See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/245410543

The critical state of sand

Article  in  Géotechnique · January 1991

DOI: 10.1680/geot.1991.41.3.365

CITATIONS READS

270 4,350

3 authors, including:

Mike Jefferies
Golder Associates, Vancouver
81 PUBLICATIONS   3,510 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

state parameter View project

All content following this page was uploaded by Mike Jefferies on 14 September 2016.

The user has requested enhancement of the downloaded file.

Been, K., Jefferies, M. G. & Hachey, J. (1991). GCotechnique 41, No. 3, 365-381
The critical state of sands
K. BEEN*, M. G. JEFFERIEST
There is conflicting evidence regarding the critical
state and steady state of sands. It seems that in
some cases the critical state and steady state are
taken to be equal and in others they differ signili-
cantly. The steady state is shown to be dependent
on stress path, initial fabric and initial density in
some cases and not in others. This Paper examines
a fine to medium, uniformly graded quartzitic
sand. The critical and steady states are shown to be
equal and independent of stress path, sample prep
aration method and initial density. Data suggest
that the critical state friction angle in sands may
be a function of the critical state void ratio. The
steady/critical state line is shown to be non-linear,
and to have a marked increase in slope at stress
levels greater than about 1000 kPa. The critical or
steady state point must be selected from triaxial
test data with care to ensure that pseudo-steady
states are not compared with true steady states.
Test procedures and the intrinsic properties of the
material being tested must be controlled well. The
overall scatter in the steady/critical state line for
the sand studied is &O-O1 in terms of void ratio at
stress levels less than 1000 kPa. For real sands
with variable intrinsic properties, the scatter may
be much larger.
KEYWORDS: fabric/structure of soils; liquefaction;
sands; triaxial tests.
INTRODUCTION
The critical void ratio of sands has been the
subject of much discussion since the pioneering
work of Casagrande (1936). A major step forward
was the development of critical state soil mech-
anics, starting with the classic paper ‘On the
yielding of soils’ by Roscoe, Schofield & Wroth
(1958). While many critical state concepts concen-
Discussion on this Paper closes 2 January 1992; for fur-
ther details see p. ii.
* Colder Associates Ltd, Calgary, Alberta.
t Klohn Leonoff Consultants, Richmond, British
Columbia.
365
and J. HACHEY*
I1 existe des indications contradictoires concernant
Mat critique et Mat stationnaire des sables. I1
parait que quelquefois Mat critique et Mat sta-
tionnaire sent pareils, tandis que dans d’autres cas
ces Ctats semblent differer de faGon significative.
Quelquefois on peut dCmontrer que Mat station-
naire d+end du chemin de contrainte, de la pr6-
paration initiate et de la densiti initiale, mais pas
dans d’autres cas. L’article considtire le cas d’un
sable fin/moyen quartzitique de classement uni-
forme dans lequel les Ctats critique et stationnaire
sont 6gaux et indkpendants du chemin de con-
trainte, de la mCthode de preparation de
Echantillon et de la densiti: initiale. Les don&s
cr&nt l’impression que I’angle de frottement
interne de Mat critique dans les sables peut &tre
une fonction de I’indice des vides dans cet &at. On
dCmontre que la ligne de IStat stationnaire/critique
est non-IinCaire et qu’elle subit un accroissement
marqui! de pente i des niveaux de contraintes
sup&ieures a 1000 kPa approximativement. II faut
sClectionner avec soin le point d’itat stationnaire/
critique obtenu i partir de don&es d’essais tri-
axiaux afin d’btre sirs que les etats
pseudo-stationnaires ne sent pas comparb avec de
vCritables &a& stationnaires. Les prockdures des
essais et les propri&s des mat&es test&es doivent
&tre extr&mement hien contr&Ses. La dispersion
totale dans la ligne d’btat stationnaire/critique pour
les sables ktudi& a une valeur de + OM exprim&
en indice des vides B des niveaux de contrainte in-
fbrieurs P 1000 kPa. La dispersion peut @tre
beaucoup plus grande pour de vrais sables avec des
propr%tb intrinkques variables.
trated on the behaviour of remoulded clays,
attention was also given to sands (e.g. Wroth &
Bassett, 1965; Stroud, 1971). Application of criti-
cal state soil mechanics to sands was less suc-
cessful, mainly because a virgin consolidation line
could not be defined and measurement of the
critical state line for sands was problematic.
With the development of modern laboratory
techniques, the measurement problems were
apparently resolved by Castro (1969), who used
undrained, stress-controlled triaxial tests on very
loose sands to obtain a steady state line.
However, there has been discussion as to whether
the critical and steady state lines are in fact the
366 BEEN, IEFFERIES AND HACHEY

same (Casagrande, 1975; Poulos, 1981; Sladen, very high stress levels in the sand, shows anom-
D’Hollander, Krahn & Mitchell, 1985; Alarcon- alies which can be explained by curvature of the
Guzman, Leonards & Chameau, 1988). critical/steady state line. Also, the accuracy with
Intellectually, for sands there appears to be which it is possible to determine the critical/
little distinction between the critical and steady steady state line has been brought into question
state and the differences appear to lie in the by Sladen & Handford (1987). They showed that
methods of measurement. Critical state workers if volumetric strain during back-pressure satura-
have generally relied on drained, strain-rate-con- tion of sands is not properly measured, large sys-
trolled tests on dilatant samples to determine the tematic errors can occur in void ratio
critical state. The steady state is measured in calculations, especially with silty sands.
undrained tests, usually on loose (contractive) The above discussion does not include the
samples. In this context, Casagrande (1975) equally important issue of potentially large
defined the S line based on drained tests, which changes in the critical/steady state line due to
can be considered the critical state line, and the F apparently small changes in the nature of the
line based on the Castro-type undrained stress- sand. In this Paper, the assumption is made that
controlled tests, which is the steady state line. each sand sample tested has the same intrinsic
The critical (or steady) state of sands is worthy properties.
of detailed appraisal as it provides a basis for Several basic questions about the critical (or
much engineering of sand fill structures. The steady) state of sands have not been satisfactorily
widely used liquefaction evaluation procedure answered, including the following.
developed by Poulos, Castro & France (1985) is
(a) From a practical standpoint, are the critical
based on the undrained strength at the steady
and steady state the same?
state. Design of hydraulic sand fills for artificial
(b) Is the critical/steady state unique for any par-
islands in the Beaufort Sea has also relied on
ticular sand, or does it depend on initial fabric
steady state strength concepts (e.g. Sladen et a[.,
and stress path of the test?
1985; Shinde, Crooks, James & Wilhams-
(c) What is the general shape of the critical/
Fitzpatrick, 1986; Jefferies, Rogers, Stewart,
steady state line over a wide range of stresses?
Shinde, James & Williams-Fitzpatrick, 1988).
(d) How precisely and repeatably can the critical/
A further reason to explore the critical (or
steady state be measured?
steady) state in detail is that Been & Jefferies
(1985) proposed the state parameter Y as a semi- The purpose of this Paper is to explore the
empirical normalizing parameter for sand behav- above issuesin some detail, using an extensive set
iour. This concept requires knowledge of the of test data for sand from the Erksak borrow area
critical/steady state line, which provides a refer- in the Beaufort Sea. These data have been
ence state from which the state parameter and the obtained by the Authors as part of engineering
most important sand behaviours are derived. projects over three years. Additional tests were
Uniqueness of the critical/steady state line is carried out on Toyoura and Leighton Buzzard
therefore required for this and many engineering sand to supplement these data. It is recognized
approaches. Kuerbis, Negussey & Vaid (1988) that the information presented is not sufficient to
and Vaid, Chung & Kuerbis (1990) showed test provide unquestionable conclusions, but the
data and stated that testing in extension resulted objectives will be satisfied if the important issues
in a different steady state from compression are clarified and identified for discussion.
testing. In addition, Alarcon-Guzman et al. (1988)
made a case that the particle matrix structure and
test type affect whether the sand will reach the S TERMINOLOGY
or the F line. Both of the above-referenced papers To avoid confusion in the terminology used in
imply non-uniqueness of the critical/steady state. this Paper, a brief discussion and background to
Clearly, if the chosen reference state is not unique, the terminology are given.
the definitions of critical and steady states must The state of a sand is the description of the
be re-evaluated and the application of these con- physical conditions under which it exists. Void
cepts becomes more difficult. ratio (or density) and stress are the primary state
The critical/steady state line was assumed to be variables for soils. Fabric is also a state variable
straight in previous work by Been & Jefferies of importance, while temperature, for example, is
(1985), but evidence has since accumulated that of little importance for sands.
the line is curved. Friction angles and dilation Intrinsic material properties of a sand can be
rates at stress levels beyond the range of previous uniquely defined and are independent of the state
good-quality steady state data deviated from the of the sand. Examples are grain size distribution,
expected trends. In addition, interpretation of the mineralogy, grain shape, mineral to mineral fric-
cone penetration test, which generally induces tion angle 4, and specific gravity.
 CRITICAL STATE OF SANDS 367

Behavioural properties of a sand are measured

in specific tests, and will depend on the test type,
initial state and intrinsic properties of the sand.
Behavioural properties include peak friction
angle, dilation rate, pore pressures at failure,
undrained strength and shear modulus.
Constitutive parameters may be intrinsic
behavioural properties, but they require a state-
ment regarding the assumed constitutive model
for the sand. Shear modulus and Poisson’s ratio
are examples for an isotropic linear elastic model.
Fabric is the term used to describe the arrange-
ment of sand grains on a particulate scale. This
would include a description of particle contacts,
their orientations and distribution, and cementa-
tion at the contacts, if any.
Anisotropy of a sand is taken in this Paper to
be an expression of fabric anisotropy in which
there will be unequal particle contacts and forces
between particles in different directions. Ani-
sotropy may be present initially in a sand, and
may be changed or induced by stress or strain
changes.
The steady state of a sand is defined by Poulos
(1981) as follows.
The steady state of deformation for any mass of
particles is that state in which the mass is con-
tinuously deforming at constant volume, con-
stant normal effective stress, constant shear
stress, and constant velocity. The steady state
of deformation is achieved only after all par-
ticle orientation has reached a statistically
steady state condition and after all particle
breakage, if any, is complete, so that the shear
stress needed to continue deformation and the
velocity of deformation remain constant.
This definition of the steady state does not state
which velocity should be constant or what the
value of velocity is.
The steady state line is the locus of steady state
points in void ratio/stress space. For convenience,
the stress invariants p’ = (a,’ + 02’ + a,‘)/3 and
q = o1 - o3 are used to describe the stress state
of a sand following the logic of Wood (1984), and
the projection of the steady state line on to the
e-p’ plane is used for data presentation and defi-
nition of the state parameter Y.
The state parameter Y defines the void ratio (e)
and mean stress level (p’) of a sand relative to a
reference state (the steady state) as shown in
Fig. 1 (Been & Jefferies, 1985). Additional param-
eters describing the fabric and stress anisotropy
are necessary to define the state of a sand
uniquely, but in practice these parameters are
generally unknown.
The critical state has been defined as the state
at which the soil ‘continues to deform at constant
stress and constant void ratio’ (Roscoe et al.,
or
LOcIi P
Fig. 1. Definition of state parameter (after Been &
Jefferies, 1985)
1958). The steady state has traditionally been
measured using undrained tests on loose sand
samples, while the critical state is generally
inferred from drained tests on dense sands.
Dilation rate is defined as the ratio of the volu-
metric strain rate to the axial strain rate in a
drained triaxial test. Dilation rate is assumed to
be positive for expansion of the soil. In undrained
tests the volumetric strain is zero, and dilation
and contraction are used in a loose manner to
describe negative and positive rates of pore press-
ure change. It is emphasized that dilation is a rate
of change of volumetric strain, not the instantane-
ous value of the volumetric strain.
Membrane penetration is the phenomenon
where the pressure difference across the mem-
brane around the sample causes it to penetrate
partially the voids between the particles. A mem-
brane penetration correction is commonly
applied to measured volume changes when the
pressure difference changes. However, this correc-
tion cannot be applied to the measured pore
pressure response in undrained tests.
LABORATORY TESTING
Test programme
Much of the discussion in this Paper is based
on a series of 56 triaxial tests carried out on
Erksak 330/0.7 sand. The index properties of this
Erksak sand are presented in Table 1. It is a uni-
formly graded, medium- to fine-grain sand (Fig.
2), dredged from the Erksak borrow area in the
physiographic region of the Beaufort Sea known
as the Tingmiark Plain. The particles are mainly
subrounded as shown in Fig. 3. Based on a semi-
quantitative X-ray diffractometry analysis, the
mineralogy of the material is mainly quartz (73%)
368 BEEN, JEFFERIES AND HACHEY

Table 1. Index properties of sands tested

Erksak Leighton
33010.7 Toyoura Buzzard

Mineralogy
Quartz: % 73 75
Feldspar : % 22 25
Other: % 5 0
Median grain size D,,: mm 0.330 0.160 0.120
Effective grain size D,,: mm 0.190 0.120 0.095
Uniformity coefficient D,,/D,,, l-8 1.5 1.5
Passing 200 sieve: % 0.7 0 5
Specific gravity 2.66 2-65 2.65
Minimum density: kg/m3 1517 1338 1310
Maximum density: kg/m3 1742 1648 1592

* X-ray diffractometry was not carried out on Leighton Buzzard sand in

this study. The sand consists mainly of quartz, by visual inspection.

with some K-feldspar (7%) and plagioclase (15%) samples. Lubricated end plates were used to mini-
and other minerals. mize platen restraint.
Table 2 is a summary of the test programme on Samples were prepared in membrane-lined split
Erksak sand. Test conditions were varied con- moulds mounted on the lower platen of the
siderably and included drained (16) and apparatus. Moist compacted (MC) samples were
undrained (40) conditions, moist compacted (39) tamped into the mould at the desired void ratio
and wet pluviated (17) sample preparation, exten- in six equal layers. The material forming each
sion (8) and compression (48) stress paths, stress- layer was first dried and weighed, and about 5%
controlled (7) and strain-controlled (49) loading water was added. Wet pluviated (PL) samples
and a wide range of initial void ratio and stress were formed by pluviation from a long-necked
states. All tests were carried out on approxi- flask into the mould, which was initially filled
mately 152 mm high by 76 mm dia. cyclindrical with water. The mould was gently tapped to
densify the sand to the required void ratio. The
main objective of using different preparation
SlE of sew opemng methods was to produce different initial fabrics,
mm ,rm as this results in different stress-strain curves. The
samples were saturated by flushing with de-aired
.25630315
water and by increasing the back pressure until a
B value greater than 0.97 was obtained. During
this process an effective stress of approximately
20 kPa was maintained on the sample. In some
cases, high back pressures were used to avoid
cavitation of the pore fluid during the test due to
large negative pore pressures.
$ 60 After saturation, samples were isotropically
E consolidated, with effective stress increments
E
8 approximately doubling in each consolidation
$ step. Samples were tested within an hour of con-
a 40
solidation, so that secondary compression effects
can be neglected. Table 2 indicates whether the
shearing was load or strain rate controlled. Data
were collected and processed using micro-
computer logging systems.

Gran s,ze: mm
Void ratio calculation
Provided that good laboratory practice is fol-
Med. Fine oarse~ Med. 1 Fine
I _ lowed, the tests are straightforward, except for the
calculation of sample void ratio. Sladen & Hand-
ford (1987) have suggested that potential errors in
Fig. 2. Grain size distribution of Erksak 330/@7 sand void ratio calculation can be large. The main
Table 2. Summary of testing programme

Initial conditions Test conditions1 End of test Remarks

T-
Test Preparation* Void pr: Drainage Strain Stress Steady rJ3’: p’: 4: eII
ratio kPa rate : path state@ kPa kPa kPa
%/hr

L-601 MC 0.757 499 0.025 U L Yes 64 100 108 0.754 27.2 Load controlled
L-602 MC 0.712 500 - 0.020 U L Dil 285 460 500 0.711 test series
L-603 MC 0.787 300 0.048 U L Yes 4 6 5 0.780 22.6
L-604 MC 0.772 699 0.044 U L Yes 55 72 50 0.768 18.2
L-605 MC 0.771 500 0.039 U L Yes 16 19 8 0.766
L-606 MC 0.763 701 0.035 U L Yes 52 74 65 0,759 22.6
L-607 MC 0.751 701 0.023 U L Yes 170 260 270 0,748 26.3

C-641a PL 0.666 140 - 0.084 U 4 C Cav 560 1060 1500 0.668 Test series with
C-609 MC 0.800 500 0.068 U 54 C Yes 8 10 6 0.793 15.8 strain control
C-610 MC 0.754 1200 U 53 C Yes 220 350 400 0.751 28.4
C-61 1 MC 0.738 1450 U 51 C Mdil 660 1107 1340 0.737
C-612 MC 0.773 500 0.041 U 51 C Yes 40 47 22 0.769 12.5
C-613 MC 0.740 1300 U 49 C Mdil 340 540 601 0.737
C-614 MC 0.740 1200 U 51 C Yes 323 520 595 0.738 28.6
C-616 MC 0.703 1200 U 48 C Dil 970 1664 203 0.703

C-620 MC O-616 8000 U 49 C Yes 1500 2350 3OGO 0.613 Test series in
C-621 MC 0.618 8000 U 47 C Yes 1850 3250 4500 0,616 high pressure
C-622 MC 0.659 8000 U 47 C Yes 1300 2000 2400 0,656 apparatus
C-623 MC 0.662 8100 U 49 C Yes 1146 2207 3184 0,659
C-624 MC 0.745 3000 U 47 C Yes 256 460 613 0.741 33.0
C-625 MC 0.703 7000 U 48 C Yes 910 1570 1960 0.700 31.2
C-626 MC 0.751 4000 U 48 C Yes 401 708 920 0.748 32.3

C-63 1 MC 0.694 200 PO.051 U 9 C Dil 424 826 1205 0.695 Examination of
C-632 PL 0.652 200 - 0.092 U 10 C Yes 860 1650 3022 0.655 sample preparation
C-633 MC 0.655 200 - 0,089 U 10 C Dil 387 737 1064 0.656 effects and dense
C-634 PL 0.667 200 - 0.077 U 10 C Yes 1240 2137 2692 0.670 states
C-635 MC 0.588 200 -0.156 U 10 C Yes 1048 2351 3900 0.591
C-636 PL 0.618 200 -0.126 U 9 C Yes 1529 2767 3714 0,621
C-637 MC 0.580 50 -0.183 U 10 C Yes 895 2232 4011 0,584
C-639 MC 0.596 800 -0.130 U 10 C Yes 1323 2768 4335 0.597
C-641b PL 0.687 200 - 0.058 U 10 C Yes 808 1476 2010 0.689
c-642 PL 0.566 800 -0.160 U 10 C Yes 1500 2799 3897 0.567
E-641~ MC 0.732 500 0 U 10 E Mdil 39 90 75 0.728 29.4 Extension test
E-642 MC 0.767 500 0.035 U 14 E Mdil 55 100 70 0.764 22.9 series (unloading)
E-643 MC 0.747 500 0.015 U 10 E Mdil 111 207 144 0.745 23.2
E-644 MC 0.783 so0 0.05 1 U 10 E Yes 11.5 13 4.5 0.777
E-645 MC 0.766 500 0.034 U 9 E Dil 112 234 183 0.764
E-646 MC 0.750 500 0.018 U 9 E Yes 50 100 78 0.746 26.0
E-647 MC 0.776 500 0.044 U 9 E Mdil 24 51 41 0.771 27.4
E-648 MC 0.702 500 - 0.030 U 10 E Dil 167 349 273 0.700

D-661 PL 0.680 140 - 0.069 D 5 C Mdil 140 240 301 0,735 31.2 Drained tests
D-662 PL 0.677 60 - 0,084 D 6 C Yes 60 104 131 0,752 31.5 on dense sands
D-663 PL 0.675 300 - 0.064 D 6 C Dil 300 565 790 0,714
D-664 PL 0.635 300 -0.104 D 6 C Maybe 300 473 520 0.691 27.7
D-665 PL 0.691 130 - 0,060 D 5 C Dil 130 240 323 0.737
D-666 PL 0.710 60 -0.051 D 5 C Dil 60 114 151 0,778
D-667 PL 0.590 130 -0.161 D 5 C Dil 130 253 355 0.702

DR-668 PL 0.680 140 - 0.069 D CR Dil 180 504 486 0,608 Radial loading
D-760 PL 0.698 250 -0,044 D 5 C Dil 250 356 620 0.742 OCR=4
D-761 PL 0.657 250 - 0.085 D 5 C Dil 250 356 620 0.716 OCR=4
D-762 PL 0609 250 -0,133 D 5 C Dil 250 360 630 0,708 OCR=4
D-874 MC 0.798 200 0.053 D 5 C Con 200 323 370 0,753 28.7 OCR = 8.8

D-681 MC 0.775 1000 0.652 D 10 C Yes 1000 1542 1613 0,703 26.5 Drained tests on
D-682 MC 0.776 500 0.044 D 8 C Yes 500 812 938 0.725 28.9 loose sands
D-684 MC 0.820 200 0.075 D 10 C Yes 200 308 321 0,775 26.4
D-685 MC 0.812 200 0.067 D 10 C Con 200 283 273 0.749 23.9

* MC, moist compacted; PL, wet pluviated.

t Values of state parameter, Y, based on mean steady state line shown in Fig. 17, for samples with p’ < 1000 kPa only.
1 U, undrained conditions; D, drained conditions; L, load-controlled compression; C, strain-controlled compression; E, strain-controlled extension; CR, strain-controlled,
radial stress increasing.
Q Yes = steady state apparently reached; Dil/Con = sample still dilating or contracting at end of test; Mdil = small amount of dilation, sample close to steady state at end of
test; Cav = cavitation of pore fluid may have occurred.
11Void ratios corrected for membrane penetration.
312 BEEN. JEFFERIES AND HACHEY

sand and membrane penetration. Assuming that
consolidation volume change is negligible during
the unload/reload loop, the total volume change
can be used to calculate a potential error in void
ratio of a standard-size sample if membrane pen-
etration were negligible. This error is estimated to
be about 0.007 at 1000 kPa (Fig. 5). Also shown
in Fig. 5 is the membrane penetration correction
applied in this study, which is about 75% of the
upper bound. This value was selected on the basis
of published empirical data (Baldi & Nova, 1984;
Vaid & Negussey, 1984). The likely errors in void
ratio due to membrane penetration in this study
are smaller than 0.003. The void ratio before and
after each test, including the membrane correc-
tion, is shown in Table 2.
Results of laboratory tests
Because of the large number of tests carried
out, it is not possible to reproduce each test
result. Results for a selection of typical tests
showing different behaviours are illustrated
Fig. 6. Fig. 6(a) shows liquefaction with strain
softening for a stress-controlled undrained test
(L-605) on a loose sand. The steady state is
clearly identified in this test. The drained behav-
iour of a similar loose sample is shown in part (b)
(test D-682). There is a large volumetric contrac-
tion in this test. At the end of the test the sample
is very close to the critical state.
The undrained behaviour of a dilatant sample
which reaches a steady state at large strains is
shown in Fig. 6(c) (test C-636). There is an abrupt
change in behaviour shortly before the steady
state is reached. The drained behaviour of a
similar dilatant sample is shown in Fig. 6(d) (test
D-665). In this case the sample has not reached
the critical state at the end of the test and is still
dilating strongly.
An undrained test on a dense sample consoli-
dated to a high confining stress of 8 MPa is
shown in Fig. 6(e) (test C-621). This sample has
the same void ratio as C-636 (Fig. 6(c)) but is con-
tractive because of the high confining stress. The
slight drop in stress at the end of the test is con-
sidered to be the result of tilting of the top platen.
An extension test on a loose sample initially
close to the critical state is shown in Fig. 6(f) (test
E-641). This test shows strain softening followed
by recovery of the sample and dilatant behaviour
at the end of the test. This behaviour is typical of
undrained triaxial extension tests, but also occurs
to a lesser extent in compression tests.
An extension test on a loose sample that does
reach a steady state is shown in Fig. 6(g) (test
E-646). In this test there was contraction of the
in sample, followed by slight dilation to the steady
state. The drained behaviour of a dilatant sample
that reaches a critical state is shown in Fig. 6(h)
(test D-662).
The test results for the entire programme are
now discussed in detail.
Selection of steady/critical state points
It is important that the method of selection of
the critical or steady state for any particular test
is clearly defined. As noted previously, the terms
S line and F line have been used in the past for
critical and steady states from ‘drained dense’ and
‘undrained loose’ samples, respectively.

Upper bound on void ratio error

for 76 mm dia. -

Correction used
0.007 at 1000 kf’a
in this study

Pressure across membrane. kPa

Fig. 5. Membrane penetration effects for Erksak sand; e, = O-6(void ratio error is estimated
assuming that all volume change is due to membrane effects, i.e. that consolidation is
negligible)
 CRITICAL STATE OF SANDS 373

1000
r

I I I I
OO a 16 24
strain: %
0
s

Awal strain %
200 I-

S’. kPa
(a)

5OOOr 7 200

Steady S-‘-
kilt2

-1600 50 -
t I t I I
OK
0 8 16 24 0 8 16 24
stram. % stram %

a 16 24
Axial strain %
(4

Fig. 6. Typical test results (S’ = (a,’ + a.‘)/2, f’ = (a,,’ - a,‘)/Z): (a) undrained test L-605 (loose sample: e, = O-771,
a3=’ = 500 kPa); (b) undrained test D-682 (loose sample; e, = @776, use’ = 500 kPa); (c) undrained test C-636 (dense,
dilatant sample; e, = 0618, uJe’ = 200 kPa); (d) drained test D-665 (dilatant sample; e, = 0691, aae’ = 130 kPa)
374 BEEN, JEFFERIES AND HACHEY

1
11000 1000

r---
5000

3000 160 600 d

Y
4000 ??
iii
4 7000 600 2
L?
w 3000 I a
iz
VI
A---‘-- 5000 400 $
. a
2 2000 :
& 3000 200 1
0
1000

0
8
I

stran %
I
16
i
24
1000
I
0 8
I

stram %
16
I A’ 0

24

4000
r

: *ooot----)
I I I I I I
0' 200;
0 4000 8000 12000 200 400 600
S' kPa S' kPa

(e) (V

50-

OO 81 16
I 24
1

Axd stratn:%

m2001
I I I I I I 1
0 200 400 600 0 8 16 24
stran %
S'. kPa
(!a (h)

Fig. 6 (continued). Typical test results S = (0: + Q/2, t’ = (a: - 4)/2): (e) undrained test C-621 (dense sample, con-
tractive due to high stresses; e, = 0618, oj,’ = 8000 kPa); (f) undrained extension test E-641 (loose sample close to
critical state line; e, = W732, uJE’ = 500 kPa); (g) undrained extension test E-646 (e, = W750, ulc’ = 500 kPa; sample
reached steady state); (h) drained test D-662 (P, = 0677, us.’ = 60 kPa; sample reached steady state)

Figure 7 shows a typical test result (E-645) for continue in that state without stress changes. This
an undrained test which does not reach a well is not the case in Fig. 7. The quasi-steady state is
defined critical/steady state. Many workers will in fact more similar to the phase transformation
select the quasi-steady state shown during this noted by Ishihara & Okada (1978). A shear plane
test as a conservative estimate of the steady state developed in this sample at about 6% strain.
or the F line (e.g. Alarcon-Guzman et al., 1988; There was also necking of this sample which was
Chern, 1985). This is inconsistent with the defini- tested in extension. It is possible that these
tion of either the critical or the steady state. The sample non-uniformities are responsible for the
definitions require that the deformation should observed response.
 CRITICAL STATE OF SANDS 375
500 1-o

e I,,< -•\

m
B Quast-steady ?? 0.9
state - 300 2
> -__
. ..-
,,= Curvatvre (?) _,->‘
I\
,/’ ‘;i---_,,. 2oo ; m
I POW
/I pressure
I
:’ d O-8 - .
Y

, ,\,
Axial stran. % o-7 ’
10 100 1000 10 000
p’ kPa

Fig. 8. Steady state line for Leighton Buzzard sand

showing curvature similar to that of line for Erksak
330/@7 sand (initial state of all samples was ahove the
m stress path
& -150- ,/, \
steady state line; all tests triaxial compression)
,I 1 I
Total stress path
i -200 I
0 100 200 300 400 500
S’ = (a’, + o’Ji2: kPa
Curvature of the steady state line
The measured steady state line for sands is fre-
Fig. 7. Typical test showing a quasi-steady state condi- quently approximated by a straight line in
tion (test E-645, e, = 0766, (r3<’= 500 kPa, I = 0936)
e-log p’ space. This is generally a reasonable
approximation for sub-angular or subrounded
In this Paper, only the final steady state which quartz sands in the stress range l&500 kPa.
appears to continue indefinitely will be used for Some second-order curvature of the line is often
either a critical or steady state point. The critical apparent, as shown in Fig. 8 for an associated
or steady state is thus an ultimate state to which series of tests on Leighton Buzzard sand by the
the sample will go after large strains under Authors. However, for comparison between
monotonic loading. various tests, the shape of the line over a wider
Even at 20% strain in triaxial tests where a stress range must be considered.
continuous steady state is reached, it is not Figure 9 shows the steady state data for 17
known whether this is a true ultimate state or undrained compression tests on contractive
whether further changes would have occurred at Erksak sand samples. The most striking feature of
larger strains. For example, Poulos, Castro & the figure is that the steady state line curves
France (1988) indicate that the steady state can be abruptly at a stress level of about 1 MPa. Similar
achieved in the laboratory only for uniform clean
sands, and probably only for highly contractive
samples.
In many tests there was a mild tendency for the
sample to dilate at the end of the test (e.g. E-641,
Fig. 6(f)), which has been shown as ‘Mdil’ in
Table 2. It is possible that this apparent dilatancy
is the result of non-uniformities and other testing
effects, rather than real behaviour. In these cases,
o’8P--i
the conditions at the end of the test are assumed
to be representative of the steady state. (Table 2
shows Con or Dil for samples which are clearly
contracting or dilating at the end of the test.) In
the remainder of this Paper, those points which
1 I I I
actually reached a steady state are plotted 10 100 1000 10000
directly on the state diagram. In some cases, end p kPa
points of tests are plotted with arrows indicating Fig. 9. Steady state line for undrained compression tests
the direction of state change, according to the on contra&e samples of Erksak 330/@7 sand with
evaluation of state shown in Table 2. initial states ahove the critical state line
376 BEEN, JEFFERIES AND HACHEY

behaviour was observed for Leighton Buzzard
sand (Fig. 8).
The break in the steady state line is indicative
of a change in mechanism of shearing at high
stress levels. For Erksak sand in the O-500 kPa
stress range, the behaviour is probably that of a
granular material in which the grains essentially
retain their shape except for elastic deformations
and some limited particle damage at grain con-
tacts. At stress levels greater than 1 MPa there
appears to be some breakage of grains which
results in a different behaviour. This hypothesis is
supported by grain size distribution curves of
samples before and after shearing at a stress level
of 8 MPa (Fig. 10). A small increase in the fine
sand and silt sizes is evident, similar to that
observed by Lee & Farhoomand (1967) for fine
sands. Further studies on a microscopic scale also
revealed a small amount of fracturing of grains
after high-pressure testing. The break point in the
steady state lines is expected to be dependent on
the sand mineralogy, assuming it is the result of
crushing. Hardin (1987) has shown a range of
break point stresses in one-dimensional
opposed to triaxial) compression, from 400 kPa
for ground feldspar to 8.6 MPa for Ottawa sand.
The shape of the steady state line at stress
levels less than 10 kPa is difficult to determine
experimentally. Tatsuoka, Sakamoto, Kawamura
& Fukushima (1986) carried out a large number
Size Of s,eve open,ncJ
mm pm
mm
5160 66
pzz--r
before
*After
of tests at stress levels of 5 kPa, and their data
suggest that the critical state line becomes flatter
at low stress levels. For the purpose of this study
of the critical/steady state of sands, the behaviour
at very low stress levels is of little practical impor-
tance and is not discussed any further.
Critical or steady state
The end points of both drained and undrained
compression tests are shown in Fig. 11. The figure
suggests that, for Erksak sand, whether the test is
drained or undrained does not have any influence
on the location of the ultimate state. Since critical
state is associated with drained tests and steady
state with undrained tests, this is equivalent to
concluding that the critical and steady state are
the same. Alternatively, the S and F lines are
coincident for Erksak sand. However, it should
be noted that definition of the critical state with
tests on dense drained samples is poor, as the
samples are generally still dilatant at the end of
the test and shear band localization occurs.
The possibility that the initial state affects the
(as critical/steady state was also investigated, as the S
line was originally associated with tests on
samples denser than critical, and the F line with
looser states. Fig. 12 shows the initial and final
states for all compression tests that apparently
reached the critical/steady state. Clearly there are
not different lines for tests starting above and
below the line. In particular, the tests beginning
below the critical/steady state do not consistently
end above the mean line, or vice versa.
There has been much discussion regarding the
‘critical state or steady state’ question in the past
(e.g. Poulos, 1981; Sladen et al., 1985; Alarcon-
Guzman et al., 1988; Poulos et al., 1988). While
test
this data set on only one sand does not resolve
test

l Undrained comprewo” testsA

d Drained compression tests

I - Some stress Increase at

end of test
Slight incn
I Some dilation at end of
.
r in fines due to
test
I Some contractlon at .
end of test
I I ,i ,
10 100 10 000
Grar we: mm 1000
p’: kPa
Fig. 11. Comparison of critical state from drained and
undrained tests (tests on samples consolidated to
p’ > 1000 kPa are not included, to eliminate effect of
Fig. 10. Change in grain size distribution of Erksak material changes due to grain-crushing daring
330/@7 sand during test at 8000 kPa (test C-621) consolidation)
 CRITICAL STATE OF SANDS 371

Efect of stress path on the critical state

A series of extension tests was carried out on
Erksak sand to determine if the critical state was
affected by stress path. Kuerbis et al. (1988) and
Vaid et al. (1990) have suggested that very large
differences in behaviour occur in extension and
compression due to inherent anisotropy. This
implies a difference between extension and com-
pression critical states. Alarcon-Guzman et al.
(1988) also argue that the initial anisotropy affects
the location of the critical state line.
Figure 13 shows the results of strain-controlled
extension tests and strain-controlled and load-
I 1 I 1 I
controlled compression tests on Erksak sand. The
10 100 1000 10000 arrows on extension test data show that most
p’ kPa
samples were still dilating at the end of the test.
Fig. 12. Effect of initial state on the critical state (all These data suggest that the critical state line is
indeed independent of the stress path for Erksak
sand. Similar data obtained by the Authors in a
related test programme on Toyoura sand are
shown in Fig. 14. These undrained tests were gen-
the issue, it is consistent with the hypothesis that erally carried out on moist compacted samples, to
the critical and steady state are coincident.
Similar data are presented by Poulos et al. (1988)
for Syncrude tailings. These data differ from data . Compression tests
that show a significant difference in the S and F a Extension tests
- Extension tests show
lines for other sands (e.g. Castro, 1969; Alarcon-
Guzman et al., 1988).
The effect of strain rate on the critical/steady
state has also been proposed as a potential reason
for differences (Casagrande, 1975). Fig. 9 shows
data from this test programme for load-
controlled and strain-controlled testing on con-
tractive, undrained samples. The range in strain I 1
rates at steady state is from 4%/h for strain- 100 1000

controlled tests to about 300000%/h for load- p’: kPa

controlled tests. No significant difference is

Fig. 13. Effect of stress path (extension v. compression)
apparent, suggesting that the strain rate effect is
on critical state of Erksak 330/@7 sand (undrained tests,
small for the magnitude of displacements induced initial states above critical state)
in triaxial tests for Erksak sand. For finer, silty
sands this may not be the case.
Poulos (1981) has pointed out that the critical 1 .o
and steady states differ in that the steady state, by
definition, has an associated flow structure and a
requirement for a constant velocity, neither of
which are incorporated in critical state concepts. 0
Moist compacted
These differences are in definition, but the partic- .

c
I \,_
ular flow structure and applicable strain rate have
not been defined. This makes a clear distinction m 0.9
of the steady state from the critical state imposs-
ible. The Authors are not aware of any data l Compression tests
demonstrating flow structures in sands on a par- 6 Extension tests
ticular scale, or of data showing that constant
velocities actually occur in load-controlled lique-
faction tests. 0.81 I I

Based on the above discussion, it is justifiable 10 100

to use a single term, the critical state, for both the

p’: kPa
critical and the steady state of sands in the Fig. 14. Effect of stress path (extension v. compression)
remainder of this Paper. on critical state of Toyoura sand
378 BEEN, JEFFERIES
ensure contractive behaviour in compression. The
inherent anisotropy due to sample preparation
may therefore be different from that in the
samples tested by others.
The extension tests did, however, behave very
differently from the compression tests. The typical
test data presented in Figs 6 and 7 illustrate these
differences. In particular, there is apparently
greater softening in the extension tests, which
precedes a tendency to dilation at the end of the
test. The softening is attributed to the fact that
the loading is in a different direction relative to
the sample fabric. At large strains the fabric ani-
sotropy becomes aligned with the loading, and
dilation will be observed. Similar behaviour has
been reported by Symes, Gens & Hight (1984) for
undrained shear in a hollow-cylinder apparatus.
Effect of sample preparation
Sand samples were prepared by wet pluviation
and by moist compaction techniques to give dif-
ferent (but unknown) initial fabrics. The purpose
of these tests was to investigate the effect of initial
fabric on the critical state line. The effect of initial
fabric is difficult to investigate because direct
quantification of fabric is not practical. However,
Oda (1972) has clearly shown how the tamping
and pluviation techniques do in fact result in dif-
ferent fabrics. Fig. 15 shows the critical state line
data as a function of sample preparation method.
No effect of sample preparation method is appar-
ent. This conclusion agrees with the findings of
Poulos et al. (1985).
The main problem with examining the effect of
sample preparation is that it is not usually poss-
ible to prepare a pluviated sample at void ratios
above the critical state void ratio at stress levels
in the range ~1000 kPa. The critical state line for
pluviated sand in this stress range can only be
reached by drained tests on dilatant samples, as
0
0 Pluvlated
l Moist compacted
-. Dramed test results
0.6 -
:
, ,\,
I 0.L
10 100 1000
p’: kPa
Fig. 15. Effect of sample preparation on critical state
(note: pluviated specimens at high void ratios cannot be
prepared)
AND HACHEY
shown in Fig. 15. There is therefore no direct
comparison of critical states reached from
samples with loose undrained initial states pre-
pared in these two ways.
At lower void ratios, either preparation method
can be used and the critical state line is well
defined and independent of preparation method.
However, it is unlikely that initial fabric would
have any effect on the critical state line at high
pressures if the mechanism of shearing involved
breakage of grains. The fact that preparation
method does not affect the critical state at high
stress levels therefore does not necessarily mean
that it has no effect at low stress levels.
Friction angle at critical state
It is generally assumed that the friction angle at
the critical or steady state is a constant and inde-
pendent of density. For granular materials, the
accepted value of 4,, ’ is about 30”. However, it is
also recognized that the true friction angle
between mineral surfaces, $,,, is less than the
value of c$,,’ (Bishop, 1954; Rowe, 1962). Intu-
itively, there is therefore no reason why the fric-
tion angle in extremely loose sands may not drop
below 30”, depending on the shearing mechan-
isms at a particulate level.
Table 2 includes measured values of 4,,‘, which
are plotted as a function of void ratio in Fig. 16.
The data show k4” of scatter, but a trend of
decreasing 4,, with increasing void ratio is
evident. Data have not been plotted for tests
which did not closely approach the critical state,
or where the mean stress at the critical state is
Range I” 4’ based on i 1 kPa error I”
I stress measurement at steady state
. Undramed compression, dzC < 1500 kPa
A Undrained compression, CJ’~~> 3000 kPa
0 Undrained extension, (7’ 500 kPa
35r q Drained compression, CJ’~~~ 1000 kPa
%
10000 Votd ratio at critical state
Fig. 16. Critical state friction angle for Erksak 330/O-7
sand (range in $’ not shown where range is less than size
of symbol)
 CRITICAi STATE OF SANDS 379

greater than 1 MPa. It is inappropriate to
compare the high stress level data with low stress
level data because at high stress levels a different
shearing mechanism may govern. At typical criti-
cal state void ratios, the critical state friction
angle is close to 30”, but at very high void ratios,
where the number of interparticle contacts per
grain may be expected to be small, the critical
state friction angles drop rapidly. Values as low
as 12-18” are shown, but it should also be noted
that the testing procedures, which have a
resolution of + 1 kPa at best, are such that rela-
tively large errors may be induced in friction
angle measurements at very low stresses.
Previous studies have possibly not identified
the dependence of &,, on density because it is
normal practice to plot the critical state points on
a p-q diagram, draw a line through the data and
measure the slope of this line. The same pro-
cedure in this case yields a 4,, of 31”. However,
the high void ratio/low stress points all lie in a
cluster near the origin of the pq diagram. The
line becomes heavily weighted towards the higher
stress level points because the data points near
the origin appear to be close to the critical state
line, although the q/p ratio could be very differ-
ent.
IMPLICATIONS OF THE TEST DATA
On the basis of an extensive testing programme
on Erksak sand, some potentially important
aspects of the critical state of sands have been
identified. In particular, the Authors have been
advocating the use of the steady state line as the
reference state for determining a density/stress
level state parameter Y for sands (Been & Jef-
feries, 1985). Steady state concepts have also been
applied widely to sand liquefaction evaluation
(Poulos et al., 1985). Differences between the so-
called critical and steady state lines and the infer-
ence that the stress path and initial fabric have a
significant influence on the critical/steady state,
to be the same as the ultimate state (Vaid et al.,
1990; Alarcon-Guzman et al., 1988). The defini-
tion of the critical or steady state as an ultimate
state to which the sample will go under mono-
tonic loading must be applied strictly.
Figure 17 shows the entire set of critical state
points for Erksak sand (i.e. those tests in Table 2
that reached a steady state, or close to it as iden-
tified by Mdil). A mean line has been drawn
through the data. This line differs from the indi-
vidual best fit lines drawn for Figs 8-15. The
scatter about the mean critical state line is about
f0.01 in void ratio at typical civil engineering
stress levels. This scatter band encompasses the
best-fit lines in the previous figures. It is unlikely
that this scatter can be significantly reduced by
enhanced test procedures. There will therefore be
an inherent imprecision of this magnitude in all
engineering applications based on the state
parameter or the critical/steady state. The scatter
is small in this study in comparison with other
studies because careful control was maintained on
the intrinsic material properties and the selection
of the steady state point (i.e. samples which were
dilating or contracting are not included in Fig.
17). Awareness of this imprecision is especially
important when undrained steady state strengths
are used, because an error of 0.01 in void ratio
may represent a factor of 2 in shear strength.
Curvature of the critical state line on an
e-log p’ state diagram must be recognized if the
precision of about 0.01 is to be maintained over a
wide stress range. This is particularly important if
stress levels may be greater than that correspond-
ing to the sharp curvature in the critical state line.
Some critical state concepts are still valid in this
Yf>o Y’<O
Undramed
Load-controlled, compress~o” 0
Stm-controlled, compressfon A A
Strancontrolled. extension 0
amed, compressIon 0 .

imply that no unique critical state line exists for

sands. This supposition makes the application of
critical state soil mechanics to sands, use of Y as
a normalizing state parameter and the steady
state liquefaction evaluation procedures very
complex. An extensive re-evaluation of the defini-
tions and procedures would be required if this
were true.
Within the limits of the testing procedures,
existence of a singular critical/steady state line
has been demonstrated for Erksak sand. Stress
path, drainage, initial state, sample preparation
I I I J
and loading method were all varied in the test 10 100 1000 10000
programme. In order to obtain a unique critical p’: kPa

state, it is important that quasi-critical states or Fig. 17. Overall scatter in critical state line for Erksak
the phase transformation point are not assumed 330/O-7 sand (38 tests)
380 BEEN, JEFFERIES AND HACHEY

range of stresses, in that a well defined ultimate has been carried out to examine the critical state
state exists at which shearing takes place at con- in a variety of tests on Erksak 33010.7 sand.
stant volume and constant stress. It is usual to use drained tests on dense sand to
A secondary issue is the observation that $,,’ is get the critical state line and undrained tests on
not a constant, but depends on the critical state loose sand to get the steady state line. For Erksak
void ratio and stress level. Further data are sand, these lines are coincident. In addition, the
required to verify this hypothesis and to explain effects of stress path, sample preparation method
the mechanism for it. and stress-controlled loading were examined.
The conclusion that a unique critical state line It is concluded, based on these tests, that for
exists which is independent of stress path, sample practical purposes a unique critical state line can
preparation or drainage conditions, may be sur- be defined for Erksak sand. This critical state line
prising. Different researchers have reached differ- can be obtained with a precision of about +O.Ol
ent conclusions in this regard. Hird & Hassona in terms of void ratio for p’ < 1000 kPa. For
(1990) show a consistent difference between load- mean stresses greater than 1000 kPa, the slope of
controlled and strain-controlled tests for Leigh- the critical state line is much steeper, and
ton Buzzard sand. Casagrande (1975) and Castro although the bandwidth of scatter about the line
(1969) showed similar data for Banding sand. does not increase, the precision in terms of void
Poulos et al. (1988), however, show no effect of ratio is significantly decreased. To achieve this
load-controlled v. strain-controlled testing for level of precision, testing procedures must be well
Syncrude tailings. controlled so that intrinsic material properties are
Vaid et al. (1990) examined the effect of stress constant, the void ratio is accurately determined
path (i.e. extension v. compression) on the steady and sample non-uniformities are minimized. In
state and reached the conclusion that a significant addition, one must adhere strictly to the defini-
effect was present. However, in their paper they tions of critical or steady state, and make sure
refer to the steady state and phase transformation that an ultimate state of steady shearing is
interchangeably. In this study on Erksak sand, achieved. The behaviour before this ultimate state
the phase transformation point was also found to is reached is, however, strongly dependent on
be strongly dependent on stress path, sample factors such as stress path, drainage conditions,
preparation, etc. Alarcon-Guzman & Leonards initial state parameter and initial fabric.
(1988) have postulated a mechanism of collapse The critical state line for Erksak sand is
which is consistent with differences in phase approximately bilinear, with the break in the line
transformation for extension and compression occurring at a stress level of about 1 MPa. There
tests and could explain the load-controlled v. appears to be a change in shearing mechanism at
strain-controlled loading difference. A method of this stress level, to one in which breaking or
examining the structure of sands as they collapse crushing of sand grains becomes significant.
and flow is needed to examine this mechanism Leighton Buzzard sand, with a similar mineral-
further. The flow structure postulated by Casag- ogy, shows a similar behaviour. It is expected that
rande (1975) and Poulos (1981) requires similar other sands with other mineralogies will be differ-
inspection. ent.
A method to quantify initial fabric or inherent The conclusions described in this Paper have
anisotropy of sand samples is also required. It is been based on extensive (but not exhaustive) test
possible that conflicting experimental evidence is data on Erksak sand, supplemented by limited
simply the result of different initial fabrics that data for Leighton Buzzard and Toyoura sand.
affect the tests in an unknown manner. If this is Although it is possible that the conclusions are
so, it is important to reproduce the field fabric generally applicable to sand, this remains to be
and loading path in the laboratory. This is not demonstrated. The influence of fabric on sand
currently possible, because the field and labor- behaviour is particularly important but has not
atory fabrics are unknown. For example, it yet been satisfactorily addressed because there is
cannot simply be assumed that one-dimensional, no simple and practical method to quantify
quiescent laboratory pluviation results in the fabric. It is a well established fact that stress
same fabric as hydraulic deposition of sand fills in strain response is strongly fabric-dependent, but
the field. this Paper suggests that the critical state is inde-
pendent of fabric for the range of test conditions
and sands examined.
CONCLUSIONS
Determination of the critical or steady state
line of sand is important to several aspects of the ACKNOWLEDGEMENTS
engineering of sand fills or natural sand deposits. The Authors thank Gulf Canada Resources
A relatively extensive series of laboratory tests Ltd, Ishikawajima-Harima Heavy Industries Ltd
 CRITICAL STATE OF SANDS 381

and Esso Resources Canada Ltd for funding to Lee, K. L. & Farhoomand, I. (1967). Compressibility
carry out the tests and for permission to publish and crushing of granular soil in anisotropic triaxial
this Paper. compression. Can. Geotech. J. 4, No. 1, 68-86.
Oda, M. (1972). Initial fabrics and their relations to
mechanical properties of granular material. Soils
REFERENCES Fdns 12, No. 1, 17-36.
Alarcon-Guzman, A. & Leonards, G. A. (1988). Dis- Poulos, S. J. (1981). The steady state of deformation. J.
cussion: Liquefaction evaluation procedure. J. Geotech. Engng Am. Sot. Civ. Engrs 17, GT5, 553-
Geotech. Engng Am. Sot. Civ. Engrs 114, No. 2, 232- 562.
236. Poulos, S. J., Castro, G. & France, J. W. (1985). Lique-
Alarcon-Guzman, A., Leonards, G. A. & Chameau, J. L. faction evaluation procedure. J. Geotech. Engng Am.
(1988). Undrained monotonic and cyclic strength of Sot. Civ. Engrs 111, No. 6, 772-792.
sands. J. Geotech. Engng Am. Sot. Civ. Engrs 114, Poulos, S. J., Castro, G. & France, J. W. (1988). Closure
No. 10, 1089-l 109. to discussion: Liquefaction evaluation procedure. J.
Baldi, G. & Nova, R. (1984). Membrane penetration Geotech. Engng Am. Sot. Ciu. Engrs 114, No. 2,251-
effects in triaxial testing. J. Geotech. Engng Am. Sot. 259.
Civ. Engrs 110, No. 3,403420. Roscoe, K. H., Schofield, A. N. & Wroth, C. P. (1958).
Been, K. & Jefferies, M. G. (1985). A state parameter for On the yielding of soils. Gtotechnique 8, No. 1,
sands. Gtotechnique 35, No. 2,99-l 12. 22-53.
Bishop, A. W. (1954). Correspondence on shear charac- Rowe, P. W. (1962). The stressdilatancy relation for
teristics of a saturated silt, measured in triaxial com- static equilibrium of an assembly of particles in
pression, by A. D. M. Penman. Gbotechnique 4, No. contact. Proc. Roy. Sot. A 269,50&527.
1, 43345. Shinde, S. B., Crooks, J. H. A., James, D. A. &
Casagrande, A. (1936). Characteristics of cohesionless Williams-Fitzpatrick, S. (1986). Geotechnical design
soils affecting the stability of earth fills. J. Boston for Beaufort Sea structures. Proc. 3rd Can. Conf.
Sot. Civ. Engrs. Reprinted in Contributions to soil Marine Geotechnical Engineering, St Johns, vol. 1,
mechanics 1925-1940. Boston Society of Civil En- pp. 347-362.
gineers. Sladen, J. A., D’Hollander, R. D., Krahn, J. & Mitchell,
Casagrande, A. (1975). Liquefaction and cyclic deforma- D. E. (1985). Back analysis of the Nerlerk berm
tions of sands. Proc. 5th Pan Am. Cant Soil Mech. liquefaction slides. Can. Geotech. J. 22, No. 4, 579-
Buenos Aires, vol. 5, pp. 8@133. 588.
Catro, G. (1969). Liquefaction of sand. PhD thesis, Divi- Sladen, J. A. St Handford, G. (1987). A potential system-
sion of Engineering and Applied Physics, Harvard atic error in laboratory testing of very loose sands.
University. Can. Geotech. J. 24, No. 3,462466.
Chern, J. C. (1985). Undrained response of saturated Stroud, M. A. (1971). The behaviour of sand at low stress
sands with emphasis on liquefaction and cyclic mobil- leuels in the simple shear apparatus. PhD thesis, Uni-
ity. PhD thesis, University of British Columbia, versity of Cambridge, UK.
Canada. Symes, M. J. P. R., Gens, A. & Hight, D. W. (1984).
Hardin, B. 0. (1987). 1-D strain in normally consoli- Undrained anisotropy and principal stress rotation
dated cohesionless soils. J. Geotech. Engng Am. Sot. in saturated sand. Geotechnique 34, No. 1, 1 l-27.
Civ. Engrs 113, No. 12, 1449-1467. Tatsuoka, F., Sakamoto, M., Kawamura, T. & Fuku-
Hird, C. C. & Hassona, F. A. K. (1990). Some factors shima, S. (1986). Strength and deformation charac-
affecting the liquefaction and flow of saturated sands teristics of sand in plane strain compression at
in laboratory tests. Engng Geology 28, 1499170. extremely low pressures. Soils Fdns 26, No. 1,65-84.
Ishihara, K. & Okada, S. (1978). Yielding of over- Vaid, Y., Chung, E. K. F. & Keurbis, R. H. (1990).
consolidated sand and liquefaction model under Stress path and steady state. Can. Geotech. J. 27,
cyclic stresses. Soils Fdns 18, No. 1, 57-72. No. 1, l-7.
Jefferies, M. G., Rogers, B. T., Stewart, H. R., Shinde, S. Vaid, Y. P. & Negussey, D. (1984). A critical assessment
B., James, D. A. & Williams-Fitzpatrick, S. (1988). of membrane penetration in the triaxial test.
Island construction in the Canadian Beaufort Sea. Geotech. Test. J. Am. Sot. Test. Mater. 7, No. 2,
ASCE Specialty Conf: Hydraulic Fill Structures, Fort 7&76.
Collins, CO., pp. 81G883. Wood, D. M. (1984). On stress parameters. Geotech-
Kuerbis, R., Negussey, D. & Vaid, Y. P. (1988). Effect of nique 34, No. 2, 2822287.
gradation and fines content on the undrained Wroth, C. P. & Bassett, R. H. (1965). A stress-strain
response of sand. ASCE Specialty Conj: Hydraulic relationship for the shearing behaviour of a sand.
Fill Structures, Fort Collins, CO., pp. 33&345. Geotechnique 15, No. 1, 32-56.
Been, K., Jefkries, M. G. & Hachey, J. (1992). GPotechnique 42, No. 4, 655-663

DISCUSSION

The critical state of sands

K. BEEN. M. G. JEFFERIES and J. HACHEY (1991). Gdotechnique 41, No. 3, 3655381

J. Chu, Nanyang Technological University, Singa- loose Sydney sand. These tests were conducted
pore, and S.-C. R. Lo, University of New South with deformation controlled loading. The void
Wales, Canberra ratios after consolidation for the three tests are
We would like to ask whether the critical state 0.8764, 0.8493 and 0.8380. Test U3 showed quasi-
parameter of dense sand can be measured by a steady behaviour. When quasi-steady flow
drained or an undrained triaxial test. occurred, no sign of non-homogeneous deforma-
For drained da,’ = 0 tests there is evidence to tion was observed. A comparison of tests U2 and
indicate that the observed strain softening (as in U3 indicates that whether a sample will manifest
Fig. 6(h)) is not a material behaviour, but the a steady or a quasi-steady behaviour in an
result of non-homogeneous deformation (Hettler undrained test depends on the void ratio of the
& Vardoulakis, 1984; Chu, Lo & Lee, 1992a, sample. Fig. 18 also shows that all the tests end at
1992b). It is therefore highly questionable to an ultimate line, irrespective of whether the
measure the critical state parameter of a dense behaviour is steady or quasi-steady. This line is
sand by a drained test. For undrained tests, the the so-called steady state line in q-p’ space.
Authors observed a steady state for a dense For loose sand, the critical state is the failure
sample (Fig. 6(c)). We wonder whether this is also state measured in a drained test and the steady
due to non-homogeneous deformation, just as in (or quasi-steady) state is the ultimate state in an
a drained test. We have conducted a number of undrained test. The equivalence of the critical
undrained tests for dense Sydney sand, which is a state and the steady state reflects the fact that the
quartz sand with mean grain diameter of 0.3 mm. failure stress ratio measured from a drained test is
None of these tests exhibited a steady state. It the same as the ultimate stress ratio measured
was observed that an undrained test led to a from an undrained test for loose sand. However,
stress path approaching a constant stress ratio for dense sand, the critical state is different from
line (Chu, 1991). This agrees with the t versus S the failure state. The relationship becomes more
curve in Fig. 6(c). If a steady state can be assumed complicated.
to occur in an undrained test, the stress ratio at
the steady state will be the same as the slope of
the resultant stress path. The principal stress ratio R. Verdugo, University of Tokyo
determined from an undrained test is 3.55. This is The Authors found that for a wide range of
larger than the stress ratio at critical state- pressure, the representation of the steady state
3,28-which was measured for loose sand along a line in terms of void ratio and log p’ is almost
drained du,’ = 0 test at a similar effective confin- bi-linear. This can be a simple and good approx-
ing stress. The Author’s data indicate the same imation in some cases. However, the physical
fact. Unless there is reliable and sufficient data to interpretation given by the Authors to the break
show that the same sand but with different initial point seems to be obscured by the use of a semi-
densities will approach a different critical state, logarithmic plot.
the critical state parameter of dense sand mea- The steady state line of Toyoura sand evalu-
sured in undrained tests will not be suitable for ated by Verdugo & Ishihara (1991) is indicated in
verification purposes. Fig. 16 may therefore need Fig. 19, together with the steady state line of
to be re-examined. Erksak 33010.7 sand using the Authors’ data.
Is the quasi-steady state a result of non- This semi-logarithmic plot shows that both sands
homogeneous deformation? have relatively linear and flat steady state lines
The Authors indicate that the quasi-steady for low pressure. For a medium level of pressure
state may be due to the non-homogeneous defor- the steady state lines are highly non-linear and
mation of samples. However, this is not generally steeper, and for a high level of pressure they are
true. Quasi-steady behaviour in an undrained test nearly linear and much steeper than before. This
can be a material property. Fig. 18 presents the trend is consistent with results given by Castro,
effective stress paths of three undrained tests for Poulos, France & Enos (1982) and Seed, Seed,
6.55
656 DISCUSSION

0
0
/’
100
I
200
/
300
) 400
1 I
500
I
600
I
700
p’ = (d, + 2n’,)/3: kPa
Fig. 18

Harder & Jong (1988). The results shown in Fig. before and after the appearance of the break
19 are replotted in Fig. 20 using an arithmetic point in the semi-logarithmic plot. Also in the
scale. For Toyoura sand the steady state line is arithmetic plot, a kind of break point where the
curved from p’ = 0 to about p’ = 500 kPa and steady state line changes from a non-linear to a
thereafter is linear up to approximately p’ = 3200 linear relationship is observed, but it is located at
kPa, the extent of my data. In spite of the scatter low stress level, typically 100-500 kPa (Verdugo,
at high stress, a similar pattern is observed for 1989), and probably it is not related with particle
Erksak 33OjO.l sand using the Authors’ data, but breakage due to the low level of pressure
in this case the linear behaviour starts earlier involved. Perhaps, the particle breakage caused
from p’ z 200 kPa. It is evident that the break by the usage of high pressure can explain the
point at p’ = 1000 kPa found by the Authors scatter observed in the steady state line on
using a semi-logarithmic scale is not manifested Erksak 330/0.7 sand. If the particle breakage is
in the arithmetic plot. This fact probably indi- important, the grain size distribution of the
cates that the same mechanism of shearing exists material would be modified, and because the

1.0

o-9
(Verdugo 8 Ishlhara. 1991)

m o-a - .
B .
m Erksak 33010.7 sand
9 (Authors’ data)
2 0.7

5 10 20 50 100 200 500 1000 2000 5000 10000

Log p’: kPa
Fig. 19. Steady state lines on semi-logarithmic scale
 DISCUSSION
P kPa
Fig. 20. Steady state lines on arithmetic scale
steady state condition is sensitive to grain size
(Poulos, Castro & France, 1985) it would be
affected.
I believe that when a small range of pressure is
under study, the steady state line can be rep-
resented by a linear relationship in e-log p’ space,
but when the range of pressure under consider-
ation is not small a better representation can be
obtained by a bi-linear relationship in e - p’
space (Verdugo, Ishihara & Towhata, 1989).
There is a difference between the fabric and
structure of a soil mass which can help the inter-
pretation of experimental results with regard to
the effect of sample preparation. For this dis-
cussion the fabric may be defined as the micro-
structure of the granular mass that involves only
the orientation of individual particles and the dis-
tribution of the normal contacts between par-
ticles; the structure of a soil mass includes the
particle size, distribution of particle size and voids
throughout the soil mass and interparticle forces.
The definition of fabric given by the Authors
includes cementation at contacts, whereas in my
definition this is included in the structure. Using
this distinction, the bond between grains created
by ageing, for example, should be reflected in the
structure of the soil, but not in its fabric. Sample
preparation methods that generate homogeneous
samples, such as moist tamping, air dry pluvia-
tion and wet pluviation, affect only the fabric of
the soil mass, but methods such as water pluvia-
tion can also affect the structure of the soil mass
because they create a different distribution of the
particle size throughout the soil mass, especially
when the soil contains fines. In my opinion the
initial fabric of the soil does not affect the loca-
657
tion of the steady state line because the level of
strain required to achieve the steady state condi-
tion erases the initial fabric. In this sense I agree
with the Authors’ conclusion regarding the effect
of moist tamping and wet pluviation methods of
sample preparation on the location of the steady
state line. However, I believe that the initial struc-
ture of the soil affects the position of the steady
state line. For example, Fig. 21 (Vasquez &
Dobry, 1988) shows the steady state lines for
samples prepared by moist tamping and by sedi-
mentation under water. Sample preparation has a
clear effect on the location of the steady state line.
This is because the sedimentation under water
creates a heterogeneous specimen with stratifi-
cation which cannot be erased even at large
deformation.
Further evidence is given in Fig. 22 (Verdugo,
Ishihara, Daud & Towhata, 1989) concerning the
effect of ageing on the steady state line. Reconsti-
tuted and undisturbed specimens taken from an
archaeological excavation with clear evidence of
boiling sands were tested. The samples were
taken from the trace left by the boiling materials.
According to archaeological studies the liquefac-
tion of the underlying layers that created the
trace of boiling sand probably took place about
400 years ago. The undisturbed samples had an
average fines content of 8%. The reconstituted
samples were prepared from two batches of soil:
the original sandy soil with 8% of fines and a
clean sand free of fines obtained by the original
soil being washed through a number 200 sieve.
The undisturbed specimen developed the steady
state condition above and to the right of the
steady state line of reconstituted specimens with
658 DISCUSSION

0.7

&moulded
:
sedunented
a \ \ steady state line
Sand SF7
.s
m
% 06 -
>

o Moist tamplng
l Sedlmen!ed

0.5 L
0.01 0.1 1 10
uj kg/cm2

Fig. 21. Steady state lines for sedimented and moist tamped speci-

8% of fines. Moreover, one undisturbed sample mens under small confining pressure do not show
developed a steady state condition above the any significant drop in the internal friction angle
steady state line for clean sand. This indicates mobilized at large deformation close to the steady
that the structure developed after 400 years state condition. This observation, together with
clearly affects the steady state condition. This the Authors’ conclusion that the same steady
pattern was consistently observed in four addi- state line can be obtained from undrained and
tional sandy soil deposits (Verdugo, 1989). drained tests, and the experimental complications,
To evaluate the friction angle $,, from lead one to question the dependence of 4,” on
undrained tests on very loose specimens is com- void ratio and stress level observed by the
plicated because of the membrane compliance. Authors. Furthermore, for extremely loose speci-
My experience is that usually in these tests, after mens tested under undrained conditions, I have
some relatively small level of axial deformation, not found 4,” to be dependent on the initial state
the membrane is crumpled and I am not aware of of the soil (Verdugo & Ishihara, 1991; Verdugo,
a reliable membrane force correction under this Ishihara & Towhata, 1991).
condition of the membrane. If the ordinary mem-
brane force correction is used, the actual axial
load on the specimen is underestimated and Y. P. Vaid, University of British Columbia, Van-
therefore 4,” would also be underestimated. couver, and V. Sithan Pillai, British Columbia
However, experimental evidence (Tatsuoka, Saka- Hydro & Power Authority, Vancouver
moto, Kawamura & Fukushima, 1986), indicates We would like to present experimental and
that drained tests on initial medium-dense speci- conceptual arguments contrary to the Authors’

1-l
, oi\-,_ Klyosu sand
A Undisturbed sample

!:ill
Clean sand

0.7 _L-~ A- .-- I /I

c 200 400 600 800 1000
p’. kPa

Fig. 22. Steady state lines for reconstituted and undisturbed samples
 DISCUSSION 659

conclusions that the steady/critical state is unique

for a sand (specifically that it does not depend on
stress path or initial state), and that the friction
angle at the critical state decreases with increas-
ing initial void ratio. We would also like to point
out that most steady state concepts have been
developed using void ratio states of sands that are
inaccessible to them in the water pluviated state,
and hence are of questionable relevance to the
design of hydraulic fill earth structures.
The loosest void ratio for Erksak sand reported
is 0.735 (Table l), but the flatter section of the
steady state line (Fig. 9) corresponds to void
ratios which are looser than the loosest. Clearly,
(n’, + rr’,)/Z kPa
such void ratio states are not possible for sands
that are water/air pluviated, and hence are not
relevant to the definition of steady state in practi-
cal design. Recent studies on Duncan Dam foun-
dation sand that was sampled after freezing of the
ground, and which is considered very loose (D, =
30%, (N,),, = lo), show in situ void ratios
nowhere greater than the ASTM loosest condi-
tion (with suitable account taken of densification
under the current overburden stress). Similar
deposition void ratio states in relation to the
ASTM loosest should apply to other water
deposited hydraulic fill and fluvial sands.
Loose water deposited sands have been shown
to be anisotropic (e.g. Symes, Shibuya, Hight &
Gens, 1985; Arthur & Menzies, 1972). Their
undrained response should therefore depend on
the direction of or during loading in relation to
the deposition direction. A typical example of
such anisotropic undrained behaviour of a tail- Fig. 23. Undrained compression and extension behaviour
ings sand water pluviated in the loosest state in of a uniform tailings sand (after Kuerbis & Vaid, 1989)
triaxial compression and extension loading is
shown in Fig. 23 (Kuerbis & Vaid, 1989).
Whereas contractive behaviour is noted in exten- is likely that moist tamping, notwithstanding
sion with clear evidence of approach to the steady yielding void ratios which are inaccessible to plu-
state, compression response is dilative at the same viated materials, does not give rise to pronounced
initial void ratios. Similar evidence for other inherent anisotropy as it occurs on pluviation,
sands has been presented by Bishop (1971) Miura and thus masks the differences in compression
& Toki (1982) Hanzawa (1980) and Kuerbis & and extension response.
Vaid (1989). Thus, at a given void ratio the steady Moist dumped sands are, in fact, particularly
state may be reached in extension through con- prone to liquefaction due to their metastable
tractive deformation. However, no such state may honeycomb structure (Casagrande, 1976). Model-
occur in compression except through an intense ling loose water deposited sands by moist
dilative process with continued shearing. This will tamping might unjustifiably condemn them as
cause 03’ to increase to very high values by a con- being potentially liquefiable in triaxial compres-
tinued decrease in pore pressure so that the sand sion. However, they could be liquefiable under
gets transformed into a contractive one. Behav- other loading modes, such as extension.
iour of this type has been shown by Poulos (1981) The Authors conclude uniqueness of the steady
and Negussey, Wijewickreme & Vaid (1988) and state line of Erksak sand by looking at its trace in
it clearly implies non-uniqueness of steady state e-p’ space only. Uniqueness cannot be accepted
in e-p’ or e-cr3’ space. unless its trace in p’+ space is also unique. This is
The Authors’ conclusions regarding the clearly not the case, as evidenced by there not
uniqueness of the steady state line in e-p’ space of being a unique value of 4’ at the critical/steady
Erksak sand are based on compression and state. A large range of +,,‘/4,,’ (between 13” and
extension tests on moist tamped samples only. It 34”) is shown in Table 2 and Fig. 16. This is
660 DISCUSSION

clearly in conflict with critical state concepts. samples looser than critical in undrained tests.
Uniqueness of the critical state in both stress- Fig. 24 shows the results of a series of drained
void ratio and stress spaces has been demon- and undrained Nor-Sand simulations, using iden-
strated by Wroth (1958) in simple shear tests, and tical material parameters, but starting from differ-
Castro, Poulos, France & Enos (1982) have ent initial states. In real triaxial tests the
shown uniqueness of the steady state in stress maximum axial strain is limited to about 20%, at
space-ontrary to the Authors’ findings. which point the sample has undergone large
deformations. Therefore, for these simulations,
the ‘critical state’ was assumed to occur at the
Authors’ reply end of the test, i.e. 20% axial strain. There is a
Two different issues are raised in the dis- clear distinction between the S and F lines
cussion: practical design using sands and a more derived in this way, despite the fact that the
fundamental exploration of the critical state model explicitly includes a unique critical state
concept for sands. The intention of the Paper was locus. The discrepancy is due to the fact that at
to examine the critical state of sands, but both 20% strain the dense samples have not yet
issues are addressed here. reached an ultimate steady state. Fig. 24 also
Much of the Paper and discussion revolves shows that the F line, determined from undrained
around the question of the uniqueness of the criti- tests on loose samples, lies close to the CSL and
cal state line, as measured in the triaxial appar- may therefore be used in practice. In contrast, the
atus. It is interesting to invert the problem, and S line is below the CSL. This numerical example
pose the question as follows: If a unique critical illustrates that the interpretation of laboratory
state line for a sand is assumed to exist, and this experiments to determine the critical state of
line is assumed to be the ultimate steady state of sands requires care and may be problematical
all distortional processes to the soil under mono- for dense sand, even leaving aside questions
tonic loads, what would be measured in the tri- such as non-homogeneous samples and strain
axial test under different conditions? An answer localization.
to this question requires a constitutive model for Chu and Lo suggest that strain softening of
the triaxial test based on these assumptions to dense sand is not a material behaviour but is the
predict the result of tests. Such a model, called result of non-homogeneous deformation. This
Nor-Sand, has been developed by Jefferies (1992) suggestion can interpreted in two ways, and the
for drained loading and extended to undrained difference is important. First, non-homogeneous
loading by Jefferies & Been (1992). deformation could be caused by limitations of the
Nor-Sand was used to determine the location triaxial apparatus (e.g. end platen restraint); in
of the S line that would be determined from con- this case the softening is an artefact of the experi-
ditions in drained triaxial tests on samples ini- ment. The tests in this study used the convention-
tially denser than critical and the F line from al 2 : 1 sample height to diameter ratio as well as

M = 1.2 /- = 0432
0 Initial state
N = 0.2 A = 0.031
0 State at ~~ = 20%. undrained Nor-sand
h from LHL
X State at ~~ = 20%. drained Nor-sand

,
1000
p: kPa

Fig. 24. State diagram showing results of Nor-Sand simulations for drained and
undrained tests with S and F line criteria from limits of triaxial apparatus
 DISCUSSION 661

lubricated end platens and it appears unlikely may stem from the quasi-steady state being taken
that there was anything particular to the Erksak as the critical state (which it is not) and limi-
test programme to cause unrepresentative behav- tations of the triaxial tests to determine the CSL
iour. The alternative interpretation is that non- for dense sands.
homogeneous deformation could be a result of Alternatively, the difference could revolve
strain softening; this is predicted by plasticity around sample preparation technique. Vaid and
theory (Drucker’s stability postulate) under work Pillai suggest that moist tamped samples do not
softening conditions. We therefore regard non- give rise to pronounced anisotropy as do pluviat-
homogeneous deformation as a fundamental ed samples. This argument is rather speculative,
aspect of the sand behaviour. and we are unaware of any physical data on par-
The proposition that d,, may be a function of ticle packing arrangements to support this asser-
density has been questioned. As was pointed out tion. As noted in the Paper, there is an urgent
in the Paper and by Verdugo, the experimental need for a technique to quantify inherent aniso-
evidence is not convincing as there are extreme tropy properly so that this question can be exam-
problems with measuring very small deviatoric ined.
stresses in loose samples. Although there may be Until such time as soil structure can be mea-
intellectual grounds for a reduced +,, on very sured, Casagrande’s concept of a metastable
loose sands, for practical purposes it is likely to structure cannot be accepted as fact and used to
be sufhcient to use a single value of 4,, support the existence of a particular packing
Vaid and Pillai assert that non-uniqueness of arrangement in sands. In fact, the term metasta-
the CSL follows from a range in b,, quoted in the ble is contentious and may even convey an incor-
Paper. Notwithstanding that b,, may be constant, rect concept. The stress strain behaviour of a
a variable 4,, does not conflict with a critical drained test (D682) is shown in Fig. 6(b). This
state framework for soil provided that 4,, is a sample was as loose as samples which liquefied
single valued function of void ratio (which is what under undrained loading. The drained sample
was suggested in the Paper). In general, the CSL does not show a metastable collapse but moves
is unique provided it forms a single unfolded monotonically and steadily from the initial to the
surface in four-dimensional o,, (TV, 03, e space. final states. The behaviour of contractive or lique-
There is no requirement for this surface to have fiable sands can be explained adequately in terms
any particular shape, or indeed a linear projection of either the Nor-Sand critical state model
of transformed conical form in simplified plots. (Jefferies & Been, 1992) or a double hardening
A view that does emerge from the questioning model (Molenkamp, 1981) without invoking con-
of q-p’ uniqueness, but which was not discussed tentious concepts such as metastable particle
in the Paper, is the mapping between compres- arrangements.
sion and extension tests in q-p’ space. In general, It is interesting to note that Verdugo has a dif-
4,, is approximately the same under conditions of ferent view from Vaid and Pillai, and possibly an
triaxial extension and compression, and therefore explanation for differences between pluviated and
M = q,JP,,’ on the CSL in extension and com- moist tamped samples. He presents evidence that
pression space must be necessarily be different sample non-uniformity in terms of the distribu-
(Wood, 1990). Specifically in the case of Erksak tion of grain sizes (defined by him as ‘structure’)
sand, for all but the loosest samples, M = 1.2 in causes very different behaviours. This is also con-
compression and M = 0.85 in extension. It sistent with the well-known fact that small
follows that the CSL uniqueness in the e-log (p’) changes of grain size distribution can alter the
plane implies quite different critical state shear CSL of sands significantly. It would be interesting
strengths depending on stress path (as suggested to see more detailed results of grain size distribu-
by Vaid and Pillai). tions in test samples, and how shear strains
Vaid and Pillai argue that sand samples are within the samples relate to the grain size dis-
almost always anisotropic and therefore behav- tribution.
iour will be different in extension and compres- Verdugo also raises the point that the curva-
sion. We agree with this proposition; in fact the ture of the CSL is dependent on the axis chosen
data in the Paper strongly support this. However, to plot the line, and therefore little importance
this proposition cannot be extended to imply should be placed on the point where the curva-
non-uniqueness of the CSL. The Erksak test pro- ture changes. We agree that the approximation
gramme explicitly included several undrained used to describe the critical state (e.g. linear, bi-
extension tests and Fig. 13 compares the CSL of linear or curved) should be selected based on the
undrained extension and compression tests in the engineering problem to be solved. However, the
e-log (p’) plane. No substantial difference was reason for using a semi-logarithmic plot is rooted
found. The apparent difference between our tests in the empirical observation that soil stiffness is
and the results of Vaid, Chung & Keuris (1990) generally proportional to pressure. In fact, a
662 DISCUSSION

log-log plot may be more appropriate in avoid- Proceeding from a concern of localization, Chu
ing certain mathematical ambiguities (Butterfield, and Lo imply that unless it can be shown that a
1979). Further, crushed or sheared grains were dense drained sand approaches the same CSL as
observed in samples sheared at stresses greater encountered with loose undrained tests, then the
than 2 MPa and no such crushed grains were undrained CSL is inappropriate for character-
found in samples tested at 500 kPa or less. There izing sand behaviour. This concern is not born
is, therefore, some justification for a mechanistic out by experimental data. The critical state is
change in behaviour of sands in the stress range characterized by a sustained zero dilation rate.
of 1 MPa as suggested by the conventional semi- The closer the sample approaches the critical
logarithmic plot. state, the smaller the dilatancy. If one plots the
Whether or not the critical state concept is peak dilatancy with offset from the undrained
applicable to sands, and whether or not it can be CSL when this dilatancy is measured ($,), then a
used in practice, have been the subject of dis- plot such as that in Fig. 25 is obtained. Clearly,
cussion for some years. Our experience is that it is the trend of the drained tests is for zero dilatancy
very useful in design practice. However, some at the undrained CSL ($, = 0). This implies that
modifications to the original concepts are the drained and undrained CSL are the same.
required when dealing with sands. The irrelevance of the CSL to real sand behav-
The CSL forms a stable reference condition for iour is argued by Vaid and Pillai on the basis that
the quantification of sand behaviour and this is naturally pluviated sands are always denser than
useful for characterizing sand behaviour regard- the CSL measured by undrained compression
less of initial density (e.g. Been & Jefferies, 1985). tests. This raises another question about the
A general theory of sand behaviour predicated on application of critical state principles to sands:
the CSL has also been developed by Jefferies what constitutes normal, or virgin, consolidation
(1992). Whether the void ratio is greater or less of sands and where is the normal consolidation
than emax determined by an arbitrary laboratory line in relation to the CSL? This issue is not dis-
procedure is irrelevant. However, practical design cussed in detail here; field evidence shows that a
should take into account the major differences of wide range of states of the same sand is possible
behaviour observed in sands deposited under dif- when different pluviation techniques are used
ferent conditions and whether loading is in exten- during the construction of underwater sand fills
sion, compression or any other stress path. Use of (Jefferies, Rogers, Stewart, Shinde, James &
the CSL as a reference condition allows a more Williams-Fitzpatrick, 1988). The implied one-to-
precise definition of the state of the sand so that one mapping between a particular laboratory
other important aspects of sand behaviour, such procedure and full-scale construction experience
as stress path and anisotropy, can be compared does not exist and therefore definitive statements
on a rational basis. may not be made on the unstated hypothesis that

2.0
+
+
+
%
1-5 1 + + 29 sands

+
* A:+;+
+++ +$ ++
+.- +**++
+++++AN+++ +
+
++%+?$:
++
++p+iw
+!i &i2-yr+
t++*+
_~_____-.-.___

-0-5 -I I
-03 -0-2 -0-l 0 0.1
$‘m at peak dllatancy

Fig. 25. Dilatancy (d&,/d&,) in drained triaxial tests plotted as a function

of state (void ratio difference relative to CSL, JI,) from undrained tests on
loose samples
 DISCUSSION 663

such a mapping exists and controls sand behav- Proceedings of specialty conference on hydraulic $11
iour. structures, Fort Collins, 810-883.
We note that the critical state (or steady state, Kuerbis, R. & Vaid, Y. P. (1989). Undrained behaviour
or critical void ratio) of sands remains an impor- of clean and silty sands. Proc. 12th Int. Conf Soil
Mech., Rio de Janeiro, 91-100.
tant area of soil mechanics. However, further
Molenkamp, F. (1981) Elasto-plastic double hardening
work may be handicapped by model MONOT. LGM Report CO-218595, Delft
(a) the limitations of the triaxial apparatus to Geotechnics.
Miura, S. & Toki, S. (1982). A sample preparation
determine the ultimate state of samples which
method and its effect on static and cyclic
frequently occurs at strains greater than 20%
deformation-strength properties of sand. Soils &
(b) the absence of a practical method to quantify
Fdns 22, No. 1,61-77.
the packing arrangement (fabric) of particles Negussey, D., Wijewickreme, W. K. D. & Vaid, Y. P.
in a sand mass. (1988). Constant volume friction angle of granular
materials. Can. Geotech. J. 25, 50-55.
It seems that critical state concepts have much to
Poulos, S. J. (1981). The steady state of deformation. J.
offer to real civil engineering projects. Significant Geotech. Engng Am. Sot. Civ. Engrs 17, GT5, 553-
progress has been made in the development of 562.
critical state concepts for sands, but more work Poulos, S. G., Castro, G. & France, J. W. (1985). Lique-
needs to be done to clarify some of the remaining faction evaluation procedure. J. Geotech. Engng Am.
issues discussed. Sot. Civ. Engrs 111,No. 6, 772-792.
Seed, H., Seed, R., Harder, L. & Jong, H. (1988). Re-
evaluation of the slide in the Lower San Fernando
Dam in the Earthquake of Feb. 9, 1971. Report UCB/
REFERENCES
EERC-88/04.
Arthur, J. R. F. & Menzies, B. (1972). Inherent aniso-
Symes, M. J., Shibuya, S., Hight, D. W. 8c Gens, A.
tropy in a sand. GCotechnique 22, No. 115-128.
(1985). Liquefaction with cyclic principal stress rota-
Been, K. & Jefferies, M. G. (1985). A state parameter for
tion. Proc. 11th Int. Conf. Soil Mech., San Francisco
sands. Gtotechnique 35, No. 2, 99-l 12.
4, 1919.
Bishop, A. W. (1971). Shear strength parameters for
Tatsuoka, F., Sakamoto, M., Kawamura, T. & Fuku-
undisturbed and remolded soil specimens. In Roscoe
shima, S. (1986). Strength and deformation charac-
Memorial Symposium, Cambridge, pp 3-58.
teristics of sand in plane strain compression at
Butterfield, R. (1979). A natural compression law for
extremely low pressures. Soils Fdns 26, No. 1,65-84.
soils (an advance on e-log p’). Gtotechnique 29, No.
Vaid, Y., Chung, E. F. K. & Keurbis, R. H. (1990).
4,469-482.
Stress path and steady state. Can. Geotech. J. 27,
Casagrande, A. (1976). Liquefaction and cyclic deforma-
No. 1, 1-7.
tion of sands-a critical review. Harvard Soil Mecha-
Vasquez, A. & Dobry, R. (1988). The behaoior of
nics Series No. 88. Cambridge: Harvard University.
undrained contractive sand and its effect on the
Castro, G., Poulos, S., France, J. & Enos, J. (1982).
seismic liquefaction J?OW failure of earth structures.
Liquefaction induced by cyclic loading. Report to
Report of Rensselaer Polytechnic Institute.
National Science Foundation, Washington, DC.
Verdugo, R. (1989). Eflect offine content on the steady
Chu, J. (1991). Strain softening behaviour of a granular
state of deformation on sandy soils. Master thesis,
soil under strain path testing. PhD thesis, University
University of Tokyo.
of New South Wales.
Verdugo, R. & Ishihara, K. (1991). Characterization of
Chu, J., Lo, S.-C. R. & Lee, I. K. (1992a). Strain soften-
the undrained behavior of sandy soils. Int. Symp.
ing behaviour of granular soil in strain path testing.
Natural Disaster Reduction & Civ. Engng. Osaka:
Submitted for publication.
Japan Society of Civil Engineers.
Chu, J., Lo, S.-C. R. & Lee, I. K. (1992b). Strain soften-
Verdugo, R., Ishihara, K. & Towhata, I. (1989). Steady
ing behaviour of granular soil in triaxial and multi-
state on sandy soils. Proc. 25th Nat. Can. Soil Mech.
axial tests. Proc. 6th Aust.-NZ Co& Geomechanics,
Tokyo.
Christchurch.
Verdugo, R., Ishihara, K. & Towhata, I. (1991). Steady
Hanzawa, H. (1980). Undrained strength and stability of
state line as a reference state. Proc. 9th Panam. Conf:
quick sand. Soils & Fdns 20, 17-29.
Soil Mech., Vina de1 Mar, Chile.
Hettler, A. & Vardoulakis, I. (1984). Behaviour of dry
Verdugo, R., Ishihara, K., Daud, A. & Towhata, I.
sand tested in a large triaxial apparatus.
(1989). Effects of aging and fine content on the
Giotechnique 34, No. 2, 183-198.
undrained behavior of sand. Proc. 44th An. Conf
Jefferies, M. G. (1993). Nor-Sand: a simple critical state
Japan Sot. Ciu. Engrs 3,484-485.
model for sand. Gtotechnique 3, to be published.
Wood, D. M. (1990). Soil behaviour and critical state soil
Jefferies, M. G. & Been, K. (1992). Undrained response
mechanics. Cambridge: University Press.
of Nor-Sand. Proc. 45th Can. Geotech. Co&
Wroth C. P. (1958). Soil behaviour during shear-
Toronto.
existence of critical void ratios. Engineering 186,
Jefferies, M. G., Rogers, B. T., Stewart, H. R., Shinde,
409-413.
S. B. James, D. A. & Williams-Fitzpatrick, S. (1988).

View publication stats