LIQUEFACTION AND POSTLIQUEFACTION BEHAVIOR

OF SAND

By Y. P. Vaid,l Member, ASCE, and J. Thomas2

ABSTRACT: The static undrained behavior of saturated sand is shown to be dilative in

triaxial compression, even in the loosest deposited state. However, the behavior in triaxial
extension is contractive for relative densities of up to 60%, implying a profound anisotropy
of response to undrained loading. On monotonic loading, following liquefaction, the sand
always responds in a dilative manner even though it is contractive under static loading.
The postliquefaction response represents continuously stiffening behavior and an ap-
proach to any residual strength is not observed, regardless of density or effective stress
Downloaded from ascelibrary.org by East Carolina University on 09/29/14. Copyright ASCE. For personal use only; all rights reserved.

conditions prior to cyclic loading.

INTRODUCTION

Liquefaction of saturated sands has been the topic of extensive laboratory research over the
past 25 yr. The terms liquefaction and liquefaction failure encompass all phenomena involving
excessive deformations of saturated cohesionless soils (National 1985). Under static loading the
term liquefaction is associated with a strain softening type of undrained response, resulting in
either unlimited or limited flow deformation (see inset Fig. 1). Sand exhibiting such response
is termed contractive (Castro 1969). If strain softening does not occur the sand is called dilative.
Under cyclic undrained loading liquefaction can manifest itself either as a strain softening
response, much in the same manner as under static loading, or by the development of cyclic
mobility. Cyclic mobility is associated with excursions during the cyclic loading of the stress
state of sand through transient states of zero effective stress (<T~ = 0). The first time this <T.~ =
o occurred it was termed initial liquefaction (Seed 1979). On the conclusion of cyclic loading,
following liquefaction, residual conditions in the sand are normally assumed to correspond to
<T~ = O.
Until recently, the major concern during earthquake loading of saturated sands was to safe-
guard against the occurrence of liquefaction. If liquefaction under level ground was a possibility,
depending on the initial stress and density state of the sand together with the characteristics of
cyclic stresses imposed by the earthquake, remedial measures consisted in the densification of
sand. No attempt was generally made to estimate earthquake-induced displacements, in case
the liquefaction domain in a given sand structure stayed contained by the surrounding nonlique-
fied sand. However, during the last few years several researchers have emphasized a need for
the estimation of such displacements. As a result some empirical and analytical techniques,
toward this goal, have been proposed (Finn et aI. 1986, Hamada et al. 1987, Byrne 1990, Bartlett
and Youd 1992).
The key information required to estimate earthquake-induced displacement is the postlique-
faction stress-strain response of sand. Specifically, sand response, when it undergoes excursions
through states of zero effective stress, is needed when modeling the spatial progress of lique-
faction in a given earth structure. Little research has been carried out on this aspect of sand
behavior. If the sand is contractive under static loading, it has been assumed that its steady-
state (or residual) strength remains unaltered on monotonic loading following liquefaction in-
duced by cyclic loading (Bryne et al. 1992). The writers know of no published experimental
data that supports this assumption.
This paper presents an experimental study of the postliquefaction behavior of sand in the
triaxial test. Clearly, this requires a comprehensive investigation of the static and cyclic behavior,
which takes the sand to the liquefied state prior to the assessment of its postcyclic behavior.
The study encompasses static undrained behavior over a range of deposition densities, from
loosest to dense, and a range of confining stresses. Both triaxial compression and extension
loading paths are included to assess the possible path dependence of behavior. Cyclic loading
leading to liquefaction is studied at specific targeted densities, from loose to dense, and a range
of confining stress levels is used. Finally, postliquefaction monotonic undrained response is
studied as it is influenced by factors such as the maximum shear strain due to cyclic loading,
relative density, mode of loading, and the level of confining stress prior to cyclic loading.

'Prof., Dept. of Civ. Engrg., Univ. of British Columbia, Vancouver, Be, Canada.
°Grad. Student, Dept. of Civ. Engrg., Univ. of British Columbia, Vancouver, BC, Canada.
Note. Discussion open until July 1, 1995. To extend the closing date one month, a written request must be
filed with the ASCE Manager of Journals. The manuscript for this paper was submitted for review and possible
publication on June 8, 1993. This paper is part of the Journal of Geotechnical Engineering, Vol. 121. No.2,
February. 1995. ©ASCE, ISSN 0733-9410/95/0002-0163-0173/$2.00 + $.25 per page. Paper No. 6322.

JOURNAL OF GEOTECHNICAL ENGINEERING 163

J. Geotech. Engrg. 1995.121:163-173.

 -..... 250
C'Cl
c.. -unvr.d UqumCllon
~
"-

~ E.
~ 40
b 0
I
b
"-
Downloaded from ascelibrary.org by East Carolina University on 09/29/14. Copyright ASCE. For personal use only; all rights reserved.

200 (a)
- 250-+----,----t-------r----j
-10 -5 0 5 10
Axial Strain, E:.(S15)
-.....
~ 300
C. 150
~
~ 0-i=:t;::nmr;f- ;....----~---+_--_1

b
1-150 (b)
b
"--300-j-----,-----,-----,----.I I

o 400 800 1200

(cr'v+cr'H)/2 (kPa)
FIG. 1. Static Undrained Behavior of Loosest Deposited Fraser River Sand

The possible similarity of the postcyclic behavior of sand brought to the liquefied state by
cyclic loading and by a static load/unload cycle is also investigated.

EXPERIMENTATION
Tests were performed on Fraser River sand, which underlies the highly seismic populated
Fraser Delta in British Columbia, Canada. The sand used has grain sizes ranging from 0.074
mm to 0.6 mm and comprises about 98% of the original material dredged from the river.
Maximum and minimum void ratios, in accordance with ASTM 02049, are 1.00 and 0.68, and
D so = 0.3 mm and e" = 1.5. The average mineral composition is 40% quartz II % feldspar,
45% unaltered rock fragments, and 4% other minerals.
Triaxial specimens used were were 63 mm diameter x 126 mm high. They were reconstituted
saturated by water pluviation. This reconstituting technique yields homogeneous replicable
specimens (Vaid and Negussey 1988) that possess fabric similar to those of water deposited
natural or tailings sands (Oda 1972). Laboratory studies on pluviated sands should, therefore,
duplicate the characteristic behavior of natural and hydraulic fill sand. Specimens were deposited
in the loosest state and then densified, as needed, by low-energy high-frequency vibrations.

TEST RESULTS AND COMMENT

Static Loading Behavior

Behavior of Loosest Deposited Sand

Fig. 1 shows the triaxial compression and extension responses of isotropically consolidated
sand at several levels of confining pressure. Deviator stress is defined as a,. - all to facilitate
distinction between compression and extension loading modes. Compression behavior may be
seen to be dilative except under the highest consolidation stress a;, = 1,200 kPa, when it
manifests a slightly contractive response. In contrast, the behavior in extension at each confining
stress is contractive. At lower levels of a;" including 200 kPa, the response is of the liquefaction
type (see inset Fig. 1) with steady-state conditions realized at axial strains of 2%-3%. At higher

164 JOURNAL OF GEOTECHNICAL ENGINEERING

J. Geotech. Engrg. 1995.121:163-173.

 a;, levels, the behavior changes to the limited liquefaction type (Castro 1961J). The specimens
did not show any tendency towards necking, which is regarded as the onset of stress/strain
inhomogeneities within the specimen, even after being loaded to a strain level of up to lJ{'!r, in
extension. Steady or phase transformation states were reached well before this strain level.
The test results in Fig. 1 illustrate the effect of confining stress on undrained response. The
response is dilative in compression, and, by contrast, is contractive in extension. In compression,
an increasing a;, results in a less-dilative behavior, eventually turning into a contractive response
at the highest a;, used; whereas the effect of increasing a;, in extension is to cause a less-
contractive response. At a given placement relative density Dr;' consolidation to a higher (J;,
implies increasing the consolidated relative density D". Higher density promotes less, and a
larger a;, promotes a more contractive response. The results in Fig. I suggest that the effect of
increasing a;, offsets the effect of densification in compression, whereas the opposite is true
during extension loading.
Pluviated sand is contractive in extension over a much larger range of consolidated relative
Downloaded from ascelibrary.org by East Carolina University on 09/29/14. Copyright ASCE. For personal use only; all rights reserved.

density Dr' than it is in compression; this has been demonstrated by several investigators [e.g.,
Bishop (1971), Hanzawa (1980), Miura and Toki (llJ82), Kuerbis and Vaid (llJiN). and others].
The loosest Fraser River sand is dilative in compression, except under high confining stresses.
Inherent anisotropy in pluviated sand [e.g., Arthur and Menzies (llJ72) 1is considered responsible
for the differences in compression and extension behavior for a given initial state.

Behavior at Fixed Density States

The undrained static response of the sand at three fixed consolidated relative densities, D"
of 19%, 40%, and 59%, was assessed (Fig. 2). This enables the isolation of the effect of confining
stress alone on the stress-strain response. A clear evidence of a less dilative behavior may be
noted in compression, with an increasing confining stress at each density state. In extension,
however, the contractive response of the liquefaction type at a lower a;, changes to a limited
liquefaction type, with an increase in confining stress. This amounts to a decreased contractive
tendency not commonly associated with increasing confining stresses. The contractive behavior
is exhibited by Fraser River sand over a large range of (loosest to more than 50%) placement
density in extension loading. But in compression loading, the contractive response was associated
with only the loosest depositional density, and when consolidation occurred at high confining
stresses.

Triggering Contractive Deformation

The effective stress ratio a; la~ at the peak of deviator stress is referred to as the critical stress
ratio (CSR) in this paper. The friction angle <PCSR mobilized at the peaks of deviator stress, for
the states that showed contractive response, is presented in Fig. 3 as a function of the placement

200

100

-100

?
ll. 300
..Ill
' - 150

~ 0
~
I -150 Drc =40l1! (b)
6"
'--300 45 Compreslion ••••• k.=1

300
'..=400kPa rn
ill
ill 30 / 000001<.=2

150
0
~
bll
ill
"t:I
15
• 0

,?, tt.
..
0:
V1
u Enenolon
-150 Drc =59l1l (c) -iT
-300-!-----,----!-----,-----I 0
-10 -5 0 5 10 0 10 20 30 40 50 60
Axial Strain, e.{lIl) Relative density. Dri ('Ki)
FIG. 2. Static Undrained Behavior at Severai Relative Densi- FiG. 3. Effective Stress Conditions at Triggering of Contrac-
ties tive Deformation

JOURNAL OF GEOTECHNICAL ENGINEERING 165

J. Geotech. Engrg. 1995.121:163-173.

 void ratio ei . Some data on specimens that were initially anisotropically consolidated under K,
= u{,1u;, are also shown in the figure.
In compression cl>CSR is observed to be essentially constant at about 26 degrees, not dependent
on initial stress conditions. Extension loading cl>CSRS, however, are much smaller than the
compression value and tend to increase with an increase in placement density. Like compression,
static shear stress prior to straining does not influence cl>CSR' A lower cl>CSR in extension loading
implies that triggering of liquefaction will occur in the extension phase during cyclic loading
(Vaid et al. 1989). The contractive behavior is exhibited by Fraser River sand over a large range;
from the loosest to more than 50% placement density in extension loading. But in compression
loading, the contractive response was associated with the loosest depositional density only.
Evidence in support of an unique cl>CSR' at which contractive deformation is triggered, also
exists for other sands (Bishop 1971, Vaid et al. 1985). However, Vaid et al. (1990) showed that
extension cl>CSR values are much lower than compression values for Ottawa and tailings sands.
These values increase with an increase in the initial sample preparation density, in a manner
Downloaded from ascelibrary.org by East Carolina University on 09/29/14. Copyright ASCE. For personal use only; all rights reserved.

similar to that observed for Fraser River sand.

Steady State and Phase Transformation (PT) States

Effective stress conditions at the steady state or phase transformation (PT) state [Ishihara
et al. (1975), see Fig. 1 inset], in the modified Mohr diagram, may be seen to lie on unique
straight lines passing through the origin regardless of relative density, initial stress state ((T~, and
KJ, type of response (contractive or dilative), and mode of loading (Fig. 4). These lines have
equal slope, implying that the friction angle at the steady state and PT state for both contractive
and dilative responses are equal (at 32°) and independent of the mode of loading, as was earlier
observed by Vaid et al. (1985, 1990) for other sands.
Fig. 5 shows the relationship between the minimum undrained shear strength SPT( = u)2) at
PT [termed quasi-steady state by Ishihara (1993)] or steady state and the consolidated void ratio
en for a range of initial confining stresses for contractive response in extension loading. As
noted earlier, contractive response did not occur in compression, except under very high stresses.
At a given e" the shear strength at PT or steady state, SPT or Sss, increases with the initial
confining stress. This is contrary to the common belief, based only on compression loading, that
this shear strength is a function of the consolidated void ratio only. For Fraser River sand no
such unique relationship between SPT or Sss and e, is noted; but different relationships exist,
each characteristic to a given u;,. For a given u~, the initial static shear stress does not influence
the SPT or Sss versus e, relationship.

Static Load/ Unload Behavior

Fig. 6 shows the static undrained loading-unloading behavior of the sand at Dr' = 19% and
(T~, = 100 kPa. Its response is dilative in compression but contractive, of the liquefaction type,
in extension. On unloading of the shear stress both specimens liquefy; i.e., they end up in a
state of u;, = O. This type of behavior will later be referred to as liquefaction induced by a
static load/unload cycle, as opposed to liquefaction induced by cyclic loading.

400..,------------------,
00000 LlquefacUon
A<lA<lA LImited liquefacUon
cccoc DilaUve
••••• k.s::2

O-liE=----------------j 200 . . . . - - - - - - - - - - - - - - ,
u'so
Extension ~ ••••• 100kPa
--;-150- """"" 200
00000300
l:l.. ccccc 400
-200 C100- ~ 00000 1200
••••• k.=2

J50-~
a I
...."
-400+-.....--.,--...,.----.--,---.----.---1 1.00
o 200 400 600 800 0.80 0.85 0.90 0.95
(cr'v+cr'aJ/2 (kPa) Void ratio, eo
FIG. 4. Effective Stress Conditions at Phase Transformation FIG. 5. Undrained Steady State or Phase Transformation
and Steady States Strength in Extension

166 JOURNAL OF GEOTECHNICAL ENGINEERING

J. Geotech. Engrg. 1995.121:163-173.

 ? 200
c.. 150
C- 100
~ 50
1I
O+--_-===d----:--:-:-::-------L---j
cr'3c=100
-50 D~- -In... (a)
b> ~
--100-+----4-----.----..,--------1 1.0 Je~ 0
o
-~
-5 5 10 15
Axial Strain, f:. (95)
o.•
~t
? 200 0.6 -
o
c.. b
Downloaded from ascelibrary.org by East Carolina University on 09/29/14. Copyright ASCE. For personal use only; all rights reserved.

::::.::: ••• ~ ~ Thermalito after bay dam foundation

C- 100 0.4 - .......... Fort Peck dam shell
00000 Saoramento river sand. D.=38.80.78.10llJl
~ 00000 Monterey 0 sand. D.=5011
a::n:IJ Reid Bedford sand. D =.0,8011
~ 0-+.-...::::::..-+--------------1 0.2 - ••••• New Jersey baokfill, W[ RC=9511
b ••••• huer river sand. D..=19,.0d:5911
I (b)
,>
~ -100-+-----.,--------,.------,---'
o. 0 ~01-"'---4-0rO--r-8-0rO---r-1-2"O-O"'---1-6r'OO---r-2-0'IO-O"'-2-4'IO--jO
o 200 400 600 Effective confining stress, u'3.(kPa)
(cr'v+cr'aJ/2 (kPa) (Mter Seed and Harder. 1990)

FIG. 6. Undrained Static Load/Unload Behavior FIG. 7. Confining Stress Effect on Resistance to Liquefaction

A static load/unload cycle did not always result in the state (J"~ = 0, upon unloading. This
situation occurred in dense sand (D re = 59%) consolidated to (J"~c = 400 kPa and higher in the
compression mode of loading, and was apparently due to an insufficient level of strain prior to
unloading. It was found that straining on static loading must surpass a minimum level beyond
the PT state before a state of zero effective stress state is realized on unloading.

Cyclic Loading Behavior

This behavior was assessed for three targeted relative densities, and at three levels of confining
stress (J"~c on isotropically consolidated sand. Cyclic loading resistance (or resistance to lique-
faction) is defined here as the cyclic stress ratio (J"dcy/2(J"~c' which causes a maximum compressional
or extensional axial strain c" amplitude of 2.5% in 10 cycles. Strain development due to cyclic
loading occurred as a consequence of either contractive deformation or cyclic mobility, depending
upon the initial state Dr" (J"~c' and the associated SPT or Sss together with the amplitude of cyclic
stress T" = (J"d".12 [Castro et al. (1982), Vaid et al. (1985)].
The K" value (the ratio of cyclic resistance at (J"~c divided by this resistance at (J"~c = 100 kPa)
for Fraser River sand is shown in Fig. 7 as a function of confining stress, together with data
from the literature on other sands (Seed and Harder 1990). The correction factor K" may be
seen to be a function of both confining stress and relative density. The resistance to liquefaction
reduces by a maximum of about 20% at the highest confining stress of 1,200 kPa for the dense
59% relative density. Most of the reduction occurs between (J"~c = 100 to about 600 kPa, and
it seems that much further drop in K" is not likely for confining stresses in the excess of about
600 kPa.
There is a wide range of reported K" values for different soils at a given confining stress (Fig.
7). Lumping data without regard to the relative density state may contribute to this large K"
range. For granular materials in which the relative density has been specified, such as Sacramento
River sand, a clear decrease in K" with the relative density may be seen. Similar behavior was
noted for a tailing sand and Ottawa sand under simple shear conditions (Vaid et al. 1985).
It may be pointed out that for the loose sands tested, which have the largest susceptibility of
liquefaction, K" is approximately unity regardless of the confining stress level. The adoption of
lower values in design, based on an average K" value from the body of data in Fig. 7, may lead
to a conservative design. Substantial economic advantage could accrue in design by evaluating
K" on a site-specific basis, if other loose sands demonstrate behavior similar to that of Fraser
River sand.

Residual Condition on Conclusion of Cylic Loading

When cyclic loading was terminated, a state of zero effective stress «J"~ = 0 or 100% pore
pressure ratio) was realized in most cases, except for specimens at medium and dense relative
densities (40% and 59%), under confining stresses of 400 kPa and higher. This state of (J"~ =
o occurred for the first time on conclusion of the last loading cycle in which the specified c" 2::
JOURNAL OF GEOTECHNICAL ENGINEERING 167

J. Geotech. Engrg. 1995.121:163-173.

 150

~
"?
ll. 100 - Drc = 1958'
C 50 - cr'3c=100 kPa
~ o ----------------
~
b -50 - .----- Cyclic loadi Ilg
I - - Post cylcic mono. loading
-~
b-l00
"- -16 -12 -6 0 6 12 16
Axial Strain, t. (If)

150
(b)
""' 150
Downloaded from ascelibrary.org by East Carolina University on 09/29/14. Copyright ASCE. For personal use only; all rights reserved.

ell ,
I:l. 100 "? cr'3c=100 kPa \9 ,,
~
"- I:l. I ,,
~ I
50 "- 100 ,,'40
~ I

~
b 0 -.' ~ I "
~ 50 I
,"
"
I
b -50
b
I :/' I

~
"- 0 100 200 300 400 0
( cr'v+ cr 'aJ/2 (kPa) -10 -5 0 5 10
FIG. 8. IT'; = 0 States Following Liquefaction by Cyclic Load- Axial Strain, t. (If)

ing FIG. 9. Relative Density Effect on Postliquefaction Response

2.5% developed. A typical example is shown in Fig. 8(a). The last half cycle of extension loading
may be seen to cause liquefaction by contractive deformation. and the unloading phase brings
the specimen to the a~ = 0 state. Thus excursion through a transient state of (J~ = 0 did not
take place before the specified level of strain development, defined as liquefaction, occurred.
Even for cyclic tests in which the residual finite effective stress remained at the conclusion of
cyclic loading [Fig. 8(b)], strain development as a result of cyclic loading was not due to any
excursion through a transient a~ = 0 state. Such states would have been achieved only if further
loading cycles were applied. In that case 10", much larger than the specified 2.5%, would have
developed.
A state of zero effective stress was also not realized by a static load/unload cycle in medium
and dense sand unless the strain during loading exceeded a certain minimum value. On conclusion
of cyclic loading. dense and medium dense specimens at a;, 2'" 400 kPa developed a strain level
of only 2.5%-5%. Static load/unload tests indicated that a minimum strain level of about 7%
was needed during static loading to achieve a a~ = 0 state on unloading.

Postcyclic Undrained Monotonic Behavior

The residual strain at the end of cyclic loading was always extensional. This was because the
specified strain, 10" 2'" 2.5%, during the last cycle developed on the extension side at the instant
of peak stress amplitude or contractive deformation in extension, and little strain recovery
occurred when the shear stress was unloaded to zero. The postliquefaction behavior of test
samples, which ended with the residual states a~ = 0 on conclusion of cyclic loading, is presented
first.

Stress Strain Response

Fig. 8(a) showed the response during the last loading cycle along with the postcyclic monotonic
stress-strain response of loose sand under a~, = 100 kPa prior to cyclic loading. The axial strains
shown are based on the sample configuration prior to cyclic loading. During postliquefaction
undrained compression loading. the sand deformed at virtually zero stiffness over a large range
of axial strain (almost 20%). With further straining, the stiffness increases with an increase in
strain. This stress-strain response in which the modulus increases with an increase in axial strain
is opposite to the usual response of soil in which the modulus decreases with an increase in
strain. The unusual stress-strain response of the liquefied soil results from the fact that upon
shearing the soil dilates all the way causing the effective stress to increase. The deformation
progresses at a mobilized friction angle (36°), which equals the angle of maximum obliquity
under static loading. After some axial strain the stress-strain curve essentially becomes linear
and there is no tendency towards approaching a residual strength, even after a postliquefaction
strain of about 32%. Under static loading. this sand would be contractive in extension. with a
residual (steady state) strength of only 5 kPa. In compression, however, it would be dilative.

168 JOURNAL OF GEOTECHNICAL ENGINEERING

J. Geotech. Engrg. 1995.121:163-173.

 Dependence on Relative Density
Fig. 9 shows postliquefaction monotonic response at three relative densities for a fixed a;,
100 kPa prior to cyclic loading. To facilitate comparison the response of dense sand is taken as
the reference curve and the stress-strain curves at other relative densities are horizontally shifted
to match the ad = 5 kPa point on each curve. The rate of stiffness increase beyond this 5 kPa
point increases as the relative density increases, and the axial strain, at which the curves become
practically linear, decreases with an increase in relative density. The postliquefaction deformation
proceeds along the average line of maximum obliquity observed under static loading (</>' = 36°),
regardless of the relative density state [see Fig. 8(b)]. The region of strain over which the stiffness
is close to zero is less for dense than for loose sand.

Comparison with Behavior Following a~ = 0 Induced by Static Loading and Unloading

Downloaded from ascelibrary.org by East Carolina University on 09/29/14. Copyright ASCE. For personal use only; all rights reserved.

In Fig. 10 the typical response of the sand following the state of a~ = 0, induced by a static
load/unload cycle, is compared with that of the sand brought to a state of a~ = 0 by cyclic
loading. As in the preceding section, the response of the cyclically loaded sample is taken as
the reference curve and the response curves of statically loaded samples are horizontally shifted
to match the ad = 5 kPa point on each curve. The postliquefaction responses are noted to be
essentially similar at each relative density regardless of the manner by which the state a~ = 0
was brought about. Thus, a convenient way of obtaining the postliquefaction response would
be to use a static load/unload cycle, instead of cyclic loading, to induce liquefaction (a~ = 0).

Comparison of Precyclic and Postcyclic Behavior

The preliquefaction and postliquefaction stress-strain response of Fraser River sand at D
40% and a;, = 100 kPa is compared in Fig. 11. The stiffness of the sand decreases with" an
increase in strain until the phase transformation state for the sand subjected to preliquefaction
static loading, but it increases continuously with strain during postliquefaction monotonic load-
ing. The contractive behavior manifested during precyclic static loading was eliminated as a
consequence of liquefaction induced by cyclic loading. The postcyclic stress-strain response is
always dilative, the stiffness is very small during the initial phase of loading, but with an increase
in the strain level the stiffness essentially becomes equal to the precyclic value in the post-PT
region. No indication of any residual strength condition on postcyclic loading is apparent,
regardless of the density state or the mode of loading.

l 150 -,---(1-'3o-=-10-0-k-P-a-----:"rr---

~ 100 - - Statically liquefied V

---- Cyclically liquefied
~
1 I
50
(a) D.. =l~
b
'- O+----,......=:...---.-----.-------j
o

? 150
c.. (1'30=100 kPa
~ 100
- - Statically liquefied
- - Cyclically liquefied

50
(b)
O+----r---""'T-----.-------j
-10 -5 0 5 10
Axial Strain, £.(~)
=-c.. 200

~'
? 150 - , - - - - - - - - - - - - - , , . . . . - - - , .----- Post liquefactio
c.. (1'30=100 kPa ~ 100 - - - Pre liquefaction
~ 100 - - Statically liquefied
- - Cyclically liquefied
~ . .-
50
(c)
0
..
11-100 _.' .. .---- ---
~ (1'30=100 kPa Drc=40sg
O+----r------r--'::::O'::~--~ '-
-15 -10 -5 0 5
-200
Axial Strain, £.(!Ifl)
-15 -10 -5 0 5 10
FIG. 10. Comparison of Monotonic Response Following Liq-
Axial Strain,e a (sg)
uefaction by Cyclic and Static Load/Unload Cycle FIG. 11. Comparison of Static and Postliquefaction Response

JOURNAL OF GEOTECHNICAL ENGINEERING 169

J. Geotech. Engrg. 1995.121:163-173.

 Range at' Postliquefaction Behavior
The postliquefaction undrained stress-strain curve can be characterized into three distinct
regions [Fig. 8(a)]. Region 1 spans from the state at which a; = 0 at the end of cyclic loading
until a measurable ad = 5 kPa develops on some shearing. This is the region with very small
stiffness. The size of this region decreases with an increase in the relative density. Region 2
commences from ad = 5 kPa and can be approximated as a parabolic curve representing
continuously increasing stiffness with axial strain. Region three corresponds with the linear
segment of the stress-strain curve, representing a constant modulus.
An approximate length of region I may be taken as the axial strain increment ~E from the
point at which E max developed in bringing about liquefaction until the point at which a measurable
ad = 5 kPa is recorded. As shown in Fig. 12, ~E is a function of the maximum strain amplitude
E max developed during cyclic loading and decreases with increasing density. Average values of
~E for the tested sand are 20%, 3.5%. and 2% for the 19%.40%, and 59% relative density
states. respectively.
Downloaded from ascelibrary.org by East Carolina University on 09/29/14. Copyright ASCE. For personal use only; all rights reserved.

Fig. 13 shows the range of the postliquefaction compression stress-strain response at relative
densities of 19%, 40%. and 59%. Only the curved portions of the response. which start at ad
= 5 kPa and extend to the beginning of the linear segments. are shown. As before, all curves
are horizontally translated to merge at ad = 5 kPa. There is a definite trend for the response
to become stiffer with an increase in confining stress for the sand at a loose relative density.
But as the relative density increases, the effect of confining stress is not so apparent.
The stress-strain response in region 3 is a straight line in all cases. Its slope increases with an
increase in the relative density. For the loose sand the slope of the line is somewhat larger for
higher confining stress. but no such dependence seems to exist at higher relative densities.
Fig. 13 also shows the range of post liquefaction monotonic behavior in extension. Both
deviator stress and axial strain are taken as positive for a direct comparison with the compression
behavior. Unlike the compression behavior. loose sand does not appear to show the specific
dependence of the extension behavior with the level of confining stress prior to liquefaction.
The spread of stress-strain curves at each relative density is of the same order as that observed
under compression loading. However. at each relative density the extensional response is sub-
stantially softer than the compressional response. implying a stress-path dependent postlique-
faction behavior as was the case with the preliquefaction behavior.
Behavior if Residual State a; 0/= 0
Clearly. the postcyclic behavior for a given a;, prior to cyclic loading will depend on the
magnitude of residual a; remaining after cyclic loading. This is illustrated in Fig. 14 for D,e =

-;' 250
a.. 200 - - 100 kPa
~ - - 400kPa
150
~ 100
1I
~
5 10
Axial Strain, e.(~)

-;.;' 250 _ _ 100 kPa

a.. 200 400 kPa
~ 150 - - - 600kPa
~ ~ 1200 kPa

1bI 50
100 J~~~~~=~.:zt:.:"'~lo~n~~=~~
= D.. 4()sg

25 '-- 0
o 5 10
20 Axial Strain. e.(~)

--.. 15 -;' 250

~ a.. 200
10 C.
eztenslon------
c..l
<l 150
/ - - - 100 kPa
5 ~ 100 400 kPa
0 1 I 50
0 5 10 15 20 b
'-- o...j::::~=:=----,.-------l
Maximum strain. emu: ($f») o 5 10
Axial Strain. e.(~)
FIG. 12. Strain Required to Mobilize Finite (J'd = 5 kPa on
Postliquefaction Monotonic Loading as Function of Maximum FIG. 13. Range of Postliquefaction Monotonic Response for
Cyclic Strain Residual (T; = 0 States

170 JOURNAL OF GEOTECHNICAL ENGINEERING

J. Geotech. Engrg. 1995.121:163-173.

 300

-..... ~
",0 *'
f!)0

(ll 11..
a.. 200 fSI'
~ q JS
'--
6"'
~
25 -l---""'2!1l"/""----,-...,

?I 100
b
'--
Drc =4()s'g 2

0
Downloaded from ascelibrary.org by East Carolina University on 09/29/14. Copyright ASCE. For personal use only; all rights reserved.

0 5 10 15 2
Axial Strain, e.(%)
FIG. 14. Postliquefaction Monotonic Response for IT; *- 0 Residual States

40%, at (J;, = 400 kPa. For postcyclic residual states of (J~ = 8.25. and 45 kPa, the sand had
liquefied (developed axial strain between 3.5% and 3.7%). However. the maximum strain
developed for the residual (J~ states of 105 and 175 kPa was less than 0.4% and. according to
definition, the sand had not liquefied. An exploded view of the stress-strain response in the
earlier stage is shown in the inset.
In Fig. 14, regardless of the level of residual (J~. the postcyclic stress-strain response is similar
to the precyclic response, in that the modulus first decreases with strain before it starts to
increase, following the initiation of dilation corresponding to the PT state. The stress-strain
curves move progressively higher and the axial strain, until the phase transformation state
decreases, as the level of residual (J~ increases. Thus the undrained stress-strain behavior of
sand. at deformation levels typical of concern during earthquakes. does not correspond to moduli
that continually degrade with strain as assumed by most researchers during the effective stress
analysis of earthquake problems [e.g., Finn et a!. (1986), Prevost (1981)J. Depending on the
residual effective stress state, the modulus on postcyclic loading does decrease initially. but
starts increasing once dilation commences.

PRACTICAL IMPLICATIONS
The realistic characterization of saturated fluvial and hydraulic fill sands and silty sands,
regarding their susceptibility to liquefaction and the nature of postliquefaction stress-strain
response, must consider:

1. Reconstituting specimens in the laboratory by water deposition to simulate fabric iden-

tical to that of the real materials. This will be necessary only if undisturbed specimens
are not available.
2. Undrained response may be strongly affected by the direction of loading (e.g.. inclination
a of (J, toward the deposition direction) due to their inherently anisotropic nature. on
account of the pluvial mode of deposition.

The results presented in this paper show that water-deposited Fraser River sand like many
other sands is not contractive in triaxial compression, even in the loosest deposited state. How-
ever, laboratory techniques that use sample reconstitution by moist tamping [e .g .. Castro et al.
(1982)] show sands contractive in compression over a range of void ratios. Moist tamping must,
therefore, impart to the sand a fabric that promotes contractive behavior. Casagrande (1976)
pointed out that moist dumped sands are particularly prone to liquefaction because of their
metastable honeycomb structure. Most tamped specimens are frequently prepared in the lab-
oratory at void ratios not even accessible to the sand; i.e., looser than the ASTM loosest value.
Recent studies on Duncan Dam foundation sand (Vaid and Pillai 1992) that was sampled after
the freezing of the ground (Duncan Dam foundation sand is considered very loose) show that
the in-situ void ratio is nowhere greater than the void ratio obtained after consolidation to in-
situ stresses following water deposition to the ASTM maximum void ratio state. Modeling loose
water-deposited sands by moist tamping may then unjustifiably condemn them as being poten-
tially liquefiable in compression.
Though dilative in compression (a = 0), Fraser River sand is contractive in triaxial extension
(a = 90°). confirming dramatic undrained anisotropic behavior, similar to that of other water-
deposited sands. Studies by Symes et al. (1985) and Shibuya and Hight (1987) on water-deposited
sands provide further comprehensive evidence of their anisotropic behavior in that a systematic

JOURNAL OF GEOTECHNICAL ENGINEERING 171

J. Geotech. Engrg. 1995.121:163-173.

 transformation of dilative behavior to a pronounced contractive response occurs as ex increases
from 0° to 90°. This clearly has serious implications in the steady-state concepts of design (Castro
et al. 1985), in which a unique steady-state line obtained only from compression tests is the key
assumption. If the surface being analyzed through a hydraulic fill embankment (for its potential
for flow failure due to liquefaction) is relatively flat, as was the case for the San Fernando dam
(Castro et al. 1985), the use of compression steady-state strength will be totally inappropriate.
The surface analyzed by Castro et al. (1985) corresponded with ex ~ 45 + <1»2, in which <1>.\ is
the friction angle at steady state; with <1>.\ = 37°, ex = 63°, which represents a loading mode
closer to triaxial extension than triaxial compression. Consequently, in the back analysis a
reduction factor of 20, to the measured steady-state strength in compression, to obtain the
strength operating at the instant of failure may not have been necessary to obtain agreement
with the field event if the appropriate ex dependence of steady-state strength was taken into
account.
The postliquefaction stress-strain response of the sand is always dilative and the modulus
Downloaded from ascelibrary.org by East Carolina University on 09/29/14. Copyright ASCE. For personal use only; all rights reserved.

increases with strain for cases with residual u; = O. The current methods of modeling sand in
this postliquefaction domain do not recognize this feature of sand behavior. Instead, in most
cases the hyperbolic type of response, in which the modulus continuously decreases with strain,
is assumed [e.g., Finn et al. (1986)]. Clearly, this assumption would predict unrealistic post-
liquefaction deformation. The test results of Fraser River sand further show the postcyclic
behavior is very much affected by the level of maximum strain developed during cyclic loading,
along with the magnitude of the effective confining stress remaining following the cessation of
cyclic loading.

CONCLUSIONS
Fraser River sand, a medium uniform sand, responds in static undrained loading in a dilative
manner even in the loosest deposited state under confining stresses of up to 1,000 kPa. However,
in extension the sand behaves in a contractive manner over a range of densities, from loose to
just under 60% relative density, implying a strong direction-dependent behavior. This has serious
implications in the steady state concepts of design, which assume a unique steady-state line for
sand obtained from compression tests. In fact, the sand may not even be contractive in compres-
sion.
K", the factor that represents the dependence of cyclic loading resistance on the level of initial
confining stress, was found to be less than one only when the sand was not loose. Even then,
the K" values noted were never less than 0.7 for confining stresses of up to 1,200 kPa.
On postliquefaction monotonic loading the sand that developed a 100% pore-pressure ratio
deformed initially with an essentially zero stiffness, which then increased with the level of strain,
until at some strain level it became constant. The rate of buildup of deviator stress with strain
increased with relative density, and the strain at which stiffness became essentially constant
decreased with an increase in the relative density. The precyclic contractive response was elim-
inated as a consequence of the cyclic loading phase. No indication of any residual strength
condition on postliquefaction loading was apparent, regardless of the density state or the mode
of loading.
A convenient way to assess postliquefaction behavior is to liquefy sand by a static load/unload
cycle instead of by cyclic loading.

ACKNOWLEDGMENTS
The research for this paper was supported by a grant from the Natural Scienccs and Engineering Research
Council of Canada. Kelly Lamb's assistance in the preparation of the manuscript is gratefully acknowledged.

APPENDIX I. REFERENCES
Arthur. J. R. F., and Menzies, B. (1972). "Inherent anisotropy in a sand." Ceotechnique, London. England.
22(1),115-128.
Bartlett. S.. and Youd. L. (1992). "Empirical prediction of lateral spread displacement." Proc.. US-Japan Work-
shop, Nat. Ctr. for Earthquake Engrg. Res. (NCEER), SUNY, Buffalo, N.Y.
Bishop. A. W. (1971). "Shear strength parameters for undisturbed and remolded soil specimens." Proc.. Roscoe
Membership Symp., Cambridge Univ.. Cambridge, England, 3-58.
Byrne, P. M. (1990). "A model for predicting liquefaction induced displacements." Soil Mechanics Sa. No. 147,
Dept. of Civ. Engrg., Univ. of British Columbia, Vancouver, Canada.
Byrne, P. M., Jitno, H., and Jeremy, H. (1992). "A procedure for predicting the seismic response of tailings
impoundments." Proc., Ceotechnique and Natural Hazards; Ceo Hazards 92, Vancouver Geotech. Soc.. Van-
couver, Canada, 281-289.
Casagrande, A. (1976). "Liquefaction and cyclic deformations of sands-a critical review." Harvard Soil Mech.
Ser. No. 88, Harvard Univ., Cambridge, Mass.
Castro, G. (1969). "Liquefaction of Sands," PhD thesis, Harvard Univ., Cambridge. Mass.
Castro, G., Poulos, S.. and Leathers. F. F. (1985). "Re-examination of slide of lower San Fernando Dam." 1.
Ceotech. Engrg., ASCE, 111(9), 1093-1107.

172 JOURNAL OF GEOTECHNICAL ENGINEERING

J. Geotech. Engrg. 1995.121:163-173.

 Castro. G .. Poulos, S., France, J. W .. and Enos, J. L. (1982). "Liquefaction induced by cyclic loading." Rep. 10
NSF, Geotech. Engr.. Winchester. Mass.
Finn, W. D., Liam. Yogendrakumar, M.. and Yoshida. N. (1986). "Response of 2D embankment systems to
seismic loading-program TARA-3." USC Soil Mech. Ser.• Univ. of British Columbia, Vancouver. Canada.
Hamada. M.. Towhata, I.. Yasuda, S., and Isoyama, R. (1987). "Study of permanent ground displacements
induced by seismic liquefaction." CompUlers and Geotechnics. Vol. 4. 197-220.
Hanzawa. H. (1980). "Undrained strength and stability of quicksand." Soils and Found.. Tokyo, Japan. 20(2).
17-29.
Ishihara, K., Tatsuoka, F .. and Yasuda. S. (1975). "Undrained deformation and liquefaction of sand under cyclic
stresses." Soils and Found.. Tokyo, Japan, 15(1).29-44.
Ishihara, K. (1993). "Liquefaction and flow failure during earthquakes." Geotechnique, London, England. 43(3),
351-415.
Kuerbis, R. H .. and Vaid. Y. P. (1989). "Static and cyclic undrained behaviour of silty sands." ?roc.. 12th
1CSMFE, Session 27. ICSMFE, Paris. France, 91-100.
Miura. S.. and Toki, S. (1982). "A sample preparation method and its effect on static and cyclic deformation-
strength properties of sand." Soils and Found.. Tokyo. Japan, 22( I), 61-77.
Downloaded from ascelibrary.org by East Carolina University on 09/29/14. Copyright ASCE. For personal use only; all rights reserved.

National Research Council (NRC). (1985). "Liquefartion of soils during earthquakes." Rep. No. CETS-EE-001,
Nat. Acad. Press, Washington, D.C.
Oda. M. (1972). "Initial fabrics and their relations to mechanical properties of granular material." Soils and
Found., Tokyo. Japan. 12(1), 17-36.
Prevost, J. H. (1981). "DYNA-FLOW: a nonlinear transient finite clement analysis program." Rep. No. 81-SM-
1, Dept. of Civ. Engrg.. Princeton Univ., Princeton, N.J.
Seed. H. B. (1979). "Soil liquefaction and cyclic mobility evaluation for level ground during earthquakes." J.
Geotech, Engrg.. ASCE. 105(2), 201-255.
Seed. R. B., and Harder, L. F. (1990). "SPT-based analysis of cyclic pore pressure generation and undrained
residual strength." Seed Memorial Symp., BiTeeh Publ.. Vancouver. Canada, 351-376.
Shibuya. S., and Hight. D. W. (1987). "A bounding surface for granular materials." Soils and Found.. Tokyo.
Japan. 27(4), 123-136.
Symes. M. J .. Shibuya. S., Hight. S. W., and Gens, A. (1985). "Liquefaction with cyclic principal stress rotation."
?roc., II 1CSMFE, Int. Conf. on Soil Mech. and Found. Engrg. (ICSMFE). San Francisco, Calif.. Vol. 4.
Vaid. Y. P.. Chern. J. c.. and Tumi. H. (1985). "Confining pressure. grain angularity and liquefaction." 1.
Geolech. Engrg., ASCE. II( 10). 1229-1235.
Vaid. Y. P., Chung, E. K. F., and Kuerbis. R. (1989). "Preshearing and undrained response of sand." Soils and
Found.. Tokyo, Japan, 29(4), 49-61.
Vaid. Y. P., Chung. E. K. F., and Kuerbis, R. H. (1990). "Stress path and steady state." Can. Geotech. J"
27(1).1-7.
Vaid, Y. P.. and Negussey. D. (1988). "Preparation of reconstituted sand specimens." Advanced Triaxial Testing
of Soils and Rock, ASTM STP 977. ASTM, Philadelphia. Pa .. 119-131.
Vaid. Y. P., and Pillai, S. (1992). "Discussion of the 'Critical state of sand' by Been et al." Geotechnique, London.
England. 42(4). 655-663.

APPENDIX II. NOTATION

The following symbols are used in this paper:

coefficient of uniformity;
relative density at end of consolidation;
relative density of sample at preparation;
mean grain size;
void ratio at end of consolidation;
ratio of liquefaction resistance at a{, to resistance at a{, 100 kPa;
phase transformation state;
phase transformation strength a)2;
steady-state strength a)2;
inclination of a l to vertical deposition direction;
Ea axial strain;
c max maximum axial strain during cyclic loading;
ad deviator stress;
a(/n cyclic deviator stress;
a,,, all
, , vertical and horizontal principal stresses;
, O:1r,
(TIc' major and minor principal consolidation stresses;
ai' a.1 major and minor principal effective stresses;
4>' friction angle at maximum obliquity;
4>CSR friction angle at initiation of contractive deformation (peak of stress-strain curve); and
4>: friction angle at steady state.

JOURNAL OF GEOTECHNICAL ENGINEERING 173

J. Geotech. Engrg. 1995.121:163-173.