Dilatancy in Slow Granular Flows

Alexandre J. Kabla
*
Department of Engineering, University of Cambridge, Trumpington Street, Cambridge CB2 1PZ, United Kingdom
Tim J. Senden
Department of Applied Mathematics, Research School of Physics and Engineering, The Australian National University,
Canberra, 0200 Australia
(Received 8 October 2008; published 3 June 2009)
When walking on wet sand, each footstep leaves behind a temporarily dry impression. This counter-
intuitive observation is the most common illustration of the Reynolds principle of dilatancy: that is, a
granular packing tends to expand as it is deformed, therefore increasing the amount of porous space.
Although widely called upon in areas such as soil mechanics and geotechnics, a deeper understanding of
this principle is constrained by the lack of analytical tools to study this behavior. Using x-ray radiography,
we track a broad variety of granular ow proles and quantify their intrinsic dilatancy behavior. These
measurements frame Reynolds dilatancy as a kinematic process. Closer inspection demonstrates, however,
the practical importance of ow induced compaction which competes with dilatancy, leading more
complex ow properties than expected.
DOI: 10.1103/PhysRevLett.102.228301 PACS numbers: 83.80.Fg, 47.57.Gc, 83.50.v
The impact of granular ows on our daily lives is often
understated, yet grains represent the second most common
material displaced by humans. Consideration of just a few
of these activities such as the efcient transport of grain,
the optimal compaction of building materials and the pre-
diction of rock slides demonstrates the strikingly complex
behavior of these materials. Many of these global phe-
nomena have been studied in detail; to mention a few, a
nite angle of repose and avalanching [1], spontaneous
jamming [2], dilatancy [3], compaction [4], shear banding
[5], or grain segregation [6]. However, theoretical advances
are hampered by fundamental problems such as history
dependent effects or heterogeneous ow proles, stressing
the need for new tools to dene and measure intrinsic local
parameters. Tackling such issues in granular systems rep-
resents an important step towards a general understanding
of the dynamics and rheology of athermal and soft glassy
systems [7].
The phenomenon of dilatancy [3] corresponds to the
decrease of a grain piles density upon shear. The intuitive
interpretation is that free volume is required for grains to
rearrange. This property is typically quantied at small
deformation by shearing a sample and measuring the ratio
between imposed boundary displacement and resulting
thickness change at constant pressure [8]. However, ef-
cient ow predictions rely on the determination and testing
of constitutive equations valid up to large deformations,
where ow proles need to be characterized. Along that
line, dilatancy during shear band formation has been re-
cently highlighted [9,10], but a general framework for
dilatancy in granular ows is still needed.
We introduce here an experimental approach which
combines real time x-ray radiography and DIC (Digital
Image Correlation) techniques to locally quantify both
density variations and strain tensors in a range of 3D
granular ows. In particular, we have studied gravity ows
and forced shear ows. Flows are generated in rectangular
containers, typically 5 cmwide, 3 cm thick, 20 cmhigh and
lled with millimetric grains. In the forced shear experi-
ment, we use a particular geometry where a rigid bracket
moves slowly upward while a small wall prevents grains
from falling below the bracket [Fig. 1(a)]. The bracket and
the opposite wall are fully covered by bound grains to
maximize friction. As discussed later, the complexity of
the resulting ow prole is a key element to probe the
generality of our observations. The gravity driven experi-
ment consists of a large smooth wall moving upward in a
similarly dimensioned container as before [Fig. 1(b)].
Withdrawing the partition progressively frees space below
it such that grains are free to ow into the void. Unlike the
rst conguration, the grains freely slip against the walls.
In both situations, ows are invariant along the cell thick-
ness and studied in the quasistatic limit, the walls being
moved incrementally at speeds of about 0.1 mms
1
, i.e.,
one grain every ten seconds.
X-Ray absorption is classically used to quantify minute
density variations in 3D piles; for instance, to study the
onset of shear banding [11] or compaction [12]. Recent
progress in real time x-ray imaging allows now to study
more dynamic processes such as preavalanche dynamics
[13]. We use here the same setup as in [13] [Fig. 1(c)] to
acquire series of x-ray images while grains are owing.
The intensity in a radiograph takes the following form:
Ix, y, i I
0
x, y, i exp

L
g
x, y, i

(1)
where L
g
x, y, i represents the local length of grain
crossed by the x-ray beam at the coordinates (x, y)
PRL 102, 228301 (2009)
P HYS I CAL RE VI E W L E T T E RS
week ending
5 JUNE 2009
0031-9007,09,102(22),228301(4) 228301-1 2009 The American Physical Society
[Fig. 1(d)],
g
corresponds to the absorption coefcient of
the grain material (assuming the absorption is uniform in
the beam spectrum); it is measured independently.
I
0
x, y, i accounts for the angular variation in x-ray source
intensity, the source temporal intensity uctuations, as well
as the absorption of the container itself and variations in
detector sensitivity across the face of the detector. To
determine I
0
x, y, i, we record radiographs I
vg
x, y, i
niI
0
x, y, i of an empty container in exactly the same
geometrical conguration. The scalar ni is due to random
intensity uctuations of the source; we use a region S
without grain, i.e., where Ix, y, i I
0
x, y, i, to quantify
it: ni hI
vg
x, y, i,Ix, y, ii
S
. The local grain thickness
is then provided by
L
g
x, y, i
g
ln

I
vg
x, y, i
niIx, y, i

. (2)
The average grain density over an arbitrary domain D
which spans the sample thickness is directly obtained
from hx, y, ii
D
hL
g
x, y, ii
D
,L where L is the gap
between the front and rear walls of the container.
DIC is now a standard tool to quantify strain elds in
complex or composite media such as ber structures [14],
biological tissues [15], and granular materials [9]. Here, we
use the fact that the density images exhibit a noise pattern
at the grain scale [Fig. 1(e)] to reconstruct material point
trajectories during whole experiments. This spatial pattern
is a signature of the local 3D structure of the packing and is
advected with the grains, although it progressively evolves
with the strain. In practice, displacement increments
around half a grain diameter and total pile thickness of
up to 50 grains provide robust measurement conditions for
cross-correlation methods.
Using the ow trajectories, we determine the evolution
of local domains arranged on a grid, and quantify for each
of them, at each instant, the local density and total strain
from the initial state. Figure 2 shows the typical ow and
density proles for both chambers. The initial congura-
tion was prepared by slowly pouring grains into the con-
tainer while mechanically tapping the cell, providing initial
packing fractions about 0.64 with a suitable homogeneity
across the cell. Both experiments exhibit strong correla-
tions between the local deformation and the packing frac-
tion variation. The local shear (deviatoric component of the
2D strain tensor) is accounted for by a scalar s j
1

2
j,2 where
1
and
2
are the eigenvalues of the logarith-
mic strain tensor for nite deformations [16]. Figure 3
shows in a scatter plot the local values of the packing
fraction and shear s for a broad range of different shear
histories and stress states. The data allow us to dene a
characteristic shear strain s
c
after which the steady state
(loose limit) is locally reached. An empirical exponential
relationship between the local strain and density, as sug-
gested in [17], is consistent with our results:
e
min

max

min
exps,s
c
. (3)
The pile response remains qualitatively unchanged with
rougher grain shapes and/or larger polydispersity [Fig. 4(a)
and 4(b)], the value of critical deformation s
c
being sys-
tematically about 0.20.3. These results indicate that
Reynolds dilatancy can be summarized by a local kine-
matic relationship, supporting a number of theoretical and
numerical approaches [18,19]. Our setup provides a ge-
neric and robust way to extract empirical parameters such
as limiting densities and characteristic shear scales. In
contrast with traditional shearbox tests where the shear
zone remains undetermined [8], taking into account the
heterogeneity of the ow prole produces well dened
material properties.
If the same experiments are repeated with loose sphere
piles [see Fig. 4(c)], the material compacts as expected at
small strains [20]. However, it rapidly reaches a steady
density similar to that previously reported, the values being
scattered mostly between 0.59 and 0.6. Our experiments
FIG. 1 (color online). (a) Shear ow cell. (b) Gravity ow cell.
Moving parts, shown with darker sides, move over 3 cm (2050
grains) for (a) and 8 cm (50100 grains) for (b). The beam enters
the chambers from the front side. (c) X-ray images acquisition
setup. Source settings: 80 kV and 150 A. A 2 mm Al lter
reduces beam hardening. The cell is 2.3 m away from the source,
ensuring a quasiparallel beam. (d) Illustration of the path of an
x-ray and (e) resulting material length image L
g
x, y. Circle
indicates the grain size.
PRL 102, 228301 (2009)
P HYS I CAL RE VI E W L E T T E RS
week ending
5 JUNE 2009
228301-2
suggest that the critical state picture [8] is valid, i.e., that a
particular packing fraction at which shear can be sustained
at constant volume exists and is independent of the initial
state and deformation mode.
The generality of this statement is, however, tempered
by closer inspection of the data variance. In the gravity
ow experiment, nearly all the regions experience a sig-
nicant increase of the density following the shear-induced
expansion (see inset in Fig. 3). This process is known as
compaction and corresponds to the intrinsic response of a
granular pile to mechanical taps or vibrations [4,21]. Such
uctuations arise in our experiments in the vicinity of
owing regions and cause frequent reorganization of the
force network. A similar yet more subtle effect is observed
in the forced shear experiments. The packing fraction for a
given shear value increases with the distance to the shear
band. This is highlighted on Fig. 4(b) but could be seen on
all studied systems. This fact again contradicts the exis-
tence of a well dened relationship between shear strain
and density. Although different regions experience differ-
ent shear rates, this cannot account for the difference in
behavior in a quasistatic experiment. The caution comes in
considering the competition of local compaction and dila-
tancy during shear. The gravity driven experiment, for
example, shows that a localized ow can generate stress
uctuations able to cause compaction in remote regions. In
the shear band, compaction effects are negligible because it
deforms over a time scale where the rest of the sample is
mostly static. Away from the shear band, we observe a
range of situations where material is being sheared and
simultaneously mechanically shaken by the stress uctua-
tions coming from the shear band. The effect is, as ob-
served, an increase of the resulting density for a given shear
strain. The relevance of a critical state has been recently
questioned in systems where grain deformation or crushing
FIG. 3 (color online). Evolution of the local density as a
function of the local shear for both ow chambers and various
bead sizes; all domains and all times are shown. Piles have been
compacted before the experiment starts. The scale for the density
(i.e., the value of the absorption length) is measured with 0.5%
accuracy. The inset and dashed line highlight a typical single
domain evolution for the gravity ows. Note that these data are
unltered and show the characteristic noise in the system and
measurement.
FIG. 2 (color online). Series of radiographs describing the ow and density proles for a shear ow (top) and gravity driven ow
(bottom). Material: monodisperse acrylic spheres, 1 mm in diameter. Numbers on the top right corner of each image indicate the
upward displacement of the moving wall. See EPAPS [24] movies Nos. 1 and 2 for the complete sequence of images. Letters designate
specic regions discussed in the text and Fig. 4(b).
PRL 102, 228301 (2009)
P HYS I CAL RE VI E W L E T T E RS
week ending
5 JUNE 2009
228301-3
is involved [22]. We demonstrate here that even with
perfectly rigid grains, the competition between dilatancy
and compaction drives the system away from the critical
state, even after large shear.
The experimental conguration detailed here offers nu-
merous improvements to the quantication of the relation-
ship between density and ow in granular matter. We
demonstrate that, to a rst approximation, dilatancy of
dense pile can be considered as a kinematic property,
with different ow situations providing consistent behav-
ior. This can be applied to a wide range of materials and
geometries, up to very large deformations, although the
study of nonisotropic grains would require a more complex
framework. Interestingly, these experiments also unveil the
intricate interplay between Reynolds dilatancy and com-
paction in complex granular ows. The dilatancy curve
measured in shear bands only denes a lower bound to the
range of available densities. To characterize the response
of the material in other regions, one needs to understand
how mechanical uctuations are generated, propagate, and
cause compaction. Although nonlocal elastic uctuations
have been derived from the ow prole in systems like
foams [23], such a question represents a major challenge in
the context of granular ows where the mechanisms of
force propagation and microplasticity are still poorly
understood. Nevertheless, this appears to be a key to pre-
dict both the density prole and stability of granular ows.
The subtle balance between structure relaxation (compac-
tion) and shear-induced disorder (dilatancy) involved in the
quasistatic rheology of granular matter deepens further the
analogy with jammed and structurally disordered systems
such as foams, emulsions, and colloidal glasses.
The authors would like to thank Tomaso Aste, Georges
Debregeas, L. Mahadevan, and Olivier Pouliquen for dis-
cussions. T. J. S. gratefully acknowledges the Australian
Research Council for support.
*[email protected]
[1] H. M. Jaeger, C. H. Liu, and S. R. Nagel, Phys. Rev. Lett.
62, 40 (1989).
[2] K. To, P-Y Lai, and H. K. Pak, Phys. Rev. Lett. 86, 71
(2001).
[3] O. Reynolds, Philos. Mag. 20, 469 (1885).
[4] J. B. Knight, C. G. Fandrich, C. N. Lau, H. M. Jaeger, and
S. R. Nagel, Phys. Rev. E 51, 3957 (1995).
[5] J. Desrues and R. Chambon, Int. J. Solids Struct. 39, 3757
(2002).
[6] Y. Oyama, Bull. Inst. Phys. Chem. Res. Jpn. Rep. 18, 600
(1939).
[7] P. Richard, M. Nicodemi, R. Delannay, P. Ribiere, and
D. Bideau, Nature Mater. 4, 121 (2005).
[8] W. Powrie, Soil Mechanics: Concepts and Applications
(Spon Press, London, 2004), 2nd ed.
[9] Rechenmacher, J. Mech. Phys. Solids 54, 22 (2006).
[10] K. Sakaie, D. Fenistein, T. J. Carroll, M. van Hecke, and
P. Umbanhowar, Europhys. Lett. 84, 38001 (2008).
[11] J. Desrues, R. Chambon, M. Mokni, and F. Mazerolle,
Geotechnique 46, 529 (1996).
[12] P. Philippe and D. Bideau, Europhys. Lett. 60, 677 (2002).
[13] A. Kabla, G. Debregeas, J. M. di Meglio, and T. J. Senden,
Europhys. Lett. 71, 932 (2005).
[14] S. Bergonnier, F. Hild, and S. Roux, J. Strain Anal. Eng.
Des. 40, 185 (2005).
[15] C. Wiebe and G. W. Brodland, J. Biomech. 38, 2087
(2005).
[16] S. Nemat-Nasser, Plasticity (Cambridge University Press,
Cambridge, England, 2004).
[17] S. Roux and F. Radjai, Physics of Dry Granular Media,
edited by H. J. Hermann et al. (Kluwer Academic
Publishers, Dordrecht, 1998), 229.
[18] J. D. Goddard and A. K. Didwania, Q. J. Mech. Appl.
Math. 51, 15 (1998).
[19] I. F. Collins and B. Muhunthan, Geotechnique 53, 611
(2003).
[20] M. Nicolas, P. Duru, and O. Pouliquen, Eur. Phys. J. E 3,
309 (2000).
[21] A. Kabla and G. Debregeas, Phys. Rev. Lett. 92, 035501
(2004).
[22] I. F. Collins, B. Muhunthan, A. T. T. Tai, and M. J. Pender,
Geotechnique 57, 437 (2007).
[23] A. Kabla, J. Scheibert, and G. Debregeas, J. Fluid Mech.
587, 45 (2007).
[24] See EPAPS Document No. E-PRLTAO-102-071920 for
supplementary movies. For more information on EPAPS,
see http://www.aip.org/pubservs/epaps.html.
0.55
0.54
0.53
0.52
0.51
0.50
1.2 1.0 0.8 0.6 0.4 0.2 0.0
0.61
0.60
0.59
0.58
1.0 0.8 0.6 0.4 0.2 0.0
0.66
0.65
0.64
0.63
0.62
0.61
0.60
0.59
0.8 0.6 0.4 0.2 0.0
L
o
c
a
l

D
e
n
s
i
t
y
L
o
c
a
l

D
e
n
s
i
t
y
L
o
c
a
l

D
e
n
s
i
t
y
Shear Component Shear Component
Shear Component
(a) (b)
(c)
C
B
A
FIG. 4. Local density as a function of the local shear for
various materials and initial densities in the shear cell. (a) with
monodisperse pasta cylinders, 2 mm in height and diameter.
(b) with polydisperse semolina grains, diameters ranging from
0.5 to 1 mm. The dashed lines represent the typical behaviors of
the regions A, B, and C dened on Fig. 2. Because of the nature
of these materials, absorption coefcients are poorly dened, and
the density scale is largely arbitrary. Relative variations are
however precise up to 0.1%. (c) Experiment with 1 mm acrylic
beads, starting with a loose pile.
PRL 102, 228301 (2009)
P HYS I CAL RE VI E W L E T T E RS
week ending
5 JUNE 2009
228301-4