Schanz, T. & Vermeer, P. A. (1996). Geotechnique 46, No.

1, 145151

TECHNICAL NOTE

Angles of friction and dilatancy of sand

T. S C H A N Z  a n d P. A . V E R M E E R 

KEYWORDS: laboratory tests; plasticity; sands; shear

strength.

LABORATORY TESTING

Triaxial compression tests were performed on a

quartz sand (Flavigny, Desrues & Palayer, 1990).
This so-called Hostun sand has been used for
many years (Desrues, 1984; Desrues, ColliatDangus & Foray, 1988) in model tests and for
research on constitutive modelling. The material
parameters were emin = 0648; emax = 1041; rs =
265 g/cm3 . Fig. 1 shows the grain size distribution.
All samples were compacted by pluviation in a
steel cylinder lined with a rubber membrane
(t0 = 03 mm). Under a back-pressure of 0 =
50 kN/m2 , the samples ( H0 = D0 = 100 mm) were
placed in the triaxial cell, back-pressure was
removed, and the samples were consolidated under
c . To speed up saturation, the samples were
saturated rst with CO2 and then with water. The
volume change was measured by pore-water
volume change, and the specimens were axially
strained at 1% per minute.
Because the heightdiameter ratio H0 /D0 of all
the samples was unity, special means were
necessary for compensation of end restraint. The
following anti-friction system was used. Both end
plates (enlarged diameter 110 mm) were made
from polished glass with a centre hole for
drainage. A silicon greaserubber interface was
placed between the plates and the sample. Previous
tests have shown the shear parameters measured
with this system to be equal to those measured

INTRODUCTION

The strength of sand is usually characterized by the

peak friction angle p and the critical state friction
angle cv . It is generally realized that the peak
friction angel depends not only on density but also
on the stress path, including differences between
plane strain and triaxial testing conditions. Indeed,
plane strain and triaxial strain angles can differ by
more than 58 for a dense sand. For a loose sand at
the critical density it is often suggested that similar
differences occur (e.g. Stroud, 1971; Lade, 1984).
However, some authors have presented data that
suggest a unique critical state angle (e.g. Rowe,
1962, 1971; Bolton, 1986).
This technical note presents data on a unique
critical state angle. The implication is that the
failure criterion of a very loose sand is accurately
described by the MohrCoulomb condition, which
gives the known six-sided pyramid in principal
stress space.
The test data on dense as well as loose Hostun
sand are also used to study the rate of dilation.
This topic was extensively treated by Bolton
(1986), and it is now generally accepted that the
triaxial rate of dilation coincides with the rate of
dilation found in plane strain tests. Following
Roscoe (1970), Bolton used an angle of dilatancy
p for plane strain, but its denition is not
extended to cover triaxial strain. However, an
attempt at this was made by Vaid & Sasitharan
(1991). A different denition is presented in this
technical note which was previously given by
Vermeer & de Borst (1984) but is derived
differently here. Empirical evidence shows that
the denition matches data from both plane strain
and triaxial strain.

Percentage finer by weight

100

Manuscript received 11 January 1995; revised manuscript

accepted 4 May 1995.
Discussion on this technical note closes 3 June 1996; for
further details see p. ii.
 Stuttgart University.

60
40
20
0
0.063

0.125

0.25
Diameter: mm

0.5

Fig. 1. Grain-size distribution of Hostun sand

145

Article number = 632

80

SCHANZ AND VERMEER

: %

(1)

(Goldscheider, 1982) where t0 is the thickness of

the membrane and 1 is the axial stress. Also, the
effect of the lateral membrane restraint was
estimated by assuming it to be a right cylinder.
With the stiffness of the membrane Em (= 1400 kN/
m2 ), the correction stress 3 c can be calculated
according to
c3 2t0 Em 3 =D0

10
1: %

15

20

Fig. 2. Stressstrain behaviour of dense Hostun sand

(2)

where 3 is the radial stress and 3 is the radial

strain. In contrast to the bedding error, this
membrane stiffness correction had little impact on
the test results.

5
8

: %

tc =t0 0
3[1 exp ( 0
00371 )]

12

1/3

conventionally with only lter plates. The present

system ensures a near-uniform deformation of the
sample up to peak stress ratio.
The bedding error t c caused by the lubrication, which can lead to a 60% reduction in the
initial moduli of axial stiffness, was numerically
eliminated using

1/ 3

146

FRICTION ANGLES

1
0

20
10
1: %

15

Fig. 3. Stressstrain behaviour of loose Hostun sand

12
5

: %

8
1/3

Standard drained triaxial tests were carried out

on dense Hostun sand, with 0 = 163 kN/m3 and
ID = 115, and on loose Hostun sand with
0 = 139 kN/m3 and ID = 038. To check the
reproducibility of test results, four control tests
were performed at a xed cell pressure of
3 = 300 kN/m2 .
Figures 2 and 3 shows test results; stressstrain
curves are plotted with the stress ratio on the left
vertical axis and strainstrain curves are superposed by plotting the volumetric strain on the right
vertical axis. The test data show that the
reproducibility of triaxial tests is quite good.
A second step in checking the reproducibility
and thus the reliability of test data is to compare
data from different laboratories. A direct comparison can be made between the present data (IGS)
and data from the Grenoble Institute of Mechanics
(IMG) (Flavigny, Hadj-Sadok, Horodecki & Balachowski, 1991), as both laboratories have used the
same sand and the same testing procedure,
including the lubrication of end plates. The
comparison was made by using test data for the
dense sand and plotting average values for a series
of control tests, as shown in Fig. 4.
Even with comparable testing procedures, different laboratories appear to produce slightly
different curves. Some differences with classical
test data (aspect ratio of two and no lubrication)
are expected, but the deviations between IGS
results and IMG results are surprising; as yet there
is no clear explanation. However, in terms of
friction angles the differences between IGS and
IMG are smaller than Fig. 4 suggests, as peak

4
IGS
IGM
Non-lubricated

1
0

0
5

10
1: %

15

20

Fig. 4. Mean stressstrain behaviour found in three

laboratories

friction angles of about 428 and 408 degrees

respectively are found (the precise values are given
in Table 1.
Figure 4 shows that all volumetric strains
compare well up to an axial strain of about 10%,
which is well beyond peak strength. Differences
occur beyond an axial strain of 10%, when a
critical state is approached in which the sample

147

FRICTION ANGLES AND DILATANCY OF SAND

Table 1. Shear strength and dilatancy of Hostun sand

under triaxial compression
trp :
degrees

trcv :
degrees

trp :
degrees

ID = 115
IGS

419

348

133

IMG

401

357

140

Non-lubricated

418

377

126

ID = 038
IGS

344

344

VALIDATION OF THE STRESSDILATANCY THEORY

Several theories have been developed for predicting the volume strain in triaxial testing as a
function of the axial strain. In particular, the
applicability of Rowe's (1962, 1971) stress dilatancy theory has been shown by Barden & Khayatt
(1966) and Wood (1990). This is also done here,
but in addition Rowe's idea of superposition is
emphasized as this is applied when considering
angles of dilatancy. The stress dilatancy theory
starts with the expression for plane states of strain
D R=K

00

(3)

where D = 2_3 /_1 , is the stress ratio 1 /3 and K is

a coefcient representing the internal friction which
may be expressed as
K tan2 (45 f =2)

deforms with further change of volume. At the end

of the test, at an axial strain of 17%, this critical
state is not yet fully reached but softening and
dilation are clearly damping out. At 17% vertical
strain the IGS and IMG data yield friction angles
of 3488 and 3578 respectively. It is possible that a
critical state angle of almost 3448 would have
been reached on further straining. This angle is
obtained from the loose sand data in Fig. 3, and is
assumed here to be the critical state angle of
friction.
Having obtained a peak friction angle of 40428
for the dense sand and a maximum friction angle
of 3448 for the loose sand, it is interesting to
compare these triaxial angles to friction angles
measured in plane strain tests by Hammad (1991).
The latter data are listed in Table 2 for various
values of the conning stress.
Taking data for a cell pressure of 300 kN/m2 , as
was also done in triaxial testing, a peak friction
angle of 45478 is found for the dense sand and a
maximum friction angle of 3253458 for the loose
sand. A signicant difference is thus found for the
dense sand, as other studies, whereas there is very
little difference for the loose sand at the critical
state. (This nding is conrmed below by data for
other sands.) Hence it seems that a unique critical
state angle cv exists independently of strain
conditions.

(4)

For loose sands f is equal to the friction angle

cv at critical state, but values tend to be lower for
dense sands. Rowe derived these relationships by
considering the rate of energy dissipation. On
changing from plane state of strain to triaxial
testing conditions, he computed the rate of energy
by adding the effects of two mechanisms. However, his resulting equation can also be obtained
without considering energy dissipation, as is now
shown. Similarly to Rowe, sliding on planes
governed by the stress ratio 1 /2 (mechanism A)
and sliding on other planes governed by 1 /3
(mechanism B) are considered.
Figure 5 shows the A mechanism with sliding
on a 1 2 plane and the B mechanism with
sliding on a 1 3 plane. Each sliding mechanism
constitutes a planar deformation, and it is thus
tempting to apply equation (3) to each separate
mechanism. This yields
_2 =_1A DA RA =K

(5a)

_3 =_1B DB RB =K

(5b)

where RA = RB = R and _ 2 = _ 3 for triaxial testing

conditions. The basic idea that follows from these
considerations is that there are two contributions to
the axial strain, i.e.
_ 1 _ 1A _ 1B 2_3 K=R

(6)

Table 2. Angles of friction and dilatancy of Hostun sand in the biaxial test (Hammad, 1991)
3 : kN/m2

ps
p : degrees

ps
p : degrees

ps
p : degrees

ID = 095

ps
p : degrees
ID = 037

100

467475

145147

355

00

200

464470

141142

325345

00

400

451453

114121

330333

213

148

SCHANZ AND VERMEER

6
1

Kcv (cv = 34.4)

1/3

K ( = 29)
2

Fig. 5. Deviation of triaxial dilatancy from biaxial

state

D R=K

(7a)

D 2_3 =_1

(7b)

Hence the difference between the plane strain

(equation (3)) and equation (7a) concerns a factor
of two in the denition of D, as noted by Rowe
(1962). In the present derivation, the idea of
superposition is shown in Fig. 5, i.e. two localized
sliding motions in shear bands. In reality much
more diffuse pre-peak deformation patterns occur,
but this does not change the idea of superposing an
A-type mechanism and a B-type mechanism, which
leads to the above results.
The value of the angle f in the expression for
K has not yet been dened. Triaxial test data are
now considered for this purpose. The data for
dense and loose Hostun sand are plotted in Figs 6
and 7 respectively. Using equation (7) in the form
R = KD, R is plotted on the vertical axis and D is
plotted on the horizontal axis.
Nearly straight lines that pass through the
origin, as suggested by the expression R = KD,
are found. In fact the plot zig-zags around such

1/3

Kcv (cv = 34.4)

K ( = 29)

1. .
D = 23/1

1.5

lines, as the strain ratio D is computed from very

small increments of strain. When a ratio is
computed, small errors tend to have large consequences. Note that the zig-zagging would vanish
if D were computed from strain increments twice
as large. In Figs 6 and 7 lines are plotted for K ,
where f is taken to be the interparticle angle of
friction, and also for Kcv , where the critical state
angle of friction is used. Accordingly to Rowe
(1971), the former should be used for dense sands
and the latter is more appropriate for loose sands.
However, the differences between the resulting
lines is small and an average value would be
adequate for most practical purposes.

ANGLE OF DILATANCY

The angle of dilatancy is rst examined in plane

strain situations and its denition is then extended
to include triaxial compression. For plane strain
conditions, the denition is given in several
textbooks and by Bolton (1986)
sin ps

0.5

Fig. 7. Stressdilatancy plot for loose Hostun sand

or in short

0.5

1
. .
D = 23/1

1.5

Fig. 6. Stressdilatancy plot for dense Hostun sand

(mean values)

_ 1 _ 3
_ 1 _ 3

(8)

The rst minus sign should be omitted when

contractive strains are considered positive. When
considering the peak dilatancy angle rates rather
than mobilized pre-peak angles of dilatancy, one
should obviously use rates of strain as measured at
and beyond peak stress ratios. Analogously to the
extension of the stressdilatancy theory, the concept
of a dilatancy angle can be extended to include
triaxial test conditions. Again the axial strain is
considered to consist of an A mechanism in
combination with 2 and a B mechanism that
relates to the other principal strain 3
_ v _ 1A _ 1B

_ 2
_ 3

DA DB

(9)

FRICTION ANGLES AND DILATANCY OF SAND

DA DB

1 sin
1 sin

(10)

This yields for the expression

sin

_ v =_1
2 _ v =_1

(11)

Hence, a denition has been derived for the

dilatancy angle that can be used to measure this
angle in triaxial compression testing. A more
formal derivation based on concepts of the theory
of plasticity is given by Vermeer & de Borst (1984).
Applying equation (11) to the triaxial test data in
Figs 2 and 3, a (peak) dilatancy angle of 148 is
found for the dense Hostun sand and a vanishingly
small value of about zero is obtained for the loose
sand. These values correspond extremely well to
values measured in plane strain tests: Hammad
(1991) reports virtually identical values to those
given in Table 2.
The plane strain denition (equation (3)) for the
dilatancy angle is formally equal to the triaxial
denition (equation (11)). This is due to the
fact that 2 vanishes for plane strain, giving
v = 1 + 3 , and so equation (3) reduces to equation (11). Hence the latter equation is valid for
both test conditions. This supports the nding that
the same dilatancy angle is measured in plane
strain and triaxial tests. Bolton (1986) presents
numerous data to show that both tests yield the
same peak ratio of _ v /_1 .

ROWE'S THEORY AND THE ANGLE OF DILATANCY

The relationship between the dilatancy angle

and the friction angle is also given by Bolton
(1986). On combining the stressdilatancy equations (3) and (4) with the denition of the
dilatancy angle in equation (11), it is found that
sin ps sin ps
f
sin
1 sin ps sin ps
f

(12)

The superscripts ps have been added to denote

plane strain angles of friction, as this formula was
derived using the plane strain equations (3) and (4),
and plane strain angles of friction tend to be larger
than friction angles measured in triaxial tests. No
superscript is used to denote the dilatancy angle, as
this angle is considered to be independent of testing
conditions. According to Rowe ps
f coincides with
the critical state angle cv . If the data in Table 2 are
used to compute ps
f from equation (12), the dense
sand yields ps
=
368
and the loose sand yields
f
ps
f = 3458. As the difference is relatively small,
there exists a more or less uniquely dened angle
f ps which corresponds well with the critical state
angle.

149

Instead of combining the plane strain equations

(3) and (4) of the stressdilatancy theory with the
denition of the dilatancy angle in equation (11),
one might use Rowe's equation (equation (7)) for
triaxial tests with equation (11) to obtain
sin

sin tr sin trf

1 sin tr sin trf

(13)

This equation is the same as equation (12) except

for the superscripts, which mean that these angles
have to be measured in triaxial tests instead of
plane strain tests. In triaxial tests one tends to nd
smaller peak friction angles than in plane strain
tests, and Rowe reports a similar tendency for f .
Indeed, for dense Hostun sand it is found that
trf = 298, which is signicantly different from the
3458 found earlier for ps
f .
It is concluded that Rowe's stress dilatancy
theory exhibits an appealing relationship between
the friction angle and the dilatancy angle for
planar deformation, in that ps
f = cv . However,
this theory needs to be supplemented for triaxial
conditions of stress and strain in order to obtain a
relationship between the friction angle and the
dilatancy angle. For this reason, relationships given
by Bolton (1986) are now considered.
BOLTON'S FINDINGS FOR PEAK ANGLES

Bolton (1986) assumes a unique critical state

angle cv for both triaxial strain and plane strain.
This is conrmed by test data for Hostun sand.
Bolton gives a large database which leads to the
correlations for plane strain
ps
ps
p cv  5I R

(14)

and
trp trcv  3I R

(15)

for triaxial strain, where IR is a relative dilatancy

index
I R I D (Q ln m ) R

(16)

which relates density and the applied stress level. It

was found that Q = 10 and R = 1 give the best t
for different sands. Combining equations (14) and
(15) gives
trp  15(3ps
p 2cv )

(17)

Equation (17) is not mentioned directly by Bolton,

but is a direct consequence of his ndings. Fig. 8
provides data from additional sources.
There is a good deal of evidence for the validity
of equation (17). It therefore appears that differences between friction angles disappear as looser

150

SCHANZ AND VERMEER

50

Cornforth (1964)
Leussink et al. (1966)
Hostun sand (dense)
Hostun sand (loose)

trp: degrees

Equation (17)

40

30
30

cv

40
psp: degrees

50

ACKNOWLEDGEMENTS

Fig. 8. Maximum strength under plane strain and

triaxial strain

states are considered. This has implications for the

form of the limiting envelope for states of stress in
principal stress space. For dense samples the plane
strain friction is well above the MohrCoulomb
prediction, but looser samples give envelopes
according to MohrCoulomb. There are a lot of
true triaxial data to conrm the former, but few
true triaxial tests have been performed on loose
sand. Therefore it is often suggested that friction
angles are strain-dependent for both loose and
dense sands. Considering results from Bolton and
the additional data of Fig. 8, the present authors do
not agree.
Another nding by Bolton is that the rate of
dilation is strain-independent. It is found for both
triaxial strain and biaxial strain that
_v =_1 0
3I R

(18)

This supports the idea of a unique angle of

dilatancy, as this angle was related to the above
rate of dilation. Combining equations (11) and (18)
gives
0
3I R
IR
sin

(19)
2 0
3I R 6
7 I R

CONCLUSIONS

tions. The extended theory is validated by the fact

that data from plane strain and triaxial strain
conditions yield the same angle of dilatancy at
least near and beyond peak.
In contrast to the angle of dilatancy, friction
angles differ considerably when triaxial strain and
plane strains are compared. This difference basically depends on the critical state friction angle, as
by Bolton (1986) and other researchers. As yet it
is not fully clear whether or not plane strain
conditions yield slightly higher critical state angles
than triaxial strain conditions. Considering data
from Hostun sand, no such difference is observed.
There is linear relationship between angles of
maximum friction for both conditions (equation
(17)).

From the results presented, the following conclusions can be drawn concerning the angles of
friction and dilatancy of sand.
By using concepts of superposition it is possible
to relate the angle of dilatancy to triaxial strain
conditions. This yields an extended denition for
the angle of dilatancy which applies to triaxial
testing conditions as well as plane strain condi-

The authors are indebted to Dr J. Desrues and

Dr E. Flavigny of the Institut de Mecanique de
Grenoble for discussion on the triaxial testing
technique, and for their biaxial testing data on
Hostun sand.

NOTATION
D diameter
e void ratio
Em membrane thickness
H height
ID dilatancy index
IR relative dilatancy index
K internal friction coefcient
R stress ratio (1 /3 )
t0 membrane thickness
t bedding error
3 radial strain
r density
0 back-pressure
1 axial stress
3 radial stress
cv critical state friction angle
p peak friction angle
p angle of dilatancy

REFERENCES
Barden, L. & Khayatt, A. J. (1966). Incremental strain
rate ratios and strength of sand in the triaxial test.
Geotechnique 16, 338357.
Bolton, M. D. (1986). The strength and dilatancy of
sands. Geotechnique 36, No. 1, 6578.
Cornforth, D. H. (1964). Some experiments on the
inuence of strain conditions on the strength of sand.
Geotechnique 16, 193.
Desrues, J. (1984). La localisation de la deformation
dans les materiaux granulaires. DSc thesis, Institut de
Mecanique de Grenoble.
Desrues, J., Colliat-Dangus, J. L. & Foray, P. (1988).
Triaxial testing of granular soil under elevated cell
pressure. In Advanced triaxial testing of soil and

FRICTION ANGLES AND DILATANCY OF SAND

rock, STP977, pp. 290310. Philadelphia: American

Society for Testing and Materials.
Flavigny, E., Desrues, J. & Palayer, B. (1990). Le sable
d'Hostun RF. Rev. Fr. Geotech. No. 53, 6770.
Flavigny, E., Hadj-Sadok, M., Horodecki, G. & Balachowski, L. (1991). Series repetives d'essais triaxiaux
dans deux laboratoires. Archwm Hydrotech. 38, 12.
Goldscheider, M. (1982). Results of the international
workshop on constitutive relations for soils, pp. 11
54. Rotterdam: Balkema.
Hammad, W. I. (1991). Modelisation non lineaire et etude
experimentale des bandes de cisaillement dans les
sable. DSc thesis, Institut de Mecanique de Grenoble.
Lade, P. V. (1984). Mechanics of engineering materials
(edited by C. S. Desai). Chichester: Wiley.
Leussink, H., Wittke, W. & Weseloh, K. (1966).
Unterschiede im Scherverhalten rolliger Erdstoffe
und Kugelschuttungen im Dreiaxial- und Biaxialversuch. Veroff. Inst. Bodenmech. Felsmech. TH
Frideric Karlsruhe, 21.

151

Roscoe, K. H. (1970). The inuence of strains in soil

mechanics. Geotechnique 20, No. 2, 129170.
Rowe, P. W. (1962). The stressdilatancy relation for
static equilibrium of an assembly of particles in
contact. Proc. R. Soc., A269, 500527.
Rowe, P. W. (1971). Theoretical meaning and observed
values of deformation parameters for soil. Proceedings of Roscoe Memorial Symposium, pp. 143194.
Henley-on-Thames: Foulis.
Stroud, M. A. (1971). The behaviour of sand at low
stress levels in the simple-shear apparatus. Dissertation, University of Cambridge.
Vaid, Y. P. & Sasitharan, S. (1991). The strength and
dilatancy of sand. Can. Geotech. J. 29, 522526.
Vermeer, P. A. & de Borst, R. (1984). Non-associated
plasticity for soils, concrete and rock. Heron 29,
No. 3.
Wood, D. M. (1990). Soil behaviour and critical state
soil mechanics. Cambridge: Cambridge University
Press.