressure

Normalized pore pressure

Virgin sample
0.4 0.4
U U
max max
0.2 0.2
11
0 0

-0.2 -0.2
O 11
12
-0.4 -0.4
12
-0.6 -0.6
0 0.5 1 1.5 0 1 2 3 4 5 6 7 8
Stress ratio Axial strain (%)

Fig. 6 Results on compression presheared cycle, stable behaviour of Toyoura sand. a Effective stress paths. b Stress–strain relationships.
c Mobilized level. d Pore pressure generation

Figs. 7 (Toyoura sand) and 8 (Hostun sand), and also in Fig. 7c. This low end behaviour was not observed for
Table 3. Hostun sand in Figs. 8b, c.
This side-by-side graphical comparison gives the false 2. The fully stable and hardening behaviour B3 of test 12
impression of déjà vu. A close examination reveals some in Fig. 7c, characterized by a continuous dilatancy
subtle extending features that were posed in the intro- (analogous to continuous decrease in excess pore
duction of the experimental programme in Sect. 2. The pressure). Hostun sand did not experience this high end
similarity of the effective stress paths in Figs. 7a and 8a characteristic in Fig. 8c.
confirms the progressive transformation of contractive
and unstable behaviour of loose sand into dilative As expected from previous results on Hostun sand,
and stable behaviour of dense-like sand by previous Fig. 9 shows a constant initial gradient (O1) of the
deviatoric strain histories [15, 16]. Furthermore, this effective stress path of about 10.7 ± 1.0, regardless of the
behaviour can be extended in two directions with length of the drained preloading. This initial gradient,
Toyoura sand: higher than that of Hostun sand, is estimated from the
1. The contractive, fully unstable and liquefied behaviour initial linear portion of the effective stress path, with a
B1 of tests 1–4 below the thick dashed red line in deviatoric stress increment of about 25 kPa, and the cor-
Fig. 7b, together with a true softening characteristic in responding average axial strain is about 0.023 %. Higher

123
Acta Geotechnica (2013) 8:293–309 301

(a) 300 (a) 300

Toyoura Undrained failure 10
12 Hostun RF
Continuous Dilatancy 9
11
250 B3
250 10
9
8
Deviatoric stress (kPa)

= 1.41 = 1.37 8
failure TP

Deviatoric stress (kPa)

10 7
200 200 7
Phase Transformation Line 6
9 6
8 CID
150 150
7 5
B2 5
6
5
100 4 100 4 4
3 = 0.58
IL
2
2
50 50 3
CID 1
1 3
B1 2
Virgin sample
0 1
0
0 50 100 150 200 0 50 100 150 200
Effective mean pressure (kPa) Effective mean pressure (kPa)

(b) 350
12
(b) 350
Continuous Dilatancy CC-C 10
300
11 300
9
Deviatoric stress (kPa)

250
250
10 Deviatoric stress (kPa)
200 8
9 200
7
150
8
150
6
100 7
100 5
6
50
5 4
50
1 2 3 4 3
0 2 1
0 5 10 15 0
Axial strain (%) 0 5 10 15
Axial strain (%)
(c) 1
1 2 3 4
B U
max
Liquefaction 5 (c) 1
max 2
0.8 6 CC-C 1 3
4
0.8
Normalized pore pressure

C 5
0.6 7 U
max
Phase Transformation
Normalized pore pressure

0.6 6
0.4 8
7
Virgin sample 0.4
0.2 9
8
A 10
0.2
0 9
O
11
Continuous Dilatancy D 0 10
-0.2
= 1.37
TP
= 1.41 12 -0.2
-0.4 failure Drained presheared levels Pseudo-elastic locus
0 0.5 1 1.5
Stress ratio -0.4
0 0.5 1 1.5
Stress ratio
Fig. 7 Results on compression presheared cycle, Toyoura sand. Fig. 8 Results on compression presheared cycle, Hostun RF sand.
a Effective stress paths. b Stress–strain relationships. c Mobilized a Effective stress paths. b Stress–strain relationships. c Mobilized
level level

123
302 Acta Geotechnica (2013) 8:293–309

Table 3 Characteristics of Hostun RF and Toyoura loose and pre- 25

sheared sands
Characteristics Hostun RF Toyoura
20

Deviatoric stress (kPa)

Initial gradient 8.0 ± 0.6 10.7 ± 1.0
Failure in compression (g, /) 1.59, 38°9 1.41, 34°9
15
Phase transformation (g, /) 1.42, 35°0 1.37, 33°9
Global collapse H
Instability line (g, /) 0.62, 16°3 0.58, 15°3 10
Poisson’s ratio 0.167 0.143
Mean Initial Gradient: 10.65
5

initial gradient means greater stiffness in the vertical

direction due to the drained compressive preshearing.
0
More importantly, it points out one of the possible man- 94 96 98 100 102 104 106
ifestations of the induced anisotropy created by drained Effective mean pressure (kPa)
past loadings.
Fig. 9 Initial gradient of effective stress paths of Toyoura loose sand
Within the simplified framework of transverse isotropic
elasticity proposed by Graham et al. [25], the stiffness
anisotropy, defined as the ratio Ev / Eh of the vertical over 300

the horizontal Young’s modulus, can be deduced from the Toyoura

Hostun
initial effective stress gradient of undrained triaxial test. 250
With the Poisson ratio determined from classical drained
Deviatoric stress (kPa)

test on virgin sample, in Table 3, Toyoura sand has a 200

stiffness anisotropy of 1.42 ± 0.05 instead of 1.50 ± 0.13
for Hostun sand. Apparently, the initial isotropic inter-
150
granular structure of Toyoura sand is more resilient to Phase Transformation Line
change imposed by a vertical compression of drained
100
preloading. This microstructure evolution can have some
experimental supports using advanced imaging techniques, CID
Frost et al. [22], or modern micro-tomography X-ray 50

imaging, Hall et al. [26]. Note that no attempt was made to

measure the usual elastic parameters with local strain 0
0 50 100 150 200
instrumentation, since the equipment is too heavy to be Effective mean pressure (kPa)
effective for loose sands.
The common effective stress responses (O2) of Toyoura Fig. 10 Common effective stress response of Hostun and Toyoura
(red line) and Hostun (blue line) sands, taken from Figs. 7a loose sand
and 8a, shown in Fig. 10, can be practically superimposed,
despite their highly nonlinear characteristics and different SSD characteristics in Fig. 8b. The physical significance of
failure angles. Therefore, isotropic hyperelasticity [38], the hypothetical behaviour represented by the red line will
resulting in an explicit analytical equation [21], can apply be examined later in Sect. 4.4.
to both sands using slightly different parameters to assess The details in Fig. 7c concerning the unstable and fully
slightly different curvatures. However, as noticed before, liquefied behaviour strongly support the progressive
the local extremum DU small
max at small strains in Fig. 5 and appearance of the dilatancy (O3). The global extremum
the following continuous decrease in pore pressure indicate DU max at large strains exists only for temporarily unstable
a dilative behaviour within this common nonlinear behaviour and beyond. Note that at both lower end below
response. There is no way various simple models of elas- test T4 and higher end above test T11, the phase trans-
ticity can explain this intriguing observation on loose sand. formation state cannot be defined.
Toyoura sand experiences a very unstable behaviour The evolution of the undrained behaviour (O4) is also
resulting in a total collapse with spontaneous liquefaction extended especially under large drained preloadings. The
for small preloading levels below the thick dashed red line fully contractive and liquefied sand in the lower end is
in Fig. 7b. In contrast, for similar relative density range, progressively transformed into fully dilative and non-liq-
Hostun sand has a stronger stress–strain behaviour with uefied in the higher end by drained preloading. The largest

123
Acta Geotechnica (2013) 8:293–309 303

compression presheared test T-C12 in Fig. 7c of Toyoura qpeak ¼ a  expðb  gmob Þ ð1Þ
loose sand has a continuous dilatancy behaviour with The linear preloading stress paths with fixed direction
negative excess pore pressure of a dense-like sand. This from an initially isotropic stress state of this paper can be
behaviour is totally missing for Hostun RF sand, despite a represented with only one parameter, the past deviatoric
very large mobilized preshearing level with gmob = 1.36, or strain history edev . Fig. 12 also gives a unique evolution of
86 % of undrained failure of test H-C10 in Fig. 8c. qpeak of each sand as a function of deviatoric strain history,
The onset of the flow liquefaction, represented by the regardless of the length of preloading stress paths. There-
undrained strength qpeak, when it exists, is an important fore, it supports the underlying idea of dominated strain-
feature of presheared and loose sand and plotted in Fig. 11 induced anisotropy, instead of stress-induced anisotropy
as function of gmob. Surprisingly, both sands share the [15].
same type of simple exponential equation between qpeak
and gmob with excellent coefficient of correlation, irre- 4.2 Dynamic local and global collapses
spective of the length of the drained preloading stress path.
Practically, nearly the same coefficients a and b are As noticed above, all tests in the unstable series B1 of
obtained for both sands, and a represents the usual history- Fig. 4, being drained presheared up to gmob = 0.95 or 67 %
less undrained strength of virgin sample : of failure, experience an instantaneous collapse in less than
2 seconds at large strains, hence terminating the test pre-
300 maturely, as shown in Fig. 4b. As an example, Fig. 13
Hostun RF shows a close-up view of typical collapse of test C2. The
Toyoura
250
excess pore pressure jumps suddenly to the initial effective
Deviatoric stress peak (kPa)

confining pressure, resulting in a total liquefaction with no

residual undrained strength. It indicates the possibility of a
200
very high axial strain rate occurring in this collapse regime.
This high strain rate of about 96 % per minute for virgin
150
sample, although not accurately measured, contrasts shar-
ply with the constant imposed axial strain rate of 0.28 %
100 per minute in this study. The usual quasi-static regime of
constant axial strain rate differs strikingly from the very
50 short dynamic one of instantaneous collapse.
From the test results, it seems that the axial strain at the
0 onset of global collapse depends on gmob and it can be very
0 0.5 1 1.5
large, more than 7 % for tests C3 and C4. Detailed study of
Mobilized frictional level mob
Fig. 4 reveals additional features. Small drops of deviatoric
Fig. 11 Undrained strength of compression presheared loose sands stress accompany the continuous decrease in q from qpeak,
and simultaneously small jump of pore pressure occurs.

1.5
25 0.95
T-CC-2
peak

Hostun RF
Toyoura Deviatoric drop
Effective stress ratio at peak

Normalized pore pressure

20 0.925
Deviatoric stress (kPa)

1 0.9
15

Acceleration
Pore pressure jump
y = 0.9023 + 0.31496log(x) R= 0.99032
10 0.875
y = 1.056 + 0.25696log(x) R= 0.93185

0.5 5 0.85
0 5 10 15 20 25 2 2.2 2.4 2.6 2.8 3
Deviatoris strain history % Axial strain (%)

Fig. 12 Effects of deviatoric strain history on the stress ratio at peak Fig. 13 Deviatoric drop and pore pressure jump of dynamic collapse

123
304 Acta Geotechnica (2013) 8:293–309

1 0.5 the pore pressure jump and the axial strain at local collapse
Normal test T-CC-C9 of each test of this unstable series. At least one local col-
Deviatoric resolution = 0.5 kPa
0.5 0.4 lapse is observed after the peak and before the final

Excess pore pressure (kPa)

dynamic global collapse. However, no sudden increase in
Deviatoric stress (kPa)

0 0.3 axial strain at local collapse is noticed in these strain-

controlled experiments, contrasting the case of undrained
-0.5 0.2 local collapse usually performed on stress-controlled
machine [12, 23, 41].
Pore pressure resolution = 0.05 kPa Figure 13 also notices a faster rate of deviatoric stress
-1 0.1
drop and an acceleration of excess pore pressure generation
-1.5 0
just before the final collapse. However, this accelerated
phase is not always measurable for all collapse samples,
due to the limitation of the current data acquisition system
-2 -0.1
6 8 10 12 14 for quasi-static behaviour. Despite these clues, the trig-
Axial strain (%) gering mechanisms concerning the deviatoric drop and the
final sudden collapse remain unknown.
Fig. 14 Resolution of deviatoric stress and pore pressure measurement
The upper bound of global collapse behaviour has a void
ratio ec = 0.935 approaching the previously known value of
0.930, Verdugo [51], Yoshimine et al. [55]. These studies
Table 4 Local and global collapse behaviour already noticed the dynamic collapse behaviour of Toyoura
very loose and moist-tamped sand under strain-controlled
Tests N Dqdrop kPa Dujump kPa ea % gcoll DUcoll
regime with ec larger than 0.930, irrespective of the confining
Local and global collapse, B1 pressure, up to 1 MPa. However, no forerunning indicators
Virgin 3 1.97 0.49 1.49 0.821 0.942 concerning the local collapses were studied. This upper
T-CC-C1 1 0.87 0.58 1.33 0.805 0.946 bound of global collapse behaviour seems to remain valid for
T-CC-C2 3 1.10 0.62 2.08 0.985 0.940 loose sand-fines mixtures, Mizanur Rahman et al. [43].
T-CC-C3 1 1.23 0.47 3.65 1.034 0.938 In Table 4, at least one local collapse also appears in the
T-CC-C4 2 1.65 0.55 7.37 1.152 0.935 lower mobilized zone of temporarily unstable behaviour,
Local collapse, B2 below test C8. It seems that the local collapse is linked to
T-CC-C5 1 5.69 2.33 0.65 – – the steady state of deformation and to the particle size
T-CC-C6 1 7.12 2.00 1.06 – – distribution for Toyoura sand. A close look at Fig. 8
T-CC-C7 1 8.17 1.59 1.06 – – reveals the total absence of local collapse for loose Hostun
RF coarse-grained sand under the same confining pressure
of 100 kPa.
Fig. 13 shows the typical peculiar feature of test C2,
termed as local collapse, occurring at a axial strain of 4.3 Instability cone
2.077 % with a deviatoric drop Dq of 1.10 kPa accompa-
nied with a pore pressure jump DU of 0.62 kPa. Another Within the framework of rate-independent and incremental
accident affecting both Dq and DU is visible at 2.329 % of nonlinear constitutive relation, the existence of instability
axial strain. The magnitude of Dq (DU) is significantly cones or cones of unstable stress directions was predicted.
larger than the actual fluctuation measured of 0.5 (0.05) If the strain increment de can be expressed as function of
kPa of the data acquisition system in Fig. 14 over a large stress increment dr, such as de ¼ Mdr, M is the incre-
range of axial strain. These events are termed as local mental stiffness, then d2W = dr. M. dr. The positiveness
collapses; they are similar to the slip regime without the of the Hill’s second-order work criterion d2W is then
subsequent stick counterpart, [1]. equivalent to that of det M. Since M depends only on the
Moist-tamped sand specimens tend to have more irreg- stress increment direction adr ¼ dr=kdrk for a fixed stress
ular microstructures with the presence of large pores, state, the directional dependency of d2W with adr is
suggesting fragile structures, as indicated by direct obtained [11].
microscopic observations [5]. These local collapses seem Experimentally, for a typical undrained triaxial test on
to indicate a chain of successive rearrangements of the loose and presheared sand, d2W vanishes at two particular
intergranular microstructure. points shown schematically in Fig. 15a. The first point is
The number of local collapses is given in Table 4, the usual deviatoric stress peak qpeak or the first local
together with the magnitude of the largest deviatoric drop, maximum of q, and the second point the first local

123
Acta Geotechnica (2013) 8:293–309 305

(a) (a) 1.5

Toyoura
Instability range
q q T-C1
min max 1

Second-order work, 10 kPa

-1
0.5
2
Deviatoric stress

d W > 0 Stable state

failure PT peak

Mohr-Coulomb criterion T-C9

2
-0.5
d W < 0 Potentially unstable state Collapse
Deviatoric peak,

-1

T-C5
2 -1.5
d W > 0 Stable state
160 180 200 220 240 260
Effective mean pressure Stress increment direction, degree

(b) (b) 1.5

Hostun H-C1 Instability range
q 1
max

Second-order work, 10 kPa

q H-C5
-1
min
0.5
Deviatoric stress

0
2
d W<0
2
d W>0 H-C8
-0.5
Deviatoric peak,
Potentially unstable Stable state
-1

2 -1.5
d W > 0 Stable state 160 180 200 220 240 260
Stress increment direction, degree
Axial strain
Fig. 16 Second-order work for a Toyoura, b Hostun RF loose sand
Fig. 15 Second-order work criterion in a effective stress response,
b stress–strain behaviour

minimum qmin around the phase transformation state. peak of the second-order work occurs at adr = 195°8, well
Between these two points, d2W has negative values, indi- before qpeak and d2W steadily declines from this point
cating a potentially unstable state. Three non-overlapped (dotted down arrow). The sign of d2W changes from
regions separated by two lines gpeak and gPT can be dis- positive to negative value at qpeak, when adr = 215°3 for
tinguished in the q  p0 plane of this figure. The first region the first time. It can change again from negative to positive
of stable state spans from the positive p0 axis to gpeak, around the phase transformation point, when adr returns to
passing through qpeak; the second region of potentially 215°3 (dotted up arrow). Returning to a same point means
unstable state strictly inside the Mohr-Coulomb failure a loop of d2W in the negative side accompanied by a
criterion from gpeak to the phase transformation line gPT negative peak at adr = 242°2. The collapse occurs shortly
around qmin; and the third stable region from gPT to the when adr = 243°, after this negative peak, obliterating the
Mohr-Coulomb failure line gItalic. These regions are also possibility of reaching the phase transformation point. The
separated in the stress–strain plane in Fig. 15b. triggering mechanism responsible for this collapse is
For undrained triaxial test, the vanishing of d2W means unknown, although some details are given in Sect. 4.2.
that of deviatoric stress increment dq = 0 or dr1 = dr2 = dr3. In the case of test T-C9, a representative case with small
In other words, the direction of stress increment is parallel to temporarily unstable behaviour created by large preshear-
the negative p0 axis or adr = 215°3 in the stress plane ing (dotted red line), Fig. 16a indicates a flatter loop of
pffiffiffi
(r1 ; r3  2), measuring from the positive radial stress axis. d2W in the negative side, compared to that of collapse
Figure 16a shows the evolution of d2W as function of behaviour of T-C1. Intermediate behaviour of T-C5 shows
adr, in the case of Toyoura sand of test T-C1. The positive a full loop in the negative side of d2W with a return to the

123
306 Acta Geotechnica (2013) 8:293–309

5
Hostun H-C1 Instability range
q
peak

Max
Second-order work, 10 kPa

0
-1

Presheared

Axial stress
Deviatoric peak, Cones of unstable stress directions
-5

q Virgin sample
peak

-10

Isotropic axis
Min

-15
50 100 150 200 250
Stress increment direction, degree Radial stress * 2

Fig. 17 Second-order work of temporarily unstable behaviour in Fig. 18 Cone of unstable stress directions for Hostun RF loose sand
undrained triaxial test for Hostun RF sand

Table 5 Stress increment angle for potentially unstable range for

second vanishing point. The loop eventually flattens with Toyoura and Hostun RF sands
increasing size of recent history represented by gmob.
Sand adr From° adr To°
Smaller loop means less energy, which is expected within
the common domain created by recent history. This figure Hostun RF 216°3 ± 0°8 240°7 ± 4°5
identifies the range of adr in which d2W has negative sign Toyoura 216°5 ± 1°4 247°0 ± 3°5
for each test in this study. Globally and surprisingly, the
potentially unstable domain ranges from 216°5 to 247°0 for
Toyoura loose sand, between two vertical down arrows, triaxial test, only in the range adr of 75°0–250°0 for both
irrespective of the stress state at qpeak created by the studied sands, and not all stress states in the stress Rendulic
mobilized level gmob in compression. Consequently, it plane, only the undrained stress peaks situated above the
seems that Toyoura loose sand has at least one continuous instability line of virgin sample. Finally, only one insta-
experimental instability cone in triaxial compression in the bility cone is experimentally detected in this paper for two
third quadrant of the Rendulic stress plane. The depen- sands. The search of the second theoretically instability
dency of the instability cone on the stress state, suggested cone in the first quadrant, predicted in [11] using incre-
by theoretical works [10, 11], is not confirmed in the mental nonlinear constitutive equations, is still an open
present experimental study. question.
The same analysis can apply for Hostun RF sand of a It seems that a unique upper value of the potentially
previous study [16] in Fig. 16b showing similar results unstable domain is obtained for both sands, regardless of
with no dynamic global collapse behaviour. Its instability the position of qpeak in the stress plane. This experimental
domain ranges from 215°9 to 238°4 for test H-C1 with a directional dependency of d2W solely on adr, limited to
full loop of d2W in the negative side and a second van- isochoric conditions in Fig. 18, with two representative
ishing point of d2W near the PT point. Its instability cone instability cones of virgin and presheared sand supports the
has a slightly smaller range, only from 216°3 to 240°7, also theoretical instability cone on incremental stress directions
in the third quadrant. of Darve et al. [11]. Table 5 shows the probable experi-
Figure 17 shows a typical and complete evolution of mental instability cone for Hostun and Toyoura sands in
d2W as a function of adr for temporarily unstable behav- triaxial compression.
iour with partial liquefaction, showing clearly a maximum Figure 19 gives the boundary of the unstable domain in
d2Wmax (respectively minimum d2Wmin) well before qpeak the stress plane, delimited by the instability line of virgin
(respectively the PT point). These extrema suggest clearly sand and the phase transformation line. The full circles
that the tendency of unstable behaviour commences at inside this domain indicate the stress peaks of all undrained
d2Wmax, before plunging toward the zone of negative value tests in compression for Toyoura loose sand in this paper
of d2W, and the recovering of stable behaviour begins at with nearly the same instability cone. Recall that only one
d2Wmin, before attaining the PT state. Note that not all confining pressure of 100 kPa is considered for all tests.
stress increment directions are accessible from undrained Since drained compression preshearing cannot allow the

123
Acta Geotechnica (2013) 8:293–309 307

350 the transition of d2W from positive value to zero occurs for
Toyoura Compression Failure
the classical critical state.
300 In the geotechnical literature, the SSD is uniquely
Phase Transformation determined by the void ratio. However, in our case, it is
250
impossible to find a void ratio or a mobilized level gen-
Axial stress

Unstable erating this hypothetical behaviour with straight line AB. It

200
Domain can be narrowed down to ec & 0.934–0.936 and gmob &
150
0.95–0.97 of gfail, between test T-CC-C4 and T-CC-C5,
since a set of additional tests were conducted within this
Instability Line narrow zone without finding it. In other words, can we
100
create a loose and presheared sample experiencing instan-
50 Isotropic axis taneous collapse at very large strains, beyond 15 %, and a
very small residual deviatoric stress? It is probably difficult
0 to find this hypothetical sample due to the variability of the
0 50 100 150
drained behaviour, in terms of stress–strain response.
Radial stress * 2
For Hostun loose RF sand in a previous study, it seems
Fig. 19 Unstable domain and Instability cones for Toyoura loose that this hypothetical behaviour does not exist, with the
sand absence of full static liquefaction and collapse at 100 kPa
in Fig. 8c. However, the steady state is well observed with
virgin or small preshearings. Moist-tamped technique was
undrained stress peak below the IL line, additional tests unable to create very loose Hostun sand samples with full
with different drained preshearing strategies are required to liquefied behaviour at this confining pressure.
cover the whole compression side in future studies. The small range of ec or gmob and the impossibility to
have a realistic experimental test following the straight
4.4 Bifurcation characteristic thick red line AB in Fig. 7c suggests a saddle bifurcation
phenomenon [47] between the liquefied, contractive,
The full static liquefaction of Fig. 4 and the non-lique- unstable, softening behaviour and the non-liquefied, dila-
faction of Fig. 5 strongly suggest the existence of a par- tive, stable, hardening behaviour. These bifurcation char-
ticular behaviour bridging the gap between these two acteristics could potentially help explain qualitatively the
extreme behaviours. It separates the contractive and absence of very small residual strength at large strains in
unstable behaviour of loose-like sand and the dilative and the stress–strain relationship of very loose Toyoura and
stable behaviour of dense-like sand. Hostun RF sands in the geotechnical literature.
The hypothetical behaviour, shown in thick red curve in Microscopic discrete element modelling can be very
Fig. 7, should have a nearly liquefied behaviour with helpful to explore more thoroughly some subtle aspects of
effective stress path plummeting toward the origin in granular media. Indeed, this approach would allow us to
Fig. 7a, a gradual drop of q from qpeak to a constant and lead the analysis at the grain scale and should help to better
small residual strength below 5 kPa in the stress–strain understand phenomena often incompletely explained by
relation of Fig. 7b, and a continuous monotonic increase in real experiments like the instability cone and the bifurca-
excess pore pressure to a plateau at large strain. This tion characteristic. Currently, very few macroscopic con-
hypothetical behaviour can be best emphasized in Fig. 7c, stitutive modellings are able to take into account all these
where it represents the frontier between the liquefied subtle aspects. Nevertheless, in the long term, these mod-
behaviour with gpeak and the non-liquefied behaviour ellings should allow the identification of the hypothetical
without peak. In this plane, it should follow the common behaviour for example.
response up to point A and then a straight line AB sepa-
rating the liquefied domain C on the left and the non-liq-
uefied domain D on the right. It represents the onset of the 5 Conclusions
dilatancy phenomenon.
Figure 7 indicates a hypothetical behaviour with con- Laboratory undrained triaxial compression tests were pre-
stant shear stress at constant volume and constant axial formed to study the strain-induced anisotropy of loose and
strain rate. It resembles closely to the classical steady state saturated Toyoura and Hostun RF sands. Past deviatoric
of deformation (SSD) defined by Castro and Poulos [9] for strain histories were created by standard triaxial drained
sand with one additional detail : the second-order work presheared cycle in compression from an initial isotropic
transits from negative value to zero for SSD state, whereas stress state and tested on samples, prepared by moist

123
308 Acta Geotechnica (2013) 8:293–309

tamping technique, having a narrowed range of loose den- 8. Castro G (1969) Liquefaction of sands. Harvard soil mechanics
sity prior to undrained shearing. The main focus of this series. Harvard University, Cambridge
9. Castro G, Poulos SJ (1977) Factors affecting liquefaction and
paper was to shed some new lights on the deviatoric strain- cyclic mobility. J Geotech Eng ASCE 103(6):501–506
induced anisotropy of sands. The progressive transforma- 10. Daouadji A, Darve F, Al Gali H, Hicher P-Y, Laouafa F, Lignon
tion of contractive and unstable behaviour of loose sand into S, Nicot F, Nova R, Pinheiro M, Prunier F, Sibille L, Wan R
dilative and stable behaviour of dense-like sand of previous (2011) Diffuse failure in geomaterials : Experiments, theory and
modelling. Int J Numer Anal Meth Geomech 35(16):1731–1773
study was validated and extended in two directions: 11. Darve F, Servant G, Laouafa F, Khoa HDV (2004) Failure in
geomaterials : continuous and discrete analyses. Comput Meth-
1. Toward the lower end of fully unstable, contractive,
ods Appl Mech Eng 193(27):3057–3085
softening and liquefied behaviour associated with 12. Di Prisco C, Imposimato S (1997) Experimental analysis and
catastrophic collapse feature of loose sand. theoretical interpretation of triaxial load controlled loose sand
2. Toward the higher end of fully stable, dilative, specimen collapses. Mech Cohes-Frict Mater 2(1):93–120
13. Di Prisco C, Matiotti R, Nova R (1995) Theoretical investigation
hardening and non-liquefied behaviour associated with
of the undrained stability of shallow submerged slopes. Geo-
continuous dilatancy of dense-like sand, while being technique 45(3):479–496
relatively loose. 14. Di Prisco C, Nova R, Lanier J (1993) A mixed isotropic-kine-
matic hardening constitutive law for sand. In: Kolymbas D (ed)
These two additional experimental observations extend Modern approaches to plasticity. Balkema, Elsevier Science
the current knowledge concerning the effects of recent Publishers, Rotterdam, pp 83–124
deviatoric strain histories on the undrained behaviour of 15. Doanh T, Dubujet Ph, Touron T (2010) Exploring the undrained
induced anisotropy of Hostun RF loose sand. Acta Geotech
granular materials. They represent a step forward in the 5(4):239–256
efforts to understand some subtleties of the strain-induced 16. Doanh T, Finge Z, Boucq S (2012) Effects of previous deviatoric
anisotropy. Additionally, the existence of the instability strain histories on the undrained behaviour of Hostun RF loose
cone and its directional dependency on stress increment sand. Geotechnical and Geological Engineering 30(4):697–712
17. Doanh T, Ibraim E (2000) Minimum undrained strength of
direction adr, based on the second-order criterion, and the Hostun RF. Geotechnique 50(4):377–392
bifurcation characteristics of loose sand are strongly sup- 18. Doanh T, Ibraim E, Matiotti R (1997) Undrained instability of
ported by new experimental data in this study. very loose Hostun sand in triaxial compression and extension.
The observed behaviour of Toyoura sand proved much Part 1: experimental observations. Mech Cohes Mater 2(1):47–70
19. Dubujet Ph, Doanh T (1997) Undrained instability of very loose
more complex and subtle than we had anticipated and leave Hostun sand in triaxial compression and extension Part 2:
a logical question: Could Hostun RF also have these two Theoretical analysis using an elastoplasticity model. Mech
extreme behaviours? In other words, could these two Cohes-Frict Mater 2(1):71–92
extreme behaviours be the new characteristics of loose 20. Finge Z (2004) Contribution à l’tude du comportement non drain
des sables lâches surconsolids et anisotropes. Thèse de doctorat,
sand? New experiments are currently investigated to address Institut National des Sciences Appliques de Lyon
these new specific aspects of granular media in more detail. 21. Finge Z, Doanh T, Dubujet Ph (2006) Undrained anisotropy
of Hostun RF loose sand : new experimental hints. Canadian
Acknowledgements The authors would like to thank Mrs. M. C. Geotechnical Journal 43(11):1195–1212
Jean for reviewing the draft version. 22. Frost JD, Chen CC, Park JY, Jang DJ (1999) Quantitative char-
acterisation of microstructure evolution. In: Yamamuro Lade
(eds) Physics and mechanics of soil liquefaction. Balkema,
Rotterdam, pp 169–177
References 23. Gajo A (2004) The influence of system compliance on collapse of
triaxial sand samples. Can Geotech J 41(2):257–273
1. Adjemian F, Evesque P (2004) Experimental study of stick-slip 24. Gajo A, Piffer L (1999) The effects of preloading history on the
behaviour. Int J Numer Anal Meth Geomech 28(6):501–530 undrained behaviour of saturated loose sand. Soils and Founda-
2. Arthur JRF, Chua KS, Dunstan T (1977) Induced anisotropy in a tions 39(6):43–54
sand. Geotechnique 27(1):13–30 25. Graham J, Houlsby GT (1983) Anisotropic elasticity of a natural
3. Arthur JRF, Chua KS, Dunstan T, Rodriguez Del JI (1980) clay. Geotechnique 33(2):165–180
Principal stress rotation: a missing parameter. J Geotech Eng 26. Hall SA, Bornert M, Desrues J, Pannier Y, Lenoir N, Viggiani G,
ASCE 106(4):419–433 Bsuelle P (2010) Discrete and continuum analysis of localised
4. Arthur JRF, Menzies B (1972) Inherent anisotropy in a sand. deformation in sand using X-ray lCT and volumetric digital
Geotechnique 22(2):115–128 image correlation. Geotechnique 60(5):315–322
5. Benahmed N, Canou J, Dupla JC (2004) Structure initiale et 27. Hight DW, Gens A, Symes MJ (1983) The development of a new
proprits de liqufaction statique d’un sable. Comptes Rendus hollow cylinder for investigating the effects of principal stress
Mcanique 332(11):887–894 rotation in soils. Geotechnique 33(4):355–384
6. Bjerrum L, Krimgstad S, Kummeneje O (1961) The shear strength 28. Hill R (1958) A general theory of uniqueness and stability in
of a fine sand. In: Proceedings of 5th International Conference Soil elastic plastic solids. J Mech Phys Solids 6(3):236–249
Mechanics and Foundation Engineering. 1:29–37 29. Ibraim E (1998) Diffrents aspects du comportement des sables à
7. Canou J, Thorel L, De Laure E (1991) Influence d’un dviateur de partir des essais triaxiaux: des petites dformations à la liqufaction
contrainte initial sur les caractristiques de liqufaction statique du statique. Thèse de doctorat, Institut National des Sciences
sable. In: Proceedings of X ECSMFE, Firenze 1:49–52 Appliques de Lyon,

123
Acta Geotechnica (2013) 8:293–309 309

30. Ibsen LB (1999) The mechanism controlling static liquefaction 44. Poorooshasb HB, Holubec I, Sherbourne AN (1966) Yielding and
and cyclic strength of sand. In: Yamamuro Lade (eds) Physics and flow of sand in triaxial compression: part 1. Can Geotech J
mechanics of soil liquefaction. Balkema, Rotterdam, pp 29–39 3(4):179–190
31. Ishihara K (1996) Soil behaviour in earthquake geotechnics. 45. Poorooshasb HB, Holubec I, Sherbourne AN (1967) Yielding and
Oxford University Press, New York flow of sand in triaxial compression: part 2 and 3. Can Geotech J
32. Ishihara K, Okada S (1978) Effects of stress history on cyclic 4(4):376–397
behaviour of sand. Soils Found 18(4):31–45 46. Protière X (2009) Influence des chemins drain de prchargement
33. Ishihara K, Okada S (1978) Yielding of overconsolidated sand and sur le comportement non drain du sable de Toyoura. Master de
liquefaction model under cyclic stresses. Soils Found 18(1):57–72 recherche, ENSMSE - ENISE
34. Ishihara K, Okada S (1982) Effects of large preshearing on cyclic 47. Strogatz SH (1994) Nonlinear dynamics and chaos. Perseus Press,
behaviour of sand. Soils Found 22(3):109–125 Cambridge, Massachusetts
35. Konrad JM (1993) Undrained response of loosely compacted 48. Tatsuoka F, Ishihara K (1974) Yielding of sand in triaxial com-
sands during monotonic and cyclic compression tests. Geotech- pression. Soils Found 14(2):63–76
nique 43(1):69–89 49. Tatsuoka F, Sato T, Park CS, Kim YS, Mukabi JN, Kohata Y
36. Ladd RS (1978) Preparing test specimens using undercompac- (1994) Measurements of elastic properties of geomaterials in
tion. Geotech Test J 1(1):16–23 laboratory compression tests. Geotech Test J 17(1):80–94
37. Lade PV, Duncan JM (1973) Cubical triaxial tests on cohesion- 50. Vaid YP, Chung EKF, Kuerbis RH (1989) Preshearing and
less soil. J Soil Mech Found ASCE 99(10):793–812 undrained response of sands. Soils Found 29(4):49–61
38. Lade PV, Nelson RB (1987) Modelling of elastic behaviour 51. Verdugo R (1992) Characterization of sandy soil behaviour under
of granular materials. Int J Numer Anal Meth Geomech 11(5): large deformation. Phd thesis, University of Tokyo
521–542 52. Verdugo R, Ishihara K (1996) The steady state of sandy soils.
39. Lade PV, Nelson RB, Ito YM (1988) Instability of granular Soils Found 36(2):81–91
materials with nonassociated flow. J Eng Mech ASCE 114(12): 53. Wood DM (1994) General report: Evaluation of material prop-
2173–2191 erties. In: Miura S Shibuya S, Mitachi T (eds) In: Proceedings of
40. Lanier J, Di Prisco C, Nova R (1993) Etude exprimentale et Pre-failure deformation of geomaterials, Sapporo, Japan, volume
analyse thorique de l’anisotropie induite du sable Hostun. Revue 2. Balkema, Rotterdam, pp 1179–1199
Française de Gotechnique 57:59–74 54. Yamada S, Takamori T, Sato, K (2010) Effects on reliquefaction
41. Lindemberg J, Koning HL (1981) Critical density of sand. resistance produced by changes in anisotropy during liquefaction.
Geotechnique 31(2):231–245 Soils Found 50(1):9–25
42. Luong MP (1978) Etat caractéristique du sol. Comptes Rendus de 55. Yoshimine M, Ishihara K (1998) Flow potential of sand during
l’Académie des Sciences, série B (287):305–307 liquefaction. Soils Found 38(3):189–198
43. Mizanur Rahman Md, Lo SR, AbdulLahil Baki Md (2011)
Equivalent granular state parameter and undrained behaviour of
sandfines mixtures. Acta Geotech 6(4):183–194

123