You are on page 1of 55

HRVARD SOIL MECHNICS SERIES No.

88
LIQUEFACTION AND CYCLIC DEFORMATION OF SANDS
A CRITICAL REVIEW
by
Arthur Casagrande
Presented at
Fifth Panamerican Conference
on Soil Mechanics and Foundation Engineering
Buenos Aires, Argentina, November 1975
Pierce Hall
Cambridge, Massachusetts
January 1976
Reprinted (with corrections) January 1979
LIQUEFACTION A CYCLIC DEFORTION OF SAS
A CRITICAL REVIEW
by
Arthur Casagrande
Professor emritus, Harard University, Camridge, M, USA
SUMRY
Tis paper reviews investigations of two basically different phenomena
for which in literature the same term LIQUEFACTION is used . To prevent
confusion, the ter ACTUA LIQUEFACION is used in this paper for the
response of contractive ( loose) sand that leads to loss of s trength
which can cause flow slides; and the ter CYCLIC LIQUEFACTION for the
response of dilative (dense) sand when subjected to cyclic laboratory
tests. Topics covered: Hypothesis of crit ical void ratio and its
early testing . Liquefaction s lide in Fort Peck Dam and hypothesis of
flow s tructure. Investigation of actual liquefaction with load control
triaxial tes ts; definition of dilative and contractive zones; F line
and liquefaction potential . Investigation of response to cyclic
loading in various types of cyclic laboratory tes ts; conclusions (1)
that cyclic liquefaction in tes t specimens is caused by redistribution
of water content which is generated by mechanism that normally are
absent in situ, and (2) that cyclic liquefaction normally cannot
develop in dense sands in situ . Because it is unlikely that laboratory
tests can be devised to eliminate the severe stress gradients in test
specimens and to reproduce the uniform stresses that exist in a typical
element in situ, the author believes that the great gap between labor
atory and in situ response to cyclic loading will require comprehensive
field inves tigations of full scale tests that nature is perforing in
highly seismic regions. For es timting the in situ cyclic response of
mdiu dense and dense sands, an interim procedure is suggested using
cyclic triaxial tests on anisotropically consolidated specimens.
1
CONTENTS
I INRODUCTION
1
II DEFINITIONS OF "ACTUAL LIQUEFACTION" AD "CYCLIC LIQUE-
FACTION" AS USED IN THIS PAPER
1
III EY LABORATORY INVESTIGATIONS OF ACTUAL LIQUEFACTION AD
HPOTHESIS OF CRITICAL VOID RATIO (1935-1938) 2
IV INSTIGATION OF PATIAL FAILURE OF FORT PECK DA AND
HPOTHESIS OF FLOW STRUCTUR
5
V ACUA LIQUEFACTION PRODUCE IN TRIAIAL TESTS WITH MONO-
TONIC LOAD CONTROL
6
VI COMNTS ON POTENTIAL FOR ACTUAL LIQUEFACTION 10
VII CYCLIC TRIAIAL TESTS BY PROFESSORS SEED A LEE 12
VIII INSTIGATIONS WITH GYTORY SHEA APPARTUS 14
I COMARISON OF CYCLIC STRSSES INDUCED IN SITU AND IN
LBORTORY TESTS 18
X COMARISON OF RESPONSE TO CYCLIC LOADING OF DENSE SADS IN
SITU AND IN LBORATORY TESTS 20
XI LRTORY TESTS FOR DESIGN PURPOSES - INTERIM RE-
COMNDATIONS 23
XII SURY OF PRINCIPA CONCLUSIONS AND RCOMNDATIONS 24
XIII ACKOWLEDGMNTS 25
XIV RFERENCES 26
LIST OF FIGURES
1 EY HYPOTHESIS OF CRITICAL VOID RTIO EXLINE BY MNS
OF DIRECT SHEAR TESTS
2 CROSS SECTION THOUGH FLO SLIDE IN FORT PECK DA AT STATION
22+0
3 COMARISON OF THREE ISOTROPlCLLY CONSOLIDATE R TESTS A
ON S TEST, USING DEA-LOAD INCRNTS
4 STSS CIRCLES OF ISOTROPlCALY CONSOLIDATE R TEST (Sa
Test as "A" in Figs. 3 ad 5)
5 COMARISON OF THE TYPES OF R TESTS WITH ACTUAL LIQUEFACTION
6 CRTICAL VOID RTIO (F LINE) FRM i TESTS USING DE-LOA
INCRNTS OR CYCLIC LOADING
7 CRITICAL VOID RTIO (E
sc
LINE) FROM R TESTS USING STRIN
CONROL LOADING 11
8 VARIABLES CONTROLLING POTENIA FOR ACTUAL LIQUEFACTION OF
BING SAND (Based on Data in Ref. 8)
9 LIQUEFACTION IN LOOSE SAD AJACENT TO A WATERFRONT
10 RSUTS OF A TYPICAL CYCLIC TRIIA TEST ON DENSE SA
(Ref. 10)
11 RISTRIBUTION OF RELATIV DESITY IN CYCLIC TRIAIAL
SPECIMN (Ref. 8)
12 DIGRS ILLUSTRTING MCHNICS OF (a) GTORY SHER AND
(b) RECIPROCATING SHEA PRODUCED BY ROTATING AR OF GYRATORY
SH APATUS
13 SCHETIC SECTION OF GYRTORY APPATUS - LEF HAF SLIDING
PLTE FOR GYRATORY TESTS, RIGH HALF FOR RECIPROCATING TESTS
14 SCHETIC PLA OF SLIDING PLTES FOR GYRATORY AND FOR
RCIPROCATING TESTS
15 COMARISON OF TYPICAL RISTRIBUTION OF RELTIVE DENSITY IN
RCIPROCATING A GYRTORY TESTS
16 HISTOGRS OF RETIVE DENSIT DISTRIBUTION IN SPECIMNS AS
PREPAD AND AFER VAIOUS NUMERS OF CYCLES IN RCIPROCATING
TSTS
17 COMLETE RECORD OF REDISTRIBUION IN MDIU LOSE SPECIMN
AE 25 RECIPROCATING CYCLES
18 RECIPROCATING TEST ON DENSE BANDING SAND
19 INUCED PORE PRESSURES A HORIZONTAL DISPLCENTS VS NER
OF CCLES IN GYRTORY TEST ON MDIU-LOOSE BANING SAND
20 COMLETE RECORD OF REISTRIBUTION IN GTORY TEST ON MIU
LOSE BADING SAD
21 STRESSES IN A ELENT IN SITU BENEATH HORIZONTAL GROUND
SURFACE SUBJECTED TO CYCLIC SHEA STRESSES
22 STESSES IN A EENT IN SITU WHICH IS CONSOLIDATE UNDER
USY TRICAL STRESSES CAUSED BY OVERLYING LOA A WITH
CYCLIC SHER FORCES SUPERPOSE
23 STRSSES IN ISOTROPICALY CONSOLIDATED SPECIMN SUBJECTED TO
CCLIC TRIAIAL TEST
24 CYCLIC STRSSES ON THE SURFACES OF A SPECIN SUBJECTED TO
RCIPROTING SHE TEST AND RSULTING RDISTRIBUTION
t. INTRODUCTION
1
I express m sincere gratitude to the Organizing Comittee of the
5th Panamerican Conference for honoring me by their invitation to
present this keynote lecture. It is indeed a great privilege to
address m colleagues from the Aerican countries stretching from
Canada to Argentina, our host country, and also colleagues from other
continents.
Soon it will be 50 years since the day when I started assisting
Krl Terzaghi. The principal task which he assigned to me from the
first day was the improvement and development of soil testing apparatus
and techniques of testing. Already then, and many times since, I have
observed that the introduction of soil mechanics theories into design
practice was handicapped by the limitations of subsurface exploration,
sapling and testing. The need of practicing engineers for pecifica
tions, their desire to use standards and similar demands by govern
mntal regulatory agencies, were often responsible for premature
adoption of informal or foral standards that lingered on in engineer
ing practice for a long time after they were proven to be unsatis
factor. I have also leared during these decades that whenever I found
mself in disagreement with a highly experienced and respected col
league, it was for one or more of the following reasons: (1) we looked
at different aspects of the same problem; (2) we generalized too much
on the basis of different sets of empirical data; and (3) we used the
sae termnology for different phenomena. I could describe instructive
exples of such differences that had arisen also between Terzaghi and
mself; examples that would demonstrate how well such differences have
served to set the stage for further necessar investigations and
developments. A examle from recent years I will describe in this
lecture, namely the differences between my views on liquefaction and
those by Professor H. B. Seed and his collaborators which they devel
oped in connection with certain aspects of their pioneer work on the
design of foundations and dam for earthquakes. I hope that a frank
discussion of these differing views will help to clarify the issues
and to promote development of reliable methods for deterining the in
situ response of sand strata to seismic loading.
Te purpose of m talk is a critical review of two basically
different phenomena which in literature are both referred to as
"liquefaction". To distinguish between them, one related to the
behavior of loose sands and the other to the response of dense sands
uder cyclic loading in triaxial tests, I found it necessary to in
troduce in this paper differentiating adjectives, as discussed under
the next heading.
11. DEFINITIONS OF "ACUAL LIQUEFACTION" A "CYCLIC LIQUEFACION" A
USE IN TIS PAER
Until 1966, the ter liquefaction was used in literature for the
reaponse of saturated loose sand to strains or sbocksthat resulted in
flo slides. With the developmnt of the cyclic triaxial tests in
conection with research on the response of sand under earthquake
2
loading, the sam ter began to be used for a specific response of
sand in cyclic triaxial tests (Ref. lOa). In an effort to prevent con
fuion by the use of the sam term for entirely different phenomena,
a inforl comittee of several colleagues and mself tried in 1969
to find another ter for the cyclic response phenomna in laboratory
tests. Finally we agreed on the term "cyclic mobility". Hoever, by
then the term liquefaction was already so well entrenched in literature
for ue with the cyclic triaxial tests that it proved impractical to
adere to this agreemnt; and the confusion continued. Wile I would
mch prefer not to ue the term "liquefaction" for a phenomenon that
truly is not liquefaction, I decided reluctantly to use it for both
phenomena in this paper, but to differentiate between them with
appropriate adjectives as follos:
1. ACTL LIQUFACTION - what was simply called "liquefaction"
before the development of cyclic laboratory tests. It is the rsponse
of loose, saturated sand when subjected to strains or shocks that
results in substantial loss of strength and in extrem cases leads to
flo slides.
2. CYCLC LIQUFACTION - the response of a test specimen of .
dilative sand to cyclic loading in a triaxial test when the peak pore
pressure rises mmntarily in each cycle to the confining pressure.
A strong minority on that 1969 comittee supported the term
"strain softening". Superficially, the progressive softening, which
develops in cyclic tests on dense sand, may resemle strain softening.
True strain softening, however, is produced by stress cycles with
essentially uniform distribution of stresses within the material. But
the softening of a saturated sand specimn during cyclic loading is
caued by redistribution of the water content, with substantial
loosening and softening of certain zones in the specimn while other
zones are being comacted.
III. EY LAORTORY INVESTIGATIONS OF ACTUAL LIQUEFACTION A
HOSIS OF CRITICAL VOID RTIO (1935-l938)
In part by observing the volum changes of dense and loose sand
in direct shear tests and in part by intuition, I developed in 1935 the
hypothesis that when loose sad is sheared it decreases in volue, it
contracts, and eventually approaches a steady state volum (or void
ratio; or density); and that the sa sand in dense state increases in
vlu, it dilates, util it also reaches the sa steady state
codition a the loose sand. This state I called the "critical
denatty", or "critical void ratio". In m first lecture on this sub
ject, in Novemer 1935, (Ref. 1) , I was still under the imression that
the critical density is independent of the noral stress. But a few
.nths later, when performing tests using a greater range of norl
stresses, I concluded that the critical density is a fuction of the
norl stress. Becaue direct shear tests did not permit sufficiently
accurate obseration of the volu chages, I desiged in 1936 for the
Crs of Engineers a triaxial apparatu for investigation of the
critica void ratio of fine sads for the Franklin Fall. Da; (Rfs. 2,
3
, 4
)
.
3
Exlanation of the concept of critical void ratio is facilitated by
referring to the relationships as derived from direct shear tests,
Fig. 1, but plotting the norl stress on a log scale, a will be used
in susequent figures. In Fig. l(b) the vertical scale is the void
ratio, or relative density, and the horizontal scale the displacemnts.
Cure L' shos the contractive response of loose sad; curve D' the
dilative response of dense sand; ad the horizontal line M' represents
the critical void ratio or critical density which both samples would
reach if shearing could be continued far enough, or if the shear test
were started at that void ratio. In Fig. lea) are plotted on the
vertical scale the applied shear stress and horizontally again the
shear displacemnts. Curve L shows a steady increae of displacement
of loose sad with increasing shear stress util eventually an ultimte
shear strength is reached. But for the dense sad curve D develops
first a peak strength ad then, with further shearing and further
volu increase, the resistance drops and the same ultimate shear strength
is reaced as for the loose sand. Curve M represents a test started
at the critical void ratio so that in this test specimn theoretically
no volu chage should develop, a indicated by line M' in Fig. l(b) .
In Fig. l(c) are plotted the void ratio (or density) on the
vertical scale, and horizontally on a log scale the applied norl
stress a
n
. For reference, curves L
c
, M
c
and D
c
represent consolida
tion curves as obtained in one-dimensional consolidation tests. The
direct shear tests plotted in Figs. lea) and (b) are assumed to be
performed under a confining pressure of one kg/sq cm. Therefore,
points x, y ad z on the consolidation curves in Fig. l(c) are the'
starting points of the direct shear tests. By projecting these points
horizontally onto the vertical scale of diagra (b) , we find the
starting points x
'
, y' and z' of the void ratio vs displacement curves.
Tus, points z and z' represent the critical void ratio for this sand
when consolidated under a normal stress of one kg/sq e.
If we repeat the tests .at higher normal stresses, the starting
void ratios, including the critical void ratio, would all be lower.
By mking several series of such tests under different norml pressures,
we find the critical void ratio line E, the heavy line in diagram (c)
which is sustantially steeper tha the void ratio-pressure curve M
c
in that rage of void ratios.
To obtain the critical void ratio line, ideally the tests should
hav been perford on sturated specimns without allowing change of
volu, and wth pore pressure masuremnts; but such testing equipmnt
had not yet been developed 40 years ago. Terefore, the question
"wat happens when saturated sad is subjected to shearing at constant
volu" was aswered indirectly, a follows: Suppose the sand is
consolidated to the void ratio and normal pressure represented by point
p, Fig. l(c). If the specimn would be sheared with volum change
aloed, the void ratio would decrease vertically doward, at constant
ef
f
activ stress, util the critical void ratio line E is reached at
point r. Hoever, if no volum chage were permtted, the sand woud
still tr to reduce its volu; but because this is not possible, it
respond by trasferring stress from the grain structure to tha pore
vater; 1. e., we mve horizontally at constant volu util we reach the
4
critical void rat io l ine at point q. Te init ial effective stress of
about 4 kg/sq e at point p would thus be reduced to almost one
tenth of the init ially appl ied stress; ad the pore water would carry
almst the entire applied stress. Such large reduction in effective
stress ad correspond ing shear strength causes actual liquefact ion;
i.e., the sand chages from a solid mterial into a mss which flos
lk a viscous flu id wen subjected to smll shear stresses. But
beneath a level groud surface, not loaded by structures, a l iqufied
ms caot flo. The only symtom of the liquefied state would be
sad boils at the surface produced by upward drainage of excess water.
Hover, a struture uderlain by l iquefied sand would be in seriou
troule. From Fig. l(c) we ca also conclude that saturated sad
belo the critical vo id ratio line would be safe aga inst actual
liqufact ion; that if such sad is exposed to shear forces, it would
develop negat ive pore pressures, i.e., the effect ive intergraular
stresses would be increased and the sand would brace itself, so to
speak, against failure by increaing its strength aove that which
exsted before the shear stresses were applied. But cav itation would
limt this increase in strength to one atmsphere negative pore
pressure.
Te hypothesis of critical void rat io ad the early test results
led to the folloing conclus ions:
(1)
All cominations of void ratio ad effect ive noral stress
wich are located below, or to the left, of the crit ical l ine "E",
F ig. l(c), represent states which the sad would develop dilat ive
response and are safe aga inst [actual) l iquefact ion.
(2) All points to the right or aove the cr itical line "E"
woud represent cominat ions that would result in contract ive response.
In order to produce a flo sl ide, and not mrely a slum of l im ited
dimnsions, the starting point would have to be substant ially to the
right of the cr itical vo id rat io l ine so that in the l iquef ied state
the effect ive stresses would drop to a small fract ion of those that
exsted at the start of shear.
(3) Te greater the effect ive confining pressure, e.g., the
greater the depth of a sad stratu, the loer is the crit ical void
ratio; or, in other words, the denser mut be the sad to be safe
against [actual] l iquefact ion. But when heav ily loaded, even a mdiu
dense sad my be sucept ible to [actul] liqufact ion.
Essentially these were m vies when I presented the second
lecture on liquefaction in Jue 1938; (Rf. 3). Tree mnths later, in
Septemer 1938, a major l iqufaction failure developed in the Ft. Peck
d . in Mntaa which wa then nearing its comletion.
5
IV. INESTIGATIONS OF PAIAL FAILURE OF FRT PK DA
-
HYOTHSIS
OF FOW STRUCURE
In Fig. 2 is plotted a typical cross section through the slide,
with true scale in the loer section ad with the vertical scale five
times enlarged in the upper section. It can be seen that the mass that
mved was almst level after the slide. The major movement occurred in
aout 3 minutes. Som portions of the upstream toe moved to a maximum
distance of about 400 meters with the speed of a rapidly moving river.
Aout 8 million cubic mters of sands in the dam ad in the foundation
participated in this slide.
Te Corps of Engineers carried out a major investigation (Ref. 5,
6, 7) wich included the taking of large diamter udisturbed samples
of the sads by freezing colums of sand in situ and then coring them
with Calyx drills. From the results of many triaxial tests, using m
hypothesis ad testing procedure, it was concluded that these sands
could not have liquefied. Probably influenced by these test results,
a mjority of the consulting board, which was appointed to investigate
the slide, concluded that the failure was not caused by liquefaction;
(Ref. 5). A minority, including mself, concluded from the speed of
the movemnt and the topography of the slide masses after the movement,
that sad at considerable depth must have liquefied. I also was
forced to the conclusion that sand located belo the critical void
ratio line "E", as defined in Fig. 1, can also liquefy; in other words
that the type of test, which I used for determining the critical void
ratio (Ref. 2), did not define correctly the boundary between sands
that are safe ad unsafe against actual liquefaction.
In the years folloing the Ft. Peck dam slide I developed
gradually the hypothesis that when sand is liquefied and is actually
floing, it must have a structure different from that when the sand is
static; that during flow each grain is constantly rotating in relation
to all surrouding grains so as to offer a minimu of frictional
resistance. I termed this the "flow structure". I postulated that
suh a structure (1) spreads by a chain reaction, (2) exists only
during flo, and (3) that in the momnt flo stops, the grains re
arrage themselves and revert into a static structure which, after the
excess water has drained, will be slightly denser than the static
structure before liquefaction occurred. Thus, a mss of sand that has
liquefied, would end up with increased safety against actual lique
faction. In a typical alluvial sand with little or no silt, the static
structure before and after liquefaction and also the flow structure are
all of the Single-grained type. A sand fill constructed by duming
sad in mist state can have a honeycom structure because the
capillary forces between the moist grains resist the sad from
settling into a single-grained structure. Wen such "bulked" sand
becoms saturated, it is particularly prone to liquefy; but afterards
tbe mas ends up with a single-grained static structure. Wen the
struture of a sensitive clay collapses in a chain reaction, lique
faction ad a flo slide will reeult. Te original structure of the
cla, hoever, is irreversibly destroyed. Experience has sho that
after reconsolidation such a ms is no longer sensitive to liqu
faction.
6
Te toIIov1ng observat1on v1II ass1st 1n v1suaI1z1ng tbe
cbaracter ot tbe tIm stzuctuze ot sand. We obsezv1ng bydrauI1c t1I
pIaceoent ot gzaveI I once obsezved a bouIdez tbat d1scbazged tzoo tbe
bydrauI1c p1pe I1ne and appeazed to be Iarger tban tbe d1aoetez ot tbat
p1pe. Obv1ouIy tbeze vas one pos1t1on 1n vb1cb 1t couId tzaveI tbzougb
tbe p1pe pzov1ded tbat tbe oass vas tIov1ng tast enougb to oa1nta1n tbe
bouIdez 1n a pos1t1on v1tbout toucb1ng tbe vaII ot tbe p1pe and ad-
j acent pebbIes.
I tr1ed severaI t1s to acb1eve 1n tbe Iaboratory a tIm
structuze but ta1Ied. I beg to suspect tbat tbe tIm structure can
deveIop d ex1st onIy v1tb1n a Iarge oass. Tben out II years ago
a gzaduate student troo Cb1Ie ConzaIo Castro asked to suggest a
doctoraI zeseazcb top1c. I dec1ded to oake one ooze attegt and I
descz1bed to b1o tb1s probIeo v1tb tbe var1ng tbat pzev1ous atteopts
bad ta1Ied. I nt1oned tbat one reason wy be tbat dur1ng tIov 1n
natuze tbe dr1v1ng tozces are oa1nta1ned essent1aIIy constant 1.e.
tbat natuze 1s appIy1ng dead Ioad dz1v1ng torces not stza1n contzoI
a I bad tz1ed. Load contzoI ensures tbat vbe0 a tIm structuze 1s
beg1nn1ng to deveIop tbe dr1v1ng torce does not zeIax as 1n stra1n
controI. y keep1ng tbe dr1v1ng stzess constant tbe wss acceIezates
taster and tastez and oore d oore ot tbe sand gra1ns v1II rearzange
tbeeIves 1nto a tIov structure. Castzo v v1II1ng to udeztake
tb1s d1tt1cuIt task. Mov I v1II expIa1n b1s z1nc1paI t1nd1ngs vb1cb
aze conta1ned 1n b1s d1ssertat1on (Ret. 8) d as suppIeoented Iatez
by add1t1onaI tests.
V. ACJM LIJFACIOM PROJC0 IM TRIAIIM TE5T5 WT HOMOTOMIC
LOA0 COMTROL
Castzo pertord a oa]oz port1on ot b1s 1nvest1gat1on on "and1ng
5and" a m1tozo cIean guartz sand v1tb subzomded to smguIar
gra1ns a coett1c1ent ot un1toro1ty ot I. 8 and v1tb about IOI by
veLgbt soaIIer tban 0.1 m.
In F1g. 3 aze pIotted tbe stress~stza1n cuzves ot tbree typ1caI
R (consoI1dated-udza1ned) tr1axIaI tests 1n vh1cb tbe ax1aI Ioad vas
1nczeased v1tb dead-Ioad 1ncrents. In add1t1on 1s sbovn tor
coqar1son tbe stress~stra1n cuzve ot an 5 test (dra1ned test). In
all four tests tbe spec1oen vas t1rst 1sotrop1caIIy consoI1dated under
a bydrostat1c pressure ot 4 kgsq co. Tben tbe ax1aI Ioad vas 1n-
creed v1tb soaII dead-Ioad 1nczents appI1ed at out one-o1nute
IntervaIs on a banger.
In test "A" tbe spec1n bad a reIat1ve dens1ty atter consoI1~
dat1on ot 0
rc
- 30%. Attez I5 oLnutes ot 1ncrentaI ax1aI Ioad1ng
te stress~stra1n curve (upper d1agrao 1n F1g. 3) zeached a peak at a
mV1atoz stzess ot about 2 kgsq e d a stza1n ot one percent. Ten
wder the next soaII Ioad 1nczent tbe spec1n sudenIy I1guet1ed
d 1n a sII tract1on ot a second stra1ned to aout 25% vben tbe
test wa stopped autooat1caIIy. At out 5% stra1n tbe ettect1ve
devIator stzess bad dropped to 0
d
_ - 0.3 kgsq co and tben 1t re-
w 1ned constt at tb1s vaIue dur1ng turtbez tIm. In tbe Imer
7
d1agr cute "A" sbovs bov tbe pore pressure rose and reacbed dur1ng
tbe state ot tIov tbe constant vaIue ot J. 85 kgsq e vItb tbe
correspond1ng ettect1ve m1nor pr1nc1paI stress
3
f 0. I5 kgsq co.
Fr tbe reIat1onsb1p s1n = U
dt
(2
Jt
+ U
dt
) =

. 5 tbe angIe ot
IatezaI tr1ct1on ot tb1s Ioose sand spec1men can be cooputed to be J0
degrees. Th1s agrees v1tb tbe resuIts ot tbe S test carr1ed out at
tbe s reIat1ve dens1ty and vb1cb 1s sbom by tbe dasbed stress-
stra1n curve 1n tbe upper d1agrao and by tbe correspond1ng stress
c1rcIes 1n F1g. 4 tor botb tests. Bovever 1t y be a co1nc1dence
tbat tbe stresses dur1ng tbe state ot tIov retIected tbe same tr1ct1on
angIe ot J0 degrees as 1n tbe S test. I am 1ncI1ned to beI1eve tbat
at eguaI reIat1ve dens1ty tbe ettect1ve tr1ct1on angIe 1n tbe tIov
structure 1s smIIer tban 1n tbe stat1c structure. Tb1s 1s aIso
supported by tbe observat1on tbat 1n F1g. 4 tbe ettect1ve stress c1rcIe
at tbe peak vben I1quetact1on started rema1ns beIo tbe J0 degree
strengtb I1ne.
Te reIat1ve magn1tude ot tbe var1ous strengtb c1rcIes 1n F1g. 4
1s a good 1IIustrat1on tor tbe enomous d1tterence 1n strengtb vb1cb
tbe s spec1men ot saturated very Ioose sand can d1spIay atter 1t
bas been consoI1dated under a bydrostat1c pressure ot 4 kgsq cm. In
an S test 1. e. vben pore pressures are not permLtted to deveIop tbe
wor pr1nc1paI stress can be 1ncreased to I2 kgsqm 1. e. tbe
strengtb (dev1ator stress at ta1Iure) v1II be 8 kgsq cm. It no
dra1nage 1s pem1tted dur1ng ax1aI Ioad 1ncrease tbe spec1men deveIops
a peak strength ot onIy 2 kgsg cm v1tb 1nduced pore pressures ot
2. 5 kgsq cm. ut tb1s peak strengtb represents tbe tbresboId ot tbe
cbange to a tIov structure vb1cb 1s part1cuIarIy prone to deveIop vben
tbe appI1ed Ioad v1II not reIax v1tb cont1nued stra1n 1n otber vords
vben a dead Ioad 1s rest1ng on tbe spec1men. F1g. 4 and aIso curve
"A" 1n tbe upper d1agram 1n F1g.
3
, sbov tbat tbe cbange to tbe tIov
structure reduces tbe peak strengtb ot 2 kgsg cm to a strengtb dur1ng
tIov ot 0. J kgsq cm. In tbe I1guet1ed state tbe sand 1n tb1s test
bad an apparent tr1ct1on angIe ot onIy about tvo degrees. Tberetore
dur1ng a tIov sI1de a mss otsucb sand vouId spread out unt1I 1ts
surtace vouId sIope onIy a tev degrees.
In F1g. J test spec1men '" bas a reIat1ve dens1ty ot 44I at tbe
start ot ax1aI Ioad1ng. Tbe upper d1agram sbovs tbat 1t deveIoped a
peak strengtb somevbat greater tban tbe Ioose spec1n 'A' tben 1t
I1quet1ed and stra1ned to I8I 1n 0. 4 second and tben aII movement
stopped. To ettect add1t1onaI stra1n1ng tbe dev1ator stress bad to
be 1ncreased aga1n v1tb deadIoad 1ncreoents. In tb1s test dur1ng tIov
tbe strengtb bad decreased onIy I1ttIe as compared to tbe peak strengtb
nevertbeIess tb1s spec1men actuaIIy I1quet1ed. Tbe 1nduced pore
pressures 1n tb1s test are sbmn 1n tbe Imer d1agram vb1cb togetber
v1tb tbe ettect1ve dev1ator stress (strengtb) dur1ng tIov 1n tbe upper
d1agr retIects t1rst acceIerat1on and tben deceIerat1on v1tb tbe
spec1men "treez1ng" at I8I stra1n. Froo tbe sbort steady state segnt
of tIov oovement one can coopute tbe rat1o O
dt
(J
dt
+ 2J
Jt
) = 0. 5J =
s1a troo vb1cb = J2 degrees. Tb1s spec1men vas substant1aIIy
denser tban spec1men "A" d a greater ettect1ve tr1ct1on angIe dur1ng
fIm 1s reasonabIe. In tb1s test tbe ettect1ve o1nor pr1nc1paI stress
8
decreased to about one-fourth of the stress uder which the specimen
was consolidated isotropically. This large decreae is not imdiately
evident from the fact that during the state of flo the peak strength
. is almst mintained. Should one describe the response of specimn
"B" 8 "limited" actual liquefaction?
Te third tet, "C". in Fig. 3 was carried out at 47I relative
densit after consolidation, i.e., slightly denser than in test "B".
Te stress-strain curve appeared to develop into a peak at a slightly
greater deviator stress tha test ItB", with a suggestion of imending
actual liqufaction. Hoever, then the stress-strain curve reversed
its curvature ad the strength increaed rapidly a a result of a
strong dilativ response, a ca be seen by the induced pore pressures
in the loer diagra. The pore pressure dropped to zero at a strain of
14%. If this test had been continued, negative pore pressures woud
hav developed. If the S test in the upper diagra had been perfored
at the sa relative density, the intersection of curves "c" and "0"
should have occurred at aout I4I strain, when the pore pressure
dropped to zero. But the S test was made on a looser specimn. For a
denser specimn curve "0" would be higer and the intersection would
agree better with the I4I strain corresponding to zero pore pressure.
In Fig. 5 (- Fig. 1 in Rf. 9) are comared three tests on loose
specimens with relative densities of aout JI after consolidation
uder a hydrostatic pressure of 4 kg/sq cm. Test `A" i the sa test
a "A" in Figs. 3 ad 4. The loading procedures in the other to tests
were different, a described below.
Specimn "AI! wa first aisotropically consolidated with a minor
principal stress of 4 kg/sq e (the sae a for the other to tests)
and a major prinCipal stress of 8 kg/sq e, thu with a principal
stress ratio of 2.0. Oly to small dead-load incremnts were applied
on this specimen ad then it liquefied at about 0.5 strain. This
teet demonstrates that loose sand. which is susceptible to actual
liqufaction, will mre easily liquefy adjacent to a steep slope tha
beneath a horizontal surface.
Specimn "e' wa initially consolidated isotropically to 4 kg/sq
c ad then sujected to cyclic loading with a deviator stress of
aout 1 kg/sq e. Te first four cycles developed almst elastic
defortions, with the hysteresis loops practically coinciding, as
sho by the single loop in Fig. 5, ad with strains of a small
fraction of one percent. Hoever, during the 5th ccle the strains
increaed, a rouded pea developed at aout one percent strain, ad
then the sale actually liquefied.
It deseres spectal emhais that during the state of flo the
.aitude of the residual strength a
df
ad of the residual effective
aDr principal stress ali were not only constant during flo, but
the vere almst identica in all three tests. After a flo structure
v . fuly developed, the sad had lost all mmory of its pat stre.s
strain histor. Terefore, it is reasoale to aSU that the flo
structure va identical in theae three teata.
9
Ia FLg. 6 are pIotted (I) tbe startLng condLtLons and (2) tbe
coadLtLons dur1ng tIov ot aII tests perto=d by Castro 1n vb1cb actuaI
ILquetactLon deveIoped. (Ret. 8 and tests pertomed smseguentIy.)
Te vertLcaI scaLes are voLd ratLo and reIat1ve dens1ty and tbe
IogarLtboLc borLzontaI scaIe 1s tbe ettect1ve oLnor prLnc1paI stress.
Te arros po1nt1ng to tbe Iett Lnd1cate tbe startLag cond1tLons and
tbe pIotted po1nt to tbe Iett ot eacb arrm Ls tbe ettect1ve o1nor
prIncLpaI stress dur1ng tbe state ot tIov. Tbe Iarge beavy c1rcIes
are tests on 1sotrop1caIIy consoILdated specLns vbLcb deveIoped
actwI ILquetactLon v1tb Iarge drop 1n strengtb. Tbe soaII cLrcIes
are tests oa LsotropLcaIIy consoI1dated specLns dur1ng vb1cb tbere
vas onIy a reIatLveIy soaII drop La strengtb dur1ng tIo tbe tr1gIes
are tests on an1sotropLcaIIy consoILdated spec1ns d tbe sguares
represent tes
@
vLtb cycI1c IoadLng. Te borLzontaI dLstance troo tbe
startLag arro to tbe ettectLve stress 1n tbe ILguetLed state 1s tbe
reductLon 1n tbe ettectLve stress tbat deveIoped by actuaI ILquetactLon.
For exIe Ln tb1s pIot tbe tbree tests v1tb tbe Ioosest spec1oens
bad a reIat1ve densLty sI1gbtIy greater tban 20% . One test vas
coasoILdated under 4 kgJsq e and durLng tIm Lts ettectLve o1nor
prLncLpaI stress dropped to 0.02 kgsq e, a reductLon by a tactor ot
200. To otber tests at tbe sa vo1d ratLo vere consoILdated under
0.3 kgJsq e and durLng actuaI ILquetactLon tbey aIso deveIoped about
0.02 kgsq co ettectLve oLnor prLncLpaI stress a reductLon by a
tactor ot onIy I5. FLg. 6 sbms tbat no oatter to vbat o1nor
prLnc1paI stress a specLn vas consoILdated 1n1tLaIIy d no oatter
vbetber Lt vas consoI1dated LsotropLcaIIy or an1sotropLcaIIy or
vbetber Lt vas cycI1caIIy Ioaded tbe cond1tLons durLng taLIure aII
ended up aIong a ta1rIy accurateIy detLned ILne vbLcb I nov caII tbe
F ILae tbe Ietter F standLng tor crLtLcaI voLd ratLo Ln vbLcb ILgue-
tactLon vLtb a flow structure deveIoped.
Tben I asked Castro to pertoro tests vLtb strain control IoadLag
d tbe resuIts are pIotted Ln F1g. 7 v1tb tbe arrms on tbe r1gbt
agaLn Lnd1cat1ng tbe startLng condLtLons. Tbe cLrcuIar po1nts vbLcb
represent tbe condLtLons dur1ng tIov ILe cIose to a ILne vb1cb 1s
d1spIaced troo tbe P I1ne (FLg. 6) by a tactor ot aout 2.5 Ln teros
ot tbe ettectLve oLnor prLncLpaI stress. I vLII reter to tbLs ILne as
tbe
sc
ILae. Te reon tor tbe Iarge d1tterence Ln tbe strengtb
durLag tIo vben us1ng Ioad controI aad stra1n controI IoadLng 1s tbe
rate ot straIn. Tbe constt dr1vLng torce ot a dead Ioad produced
Ln tbese tests a rate ot straLn aout 20,000 t1oes taster tban tbat
Ln tbe tests v1tb straLa controL IoadLng. Tbe reIat1veIy sIm
straLnLag 1a tbe Iatter tests caues IocaIIy groups ot sand gra1ns to
Iose tegorarLIy tbe1r tIo structure. I suspect tbat one couId
acbLeve a vLde rge ot
sc
I1nes vLtb vLdeIy dLtterLng rates ot
straIa coatroI.
Tea I ked Castro to pertom tr1axLaI 5 (draLned) tests s1oLIar
to tbose I used 40 years ago d vbLcb vere used to 1nvest1gate tbe
ILquetact1on poteat1aI ot tbe sands 1n connect1on v1tb tbe 1nvestL~
gatLons atter tbe sILde Ln Ft. Peck Da. VLtb a tev except1ons tbe
resuIts teII consLderabIy above tbe I1ne1.e. even turtber ay
tro tbe F ILne d tar a tbe @
s
_
1ne Ln FLg. 8; d tbey
1
0
scattered v1deIy. In retrospect 1t 1s nm cIear wy tbe 1nvest1-
gatIons ot tbe sens1t1v1ty ot tbe sands 1n d beIov tbe Ft. Peck Da
yIeIded strengtbs about IO-t1s greater tbao tr1ax1aI R (udra1ned)
tests u1ng dead Ioads. Tbe reason 1s s1gIy tbat a tIov structue
caot deveIop 1n 5 tests.
VI. QK5 OM POTOIM FR ACJAL LIJFACI0
Rterr1ng to F1g. 8 vbere tbeand F I1nes are reproduced troo
FIgs. d 7 1t see reasonabIe to express tbe I1guetact1on
potentIaI L, by a rat1o reIat1ng tbe 1n1t1aI ettect1ve o1nor
pr1nc1paI stress

1
to tbe ettect1ve onor pr1nc1paI stress

t
on
tbe F I1ne. In or

er to wke tb1s rat1o eguaI to zero aIong tbe F


IIne I ue tbe express1on L
p
= (J

1
~ J

t
. g. tor tbe 1n1t1aI
cond1t1ons at po1nt H
1
tbe cond1t1ons durng actuaI I1guetact1on are
tomd by dr1ng a bor1zontaI I1ne to 1ts 1ntersect1on H
t
on tbe F IIne
correspond1ng to a I1guetact1on potent1aI L
p
- 4.0/0.15 - 1
= 2.
For y start1ng co1nat1ons on tbe F I1ne tbe I1guetact1on potent1aI
voud be zero and beIov toe F I1ne 1t vouId be negat1ve 1. e. vben
tbe sd 1s stra1ned beIm tbe F I1ne v1tbout aIIo1ng voIu cbacge
negat1ve pore pressures and add1t1onaI strengtb vouId be oob1I1zed.
In tbe area beIm tbe F I1ne vb1cb I des1gnate tbe "d1Iat1ve zone"
or tbe "D" zone tbe sd tor vb1cb tb1s partIcuIar F I1ne vas deter-
oIned cnot actuaIIy I1guety. Move tbe upper I1ne v1tb tbe sbort
dbes tbe
u
I1ne tb1s sd v1II aIvays deveIop a d1st1nctIy con-
tract1ve response vben stra1ned and 1n tbat zone I1guetact1on 1s
I1keIy 8 a resuIt ot acy rap1d stra1n1ng or sbocks. etveen tbe
u
d tbe F I1ne e1tber d1Iat1ve or contract1ve response 1s poss1bIe
depend1ng on tbe rate ot stra1ns d aIso depend1ng on bov cIose to
tbe F I1ne tbe start1ng po1nt 1s Iocated. Hucb reoa1ns to be Ieared
out tbe response ot sacd and tbe cbaracter1st1cs ot tbe tIov
structu v1tb1n tbat zone. . g., 1n stra1n controI tests 1t 1s
poss1bIe tbat tIm structure v1II appear and d1sappear 1n soaII
pockets or Ienses tbat as soon a 1t deveIops 1n one pocket tbe
stresses 1n tbat zone v1II drop by red1str1but1on acd tbe structure 1n
tbat pocket oay revert to a nomaI structure vb1Ie 1n aootber pocket
It oay cbge 1nto a tIov structure. Tbe overaII res1stance ot a
sIe dur1ng stra1n controI vouId tben be an average res1stacce ot a
coasttIy cbg1ng pattez ot d1str1but1on ot tIov structure v1tb1n
tbe sd.
Men tbe strengtb dur1ng actuaI I1guetact1on drops onIy sI1gbtIy
beIm tbe pe strengtb e. g. 1n test "" 1n FIg. J tbe I1que-
factIoa potent1aI rges betveen about 2 d 1. e. tbe start1ng
poInts vouId be Iocated 1n F1g. 8 approx1oateIy aIo

g tbe
sc
IIne.
In spIte of tbe drop 1n pore pressure by a tactor of 3 to 4, tbe
stzengtb drops onIy I1ttIe.
Te P IIne 1s of pr1w 1nterest tor eng1neer1ng appI1cat1ons. O
tbe bIs of m present kamIede acd ]udgnt I cons1der tbe ent1re
azea ove tbe F I1ne to be a bazard v1tb respect to actuaI IIquetact1on
i fomdatIons of d d Iportt structures 1n b1@Iy seIsmc
11
zones . Belo the F line actual liquefaction is not possible , althoug
small s trains may develop .
From the F line in Fig . 8 we can quickly estimte the s trength of
tis particular sad in liquefied state . Assuing a friction angle of
30 degrees and a relative density of 30%, during flo the effective
minor principal stress would be about 0 . 1 kg/sq em and the deviator
strength about 0.2 kg/sq e. Hoever, at a relative density of 50%,
the minor effective principal stress during flo would be about
3 kg/sq e ad this would require a deviator strength of about 6 kg/sq
e, corresponding to an effective major principal stress of about
9 kg/sq cm. Therefore, in liquefied state there is an enormous
difference in the strength of the sam sand at relative densities of
3% ad 50%. For all practical purposes , this s and with a relative
density of 50% could not develop actual liquefaction.
No let m ak this question: Is it conceivable that a mass of
sad located well below the F line , say at point A in Fig. 8 where
it is safe agains t actual liquefaction, could expand or dilate by
natural causes to point A
1
well above the F line where it would then
be suceptible to actual liquefaction? Let me describe a natural
phenomnon which in the Europea Alps is knon by the name HRN.
Occasionally large mases of rather dense graular talu will liquefy
ad flo dow a valley cauing great destruction. It is well documented
in literature that just prior to such a avalanche the mountain b rook,
which emerges from the toe of the talus deposi t , stops flowing. The
native population in moutain valleys have knon and heeded such
waring signals for centuries . I consider it possib le that a comina
tion of heavy rainfall or sno melt, perhaps comined with a thick zone
of still frozen and relatively impervious talus adjacent to the slope,
will raise the groundater level in the talus mass and create large
shear stresses which cause strong dilation, i.e. , a substantial increase
in volum which soaks up large volums of water. Such changes could
conceivably change the position of large masses adjacent to the talu
slope from below the F line to well above the F line . Perhaps the
feaibility of such a mchanism could be inves tigated at relatively
small exense by meas of laoratory tes ts .
Liquefaction slides somtimes create the imression of a succession
of shear s lides . Assu a river bank in loose sand, Fig. 9, and that
a elemnt "A" som dis tance from the slope is subject to a safe
principal stress ratio; but that an element "B" near the slope is
suject to a much greater principal stress ratio which renders it mre
vlnerale to liquefaction. This, comined with progressive s teepening
of the slope by erosion, can cause a limited zone to liquefy as
illutrated in Fig. 9 ( a) . A the hatched wedge of sand flos out, the
principal stress ratio will increae quickly in a adjacent zone in
wich, in addition, also the seepage forces will be greater a we mve
further back into the river bank; Fig. 9 (b) . In this manner lique- .
factio can progress backard ad a large volu of sad may flo into
the river, leaving behind a slope surface with a ver flat agle that
reflects the lo strength of the liquefied sad. Hoever, t o a shallo
depth belo the groud surface ad next to the face of the slope. i.e. ,
12
the zones not shaded in Fig. 9(c) , the sand is so lightly loaded that
it my lie belo the critical F line where it cannot liquefy . O the
groud surface a observer my see only a progression of shear cracks ,
conveying the imression that the entire movemnt consists of a
succession of ordinary shear s lides , whereas in fact the major portion
of the mass is moving like
a viscou mass without shear surfaces . Such
flo slides usually s tart along a short stretch of the bank . A lique
faction progresses ladward , the area widens such that in pla the
final area appears like a flak with a narro zone through which the
liqufied sad has floed out .
VII . CCLIC TRAIA TESTS 7 PRFSSORS SEED AD LEE
Wen finally , after so may years , I believed that with Cas tro's
help I had solved the problem of liquefaction of sand in a satis
factory manner, Professors Seed and Lee pulished their well-known
paper entitled LIQUFACTION OF SATURTD SAD DURING CYCLIC LADING; ,
(Rf . lOa). Professor Seed and his disciples had then already carried
out imortant pioneer work on the response of earth masses and
structures to earthquakes. To apply their theories, they needed soil
paraters that express the response of sand to earthquake loading.
It wa logical to experiment with cyclic triaxial tests . Thereby they
discovered that when a triaxial specimen of saturated sand is subjected
to cyclic loading in a undrained test in such a manner that the
principal stresses pass in each cycle through a hydrostatic s tate of
stress (which means that all shear s tresses disappear) , even a dense
ad highly dilative sand will develop cyclically high pore pressures
ad deforations .
A typical test result from the paper by Seed and Lee is reproduced
in Fig. 10 . The sand specimen was placed at a relative density of 78% .
In the upper diagra are plotted vertically the axial strains and
horizontally the nuer of cycles on a log scale . The horizontal line
in the middle represents zero s train. The horizontal lines aove the
zero line are for 5 , 10 ad 15% comressive strains; ad below the zero
line for 5 , 10 ad 15% extension. In the lower diagram are plotted
vrtically the induced pore pressures and horizontally again the numer
of cycles on a log scale. The initial hydrostatic confining pressure
va one kg/sq cm. The axial deviator stress was plu/minus 0 . 7 kg/sq
C; i.e the vertical principal stress was cycled between 1 . 7 and
0.3 kg/sq e, while the lateral s tress remained constat at 1 . 0 kg/sq
e. During the first 10 cycles the axial strains remained negligible,
Dut the pore pressures increased ad decreased with each cycle,
reacing higher ad higher values util after about 12 cycles the pore
pressure reached the confining pressure of one kg/sq cm in the momnt
wen all principal stresses were equal ad all shear s tresses in the
.pecimn beca zero. When the pore pressure equalled the confining
prssure for the first time. Seed ad Lee defined this as "initial
liquefaction. II With additional cycles the axial strains kept increasing
. 8een in the upper diagra. railure was defined in ters of the
nuer of cycles when the strain alitude in one coplete cycle
r.aced a certain mgitude. A ca be seen in Fig. 10. the pore
IJ
pzessuze dzopped substant1aIIy d tbe spec1oen d1spIayed a stzong
d1Iat1ve zesponse vbenever tbe pz1ncipaI stzess zat1o zose to a
o1om. ut 1n sp1te ot an essentiaIIy steady patter ot cycI1c
poze pzessuze inczease and deczease tbe cycI1c peak stza1ns ot tbe
spec1n kept 1nczeas1ng. Foz exaopIe at 2O cycIes tbe totaI stza1n
aqI1tude vas IOI aout eguaIIy d1v1ded beUeen cogzess1on and
extens1on.
Fzoo such tests Seed and Lee dzev tbe toIIm1ng concIus1ons:
"I. In1t1aI I1guetact1on c be pzoduced by cycI1c Ioad1ng aIso
1n wd1m dense d dense sands and 1s not I1o1ted to Ioose sands.
"2. Te I1quetact1on potent1aI deczeases v1tb 1nczeas1ng con~
t1n1ng pzessuze.
"3. Tbe I1quetact1on potent1aI deczeases v1tb 1nczeas1ng 1n1t1aI
pz1nc1paI stzess zat1o and tbezetoze sand ad]acent to a sIope 1s
Iess suscept1bIe to I1guetact1on tb beneatb a boz1zontaI suztace.
II
At tbat t1 I d1d not zeaI1ze tbat tbe tem "I1quetact1on" 1n
tbe1z papez bad an ent1zeIy d1ttezent an1ng tban tbe pbenooena tbat
I bad aIvays assoc1ated v1tb tbat tezo. Tbezetoze tbese tbzee con-
cI1ons appeazed to contzad1ct aII m past expez1ence and aIso tbe
zesuIts ot Castzo's 1nvest1gat1ons. 51nce I couId not t1nd any cIues
1n tbe1z papzs toz tbese staztI1ng concIus1ons I bad no cbo1ce but
to<az out such cycI1c tests d obsezve cazetuIIy evezytb1ng tbat
v bappen1ng 1n tbese tests. I des1gned a s1opIe azzangeoent tbat
coud be attacbed to Castzo's tz1ax1aI appazatus to appIy cycI1c
Ioads ot s1nuso1daI sbape. AIzeady tbe t1zst tev cycI1c tests vb1cb
Castzo peztozd sboved tbat tbe top ot a spec1oen vas gett1ng sott
d tbat obv1ousIy a zad1caI zed1stz1but1on ot tbe vatez content
deveIoped v1tb1n tbe spec1oen tbat gzaduaIIy tbe top ot tbe spec1oen
detomd by aIterate neck1ng and buIg1ng d tbat sot1oes even a
th1n Iayez ot vatez appeazed on tbe top ot tbe spec1oen 1n tbe oont
vben aII sbeaz stzesses beca zezo and tbe poze pzessuzes eguaIIed
t0e cont1n1ng pzessuze. Ve oeuzed tb1s zed1stz1but1on ot tbe vatez
coatent by tzeezing test spec1oens. Tb1s zegu1zed tbat tbe spec1oen
be zeconsoI1dated undez tbe cont1n1ng pzessuze by open1ng tbe vaIves.
Tezeby a cezta1n ammt ot vatez tzoo tbe sott top pozt1on vas
dzaLned ott and tben tbe saopIe couId be zeooved and tzozen. A
exIe ot tbe uzed zed1stz1but1on ot tbe vatez content expzessed
1a tem ot zeIat1ve dens1ty 1s pzesented 1n F1g. II (tzoM Ret. 8).
uz1ng tbe test tbe avezage zeIat1ve dens1ty ot tbe spec1oen vas 7ZI
but wea zeconsoI1dated at tbe end ot tbe test 1t 1nczeased to
avezage ot 78I. Tbe d1ttezence vas due to vatez dza1ned tzoo tbe sott
top pozt1on ot tbe spec1oen d tbezetoze tbe actuaI zeIat1ve
de0s1ty ot tbe top ot bave been cons1dezabIy soaIIez tba0 >0I. Tbe
muzed zeIat1ve dens1ty 1n tbe Imez pozt1on bad 1nczeased to 80 to
90%. T1s zetIects tbe substt1aI 1nczease 1n dens1ty ot tbe Imez
0aIt ot tbe spec1oen. Tb1s zad1caI zed1stz1but1on ot vatez content
d ot cozzespond1ng zeIat1ve dens1ty 1s tbe co1ned zesuIt ot tbzee
ettects: (I) tbe bomda ettects vb1O cause zedLstz1but1on aIso 1n
14
mnotonically loaded specimens , a was meaured already may years ago
at the Waterays Experiment Station; (Ref . 11) . (2) the high pore
pressure and softening which develop in the momnt the specimn is
ccled through the state of hydros tatic stress; and (3) a kind of
puing action which seem to dra water to the top that is freed in
the mmnts when cycling through the hydros tatic state of stress . The
iportant question whether such redis tribution in test specimens is
representative for wha t happens in an elemnt in situ will be dis
cused later.
A detailed inves tigation of the redistribution of the water
content in cyclic triaial tests, in which redistribution is produced
by superposition of several different mchanism, would be a difficult
udertaking. Terefore, I decided to tackle first the redistribution
produced by one mechanism only , naly that produced by non-uifor
stress dis tribution on the specimen boudaries . For this purpose I
.
concentrated on the design of a gyratory apparatus in which we could
perfor gyratory a well a cyclic direct shear tests . The design I
had in mind would permit freezing the specimen at the end of the-test
in the apparatu. (Note: At that time I could not think of a
relatively simple design for freezing the test specimen inside a
triaxial cell . In the manwhile I have developed a design which ful
fills this requirement . I have sumitted i t to the Waterays Experi
mnt Station, Vicksburg, Miss. , in the hope that it will stimulate a
detailed inves tigation of the redistribution in cyclic triaxial tests
. )
VIII. INVESTIGATIONS WITH GRORY SHEA APATUS
Investigation of redis tribution in reciprocating shear and
gyratory shear tes ts would have the following advatages: ( 1) It would
permit investigating the boudary effects more clearly than in cyclic
triaxial tests in which it would be difficult to sort out the influence
of ( a) boudary effects , (b) of the pumping action and (c) of cycling
through the hydros tatic s tate of s tress; (2) it would be easier to
freeze the test specimen in the apparatus and without the need of re
consolidation; ad (3) the redis tribution could probably be inves ti
gated with greater accuracy.
Originally I intended to confine the specimens in wire-reinforced
ruber membraes of the type developed by the Noregian Geotechnical
Institute. But I soon realized that it would be necessar to des troy
one mmrae for each tes t , when cutting the frozen speCimen , and that
suc cut ting would be difficult . Then it occurred to m that we could
support the ruber mmbrae by a type of flat coil spring which is
ued 8 a to by the na of "slink. y" . This proved a excellent
solution . We covered the surfaces of the flat coils wi th teflon to
reduce friction; but principally we relied on a s light separation of
t coils by temorarily inserting "spreaders" before the start of
.hearing.
Te first grator apparatus I desiged with the help of Catro.
But for a mjor research effort I needed aother intelligent slave.
(Tis is wha t I called m doctoral candidates only half-jokingly . They
15
are in fact a moder kind of sl avery because they slave for several
years for the cherished Ph . D . , although they may not have read a
single book about philosophy . ) Jus t in time I found a sui table
candidate , Franklin Rendon, a Mexican s tudent who was wel l endowed
with endurance , a quality particularly important for this project
becaue of the many changes in des i gn and in tes t ing techniques that
had t o be developed by trial and error.
I will des cribe the apparatus only briefly . I t is designed to be
ued for cyclic direct shear tes ts and for gyratory shear tes ts . The
mchanics of the gyratory tes ts is illustrated in the top of Fig . 12 .
A cons tant horizontal force is applied to the top of the specimen by a
pair of negator springs mounted on a rotating arm. For brevi ty I
designated this type of tes t the
f ly
test" . In the loer diagram of
Fig . 12 is shown the mechanics of the cyclic direct shear test for
which I int roduced the name "reciprocating shear test" , or brief ly the
"I tes t" . For both types of tes t the s ame sys tem is used for applying
the horizontal force except that in the reciprocating tes ts it is
trans formed into a reciprocating force of sinusoidal shape .
Fig . 13 shows a s chematic cross section of the apparatus . To the
ro tating ar shon on the lef t is mounted a pair of Negator springs
whi ch transmit a cons t ant horizontal force to the top of the specimen,
i . e . , the horizontal force remains cons tant regardless of the magnitude
of the horizontal displacement . The bushing and pis ton are mounted in
a s liding plate which ( 1) for the gyratory tes t , show on the lef t
side , glides between gyrator ball bearings ; and ( 2) for the recipro
cating shear tes ts , shown on the right side , is moving back and forth
between 3 sets of bearings : side , top and bottom guide bearings .
The specimen is enclosed in a rubber membrane which in tur is
supported by the "slinky" spring. When placing the specimen, the coils
of the flat spring are kept separated by coil spreaders which are
removed j us t before a test is s tarted . This ensures that the s linky
will not carry any vertical load during the test .
I will not describe such det ails as the electrical displacement
ad pore pressure transducers , equipment for recording the output of
these transducers , the type of motor used for rotating the arm, the
lines and equipment for saturating the specimen prior to the tes t and
the many important detai ls related to the technique of freezing the
sales , and cutting them into 64 segments . (All these de tails are
described in references 12 and 13 . )
Te upper diagra in Fig . 14 is a plan view of the gyratory
s liding plate and of the to displacement transducers , one munted on
the rotating ar in the direction of the horizontal force , and the
second one at 90 degrees to the ar. Wen using for calibration
purposes a s ilicone rubber specimen, the deflection was always s tri ctly
in the di rection of the rotating ar. Hoever , for sand specimens the
displacemnt vector lagged behind the direction of the ar, as indicated
by the la, angle A . Te interal friction in the s and causes a kind
of stick-s lip motion and also slight vibrations . Te loer diagra in
1 6
Fi g . 1 4 i s a p lan vi ew o f the rec i
p
r0cat i n
g
s l i di n
g p l at e wi t h i t s
l ateral and ve r t i c a l gui de b e ar L ngs .
Fi g . IS shuws typ i c al pat : e rns o f thl' redi s t r ib u t i on t hat deve lops
in such tes t s . Tes e t e s t spec i mens we re made ,) f al t e rn a t in
g
l ayers o f
unc olored and cu l ored : anJ , t ,) (!bst rve t h e de f onnat i cfb . At the s t ar t
o f these t es ts the re l at ive dens i ty was 1 and as un f orT as
p rac t i cab le . The alp Ue d ver t i cal :lll d shear fo rces we re the s ae in
both tes t s . fhp l e f t g ,lmp le was sub j ect ed t o 30 rec i procat i ng cycles
and the ri ght on,' t o 35 gy rato ry ey e les . Af t er f reeziag , cdch s amp le
was cut into 6 4 e l ement s , whi ch had been adop t ed as a s t andard
procedure f or "os t t e s t s i n th i s i ltve s t i g.t i on . In Fi g . 15 only the
results of a cent er s 1 1 ( ' e are shown : ( 1 ) at t he t op
,
pho t og raphs ;
( 2) below , t he re lat ive dens i t ies o f each e lemen t comput ed f rom the
water cont ent s ( af t e r adding the amount e xpe l led by f reezing) ; and
( 3) in the lowe r di agrams are indicated the compacted zones as hatched
areas and the l oos ened zones as b l ank areas . Both s amples developed
compac te d zones adj acent to the c ap and the b ase . In the lef t s amp le
the maximum re lat i ve dens i ty i n the t op layer was 5 3 % . in the b o ttom
layer 67% . I n the ri gh t s amp le the maximum in the t op was 61% and i n
the b ottom 5 5% . The greates t loosening o f the sand deve loped in the
middle pl ane , wi th the lowes t value 20% in the le f t s ample and 8% in
the right s amp le . However , in the r i ght upp e r corner of the lef t
sample one < e gmen t shoed minus 3 3% where s ome f ree water had
accumulated on top of the s pec imen and whi ch , af ter f reezing, was
included as par t o f that element .
Zones wi th relat ive densi ti es of less than 30% , for this type of
sand, were prob ab ly lique f ied, at leas t in par t . In any case , the
mob i li ty of these spe c imens was obvi ous ly centere d in the l oosened
zones . This is al s o ref le c ted by t he def ormat i ons of the s and l aye rs ,
s een in the photographs at the t op , whi ch are mos t p ronounced t hrough
out the mi dd le o f t he specimen ; and in the recip rocating tes t , in the
left ph ot ograph , al so al ong the s i des whe re we o f ten f ound the l ooses t
segment s . In gy rato ry tes t s the l ooses t segment s were
o ften
observed along the mid-p lane .
Te pat ter o f l' edis t r ib ut i on i s p rob ab ly cons t an t ly changing ,
even within a s in gle cycle , exce p t i n the comp ac ted t op and b o t tom
zones that remain essent i ally s t ab le i n shape , but become more c ompact
with increas ing number of cy cles . There i s a radical di f ference in
redis t ribut i on whi ch develops in rec i procat ing and gyrat ory tes ts as
compared to cycl i c t ri axial tes ts . In the l at ter the greate s t water
content accumul atEs in the top o f i ni t i ally uni form tes t s pec imens .
In Fig . 16 are shon typi cal hi s t ograms of the di s t ribut i on o f the
water content and re lat ive dens i t y of spec imens ( 1) as p laced ; and
( 2 ) at the end o f re ci procat i ng tes ts . The l ef t upper one is a typi cal
dis tribut i on i n a tes t spec imen as p repared , but not sheare d . In this
samle the water content ranged between 2 3 . 2 and 24 . 6% , corresponding
to a range of re lat ive dens i ty be tween 50 and 62% . The di s t ri buti on
approximat ed a normal di s t ri but i on curve . All four spec imens were
placed at about the s ame relat ive dens i t y , s l i gh t ly greater than 50% .
l
(
:
I ,
I
;
1 7
In the lower le f t , a tes t wi th only two cycles of reciprocating shear
developed s ubs tan ti al redis t ribut i on as compared to the ori ginal p lace
ment . The upper r i gh t d i agram shows the redi s t r ibut ion a f ter 2 5
cyc les , wi th a range o f relat ive dens i ty f rom 3 1 t o 76% . In the lowe r
right the redi s t ribut ion o f re l at i ve dens i tv af t er 6 4 cyc ies range d
from 29 to 74% , and the shape of the h is togram no longer resemb les a
normal dis t rib u t i on curve .
Fig . 1 7 s umari zes the redis t rib ut ion measuremen t s o f al l 6 4
elements f or a tes t af ter 25 cycles o f re c i p rocating she ar . Visual ly
the redis t ribut ion is i llus t rated b y us ing t hree ranges of re lat ive
dens i ties : ( 1 ) b lank areas for the looses t zones , wi th relat ive
densi t ies less than 48% ; ( 2 ) the dot ted areas for the range between
48 and 59% ; and ( 3) the hat ched areas f or the dense s t zones wi th
relative dens i t ies greater than 59% . The t o tal range of red i s tribution
in this tes t was between 33 and 74% relative dens it y .
Pore pressure and disp lacement trans ducer records are reproduced
in Fig . 18 for a recip rocat ing shear tes t on dense s and with an
average relat ive densi ty of 7 3% which changed i n 7 1 cyc les to a range
of 5 7 t o 8 7% . In the top diagram the pore pressure gradual ly in
creased , reaching a maximum in 2 1 cycles ; and after that almo s t the
s am pat tern repeated it sel f in all cycles , wi th the peak pore
pressure in each cycle reaching about 9 5% of the 2 . 0 kg/ sq cm vertical
confining press ure . Only three segment s of the total record are
reproduce d : the f irs t 23 cycles , cycles 30 to 34 and cyc les 68 to 71 .
In the lower record i t can b e s een tha t the horizontal def lect ions
increased much more s lowly than the pore pres sures . I fac t , when the
pore pressure reached i ts maximum af ter 21 cycles , the def lect i ons
were only about one m in each direc tion ; but they kept increas ing
almos t throughout the tes t and developed reas onab ly s teady de f lections
only during the las t 10 cyc les of about 7 mm in each di rec t ion . In
this tes t the ini t ia l e ff ec t ive confining p ressure was 2 . 00 kg/ sq cm
and the shear stres s cy
cled between
0 . 2 2 kg/sq cm wi th a frequency
of 0 . 12 cycles / sec .
The t rans duce r records of a gyratory tes t on a specimen wi th a
relative densi ty o f 50% is shown in F i g . 19 ; and the redis t ributi on
record is s umarized in Fi g . 2 0 . The e ffec tive vert i cal s t ress at the
s tart of the t est was 2 . 0 kg/ sq cm and the gyra tory shear s t ress
0 . 22 kg/sq cm; thus , T /O
c
= 0 . 11 . The f requency was f = 0 . 10 cycles /
sec .
At the top of Fig . 19 one can s ee that the induced pore pre ssure
rose s teadily to a mximum equal to about 2 / 3 of the ini t ial ef fective
confining p ressure and then i t remained fai rly cons tant . This i s
typical of all gyratory tes t s perf orme d . I n spi t e of t hi s response ,
which di ffers radi cally from the reciprocating tes t s ( and even more so
from cyclic triaxial tes ts in which the pore pressure rises eventual ly
t o the confining p ressure during the moment when the cyclic deviator
st ress passes through zero) , we observe in gyratory test s at leas t as
effec t ive redi st ribution and severe sof tening of the tes t spec imens as
in rec iprocating shear and cycl i c t riaxial t ests .
"
I
18
In gyrat ory t es t s , the def lect ion mus t be measured in two
di rec tions , to de f ine the de f lec t i on ve ctor . In Fig . 19 are reproduced
the records o f the two di spl acement t r:Ds duce rs which meas ure de
f lect i ons in t he p l ane of the rot a t i ng ar and normal t o the ar. For
this tes t the l ag angl e , \ = 84 de grees , is unus ual ly great . For
much denser specimens t he lag angl e i s gene ral ly sml l e r which I f ind
st range .
The redi s t ri but i on a t the end o f this tes t , afte r 10 cycles , is
shon in F i g . 20 . The patter of redis t ribut ion is somewhat s imi lar
to that in re ciprocating tes t s , par t icul ar ly the compacted zones
adj acent t o the bas e and the cap ; but the very loose z'.mes concent rate
in this tes t chi e f ly along the s ides and the edges of the specimen .
The his togram of redi s t ribution i s clearly unsymt rical . The s tandard
deviation is 0 = 7 . 7% .
If the development of pore p res sure equal to the confining
pressure were used as a criterion for t he ons et of cyc lic li que faction ,
we would be forced to the conclusion that one canno t achieve cy clic
l iquefaction in rec iprocating shear and in gyratory shear tes ts .
Nevertheless , radi cal redi s t ributi on of the water content and s eve re
sof tening o f the t es t specimens develop not only in cycli c t riaxial
tests but als o in reciprocat ing shear and e special ly in gyratory shear
tests . The gyratory mo tion seems to be particularly e ffective in pro
ducing redis t ribut ion of water content and density in conj unction with
relatively hi gh ef f ect ive s t resses .
A st at ist ical analys is of the redi s t ribution in al l reciprocating
and gyratory tes t s on Banding Sand showed that the s tandard deviation
decreases wi th increasing relat ive dens i ty ; and approximately in the
same patter for b o th types of t es t s when other variables are kept un
changed . The t ot al range of the s tandard devia ti on ranges between 2
ad 10% . Specimens as prepared , not s ubj ected to shear , fal l into the
narrow range o f 2 to 3% ; and dense specimens had only s light ly greater
st andard deviations , even after 100 cycles .
IX. COMARI SON OF CYCLIC STRE SSES INDUCED IN S ITU AD IN LABORATORY
TESTS
The s tresses in an e lement in s i t u, b eneath a level ground surface
in a normally cons olidated sand s t ratum, are i l lust rated in Fi g . 2 1 .
Ini tially , the princ ipal s t resses are ve rtical and horizontal , with a
principal s t res s rat io of ab out 2 . 0 , a show by element ( A) in the
left upper corner and in the Mohr diagram by the s t ress c i rcle "A" .
No we s uperpose earthquake-induced horizontal cyclic shear force s , as
sho in the e lement des i gnated ( B&C) . This causes ,the princ ipal
st resses to swing l ike a pendulum between the pos it ions shown in the
inc lined element s ( B) and ( C) . The corresponding s t ress ci rcle s
coincide i n the single s tress ci rcle des i gnated "B&C" which i s large r
than circle i t A" and concentric . During each cycle , all s t resses are
represented by al l c oncent ric circles between these two circles ,
progress ing f rom A to B to A t o C t o A. The s t resses in the element
never approach a hydros tati c s t ate of s t res s , i . e . , a s tate in which
19
shear s t resses on all p lane s s i mul taneous ly are reduced to ze ro .
In Fig . 2 2 we cons ide r an e lement i n s i tu whi ch is s ub j ect ed also
initially to s tat i c shear s t resses on hor l zont a l p lanes , as e . g . , in a
dam or in the f oundat i on of a s t ruct ure . The superpos i t i on o f the
cyclic shear s tres ses on the s tat i c she a r s t ress causes the ini t ial
s tres s c i rcl e
t i l "
t o cyc l e be twe en t he c i rc les " I I " , " I I I " and " IV" .
As in the p reced ing case , tbe s t res s es in the e lement wi l l never
app roach a hydros t at i c s tate of s t res s .
In an e lement in s i t u, with the dimens i ons of a smal l laborat ory
test specimen , at any given moment al l s t resses on the s ur faces o f the
element and in i ts interior are for al l pr ac t i cal puroses uni formly
dis t ributed and the s t ress gradi ent s w i thi n the specimen are zero .
Even in a s t ra t i fied sand wi th layers of di f fe rent compos i t ion , the
normal and shear s tres ses that are t ransmi t ted along the interface of
to layers wi l l be uniformly dist ributed over an area of a few square
inches at any given moment during cyclic loading .
Now we cons ider in Fig . 2 3 the s tresses on the boundaries and in
the interior of a t riaxial tes t spec imen which is f i rs t i s ot ropica11y
consolidated and then s ubj ected in undrained s tate to posi tive and
negative vertical deviator s t resses . The s tress circle for the ini t ial
s t resses is the pOint H on the hori zontal axis in the Mohr diagram.
Superpos i tion of deviat or s t resses ranging between + o and - o
causes s t res s ci rcles to emerge on both s i des of point H , with a
maximum diame ter o . How wi ll a dense s and respond to such cycli c
loading? Al though the s and is s t rongly di l ative , nevertheless every
t ime i t is cycled through the hydros tatic s tate of s t ress , it deve lops
a slight ly cont ractive response over a s mall range of deviator s t ress
ad s light pore p ressures will be induced . Pumping act i on of the
ver tical cyclic forces tends to move the excess water toward the top of
the specimen . Thi s action , comb ined wi th the e ffec ts of interal
s tress gradients induced by non-uni form dis t ribut ion of s tresses on the
boundaries . causes redi s t rib ut i on of water content and sof tening of the
top of the spec imen . The cyclic peaks of the pore p ressures keep in
creas ing and finally rise momentari ly to the c onfining pressure 0
c
every time the s t resses cyc le through the hydros tatic s t ate ; or, to use
the term no generally used in l i terature for this phenomenon , the
specimen suf fe rs " lique fac t ion" . As already exp lained in the int ro
duct ion, this response bears no relat ionship to actual l iquefact ion of
sad ; and t o di st inguish between thes e two fundamentally dif ferent
phenomena I am us ing in thi s paper " cycl i c l ique fact ion" when referring
to the phenomenon of redi s tribut ion and sof tening in a laboratory
specimens and when the cycl i c pore pressure moment arily equals the
conf ining p ress ure . Exactly how cy cling through the hydros tatic s t ress ,
the pumping action and the cyclic inte ral s t res s gradi ents comb ine t o
achieve cy clic l iquefact ion in such tes t s remains t o b e inves t igated .
Te mechanics of redis tribution in recip rocating direct shear tes t s
is much eas ier to comp rehend . In Fi g . 24 are shown the non-uni form
s t resses that are gene rated in such tes ts . For compari son, in the
upper l ef t hand corer are shown again the s t res ses in an el ement in
situ in whi ch app l i cat ion of cycl ic hori zontal shear s t resses
' I
i
20
automtically mob ilize the same shear s t resses on vertical planes .
All stresses on these ficti tious boundaries of the elemnt , as well as
in the interior of the element , are uniform at any given moment . In
s tark cont ras t , the boundaries of every tes t specimen are cause for
non-uiform s tress dis tributions on the boundaries and s tress gradients
in the interior . The rubber memrane transmits to the cylindrical
boundary of the specimen only normal s t resses . Because the shear
s tresses on this vertical boundary are zero , the shear s tresses along
the edges on the horizontal faces of cap and b ase are also zero .
Terefore , the applied shear force is distributed approximately para
bolically , a shown in Fig . 24 . But the overturing moment of these
shear forces is not balanced by a countermoment produced by shear
forces on the vertical boundaries . Therefore , to maintain equilibrium,
the center of gravity of the norml st resses transmit ted by the cap and
the base mus t be excent ric ; ad this excentri ci ty must cycle back and
forth. This highly non-uniform and cyclically changing pattern of
boudary s tresses and the resulting severe interal s tress gradients
tend t o compact the sand adj acent to the cap and the base and produces
the type of redis tribution that we have measured in many reciprocating
ad gyratory shear tests ; (Ref . 12 and 13).
X. COMAISON OF RSPONSE TO CYCLIC LOADING OF DENSE SADS IN SITU
AD IN LABORTORY TESTS
Depending on the type of cyclic tes t , one or more of the following
three mechanism are the maj or causes of redist ribution of water
content and development of cyclic pore pressures and cyclic s trains in
laboratory tes ts on dense sands :
1) Cycling through hydros tatic s tate of s t ress .
2) Puping action by vertically app lied cyclic loads .
3) S tress gradients within the test specimen generated by non
uiform s t ress dist ribution on the specimen boundaries .
Tis mechanism is probably aggravated by the small s i ze of
the test specimens in conventional tes ts .
In addit ion, a slight rearrangement of grains , generated by b reaking of
edges and corners of grains , may make a minor cont ribution to the
development of pore pressures and s t rains in laboratory tests . This
mchaism is normally the only one that p roduces pore pressures and
s trains within an element of dense s and in sit u. But in laboratory
tes ts this mechaism is completely overshadowed by the effects of the
three principal mechanisms listed above .
Because of the basic differences between the mchanisms that
produce cyclic pore pressures ad s t rains in the laboratory and in si tu,
I cosider it virtually an impossib le task to extrapolate or est imate
from such laboratory tes ts , with the help of emirical relationships ,
ho cclic pore pressures and strains will develop in situ.
21
Wenever one makes use of empirical relationships , there mus t be
a comon denominator. ( E . g . , the At terberg limi ts are useful in soil
mchanics for es tablishing emi rical relationships because p las ticity
ad various engineering properties are related t o the contents of clay
particles and their mineralogy . ) I cannot find a comon denominator
between the p rincipal mechanisms that control the cylic response in
laoratory specimens and the response of an element in situ. In
laoratory specimens the deve lopment of pore pressures and cyclic
deformations is the result of radical redis tribut ion which in turn is
produced by mechanisms that normally do not exi s t in an element in situ.
I n a element in s i tu normally there is no redis tribution and the
s tress es remain homogeneou, i . e . , no s t ress gradients develop wi thin
the specimen. In the conventional cyclic t riaxial tests cyclic lique
faction wi l l develop even in very dense sands , when applying
enough cycles ; but in s itu, cyclic liquefaction cannot develop , except
perhaps quite locally , adj acent to certain rigid boundaries such as
piles or piers .
Is there a pos s ib ili ty of developing a laboratory test which will
duplicate the s tresses within an element in situ? I t is conceivable
that we could shake a large mass of sand, e . g . , a 3 meter cube , with a
well instrumented "element" in the center of this mass , to measure
cyclic pore pressures and s trains . I t would be a costly research
effor t , b ut worth serious consideration .
Aother approach would be the application of the finite element
mthod . To reduce the number of variab les to a minimum, I suggest
comaris on of the two-dimensional s tress dis tribution of an element in
s i tu with that in a long prismatic tes t specimen subj ected to reci
procating shear , with zero shear s tress on the vertical sides . This
would be equivalent to ass uming that in situ a sand s tratum cont ains
( 1) frictionless , vertical , parallel planes which are aligned noral t o
the cyclic forces , ad ( 2 ) roug and inextensib le horizontal planes .
I t will require experimentation with various assumptions of the sand
properties to achieve redistribution of water content that approximates
what we have observed in reciprocating shear tes ts . When I firs t
considered this app roach , I concluded that it is too complicated for
the availab le techniques . But the vision of successful treatment is
so ver attractive that I would like to persuade the mos t resourceful
specialis ts in this b ranch of app lied mathematics to lend a hand .
Success of this approach would not only result in a convincing
demons tration of the enormou differences between the response of in
situ elements and laboratory specimens t o cyclic loading, but it would
open up the pos s ibility of inves tigating the effects of local s tress
concentrations produced by rigid bodies that penet rate into a sand
mss , when subj ec ted to earthquake loading .
Te only clearly realis tic approach t o this ent ire prob lem area
consis ts of comprehensive investigations of the full-scale experiments
which nature has made and continues to perform in highly seismic
relions of the world. But reliab le information wi ll not be derived
from such inves tigations uless they are preceded by extensive
prelimnary efforts , t o sharpen our tools so t o speak ; in particular to
imrove available methods and to develop new methods ( 1) for conducting
, '
22
in s i tu measurements of the degree of denseness o f sands , and ( 2) for
determining the relative dens ity on small specimens cut from un
disturbed block samples that are taken by hand after lowering the
groundwater . Whenever possible also the age of s and s trata should be
determined because there are indications that young alluvial sands are
much more sus ceptible to actual liquefaction or development of large
s trains than old sediments that have been subj ected already to many
severe earthquakes . Because of the great magnitude and cos ts o f such
comprehensive investi gat ions and for o ther reasons , they would best be
udertaken as cooperative efforts by several countries
with zones
subj ect to severe seismic act ivity .
Most pub lished field invest igations on effects of earthquakes on
sad deposits are deficient and make interpretation difficult . Un
warranted reliance is placed on the s tandard dynamic penetration tes t
which is far from a s t andard as pract iced today . ( E . g . , during a
comrehensive inves tigation of a rather uiform s tratu of beach sand
deposi ts for the foundation of a nuclear poer plant , four different
boring organizations made borings independently at the s ame location ,
within a 3 m radius . When comparing their result s , the blowcounts
differed by as much as a factor of three ! ) As a basis for es timating
the in situ relative densi ties by correlation with blowcounts , there is
a tendency to use the average b lowcount for an entire s tratum ins tead
of paying special at tent ion to layers or lenses with the lowest blow
counts . In alluvial deposits one often encounters large variations in
densities between adj acent zones in the same sand stratum. Dest ructive
movements at ground level may be caused by actual liquefact ion of a
loose , relatively thin layer that can be eas i ly overlooked in such erude
subsurface invest i gations ; and then overlying , much denser layers are
wrongly b lamed for having caused the movements . We need more detai led
and more accurate in situ measurements than is possible by dynamic
penetration tes ts , perhaps continuous s tatic cone penetration measure
ments , to identify the looses t zones that are responsib le for large
movements . Unfortunately the identification of such zones is
particularly dif ficult or impossible when the lique fied sand has
floed out laterally or through cracks to the surface .
Study of descriptive l iterature of earthquake ef fects on alluvial
deposits convinced me that large and des tructive groud movements are
caused by actual liquefaction or severe sof tening of contract ive sand
l ayers at some depth. I believe that the lique faction potent ial of
such layers could best be identified by means of triaxial t es ts on
udisturbed samp les , using load control tes ts of the type developed by
Cat ro ( Ref . 8) .
For typical alluvial s ands containing not more than a few percent
of silt sizes , the upper limit of relative densi ty for which I consider
actual liquefact ion to be a possibility , lies in the range be tween
aout 40 and 50% , depending on confining pressure . Medium- loose sands
in the range between 40 and 60% may be slightly contractive or slight ly
dilative ; and in s i tu they may respond to cyclic loading with s trains
of objectionab le magnitude , but rarely with actual lique faction . In
strongly dilative , anisotropically consolidated sands in s itu, with

Ij


I
23
relative dens ities greater than about 70% , I consider it norally
impossib le for cyclic pore pressures to approach or equal the confining
pressure becaue dilatency will automatically cause the grain
s tructure to offset loss of s trength by "bracing itself" so to speak ,
requiring only minute st rains ; and I doub t that induced pore press ures
will normal ly ri se as high as 50% of the confining pressure . In
contras t , in laboratory specimens of dense sand the redistribution of
water content , development of s of t zones and cyclic l iquefaction
cannot be prevented by mob i lization of a strong di lative respons e.
Special condi tions that may cause development of high pore pressures
ad excessive s trains in dense sand in s i tu are the following:
( 1) Adj acent to rigid b oundaries where severe st ress gradients can
develop and produce redis tribution of water content . I t would merit
investigation whether such response could substantially reduce the
b earing capacity of piles or piers during earthquakes . ( 2 ) I f i t were
possib le to maintain in s i tu a mass of sand in an isot ropically conso
lidated s t ate during cyclic loading (which I doubt ) , then cyclic lique
faction in dense sand would be theoretically possible .
XI . LAORATORY TESTS FOR DESIG PURPOSES - INTERIM RCOMNDATIONS
Until such time when more reliab le data will be available on the
in situ response of dil ative sands to cyclic loading , I have recomended
in recent years , as a temporary expedient , the use of cyclic triaxial
tes ts on anisotropically consolidated specimens cut from undis turbed
sales . Independent ly , other investigators have also tended to
include cyclic triaxial tes ts on anisotropically consolidated specimens .
See, e . g . , the comrehensive report on ANALYSIS OF THE SLIDES IN THE
SA FERAD DAMS DURING THE EARTH -UAE OF FEBRUAY 9, 1 971 , ( Ref . 14) .
In m recomendation and interpret at ion of such tests I a now guided
by the following observations and assump tions :
1) In cyclic t riaxial tes ts , redis tribution and buildup of cyclic
pore pressures becomes particularly severe when the peak pore pressures
rise wel l above 50% of the confining pressure . Therefore , I use as a
test paraeter the number of cycles , N
SO
' required for the pore
pressure to reach 50% of the confining pressure . At that point the
cyclic s trains are generally s ti ll small .
2) I mae the following arbi trary assumptions which I consider
to be well on the s afe side : ( a) that the mechanism which causes the
buildup of pore pressure in situ (breakage and minor rearrangement of
grains) is responsible for one-fourth of the cyclic pore pressures
induced in laboratory tes ts , with the other three-fourths produced by
the mchanism that cause redist ribution of water content and sof tening
of the tes t specimens ; and (b ) that in situ the pore pressure will
reach 50% of the confining pressure which in m j udgment is an upper
limt for medium dense and dense s ands under the wors t conditions .
3) Te pore pressure in the tes ts rises app roximately in direct
proportion to the numer of cycles up to about 50% of the confining
pressure . Wen comining this wi th the preceding assumption, it follos
I
. i
24
that the numer of cycles which would be required in s i tu to reach 50r
of the confining pressure woud be 4xN
50
.
4) I f an estimate of the cyclic s trains is required, it would
be netessary to carry out precise measurements of the cyclic s t rains
up to N
SO
and use these s trains a a basis for j udgment , keeping in
mnd that only a fraction of these s trains are caused by the mechanism
that controls the development of pore pressures and s trains in s i tu .
Again I wish t o emphasize that the strains which develop in cyclic
laboratory tests when the pore pressures rise to the conf ining pressure
( i . e . when cycl ic liquefact ion is reached ) , as well as all further
increase in cyclic strains during additional cycling , bear no relation
shi t o the strains that develop in situ for dilat ive sands , in my j udgment .
I consider the use of the bes t possible undis turbed samples
(preferably hand-cut b lock s amp les ) an essential requirement for
meaningful resul t s . I also require the use of the type of lub ricated
ends which were developed by Professors P . Rowe and L. Barden (Ref . 15) .
In general , the tes t specimens should be preconsolidated to a principal
st ress-ratio of 2 . 0 .
For average alluvial sands wi th a silt fract ion not exceeding a
few percent , ad with a relative densi ty in the range between about 40
and 60% , we deal with a transition range between loose sands that can
actually liquefy and medium dense s ands that are safe agains t actual
liquefaction . In that range it is possible that triaxial tes ts with
monotonic load increas e, using dead l oad increments , will produce
larger s trains than cyclic loading . Therefore , in that range I
recomend also performance of monotonic load control tests on specimens
that are anisot ropically consolidated with a principal s tress ratio of
2 . 0 . Such tes ts also permit a better dif ferentiation between dilative
ad contractive response than cyclic tests . Unless the results sho a
well developed dilative response , I advise performance of additional
tests to define the F line ; Figs . 6 and 8.
XII . SUMRY OF PRINCIPA CONCLUS IONS AND RCOMNDATIONS
1. General - To dif ferentiate between basically dif ferent
phenomna which are both called LIQUFACTION in literature , the term
ACUAL LIQUFACTION is used in this paper for the response of contractive
( loose) sand which results in subs tantial increase in pore pressure and
loss of s trength . that can cause a flow s lide ; and the ter CYCLIC LIQU
FACION is used for the response of dilative (dense) s and in cyclic
triaxial tes ts when the pore pressure rises momentarily in each cycle
t o the confining pressure .
2 . Cri tical void Ratio - The concept of the critical void ratio,
developed more than 40 years ago ( Fig . 1 ) had t o be modified on the
bas is of inves tigations of the f lo slide in Ft . Peck dam and Cas tro ' s
investigations ( Ref . 8) . During actual liquefaction the sand grains
,
(
25
rearrange themselves into a flow structure which ensures a minimum
resistance to flow. By using dead loads in R (undrained ) triaxial tests ,
liquefaction can be achieved at greater relat ive densities than with
s train control tests ; and the type of test used 40 years ago gave
results even more on the unsafe side . The sae relationship between
critical void ratio and effective minor p rincipal stress is obtained
with ( 1) isotropically consolidated specimens , ( 2 ) anisotrop ically
consolidated specimens , and ( 3) with cyclic load tes ts ; Fig . 6 .
3 . Redi stribu tion of Wa ter Content in Cycli c Tests - The pro
gressive increase of cyclic pore pressures and softening in test
specimens in various types of cyclic tes ts , including cyclic l iquefaction in
triaxial tests , are caused by radical redistribut ion of the water content
which is generated by mechanisms that are norally not active in situ.
4 . Recomenda tion for Field Investi
g
a ti ons - Because it is
unlikely that a laboratory tes t can be devised which will produce in
tes t specimens the type of uniform stress distribution that exists
during cyclic loading in a typical element in s itu , the author believes
that closing the great gap between laboratory and in situ response will
require primarily comprehensive field invest igations of many full-scale
tests that nature has performed in highly seismic regions . Such in
vestigations will have t o be carried out with the bes t possible
investigational tools and with meticulous attention to details , lest
they will create misleading information .
5 . Interim Recomendations for Laborator
y
Tests - For estimat ing
the in situ response of medium dense and dense sands , i . e . , sands which
are safe agains t actual liquefact ion , an interim procedure is suggested
which is based on performance of cyclic t riaxial tests on anisotropically
consolidated specimens cut from undisturbed samples .
Whenever a quest ion arises whether a sand stratum could experience
actual liquefaction , it is recomended to carry out R tests with
dead load increments , as developed by Castro ( Ref . 8) , using the best
possible undisturbed samp les .
XIII . ACKOWEDGEMNTS
The doctoral research by Gonzalo Cas tro and Franklin Rendon was
made possible with assistance of many organizations and individuals .
Detailed acknowledgement s were included in Castro ' s thesis ( 8) and
will be included in Rendon ' s thesis .
I am particularly indebted to Stanley D . Wilson for his crit ical
review of a draft of this paper and his many valuable suggest ions .
j
!
}
i
I
2
XIV. REFERNCES
(l) Casagrande, A. , "Characteristics of Cohesionless Soils Affecting
the Staility of Earth Fillslf , Journal of the Boston Soiet
y
of Civil Engineers, January 1936 . Reprinted in "COTRIBUTIONS
r SOIL MCHANICS, 1925-1940" , Boston Society of Ci vil
Engineers , October 1940 .
( 2) Casagrande , A. and Watson, J . D. , "Copaction Tes ts and Critical
Dnsity Investigations of Cohesionless Materials for Franklin
Falls Damtl , Appendix BII in Reprt to the U. S. En
g
ineer Corps,
Boston, 1938.
(3) Casagrande , A. , liThe Shearing Resistace of Soils and its Relation
to the Stability of Earth Da" , Proceedings , Soils and
Foudation Conference of the U. S. Engineer Dprtment , Bos ton,
Ms . , June 1938 .
(4) Watson, J . D. , "Stress-Deformation Characteristics of Cohesion1ess
Soils from Triaxial Compression Tests" , Sc. D. TheSiS , Harvard
University , 1940 .
(>) Cors of Engineers , U. S . Department of the Arm, "Report on the
Slide of a Portion of the Upstrea Face at Fort Peck Da. "
U.S. Governnt Printing Office, Washington, D. C. , July 1939 .
() Mddlebrooks , T . A. , "Fort Peck Slide" , ACE Transactions , Vol .
107 , 1942 .
( 7) Casagrande, A. , "Role of the Calculated Risk in Earthwork and
Foudation Engineering" , Journal of the Soil Mechanics and
Foudations Division , ASCE, July 1965 .
(8) Catro, Gonzalo , "Liquefaction of Sands" , Harvard Soil Mehanics
Series No. 81, January 1969 .
(9) Cas tro, Gonza10 , "Liquefaction ad Cyclic Mobility of Saturated
Sads " , Joural of the Gotechical Engineering Divisi o,
ACE, Jue 1975 .
(lOa) Seed, H. B . ad Lee, K. L. , "Liquefaction of Saturated Sands during
Cclic Loading" , Journal of the Soil Mehanics and Foundations
Dvision, ASCE , Novemer 1966 .
( lOb) Lee, K. L. and Seed, H. B . , "Cyclic Stress Conditions Causing
Liquefaction of Sand" , Journal of the Soil Mehani cs and
Fouations Division, January 1967 .
( 11) Shockley, W. G. and Ahlvin, R. G. , 'on-Unifor Conditions in
Triaxial Tes t Specimens" , Research Conferene on Shear strength
of Cohesiv Soils , ASCE, Boulder, Colorado, 1960 .
27
( 12) Casagrande , A. and Rendon, F . , "Reciprocating and Gyratory Shear
Apparatus - Design , Tes ting Procedures and Tes ts on Saturated
Sand , " Report to WATERWAYS EXPERIMENT STATION, Vicksburg ,
Miss . , 1976 .
( 13) Rendon, t . , Doctoral Dissertation ( in preparation) , Harvard
University , 19 76 .
( 14) Seed , H . Bolton, Lee , K. L . , Idries , I . M. and Makdisi , F. l . ,
"Aalysis of the Slides in the San Fernando Dam during the
Earthquake of Feb . 9 , 19 7 1, " Earthquake Engineering Research
Center, Report No . EERC 73-2 , University of California ,
Berkeley , California, June 1973 .
( 15) Rowe , P . and Barden , L . , "Importance of Free Ends in Triaxial
Testing, " Journal , Soil Mechanics and Foundations Di vision ,
ASCE , January 1964 .

; i ,
-

You might also like