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

net/publication/239229537

Strength and dilatancy of sands

Article in Geotechnique · January 1986

DOI: 10.1680/geot.1986.36.1.65

CITATIONS READS
2,690 13,758

1 author:

Malcolm D. Bolton
University of Cambridge
248 PUBLICATIONS 15,423 CITATIONS

SEE PROFILE

All content following this page was uploaded by Malcolm D. Bolton on 23 October 2018.

The user has requested enhancement of the downloaded file.

 BOLTON,M. D. (1986). GCotechnique 36, No. I.65578

The strength and dilatancy of sands

M. D. BOLTON*

Extensive data of the strength and dilatancy of 17 sands were more accentuated than the points in
in axisymmetric or plane strain at different densities and common. There has followed a growing agree-
confining pressures are collated. The critical state angle ment amongst those researching the strength of
of shearing resistance of soil which is shearing at con- soil
stant volume is principally a function of mineralogy and
can readily be determined experimentally within a (a) that secant, rather than tangent, 4’ values
margin of about l”, being roughly 33” for quartz and should be the basis for discussion
40” for feldspar. The extra angle of shearing of ‘dense’ (b) that dilatancy towards critical states is central
soil is correlated to its rate of dilation and thence to its to an understanding of soil behaviour
relative density and mean effective stress, combined in a
(c) that both effective stress and soil density affect
new relative dilatancy index. The data of #,,., - &, in
the rate of dilatancy of soils and thereby their
triaxial or plane strain are separately fitted within a
typical margin of about 2”. though the streneth of strength parameters.
certain sands is underpredicted ii the looO_?OOOO That this understanding has failed to permeate
kN/m’ range owing to the continued dilation of their
more widely into practice can partly be blamed
crush-resistant grains. The practical consequences of
these new correlations are assessed, with regard to both on the structure of the historic argument, which
laboratory and field testing procedures. revolved around the theoretical relationship
between strength and dilatancy. Since practition-
L’auteur analyse de nombreuses don&es concemant la ers would usually be in the position either of mea-
resistance et la dilatance de 17 sables sous deformation suring both or guessing both, this aspect of the
plane ou axisymctrique pour differentes densites et dispute must have seemed sterile.
pressions d’itreinte. L’angle de resistance au cis- The failure to bridge the gap between research
aillement dans Y&at critique d’un sol soumis au and practice has many serious consequences,
cisaillement A volume constant est principalement une however. Engineers often do not appreciate
fonction de la mineralogie et peut se determiner facile-
ment a 1” pres, comme ayant une valeur &environ 33” (4 that a secant 4’ value derived from a single
pour le quartz et 40” pour le feldspath. L’angle supple- triaxial test on a single specimen can offer a
mentaire de cisaillement dun sol dense depend a sa conservative parameter for design if the
vitesse de dilation done de sa densite relative et de la testing conditions are carefully chosen
contrainte effective moyenne, combinbes dans un nouvel
indice de dilatance relative. Les donnies de &., - &, (b) that the full range of soil strengths can be
en deformation plane ou triaxiale sent separees par une
expressed in terms of the variation of (secant)
marge de 2” approximativement, bien que la resistance 4 with density and stress
de certains sables soit sousestimee dans une fourchette (4 that the conventional tangent parameters (c’,
de 100&10000 kN/m* en raison de la poursuite de la 4’) can only describe the full range of soil
provoqute par l’bcrasement de leurs grains resistants. strengths if both are allowed to vary with
L’article evalue les consequences pratiques de ces nou- density and stress
velles correlations en ce qui concerne les mdthodes
(4 that ignorance of the foregoing can lead to
d’essai en laboratoire et in-situ. significant errors in predicting ultimate
bearing stresses, for example.
KEYWORDS: friction; sands; shear tests; soil proper-
ties; statistical analysis. The objectives of this Paper are as follows:
firstly, to clarify the concepts of friction and dilat-
ancy in relation to the selection of strength
INTRODUCTION parameters for design; secondly, to introduce a
Following the early work of D. W. Taylor, the new relative dilatancy index and to demonstrate
strength and dilatancy of soils received a great its use in the prediction of the behaviour of sands
deal of attention in the 1960s. Several schools of at failure in relation to the available published
thought developed, the differences between which data; thirdly to consider the implications of the
new correlations to the procedures of laboratory
Discussion on this Paper closes on 1 July 1986. For and field testing.
further details see inside back cover. Most of the concepts which are to be discussed
* Cambridge University Engineering Department. can be applied with equal force to clays as to
65
66 BOLTON

&j//y kc
I 2 3 4 0 2 4 6 8
E, % E,: %
(a) (b)
Fig. 1. Stress-strain behaviour of dense sand in plane compression (a)
at low stress and (b) at high stress (Barden e? al., 1969)

sands. In particular, the explanation of ‘true cohe- will dilate fully to achieve a critical state, at which
sion’ as a by-product of dilatancy in over- shear deformation can continue in the absence of
consolidated clays is particularly fruitful (Rowe, a volume change.
Oates & Skermer, 1963). The present Paper is The point of peak strength, P, is usually associ-
restricted in data, however, to those granular ated with the maximum rate of dilation defined as
materials for which ‘relative density’ is a relevant ( - d&,/de Jmax where E, is the volumetric strain
and measurable parameter. Accordingly, the and ai is the major principal strain (both defined
undrained strength problem has been set aside, positive in compression). The Mohr circles of
and the discussion is focused on fully drained effective stress, and differential strain, pertaining
behaviour in terms of effective soil stresses. at peak P are shown in Fig. 2(a). The peak angles
of shearing resistance &,,, and dilation II/,,, are
ANGLES OF SHEARING AND DILATION defined therein and given by
Data of a typical drained, plane strain, com-
pression test (Barden, Ismail & Tong, 1969) on a
dense, rectangular sample with frictionless ends is
shown in Fig. l(a). Strains were inferred from
:. &,,,, = 44.8”
boundary displacements and volume changes,
and they therefore underestimate the strains in
the rupture zone which developed between points (dsi/W,,, + 1 (,)
P and C. The achievement and accurate determi- (Wd~,L,,,, - 1 -

nation of the ultimate conditions at C are con- :. $,,, = 14.7”

siderably hampered by the non-uniformity of the
sample and the uncertainty regarding membrane It should particularly be noted that q5’ herein
correction following the formation of a rupture refers to the ‘secant’ angle of shearing obtained by
plane. Nevertheless such evidence as exists dropping a tangent from the origin on to a single
(Roscoe, 1970) suggests that soil in rupture zones Mohr circle of effective stress.
 STRENGTH AND DILATANCY OF SANDS 61

i kN/m’ T kN/m2
t

KO
dyi2

Fig. 2. Mohr circles of stress and strain increments of dense sand at maximum
stress ratio in Fig. 1 (a) at low stress and (b) at high stress

A similar plane strain test on dense sand under

extreme stresses mimics the normal behaviour of
very loose sand and appears in Fig. l(b), with the
data for point C at .a1 = 8% presented in Mohr
circles in Fig. 2(b). The peak phenomenon is effec-
tively suppressed while the sample volume con-
tracts rather than dilates. The end state of the
whole sample at C is analogous to that which can
be observed in the thin rupture zones of the dilat-
ant sample, tending towards a critical state

The mechanical significance of the angle of

dilation in a plane strain deformation can best be
appreciated by assuming that the Mohr circle of
plane strain increments in Fig. 2(a) can be applied
to the case of direct shear shown in Fig. 3. If rigid
blocks of non-failing soil are assumed to bound
the thin uniformly straining rupture zone ZZ, this
must mean that for compatibility ZZ must be a
zero extension line so that

de, = 0 (3) Fig. 3. Angle of dilation $ in plane shear

68 BOLTON

line for a given ‘upstream’ excitation at r = r. If

a relative rotation of the two supposedly rigid
zones then occurs during collapse, the ensuing
downstream displacements are simply pro-
portional to the radius and are therefore corre-
spondingly increased as tj increases. In practice, it
is known that failures can occur progressively,
with different mobilizations of strength and dilat-
ancy at various locations along a developing slip
surface.
If rupture figures comprising a mosaic of
sliding wedges with plane faces are alternatively
assumed, the effect on the stability calculations of
increasing Ic/ at constant 4’ is easily shown either
to be zero if the mosaic is still capable of sliding
in the manner intended or to cause a seizure of
the mosaic due to enhanced wedging which can
Fig. 4. Logarithmic spiral slip surface make a previously critical slip mechanism kine-
matically inadmissible. Depending on the particu-
within the rupture zone. Also lar geometry of the problem therefore, the effect
of increasing ij at constant 4’ may either be
dyy, = * neutral or beneficial to the overall stability. A safe
Y strategy must therefore be to err, if at all, by
underestimating both 4’ and $ in any subsequent
dE z-d! analysis.
Y A great deal of attention has been focused on
Y
the relation between 4’ and II/, and in particular
so in Fig. 3(b) between &,,, and Ijl_X. The stressdilatancy
theory of Rowe (De Josselin de Jong, 1976; Rowe,
tan*= _dEy=dy 1962, 1969) has proved to possess great explana-
(4) tory power, as well as a close approach to the
dyyz dz
published data. Since the intention here is prin-
The angle of dilation is then seen in Fig. 3(b) to cipally to introduce and correlate data, it is
be equal to the instantaneous angle of motion of unnecessary to expand in detail on the theory. It
the sliding blocks relative to the rupture surface. will be valuable, however, to employ a less rigor-
Figure 4 demonstrates that the consequence of ous but similar approach to develop an expres-
the assumption of a constant angle of dilation on sion which deviates only slightly from Rowe’s in
a slip surface is the replacement of slip circles by its estimate of 4’ appropriate to a plane shear
the logarithmic spiral slip surface between rigid test.
zones, since for a small angle AOB = d6 Suppose that &, is the angle of shearing
CAB=$ observed in a simple shear test on soil loose
enough to be in a critical state, with zero dilation.
CB=dr Now suppose, following Fig. 5, that the same soil
AC=rdtJ is tested dense, so that overriding at points of
contact must occur unless the particles crush.
Suppose that the particles above the overall zero-
:. tan II, = s
extension line ZZ form one rigid zone sliding
upwards at $ over the rigid zone beneath, in
so that
accordance with the external observation of a
dr dilatancy angle +. Assume that the angle of shear-
- = tan + d0 (5) ing developed on the inclined microfacets SS, on
r
which there is zero dilation, remains at &,.
On integrating Since all the sliding now takes place on surfaces
parallel to SS it is permissible to view the
r = r. exp (0 tan $) (6) observed angle of shearing 4’ on the rupture
where r = r,, at 0 = 0. At tj = 0 the well-known surface as comprising the two components &,
slip circle r-= r. is indicated. The larger is $ the and $ as shown in Fig. 5. This specifies
larger is the ‘downstream’ radius of the rupture
 STRENGTH AND DILATANCY OF SANDS 69

L I

Fig. 5. The saw blades model of dilatancy

as an elementary frictiondilatancy relation in the tendency towards dilation would be solely a

which no attempt has been made either to opti- function of the density and arrangement of the
mize the failure mechanism or to correlate the particle structure. However, experiments with
directions of principal stress and strain steel shot, which were effectively unbreakable but
increments. were deformable plastically, have shown (Bishop,
Figure 6 compares equation (7) with Rowe’s 1972) that increased confining pressure leads to
stressdilatancy relation for plane strain reduced angle of shearing. It has also been
demonstrated (Bishop, 1972; Billam, 1972; Vesic
& Clough, 1968) by tests on granular soils at ele-
vated pressures that particle crushing occurs,
thereby reducing the observed maximum angles
in the particular case &, = 33” corresponding to
of dilation and shearing, $,,, and 4&,, for a
the typical value for quartz sands, and using
given initial density. Soil particles may crush
equations (1) and (2) for 4’ and I,+respectively. It
before they override.
will be seen that equation (7) overestimates
No completely consistent treatment of both
4’ - 4&t compared with equation (8) by about
density and confining pressure has previously
20%. In other words, Rowe’s stress-dilatancy
been undertaken in producing empirical relations
relationship for plane shear, over the range of ij
by which &,,,, could be predicted. This is now
shown in Fig. 6, is operationally indistinguishable
attempted. To achieve a given angle of dilatancy,
from
it is argued, the particle structure should both be
4’ = &, + o+W (9) dense and not so highly stressed that asperities
The advantage of having developed expressions should fracture in preference to overriding. This
such as (7) and (9) is that any angle of shearing in suggests a relation in which a density parameter
excess of the friction angle of loose earth is seen is multiplied by a stress parameter. Since the
to be due solely to the geometry of the volumetric logarithm of stress has been shown (Billam, 1972;
expansion which is necessary before shearing can Vesic & Clough, 1968) to effect a linear reduction
take place. 4Ja., >an expression of the form

EFFECTS OF DENSITY AND CONFINING $,,, = AI, In y (11)

STRESS ( >
The accepted definition for the state of com-
paction of granular materials is relative density 50-

I, = emax
-e
- emin
emax
where em_ is defined as the voids ratio achieved f2
n
in quickly inverting a measuring cylinder contain-
‘s
ing the dry soil and emin is that achieved under 40-
optimal vibration of a compactive mass under
saturated conditions and without causing crush-
ing. It has generally been found (Cornforth, 1973)
that relative density offers a superior correlation
d Cl,,’
compared with voids ratio for the strength of
sands, presumably since it compensates for effects I
301
of particle grading and shape which influence emax 0 10 20
and e,,,i,. I":deg

If soil particles were perfectly strong and rigid, Fig. 6. Stress-dilatancy relations
 70 BOLTON

Table 1. Sand data

T

Identification Name 4,: 4,: emin emax d&it Reference

mm mm
-
A Brasted river 0.29 0.12 0.47 0.79 32.6 Cornforth (1964, 1973)
B Limassol marine 0.11 0.003 0.57 1.18 34.4 Cornforth (1973)
C Mersey river CO.2 -0.1 0.49 0.82 32.0 Rowe (1969)
Rowe bt Baiden (1964)
D Monterey no. 20 zo.3 -0.15 0.57 0.78 36.9 Marachi, Chan,
Seed & Duncan (1969)
E Monterey no. 0 -0.5 zzo.3 0.57 0.86 37.0 Lade & Duncan (1973)
F Ham river 0.25 0.16 0.59 0.92 33.0 Bishop & Green (1965)
G Leighton Buzzard 14/25 0.85 0.65 0.49 0.79 35.0 Stroud (1971)
H Welland river 0.14 0.10 0.62 0.94 35.0 Barden et a1.‘(1969)
Chattahoochee river 0.47 0.21 0.61 1.10 32.5 Vesic & Clough (1968)
Mol 0.21 0.14 0.56 0.89 32.5 Ladanyi (1960)
Berlin 0.25 0.11 0.46 0.75 33.0 De Beer (1965)
Guinea marine 0.41 0.16 0.52 0.90 33.0 Cornforth (1973)
Portland river 0.36 0.23 0.63 1.10 36.1 Cornforth (1973)
Glacial outwash sand 0.9 0.15 0.41 0.84 37.0 Hirschfield & Poulos (1964)
Karlsruhe medium sand 0.38 0.20 0.54 0.82 34.0 Hettler (1981)
Sacramento river 0.22 0.15 0.61 1.03 33.3 Lee & Seed (1967)
,Ottawa sand 0.76 0.65 0.49 x0.8 30.0 Lee & Seed (1967)

is implied, in which p’ is the mean effective stress by the researchers. Figures for emax and emin were
at failure, pi,,, is the elevated stress just suffkient essential to the studv. Tavenas & La Rochelle
to eliminate dilation by crushing and A is a con- (1972) have shown ;hat variations in limiting
stant. However, the quantity of data for &,,,, is voids ratios due to variations in technique
rather greater than that for $,,,, and that for between the American Society for Testing and
triaxial tests greatly exceeded that for plane strain Materials’ methods and those of Kolbuszewski
tests, and therefore an approach must be used should not usually exceed 0.02. This would imply
which can embrace these additional data. an experimental error band for relative density
It should be recognized that the angle of dilat- determinations of the order ofi& 0.05.
ancy becomes a meaningless parameter in an The critical state angle of shearing was central
axisymmetric triaxial compression test, since the to a rational portrayal of the data. The ideal
geometrical relationship of Fig. 2 applies only to method of determination for 4&t is from extrapo-
plane stiain. Indeed a Mohr circle of strain lation of a series of values of &,,, recorded in
increments for the densest sand in triaxial com- compression tests on soil samples of various den-
pression would indicate that mean strains in any sities, in which the rate of dilation at failure was
vertical plane were contractile rather than dilat- observed, so that a value consistent with zero
ant, owing to the implied distribution of lateral dilation at failure could be determined. In each
strain amongst all the available directions. It is case quoted in Table 1, it was felt that sufficient
therefore necessary to revert to (-~E,/~EJ,,~~ as information was available to make a judgement
the useful measure of triaxial dilatancy rate. on &, which should not have been in error by
Table 1 assembles the characteristics of the more than 1”. The observed range for the sands
sands used in the correlation study, comprising a was from 32.0” to 37V, the higher values always
large proportion of the relevant data available in pertaining to sands which were said to contain a
the literature and emanating from well- significant proportion of feldspar, the lower
established laboratories. The data refer to values pertaining to quartz sands. This was con-
samples tested in compression with an initial sistent with previous findings (Koerner, 1970;
height-to-width ratio of 2, unless indicated other- Lee, 1966) that &, for felspathic sands was of the
wise. Some researchers failed to disclose the order of 40”.
precise conditions of the test platens, and it must An initial perusal of the test data showed that
be assumed that some of the variations in results equation (11) was insufficient as a measure of
may be accounted for by variations in end fric- dilatancy potential, since P&, appeared to reduce
tion. The wide range of grading classifications is with reducing relative density, and at an increas-
made evident by the d,, and d,, sizes which have ing rate at very small relative densities. A relative
been quoted; estimates have been made in certain dilatancy index of the form
cases where these particular values were not given I, = 1,(Q - In p’) - R (12)
 STRENGTH AND DILATANCY OF SANDS 71

was therefore studied, since this function returns a Equation

zero value at some pressure pkri,such that (15)

D
In p:,,, = Q - p (13)
D / A

(Here, the value for Q will depend on the units

taken for p’: kilonewtons per metre squared will
be used here). This expression has the advantage
of ensuring that zero dilatancy is achieved at a
critical effective stress which itself reduces strong-
ly when the relative density I, takes small values.
Beyond these considerations, however, any pref-
erence for a particular definition of a relative
dilatancy index must be entirely empirical.
It was found that values Q = 10 and R = 1
created a definition for a relative dilatancy index

I, = I,(10 - In p’) - 1 (14)

which apparently offered a unique set of correla-
tions for the dilatancy-related behaviour of each
of the sands in laboratory element tests. The fol-
lowing correlations were found to be available in
the range 0 < I, < 4. For plane strain

&II,, - 4bri, = O.S$,,, = 51,” (15)

For triaxial strain

&n,X - &, = 31,” (16)

For both test configurations 0 Plane dram
Data for p’ i= 300 kN/m’
0 Triax~al

= 0.31, (17) Fig. 7. Data for sands in Table 1 at failure with p’

in the range 150400 kN/m*
It should be recognized that equations (15) and
(17) permit alternative predictions of the rate of
dilatancy in plane strain, although manipulation of 300 kN/m’. To trap a sufficient variety of data
of equation (2) would reveal that the difference is it was necessary to accept data falling within a
numerically insignificant in the range of interest. factor of 2 of this stress: this has been plotted
The justification for these relations now without modification. However, some data from
follows, consisting of the test data of the sands the longer test series have been omitted for parity,
listed in Table 1 presented to display where they simply confirm the trend and scatter
of results already plotted from the same series.
(a) the variation in the peak angle of shearing
The pattern of plane strain strengths lying
resistance, and the rate of dilatancy, with rela-
above triaxial strengths has long been recognized
tive density at a particular stress level
and was noted by many of the researchers listed
(b) the variation in peak angle of shearing resis-
in Table 1. The apparently linear correlation of
tance with stress level, at particular relative
particular sands on these axes has also been
densities
noted (Bishop, 1972) but the superimposition of
(c) the combined variation for particular sands in
data of such a variety of sands taking advantage
those few studies where both density and
of +Lrit as a baseline was encouraging. The
stress level effects were studied
uniqueness of data of dilatancy rate irrespective
(d) the comparison in behaviour between plane
of test mode has also been widely reported
strain and triaxial test behaviour.
(Cornforth, 1973).
Figure 7 displays the available data of both The empirical relations (14H17) appear as con-
&?I,, - $brit and dilatancy rate ( -de,/daJmaX verging straight lines in Fig. 7, offering the
versus initial relative density, at what is intended concept of zero dilation in soils with an initial
to be a constant mean effective stress p’ at failure relative density of 0.23 at this effective stress level.
 72 BOLTON

zo- p’: kN/m’ Data

Empwxil relatmn (11)
20 0
/ -I-
16-

la

0 4 8 12 16 20 24
!4,,,,. dw
(a)
Fig. 8. Angles of shearing resistance and dilataocy in
plane strain

16
It will be seen that the lines are not apparently of
optimal fit since particular weightings have been
attached to the data. Where a sequence of com-
pression tests has been carried out at various den-
sities, for example on soil A, it is usual for the
experimenter to adopt a constant value of the cell
pressure providing us’. This leads to’enhanced
values of the mean effective stress at higher rela-
tive densities, due to the attendant increase in
crl’/uj’. It is consistent with the later data of stress
dependence to suppose that data of &,,, - #brit I A ‘D
in the vicinity’ of I, = 1 could be relatively (b)
reduced by 2” on this account, with a correspond- Fig. 9. (a) Triaxial test data for Berlin sand (K) (De
ing shortfall in dilatancy rate. Beer, 1965) and (b) triaxial test data for Mol sand (J)
Furthermore, the empirical relations have been (Ladanyi, 1960)
selected taking into account the whole spread of
data depicted in Figs 7-11, the required linkages
between parameters suggested by stress-dilatancy Figure 9(a) traces the triaxial data of Berlin
theory and the desire to produce the most simple sand and Fig. 9(b) traces Ladanyi’s triaxial data
expressions in equations (14H17) which could (Ladanyi, 1960) for Mol sand, which formed part
offer a useful correlation with the data. Statistical of De Beer’s remarkable study (De Beer, 1965) of
optimization has therefore not been attempted. the separate effects of relative density and mean
Nevertheless, the maximum departure of the data effective stress. It will be found that the proposed
is 2”, and more than 80% of the data lies within empirical relations fit the totality of data at
1” of the correlation. The zone of greatest depar- various densities and stresses within a typical
ture is at low relative density, where there is evi- margin of f 1” and within a maximum margin of
dence of greater rates of dilation in plane strain &2”.
than in triaxial strain, with correspondingly Figure 10 records triaxial data of increasing
higher angles of shearing. For simplicity, this par- effective stress at failure of sands with initial rela-
ticular phenomenon was conservatively ignored. tive densities either in the vicinity (kO.05) of 0.8
Figure 8 compares the first part of equation or of 0.5. The chosen axes are similar to those
(15), which has been shown in Fig. 6 to be almost used by Vesic & Clough (1968) in their classic
indistinguishable in operation from Rowe’s report of high pressure test results, except that
stressdilatancy theory, with the data from plane advantage has again been taken of the use of
strain tests. The closeness of fit is excellent, the ~L,x - 4brit as the dilatancy-related strength
widest departure being 2” in a &,,,, of 47” on parameter. It will be seen that the data of
dense Leighton Buzzard sand, G, which was the Kl,, - &, in Fig. 10 fall well along the appro-
only soil tested in the simple shear apparatus. A priate I, = constant lines produced using equa-
departure of less than 1” was more typical. tions (14) and (16), with a typical departure of less
 STRENGTH AND DILATANCY OF SANDS 73

16- \
\
\
\
14% \
\
\ \

Empmcal relation (16)

P’ kN/m’

Fig. 10. Triaxial test data for sands in Table 1 failing at various mean effec-
tive stresses

than 2”. The logarithmic axis carries p’ beyond indicate that significant errors may arise in the
lo4 kN/m* into the region where k2” scatter in prediction of pLri,due to the logarithmic scale.
&II,, was commonly observed around the best There is some evidence in the literature that
estimate for c#&, certain uniform, rounded sands are little affected
The best interpretation of I, values which are by confining pressures less than about 1000
calculated to be negative, owing either to extreme kN/m’, after which they begin to crush relatively
looseness or to high confining stress is that a con- swiftly. This is shown in Fig. 11(a) where the tri-
siderable contraction in volume (and consequent axial data of Lee & Seed (1967) for subangular
increase in I,) will take place before &, is mobi- Sacramento river sand follow the empirical rela-
lized at I, = 0 after large shear strains. Expres- tion (16) in contrast with Ottowa sand which was
sion (14) gives an indication of the value of p’ found not to crush significantly up to pressures of
(P&it 3 say) at which I, = 0 and dilation is sup- 4000 kN/m’ and whose data lie up to 5” above
pressed for any initial relative density: In pLri,= the empirical prediction in the medium-to-high
10 - l/I,. However, a perusal of Fig. 10 will stress range. Fig. 11(b) for Karlsruhe sand

Empfrlcal relation (16)

ywth iD = 1

0 OO
L onn 0
1000 oloooo" 100 200 400 800 1600

P’ kN/m’
p’ kN/m2

(a) (b)
Fig. 11. (a) Triaxial test data for two dense sands at elevated stresses (Lee & Seed, 1967) and (b) triaxial test data for
dense Karlsrube sand (P) at various stresses (Hettler, 1981)
 BOLTON

p” kN/m2 0

Fig. 12. Plane strain data for sands in Table 1 failing at various
mean effective stresses

(Hettler, 1981) demonstrates a similar tendency in ticle diameter. This suggests that the sand was in
a uniform rounded sand noted for its particle a somewhat unstable state within its test
strength (Drescher & Vardoulakis, 1982). chamber, and may correspond to the remarks
Figure 12 introduces the rather sparse available made earlier about the sudden degradation of
information of plane strain tests conducted at some uniform sands which had displayed rela-
various stresses. Results in the middle band of tively little progressive crushing. Whether or not
stresses were broadly as expected, but the low the extra mid-range strength was due to this, or
stress test on sand G in the simple shear appar- to some other cause, it might be prudent to disre-
atus and one of the high stress tests on sand H in gard it for design purposes.
a plane compression apparatus deviated quite It should be recalled that the bulk of data so
markedly in opposite senses. Both tests must have far presented derive from compression tests with
been difficult to carry out satisfactorily, for con- H/D = 2 and some undetermined platen friction.
trasting reasons. Previous studies are ambiguous with regard to
For the low stress shear test, the effects of any the effect of reducing friction on samples with
stress and strain non-uniformities could have H/D = 2. Bishop & Green (1965), in an exhaus-
been strong. The sand was exceptionally brittle tive study of Ham sand, conclude that the effect is
and therefore subject to progressive rupture as 4’ negligible. Rowe & Barden (1964) found that,
could drop from at least 52” to 35”. Furthermore, across a full range of densities for Mersey sand,
the accurate measurement of stresses as small as f&n,, measured with lubricated platens on samples
15 kN/m’ presents severe difficulties. If the result with H/D = 1 were about 2.5” less than values
can be relied on, however, it might indicate an measured with conventional platens on samples
absolute limit to the rate of dilation of a dense with HID = 2. Tests conducted on both dense
sand irrespective of how small the stresses are. and loose Karlsruhe sand with lubricated platens
Until good low stress data are available for a on samples with H/D = l/3 indicated (Drescher
particular sand, it would be prudent to set a limit & Vardoulakis, 1982) that &,,,, was about 3” less
of 4 to the magnitude of the relative dilatancy than in the conventional test. Since the majority
index I,, irrespective of larger values calculated of engineers are used to working with test data
by equation (14). This would have the effect of derived in the conventional way these discrep-
limiting &,, - c&, to 20” in plane strain and 12” ancies are, perhaps, of interest principally to the
in triaxial strain, values which have been attained research worker.
but not exceeded in the test programmes recalled No attempt has been made here to extend the
here. discussion to tests in triaxial extension, or to tests
For the plane compression test on soil H at with arbitrary intermediate stress ratios. These
6000 kN/m2, the unexpectedly high rate of dilat- are habitually found to offer strengths that are
ancy at failure was matched by an even swifter intermediate to triaxial compression and plane
rate of softening to a critical state: the required strain values, the differences between which have
boundary displacement was about one half a par- been shown here to relate solely to dilatancy
 STRENGTH AND DILATANCY OF SANDS

effects in the soil. It may be sufficient to note that Table 2

the soil grains in a triaxial test have considerably
greater freedom to deviate laterally than those in Q Grain type
a plane strain test, and that in these circum-
10 Quartz and feldspar
stances the enhancement of strength due to dilat- 8 Limestone
ancy might have been expected to be smaller. I Anthracite
5.5 Chalk
IMPLICATIONS FOR LABORATORY F
SOIL TESTING
The concept of a relative dilatancy index I, of initially dense samples of granulated chalk,
which is coupled with rates of dilatancy, and crushed anthracite and limestone sand indicate
related to the extra angle of shearing above &,, that they each gain roughly 7” of shearing resist-
has proved powerful in organizing the data of 17 ance per tenfold stress reduction, according preci-
contrasting sands from around the world tested sely with the gradient of the I, = 1 line in Fig. 10.
independently by 12 groups of workers. The par- What was shown, however, was that reducing the
ticular numerical values quoted in expressions crushing strength of the grains reduced the criti-
(14H17) have been capable of retrospectively cal mean effective stress at which dilatancy was
fitting the conventional data of shearing resist- suppressed. This implies that parameter Q in
ance in the fO&lOOO kN/m2 range of mean effec- equation (12) should be reduced for soils of
tive stress with a typical departure of 1” or 2” weaker grains. Billam’s data, assuming that they
once a good value for $bri, has been obtained. relate to ‘soils’ with I, = 1, is roughly consistent
The data themselves make clear that a further 1” with the values in Table 2.
or 2” error in &,X might be incurred if &, were
estimated from the shearing strength of a single
very loose sample. Cornforth (1973) has reported IMPLICATIONS FOR LABORATORY
that the angle of repose of a loosely tipped heap REPORTS
of dry material subjected to excavation of the toe A minimal laboratory report could comprise
provided a simple bench test for I#&, which
(a) source of material and assessment of mineral-
should similarly offer an accuracy of about 1”.
The likely error in simply taking &, = 33” for a ogy
(b) grading curve
sand which has been identified as comprising
chiefly quartz grains is apparently also only of the (c) emax and emin by a standard method
(d) angle of repose of dry heap
order of 1” or 2”.
(e) in situ density.
Although the limiting behaviour of sands in the
range 103-lo4 kN/mZ has also been fitted, the Depending on the significance of the problem in
uncertainties are greater. Samples deform and hand, the following additional triaxial compres-
contract substantially under test, and the rele- sion tests taken in order would confirm the rele-
vance of initial relative density must be ques- vance or need for recalibration of the empirical
tioned, as must the relevance of the strength relations proposed above
parameter itself since the shear strains required to
(f) loose soil at a cell pressure circa 500 kN/m*
mobilize &, are large (Vesic & Clough, 1968).
(g) dense soil at a cell pressure circa 150 kN/m2
The analogy with the engineering problems posed
(h) dense soil at the highest available cell pressure
by soft clays in the small stress range is appealing.
(i) other relevant tests.
This suggests that further study of the behaviour
of sands under extreme pressures in terms of their It should then be possible, with a degree of con-
instantaneous (e, 4’, p’) states is necessary if more fidence that is proportional to the number of tests
accurate and complete predictions are required. conducted, to predict the peak strength of any
The present correlations underestimate the element of the soil whose density and stress is
strengths of some uniform well-rounded sands in known.
this stress range: these materials do no’t suffer An alternative consistent approach would be to
progressive crushing but tend to degrade rather test a single soil sample from each zone at a
suddenly at some very large stress. density that is lower than any which could rea-
The sands presented in this study are mainly of sonably be anticipated in the field and at a con-
quartz or feldspar. The presence of substantial fining pressure that is larger than any which
proportions of mica, calcite or other materials could be experienced. The resulting secant value
would be bound to affect both &, and the crush- of KU, could then be used conservatively in
ing which reduces I, at high stresses. Billam’s design for the whole zone, provided that pro-
(1972) triaxial data for the dilatancy and strength gressive failure could not occur. If there were sig-
 76 BOLTON

confusion generated by describing any such

tangent angle as the ‘angle of internal friction’
must be deplored as vigorously as the description
of the intercept c’ as the ‘true cohesion’. The
detailed studies brought together here tell an
unambiguous story of friction and dilation
related to secant 4’ values and $ values deter-
mined separately for each soil test.

IMPLICATIONS FOR FIELD TESTING

Although the foregoing has been concerned
with the collation of laboratory test data, it has a
’ = 33"
strong implication for the interpretation of field
Cl,,
tests. The new correlations make clear that there
0 12 3 4 5 6 7 should not be a one-to-one correspondence
between the strength of granular soils and their
relative density, even at a given stress level. A
500-
particular angle of shearing resistance (or pen-
etrometer reading) could indicate, for example,
either a dense quartz sand with low &, and high
* max values or a less dense felspathic sand with
higher &, and low $,,, values. The potential
confusion between these alternatives might be
most significant in a prediction of earthquake-
induced excess porewater pressures. The relative
density has been shown to correlate only with the
dilatant strength component, which can only be
0 100 200 300 400 500 600 700 resolved from the peak strength (or penetrometer
o' kN/m’ reading) if &, is separately determined using a
(b) disturbed sample, for example.
Fig. 13. Mohr envelopes derived from empirical relations It may therefore be considered preferable to
for triaxial tests on quartz sands measure the angle of dilation $,,, directly, using
a self-boring pressuremeter after the fashion of
Hughes, Wroth & Windle (1977). The direct cor-
nificant doubts about the loosest state in the field relation in equation (15) between $,,, and I, at a
or the possibility of strain softening, the use of given stress level should then lead to a more reli-
+Lrit would be necessary. In this case, the testing able estimate of relative density than would
programme could simply consist of items (d) and otherwise be available. The attempt to relate
(f). cyclic pore pressure generation directly to the
If required, the ‘intrinsic curve’ on a T-CT’ measured angle of dilation of the soil (Vaid,
diagram can be generated using equation (14) Byrne & Hughes, 1981) should prove equally
with either equation (15) or (16) depending on fruitful in studies of liquefaction susceptibility.
whether a plane strain or triaxial envelope is The empirical relations (14H17) provide a con-
required. It may be assumed, with sufficient accu- sistent treatment of stress level effects by which
racy, that the point of contact of a Mohr circle the data of all penetration tests can be normal-
with the intrinsic curve is (T’= p’, 7 = p’ tan &,,,, ized. Previous treatments of the problem of
Depending on the required scale, curves of the curved strength envelopes (Baligh, 1976) may
form of Fig. 13(a) or Fig. 13(b) can then be pro- thereby be recast as problems of &,, varying
duced for any given I,. It is then easy to confirm with stress (De Beer, 1970), and appropriate
that the conventional tangent parameters (c’, 4’) methods of calculation developed. The prospects
cannot be fitted without serious misrepresenta- for a consistent approach to reductions in N,
tion. At least three pairs of such values, corre- with depth or N, with foundation width are then
sponding to low stress, medium stress and high evident. The first objective should be to use the
stress ranges, would be necessary to capture the new relations to predict idealized cone penetrat-
data of sand at each relative density. The fitting of ion resistance profiles which could be compared
common tangents to Mohr circles must therefore with those obtained in practice (Mitchell &
be regarded as a quite inferior method of presen- Lunne, 1978), so that laboratory and field data
tation of soil strength results. Furthermore, the can be harmonized.
 STRENGTH AND DILATANCY OF SANDS 71

CONCLUSIONS Cornforth, D. H. (1973). Prediction of drained strength

It has been demonstrated that secant angles of of sands from relative density measurements. In
shearing are required in a rational approach to Evaluation of relative density and its role in geotech-
the strength and dilatancy of sands. &,,., - &, nical projects inuoluing cohesionless soils. ASTM
Spec. Tech. Publ. 523, pp. 281-303, American
has been shown to be a useful measure of the
Society for Testing and Materials, Philadelphia.
extra component of strength due to dilatancy in a De Beer, E. E. (1965). Influence of the mean normal
dense soil, and an expression has been derived stress on the shearing strength of sand: Proc. 6th Int.
which is operationally indistinguishable from Conf Soil Mech. Fdn Engng, Montreal 1, 165-169.
Rowe’s well-established theory. De Beer, E. E. (1970). Experimental determination of
A new relative dilatancy index I, has been the shape factors and the bearing capacity factors of
defined in terms of relative density and effective sand. Gt!otechnique 20, No. 4, 387411.
stress level. Extremely simple correlations De Josselin de Jong, G. (1976). Rowe’s stress+lilatancy
relation based on friction. GPotechnique 26, No. 3,
between 1, and K,,,, - 4Lp $,,, and WWmax
527-534.
have been supported using the data of 17 soils
Drescher, A. & Vardoulakis, I. (1982). Geometric soften-
available in the literature. A corollary is that in triaxial tests on granular material.
ing
there is not a one-to-one correspondence between GCotechnique 32, No. 4, 291-303.
&II,, and relative density, even at a given stress Hettler, A. (1981). Verschiebungen Starrer und elastis-
level, since the dilatancy-related component of cher Grlndungsk6rper in Sand bei monotoner und
strength is &,, - &, and &, is a function of zyklischer Belastung. veriifl Inst. Bodenmech. Fels-
mineralogy. Typical values for dbrit are 33” for a mech. No. 90, l-127.
quartz sand and 40” for a felspathic sand, though Hirschlield, R. C. & Poulos, S. J. (1964). High-pressure
simple laboratory tests should be used as a triaxial tests on a compacted sand and an undis-
turbed silt. In Laboratory shear testing of soils.
routine measure. It would be prudent not to
ASTM Spec. Tech. Publ. 361, pp. 329-339, Amer-
invoke I, values in excess of 4 without test cor-
ican Society for Testing and Materials, Philadelphia.
roboration of the extreme dilatant strength which Hughes, J. M. O., Wroth, C. P. & Windle, D. (1977).
would be implied. All &_ values in excess of 4bri, Pressuremeter tests in sands. Ghotechnique 27, No. 4,
are transitory with larger strains and subject in 455477.
design calculations to an assurance that progress- Koerner, R. M. (1970). Behaviour of single mineral soils
ive failure will not occur. Negative values of I, in triaxial shear. J. Soil Mech. Fdns Div. Am. Sot.
indicate that large contractile strains will occur Civ. Engrs 96, SM4, 1373-1390.
before $Lri, can be mobilized. Ladanyi, B. (1960). Etude des relations entre les con-
traintes et des d&formations lors du cisaillement des
The possibility now exists of rationalizing the
sols pulvtrulents. Annls Trau. Publ. Belg. No. 3,
various ad hoc stress level corrections which have
l-30.
emerged in penetration testing and pile design. Lade, P. V. & Duncan, J. M. (1973). Cubical triaxial
Clarifications of the relative proportions of fric- tests on cohesionless soil. J. Soil Mech. Fdns Div.
tion and dilatant strength components are likely Am. Sot. Ciu. Engrs 99, SMlO, 793-811.
to be particularly important in investigations of Lee, I. K. (1966). Stress-dilatancy performance of feld-
the liquefaction potential. spar. J. Soil Mech. Fdns Div. Am. Sot. Ciu. Engrs 92,
SM2,79%103.
REFERENCES Lee, K. L. & Seed, H. B. (1967). Drained strength char-
Baligh, M. M. (1976). Cavity expansion in sands with acteristics of sands. J. Soil Mech. Fdns Div. Am. Sot.
curved envelopes. J. Geotech. Engng Div. Am. Sot. Ciu. Engrs 93, SM6, 117-141.
Cio. Engrs 102, CT1 1, 1131-l 146. Marachi, N. D., Chan, C. K., Seed, H. B. & Duncan,
Barden, L., Ismail, H. & Tong, P. (1969). Plane strain J. M. (1969). Strength and deformation characteristics
deformation of granular material at low and high of rockJill materials. University of California Report
pressures. Gkotechnique 19, No. 4,441452. TE-69-5, Berkeley.
Billam, J. (1972). Some aspects of the behaviour of Mitchell, J. K. & Lunne, T. A. (1978). Cone resistance as
granular materials at high pressures. In Stress- measure of sand strength. J. Geotech. Engng Div.
strain behaviour of soils. DD. 69-80 (ed. R. H. G. Am. Sot. Ciu. Engrs 104, GT7,995-1012.
Parry). London: Foulis. ._ Roscoe, K. H. (1970). The influence of strains in soil
Bishop, A. W. (1972). Shear strength parameters for mechanics. Giotechnique 20, No. 2, 129-170.
undisturbed and remoulded soils specimens. In Rowe, P. W. (1962). The stress+lilatancy relation for
Stress-strain behaviour of soils, pp. 3-58 (ed. static equilibrium of an assembly of particles in
R. H. G. Parry). London: Foulis. contact. Proc. R. Sot. 269A, 5OG-527.
Bishop. A. W. & Green, G. E. (1965). The influence of Rowe, P. W. (1969). The relation between the shear
end restraint on the compression strength of a cohe- strength of sands in triaxial compression, plane
sionless soil. GPotechniaue 15, No. 3,243-266. strain and direct shear. GPotechnique 19, No. 1,
Cornforth, D. H. (1964). Some experiments on the influ- 75-86.
ence of strain conditions on the strength of sand. Rowe, P. W. (1972). Theoretical meaning and observed
Gkotechnique 14, No. 2, 143-167. values of deformation parameters for soil. In Stress-
78 BOLTON

strain behaviour of soils, pp. 143-194 (ed. R. H. G. S.S.A. PhD thesis, Cambridge University.
Parrv). London: Foulis. Tavenas, F. 8~ La Rochelle, P. (1972). Accuracy of rela-
Rowe, 6.’ W. & Barden, L. (1964). Importance of free tive density measurements. &ot&hnique 22, No. 4,
ends in triaxial testing. J. Soil Me& Fdns Div. Am. 549-562.
Sot. Cio. Engrs 90, SMl, l-27. Vaid, Y. P., Byrne, P. M. & Hughes, M. 0. (1981).
Rowe, P. W., Oates, D. B. & Skermer, N. A. (1963). The Dilation angle and liquefaction potential. J. Geotech.
stress-dilatancy performance of two clays. In Labor- Engng Div. Am. Sot. Civ. Engrs 107, GT7, 1003-
atory shear testing of soils. ASTM Spec. Tech. Publ. 1008.
361, pp. 134143, American Society for Testing and Vesic, A. S. & Clough, G. W. (1968). Behaviour of
Materials, Philadelphia. granular materials under high stresses. J. Soil Mech.
Stroud, M. A. (1971). Sand at low stress levels in the Fdns Div. Am. Sot. Civ. Engrs 94, SM3,661-688.
 5(

Average for 6 = 90”

4: S-

_ower bound for

$ Lower’?
xl
;; 4( I- bound

E
s
oj’: kN/m2 ‘\ us’: kN/m*
\ -sA
98 392
4.9 9.8 49 _ ‘\+e
cl -0 o- 90 ‘A 0 b 90
. 67 G, . 67
3: j- ’ B
0 D 45
6: deg
+ -t + 34 > 6: deg
V 623
0 Q w 23
.
A 11
A a’o
A -A PO J
3( I
6 07 0.8 0.7 0.8
e0 e0
(a) lb)

Fig. 2. Angle of shearing resistance @ma.versus initial voids ratio es measured at ur’ = O-OSkgf/cm’
(4-9 kN/m’) in plane strain for Toyoura saod: (a) IQ’ values of 49 kN/m’, 98 kN/m’ and
49-O kN/m*; (b) aj’ values of 98 kN/m’ and 392 kN/m’ (Tatsuoka, Sakamoto, Kawamura &
Fukusbima, 1986)

,”. e. \ 0
\ VU 8
Average for Lower A

1
6 = 90”

Lowe,‘\, &A
bound \+
v
+
\
\ T*
4
Us’: kN/m2 \
\ c-
4.9 9.8 49 0-P \
\
0 -cJ - 90 \+“+p”
67
. 1%
\
0 D 45 I \ 8

. 11

Ad A-0 J

‘0.6 0.7 0.8 I o-7 0.8

e0 e0

(a) (b)

Fig. 3. Angle of dilatancy at peak stress condition $,., versus e, in plane strain for Toyoura sand:
(a) 03’ values of 4.9 kN/m’, 98 kN/m2 and 490 kN/m’; (b) uj’ values of 98 kN/m’ and
392 kN/m’ (Tatsuoka, Sakamoto, Kawamura & Fukushima, 1986)

220
 DISCUSSION 221

Table 2. Test conditions for Toyoura sand

Test condition Sample dimensions: End condition Isotropic confining

cm pressure: kN/m’

Plane strain compression Height = 10.5, Well lubricated* 4.9, 9.8, 49, 98, 392 Corrected for
(Tatsuoka, Sakamoto width = 4, membrane forces
Kawamura & length (a; and others
Fukushima, 1986) direction) = 8 (method C-2-T)
Triaxial compression Height = 15, Well lubricated 2.0, 4.9, 9.8, 19.6, Corrected for
(Fukushima 8~ diameter = 7 49, 98, 196, 392t membrane forces
Tatsuoka, 1984) (solid cylinder) defined at
midheight of
sample
Triaxial compression Height = 7.8, Well lubricated 98 Corrected for
(Lam & Tatsuoka, width = 7.8 membrane forces
1986) (prismatic, square
cross-section)

* Refer to Tatsuoka, Molenkamp, Torii & Hino (1984).

t us’ at failure was slightly different.

Fig. 4. Definition of angle 6

5c

392
b 90

b ::
45 4 34
6: deg
+ 23
$ 11
7J
x ho
E
8

4c

35
15 20
hax: deg
Fig. 5. Relationship between Wm., and Jr,,, in plane strain for Toyoura sand
222 DISCUSSION

B P
p40’- TslO- e o- - 0.80
7---x----__-x_
I --X-
< -: --
--x.

I I
3OL Ol 5 10 50 100 500
us’: kNjm2

Fig. 6. Reletionsbips between I$&, &,.. and log u3’ at 6 = !W’in PlmJe SWin
compression for Toyoura saod

densities and for the entire range of angles S (i.e. The reason for this definition is as follows. It can
(X90’). A weak dependence of &_ and I(lmaXon easily be shown that a comparison of d&,/de, for
stress level at a stress range of c3’ 6 50 kN/m’ triaxial compression and plane strain is equiva-
for air-pluviated Toyoura sand has also been lent to a comparison of $,,,., when $,,, is given
obtained in both triaxial compression tests by equation (1) for triaxial compression and
(Fukushima & Tatsuoka, 1984) (also Fig. 7) and $,., = arcsin [--(d&i + de,)/(de, - de3)lmnX for
torsional shear tests (Tatsuoka, Sonoda, Hara, plane strain. It is well known that the stress-
Fukushima & Pradhan, 1986). The dilatancy dilatancy relations proposed by Rowe (1969)
angle *,,X shown in Figs 7(b) and 7(c) will be dis-
cussed later. The data for dense Leighton
Buzzard sand at low stress levels shown in fig. 12 (24
of the Paper also show the tendency that the rate
of change in both &,, and I+G,,,.~with In p’ for triaxial compression and
becomes very small at very low stress levels.
Therefore, as suggested in the Paper, it may be
quite misleading without a basis of very reliable
experimental results to assume that both &,,
and Ijl,,, change in proportion to -In p‘in low for plane strain fit most experimental data,
stress ranges. whereas the value of K for plane strain is known
Finally, the data shown in fig. 7 of the Paper to be slightly larger than that for triaxial com-
indicate that for the same density &,,, is larger in pression. Thus it can easily be shown that a com-
plane strain than in triaxial compression, whereas parison of dilatancy characteristics between
for the same density (ds,/dQ,,aX is similar in both triaxial compression and plane strain in terms of
testing methods. However, the data for air- K is equivalent to a comparison in terms of the
pluviated Toyoura sand, which were obtained at ratio K&,., . It can be seen in Fig. 8(b) that
the Institute of Industrial Science, the University * mBxis larger in plane strain than in triaxial com-
of Tokyo, show that the relationships for ~$6,~ pression at 6 = 90”. This point can also be seen
and (d&,/de 1h,,,, between plane strain and triaxial by comparing Figs 3(a) and 3(b) with Fig. 7(b).
compression are not as simple as suggested in the However, the difference decreases to a minimum
Paper owing to the anisotropic mechanical at 6 x 30” as 6 decreases as is the case for &,,.,
properties of the sand. It can be seen in Fig. 8(a) Therefore it is an oversimplification to assume
that &,,a, is similar for plane strain and triaxial that for the same density the value (ds,/dsl)maX is
compression at an angle 6 of about 30” where similar for plane strain and triaxial compression.
&,,,, becomes the minimum in plane strain. Furthermore, it can be seen by comparing Fig. 5
In Figs 7(b), 7(c) and 8(b) the dilatancy angle with Fig. 7(c) that to compare the value of K in
+,,, for triaxial compression is defined as equation (2) for triaxial compression and plane
strain the anisotropy in strength and deformation
characteristics should be taken into account.
- (de,,2 _ d )
-53 1 max
(1) In summary, the empirical relations for #,,,,
 DISCUSSION
223

45-
15-

40-
lo-
F
D
$
x D
E us’: kN/m2 L
8
y”
I 0 2.6-5 8
35 A 6 O-7 1
t V 10-12 5-
0 20-22
0 51
99
: i97 = 393 kN/m”’

I*
I
3o06 0.7 0.8 0.9 0’
0.6 0.7 O-8
e0 0.9
ia e0

(b)

45-

Average for plane

stram, 6 = 90”and
us’ = 4.9-392 kN/mz

40-

P
.o

I
s’ 0 X35.8
A 6G7.1
35- V 1 O-20

0 20-22

x 393
30
5 15 $0

Fig. 7. (a) t#mu; (b) JI,. ; (cl the dhmbip hmea Ok_ ad +_ at 6 = 90” in triaxial cempressioa (Fmkdma &
Tatsuoka, 198A)
224 DISCUSSION

us’ = 98 kN/mz

(4

0.4
e0 Qmax’@ = 90”, PS): deg ‘&,& = go”, PS): deg
0.70 46.5 17.0
0.2 0.80 41 .o 8.5

0.01 1 I I A
0 1o 20 30 40. 50 60 70 80 90
6: deg
(b)

Fig. 8. K,.. and Jr,., as a function of angle 6 at u3’ = 98 kN/m* in plane

strain (PS) and triaxial compression (TC) (Lam & Tatsuoka, 1986)

and *,, proposed in the Paper can be put to Lam, W. K. & Tatsuoka, F. (1986). The strength surface
more general use by taking into account both of sand. Proc. Japan Symp. Soil Mech. Fdn Engng 1,
315-318.
(a) the anisotropy in &,., and ICI,,,,, Oda, M. (1981). Anisotropic strength of cohesionless
(b) the low stress level dependence of &,,,, and sands. J. Geotech. Engng Div. Am. Sot. Ciu. Engrs
$ max 107, GT9, 1219-1231.
Rowe, P. W. (1969). The relation between the shear
at very low stress levels at least for such clean
strength of sands in triaxial compression, plane
sands as Toyoura sand.
strain and direct shear. Gotechnique 19, No. 1,
75-86.
REFERENCES Tatsuoka, F., Molenkamp, F., Torii, T. & Hino, T.
Arthur, J. R. F. & Assadi, A. (1977). Ruptured sand (1984). Behavior of lubrication layers of platens in
sheared in plane strain. Proc. 9th Int. Con& Soil element tests. Soils Fdns 24, No. 1, 113-128.
Mech. Fdn Engng, Tokyo 1, 19-22. Tatsuoka, F., Sakamoto, M., Kawamura, T. & Fuku-
Fukushima, S. & Tatsuoka, F. (1984). Strength and shima, S. (1986). Strength and deformation charac-
deformation characteristics of saturated sand at teristics of sand in plane strain compression at
extremely low pressures. Soils Fdns 24, No. 4, 30-48. extremely low pressures. Soils Fdns 26, No. 1,65-84.
Goto, S. (1984). Strength and deformation characteristics Tatsuoka, F., Sonoda, S., Hara, K., Fukushima, S. &
of sand in small and large triaxial apparatuses. MEng Pradhan, T. B. S. (1986). Failure and deformation of
thesis, University of Tokyo (in Japanese). sand in torsional shear. Soils Fdns 26, No. 4.
 DISCUSSION 225
Yoshimi, Y., Hatanaka, M. & Oh-oka, H. (1978). Undis- that the further reduction of cj from 50 kN/m2 to
turbed sampling of saturated sand by freezing. Soils 3 kN/m* (p’ from about 100 kN/m2 to 6 kN/m’)
Fdns 18, No. 3,59-73. does nothing to compensate for a low initial
density. Indeed &,,, for p’= 3 kN/m2 is about
1.5” smaller than that for p’= 6 kN/m2, which is
Author’s reply paradoxical. A prudent conclusion would be, in
The Writer has introduced an extensive and the absence of further verification for a particular
valuable body of recent data for Toyoura sand. sand, that the Paper’s empirical correlation using
For comparison with the empirical relations I, in equation (14) should be limited to the range
introduced in the Paper, it is necessary to select a of p’> 150 kN/m’. Any smaller value of p’
value of 4Eli,. This can, perhaps, be achieved by should be substituted by 150 kN/m’. The original
extrapolating Figs 5 and 7(c) of &,,,, against $,,,., relative dilatancy index could then be rewritten
to find &,,., at ICI,,, = 0. A range of values can be
determined from about 35” with the bedding at
6 = 90” to 33” with 6 = 30” in plane strain and In=,,[5-ln(&)]-l
about 32” in triaxial strain. In each case a scatter
of &-1” is discernible, which is typical. for p’> 150 kN/m’ and
If Rowe’s stress dilatancy theory is fitted
approximately, with lines of slope 0.8 following I, = 51, - 1
the Paper’s fig. 6, it can then be seen that dcri, is
greater for p’< 150 kN/m2. For dense soils (In = 1) this
is an identical condition to the limit I, = 4 sug-
(a) in plane strain than in triaxial strain
gested in the Paper: however, it is a stronger con-
(b) at smaller confining pressure
dition in the case of looser soils.
(c) in looser packings
Perhaps the most striking aspect of the Writer’s
(d) when compressed perpendicularly to the
data is his demonstration of the effects of inherent
bedding.
anisotropy. Strength and dilatancy in plane strain
This accords with the view that strength is reduce from a maximum for normal bedding,
enhanced somewhat by local dilation even in a 6 = 90”, to a minimum when 6 = 30”. At this
sample which is shearing at overall constant inclination of bedding, the plane strain data
volume. Shear stresses will tend to be transmitted (6 = 30”) had fallen as low as triaxial data for
through the stronger particle assemblies which strength and dilatancy on normal samples
are locally dilating rather than the weaker (6 = 90”). This interesting observation should be
assemblies which are contracting. This explana- coupled with the deterioration in peak strength
tion also satisfies the observation of Norris (1977) parameters following the induction of anisotropy
that, for a sand of quartz grains at moderate due to principal stress reversal or rotation (Wong
stresses (p’ u” 100 kN/m’), 4crit increases with par- & Arthur, 1985). Further work on the strength
ticle angularity from 29” when rounded to 40” anisotropy of other sands, with particles of
when angular, the latter value reducing at higher varying sphericity and angularity, would be
stresses. advantageous.
The 3” variation in 4cri, due to inherent aniso- Clearly, the peak strength depends on the rate
tropy is rather larger than might hitherto have of dilatancy, which in turn depends on the
been supposed. This effect adds to the problem of geometry of particle movements in relation to the
selecting a value for dcri,, but this uncertainty principal stress direction. The particle trajectories
need not be overemphasized. A value of 34” is apparently depend on the mode of strain in three
consistent with Norris’s observation and fits all dimensions, the inherent anisotropy due to
the Writer’s data to + 2”. bedding and the anisotropy induced by strain
Following this selection, the Paper’s correla- history, even for samples at a given voids ratio
tions fit the equivalent data of triaxial and plane confined under a given mean effective stress. Nor
strain compression perpendicular to the bedding should the possible effects of progressive failure in
(b = 90°) approximately as well as the average the soil mass be forgotten when applying soil
sand in the original survey. However, the Writer data to the design of geotechnical works.
properly points out that his data at small confin- The Writer has provided a valuable indication
ing pressures lead to the deduction that there is of the detailed behaviour of a sand at low stress
no appreciable further effect on soil behaviour levels, taking anisotropy into account-facets
when the mean stress p‘ is reduced below about which were omitted from the Paper. Designers
150 kN/m2. The tendency to crush must be cannot rely uncritically on strengths in sands
almost completely eliminated at these stress that are in excess of those measurable in conven-
levels. It may be seen in Fig. 7(a), for example, tional triaxial tests on normally bedded samples.
226 DISCUSSION

Indeed the uncertainties in density, bedding REFERENCES

orientation, strain localization, stress distribution Norris, G. M. (1977). The drained shear strength of
and strain path which will affect typical design uniform quartz sand as related to particle size and
natural variation in particle shape and surface rough-
decisions will often best be countered by a re-
ness. PhD thesis, University of California, Berkeley.
liance solely on the critical state strength com-
Wong, R. K. S. & Arthur, J. R. F. (1985). Induced and
ponent, sacrificing any overall dilatancy which inherent anisotropy in sand. Gtotechnique 35, No. 4,
may or may not prove to be mobilizable. 471481.

View publication stats