Granular Matter (2012) 14:433–447

DOI 10.1007/s10035-012-0348-x

ORIGINAL PAPER

On the influence of inter-particle friction and dilatancy

in granular materials: a numerical analysis
Z. X. Yang · J. Yang · L. Z. Wang

Received: 20 July 2011 / Published online: 5 April 2012

© Springer-Verlag 2012

Abstract Mechanical behavior of granular soils is a clas- of the different macro-scale response due to variations in the
sic research realm but still yet not completely understood as inter-particle friction.
it can be influenced by a large number of factors, includ-
ing confining pressure, soil density, loading conditions, and Keywords Discrete element method ·
anisotropy of soil etc. Traditionally granular materials are Inter-particle friction · Mechanical behavior · Dilatancy ·
macroscopically regarded as continua and their particulate Macro- and micro-scales · Granular materials
and discrete nature has not been thoroughly considered
although many researches indicate the macro mechanical
behavior closely depends on the micro-scale characteristics List of symbols
of particles. This paper presents a DEM (discrete element kn , ks Normal and tangential stiffness of contact
method)-based micromechanical investigation of inter-par- model
ticle friction effects on the behavior of granular materials. D50 Mean particle diameter in a particle
In this study, biaxial DEM simulations are carried out under assemblage
both ‘drained’ and ‘undrained’ (constant volume) conditions. σx , σ y Principal stresses along x and y directions
The numerical experiments employ samples having similar respectively
initial isotropic fabric and density, and the same confining σ1 , σ2 Major and minor principal stresses respectively
pressure, but with different inter-particle friction coefficient. p, q Mean normal stress and deviatoric stress
Test results show that the inter-particle friction has a sub- respectively
stantial effect on the stress-strain curve, peak strength and εa , εv Axial strain along y direction and volumetric
dilatancy characteristics of the granular assembly. Clearly, it strain respectively
p p
is noted that apart from the inter-particle friction, the shear εv , εq Plastic volumetric and deviatoric strains
resistance is also contributed to the dilation and the particle respectively
packing and arrangements. The corresponding microstruc- CN Coordination number
ture evolutions and variations in contact properties in the ϕμ , μ Inter-particle friction angle and inter-particle
particulate level are also elaborated, to interpret the origin friction coefficient
di j , δi j Second order deviatoric tensor and Kronecker
Z. X. Yang (B) · L. Z. Wang
delta
Department of Civil Engineering, Zhejiang University, Hangzhou,
Zhejiang, China ni Direction cosines of the unit vector with
e-mail: [email protected] respective to the reference axes xi
L. Z. Wang E(ϕ), E 0 A density function and mean value of the
e-mail: [email protected] density over direction respectively
d , ϕd Intensity and principal direction fabric tensor
J. Yang
in general
Department of Civil Engineering, The University of Hong Kong,
Pokfulam, Hong Kong r (ϕ), r0 Spatial distribution of quantity r and its mean
e-mail: [email protected] value over direction respectively

123
434 Z. X. Yang et al.

rd , ϕdr Intensity and principal direction fabric tensor increasing surface roughness of granular particles in triax-
in term of quantity r ial compression tests. And Procter and Barton [14] further
f n (ϕ), f 0 Normal contact force distribution and mean pointed out that the inter-particle friction is a function of min-
value of normal contact force respectively eral composition, surface chemistry, as well as the shape and
f t (ϕ) Tangential contact force distribution surface roughness of the granular particles. Recently Cavar-
nd , ϕdn Intensity and principal direction of fabric retta et al. [15] established a correlation between inter-par-
tensor in terms of contact normal force ticle friction and roughness by directly measuring particle
td , ϕdt Intensity and principal direction of fabric roughness and inter-particle friction. Interestingly, however,
tensor in terms of contact tangential force macro-scale experimental tests [16–18] indicated that the
f μ (ϕ) Mobilized inter-particle friction distribution influence of particle shape is much more noticeable than that
μ μ
d , ϕd Intensity and principal direction fabric tensor of surface roughness (or inter-particle friction).
in terms of mobilized inter-particle friction Over the past decades, DEM (discrete element method)-
φmob Mobilized friction angle based numerical analysis has prevailed in the investigations
η, ηres Mobilized stress ratio and stress ratio at of mechanical behavior of granular soils, including the influ-
residual state respectively ences of particle characteristics, typically inter-particle fric-
A.R. Aspect ratio of particle characterizing the tion and particle shape. DEM simulation treats the granular
particle’s shape material as an assembly of particles that interact via a contact
logic and can offer a thorough perspective of the response at
the particulate level as well as the overall behaviors. Jensen
1 Introduction et al. [19] demonstrated the particle rotations and the shear
strength greatly depends on the particle shape. Polygonal
Mechanical behavior of granular soils is very complex particles are also used to investigate the influence of particle
and can be influenced by various factors, including confin- shape on the mechanical behavior both in macro- and micro-
ing pressure, density, fabric, loading conditions, etc. [1–7]. scales, see [20–22]. Thornton [23] showed both the peak and
Recent investigations [8–10] also indicate the mechanical critical state friction angles increase with the coefficients of
behavior (such as shear resistance) could be significantly inter-particle friction μ when μ ≤ 0.5, while Powrie et al.
affected by the particle level’s characteristics, such as par- [24] indicated a significant dependence of peak strength on
ticle shape (angularity, roundness and sphericity), particle the inter-particle friction in the range of 0.7 ≥ μ ≥ 0.3
surface characteristics, gradations of particles, etc., which but only modest increase in the residual strength based on
reflect the chemical, physical, and biological environmental triaxial test simulation. Alonso-Marroquín et al. [25] also
evolutions in the geological process. observed independence of the mobilized friction angle at the
Among the mechanical behaviors, dilatancy is one of the critical state of the inter-particle friction, and this non-depen-
essential characteristics of granular soils and of significance dence was attributed to the spontaneous formation of rota-
in academia as well as in engineering practice. When sub- tional patterns and clusters of particles with intense rolling.
jected to shearing, a loose sand contracts and dense sand Peña et al. [22] presented the similar results of the investi-
dilates, governed by both the stress ratio and inter-particle gation of the effect of inter-particle friction on the macro-
friction in terms of Rowe’s stress-dilatancy equation [11]. mechanical behaviour of granular materials, using convex
The shear strength stems from the inter-particle friction and polygonal particles. Other DEM simulations [26–28] imply
the resistance against the volumetric dilation of the parti- that the inter-particle friction may influence the macroscopic
cle assembly. Microscopically, the volumetric deformation response, including the dilatancy characteristics of granular
of the granular assembly upon shearing (i.e. dilatancy) is materials.
largely attributed to particle motion and rearrangement in Although the macro-mechanical behavior of granular soils
the manner of sliding and/or rolling, if no particle breakage has been well examined by laboratory experiments and
occurs. From the particle scale’s perspective, it is interest- numerical simulations as summarized above, the micro-scale
ing to explore how the inter-particle friction influences the analysis has not been adequately addressed to provide a clear
motion of the particles and then the macro-mechanical behav- insight into the mechanisms and the significance of the inter-
ior of granular soils. particle friction. This paper presents extensive numerical
Experimental work has been reported to justify the depen- investigation into the influence of inter-particle frictions on
dence of mechanical response on the particle characteris- the mechanical behavior of granular assembly based on DEM
tics. Skinner [12] suggested that the inter-particle friction analysis. The numerical samples all have similar initial iso-
does not significantly change the shear strength of granu- tropic fabric and density, and the same confining pressure, but
lar materials under plane strain conditions, while Haruyama with different inter-particle friction coefficient. A series of
[13] showed that shear strength and dilatancy increased with numerical ‘experiments’ (simulations) are carried out under

123
On the influence of inter-particle friction and dilatancy 435

100
both drained and undrained conditions. The dilatancy and the
shear failure modes of samples with varying inter-particle

cumulative percentage (%)

frictions can be identified and quantified. Comparisons are 80

also made to Rowe’s stress-dilatancy formulations, to take

into account the effect of the inter-particle friction. It has 60

also shown that the micromechanical analysis can be linked

D50=0.535mm
with and interpreted into the macro-scale responses that due 40

to variations in the inter-particle friction.

20

2 DEM Model 0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
particle size (mm)
DEM assumes the material consisting of separate, discrete
particles and is particularly capable of describing the micro- Fig. 1 Grading curve of the numerical sample
mechanical behavior of granular assemblies. In this paper, a
commercial code P FC 2D (Particle Flow Code in two dimen-
sions) (Itasca, 2005) based on the distinct element method less effort in computation, well control of desired particle
proposed by Cundall and Strack [29] is employed. In the gradation and readily achieving a specified initial density.
code, an explicit time-stepping numerical scheme is imple- Generally, the expansion method can produce an initially
mented and the motion of individual particles is traced and isotropic sample while the deposition method tends to intro-
contacts with neighboring particles are updated according to duce certain inherent anisotropy into the microstructure of the
a particular contact law. A detailed description on the theory sample [34,35]. The expansion method was adopted here to
and background can be referred to Cundall and Strack [29]. generate samples such that the influence of the initial anisot-
For a granular assembly simulated with P FC 2D , the par- ropy effect can be isolated. The dimension of the sample is
ticles are assumed as rigid bodies and are interacting through 25 mm × 25 mm.
the contacts in the micro-scale. As the contacts are simulated In the P FC 2D model, the basic circular disk elements
as a soft contact model with a specified contact force-defor- range from 0.26 to 0.66 mm in diameter with d50 = 0.535 mm,
mation relationship, the deformation of the assembly can still as shown in Fig. 1. Noting that the circular particle elements
be allowed to take place at the contacts. The contact law may lead to over-loosening the restriction of particle roll-
used in this study is a spring-slider model with two portions: ing during the loading process and then inevitably affect
bi-directional (normal and tangent) linear spring model and the dilatancy and deformation of the granular assembly [36–
tangent Coulomb-type slip model. The former states a linear 38], a spindly clumped element [39] shown in Fig. 2 is used
elastic relationship between contact force and contact defor- in this study. Some previous discrete element simulations
mation, both in normal and tangential directions, while the [20–22,40,41] have shown that using non-circular(oval or
latter ensures a Coulomb type friction mechanism between polygonal) particles can also reduce the particle rolling
the normal and shear forces at contact points. In addition, greatly. The clumped element is formed by glueing and over-
as P FC 2D solves a pseudo-static problem by solving a set lapping two identical basic disk particles and behaves as a
of motion equations, damping is introduced to eliminate the rigid body that will not break apart, regardless of the forces
non-steady kinetic energy. acting upon it, and can be implemented within an existing
All the parameters used in the numerical simulation are DEM framework without the introduction of new contact or
commonly used in literature or default values of PFC, and force algorithms. A parameter, aspect ratio A.R., is defined to
are summarized in Table 1. The wall stiffness is assigned to characterize the shape of clumped particle as shown in Fig. 2.
be the same as the particles, while the friction between walls The variation of A.R. values is bounded by two extreme
and particles is set to zero, signifying the boundary particles cases: A.R. = 1.0 when two disks are fully overlapped,
can move freely along the walls. and A.R. = 0.5 when two disks only contact without any
overlapping. However, it is clarified that convex polygonal
particle with A.R. = 1.0 can also restrict the particle rotation
3 Numerical sample and simulate more realistic behavior of granular materials,
see Peña et al. [22].
In recent years, some algorithms and techniques for the gen- Once the sample has been generated, an isotropic small
eration of numerical samples have been developed for DEM confining stress 10 kPa is applied and the sample thus has
analysis (e.g. [30–33]). Among them, both expansion method an initial effective stress = 10 kPa. After that, a compac-
and deposition method have the advantage over others in tion procedure is implemented to achieve the specified stress

123
436 Z. X. Yang et al.

Table 1 Parameters used in

Mass density Normal/tangential stiffness of particle/wall Particle/wall friction Damping parameter
numerical simulations
2,600 kg/m2 109 N/m 0 0.7

aspect ratio=b /a (DEM) were presented previously by Ng [40] and Shodja and
Nezami [41] and Nouguier-Lehon et al. [20] and Peña et al.
[21,22], using oval particles and polygonal particles respec-
tively. Alternatively in the current study, to restrict particle
b rolling and simulate more realistic behavior of granular mate-
rials, clumped particles with an aspect ratio of 0.6 are used
in the simulation, as defined in Fig. 2. All the test specimens
are compacted to the same isotropic stress state with an ini-
tial confining pressure of 1,000 kPa and expected to possess
a an isotropic microstructure. This is justified by the distri-
butions of particle orientation, contact unit normal, branch
Fig. 2 A clumped particle element vector and the contact force network shown in Fig. 4 for a
specimen after the isotropic compaction, with dotted lines
σy indicating the Fourier approximation (the variables’ defini-
tions and representations can be found in the Appendix). To
investigate the influence of inter-particle friction, five dif-
Rigid walls ferent target inter-particle friction coefficients of 0.05, 0.10,
Particle assembly
0.20, 0.50 and 1.0 are considered. To highlight this influence
σx and make the simulation results more comparable, all the
numerical specimens have close initial void ratio (in a narrow
range of e = 0.15 ∼ 0.17) before shearing. Based on some
Rigid walls
trial tests to control the density of the samples by varying
y inter-particle friction coefficient μpre during the preparation,
it is noted that at the reference state p  = 10 kPa, maximum
x void ratio is emax = 0.257 (μpre = 0.5) and minimal void
0 ratio emin = 0.136 (μpre = 0.0). Thus initial void ratio at
Fig. 3 A numerical sample with boundary conditions p  = 1, 000 kPa in this test is around 0.15 ∼ 0.17 and rali-
tative density Dr =∼80 %, and the samples are considered
as very dense.
state. Then, biaxial compressive shear under ‘drained’ or In the two dimensional analysis, the mean normal stress
‘undrained’ condition is proceeded. For ‘drained’ simulation, and deviatoric stress are defined as p  = (σx + σ y )/2 and
both the upper and lower boundary walls are pushed inwards q = σ y − σx , where σx and σ y are principal values along
the sample with a very small loading rate, while maintaining x and y directions respectively (Fig. 3). Figure 5 gives the
a constant horizontal stress (lateral confining stress) by mov- biaxial shear test results with different inter-particle friction
ing the lateral walls inwards or outwards as required. The coefficients under drained condition. As shown in Fig. 5a, the
‘undrained’ simulation is conducted under constant volume samples with μ ≥ 0.5 exhibit significant volumetric dilation,
conditions during the course of shearing. Figure 3 illustrates while those with μ < 0.2 behave in a contractive fashion.
a biaxial test sample with imposed boundary conditions. This suggests the inter-particle friction notably influences the
dilatancy response for granular soils. The underlying mech-
anisms may be that: a small friction facilitates the repacking
4 Numerical biaxial shear test and rearrangement of the particles during the shearing and
results in a contractive volume change; on the other hand,
In order to investigate the influence of inter-particle friction large friction may tend to partially hinder the motion of par-
on the mechanical behavior of the granular materials upon ticles and the inter-particle sliding is less likely to occur,
shearing, a series of numerical biaxial tests are conducted therefore the specimen tends to dilate, in response to the
under both drained and undrained conditions. It is known application of the deviatoric load.
that the particle shape also plays an important role on the Figure 5b presents the evolution of the stress ratio q/ p  as
behavior of particles, and some discrete element simulations the axial strain εa increases, where the axial strain denoting

123
On the influence of inter-particle friction and dilatancy 437

90 90
120 60 120 60
20 20

150 30 150 30
10 10
Percentage (%)

Percentage (%)
0 180 0 0 180 0

10 10
210 330 210 330

20 20
240 300 240 300
270 270

(a) (b)

90
120 60
20

150 30
10
Percentage (%)

0 180 0

10
210 330

20
240 300
270

(c) (d)

Fig. 4 An isotropic specimen after compaction. a particle orientation, b contact normal, c branch vector orientation, d contact force network

strain along y direction. It is seen that for the cases with ratio. Noting that for engineering application, mobilized fric-
μ = 0.5 and 1.0, q/ p  reaches the peak values at εa ≈ 2 % tion angles at both peak and critical states are key engineer-
and then drops as εa increases, while the cases with μ ≤ 0.2 ing design parameters, Fig. 5d shows the mobilized friction
exhibit hardening stress-strain responses. However, critical angles, φmob against various inter-particle frictions, μ, where
state values of q/ p  for the cases with μ ≥ 0.2 are margin- φmob = sin−1 [(σ y − σx )/(σ y + σx )]. It is seen that the peak
ally influenced and tend to approach a common value ≈0.8. friction angles does exist for the cases with μ ≥ 0.5, whereas
It is also interesting to note that for the cases with μ ≤ 0.10, critical state friction angle is just slightly different for all
the critical values of q/ p  are not far smaller than those with cases.
higher inter-particle frictions, indicating that inter-particle Figure 6 presents the simulation results under undrained
friction is not the only constituent for granulates in resisting conditions. As can be seen in Fig. 6a, the stress paths with
the shear loading. Noting that at the critical state, the mate- different friction coefficients are different from one another.
rials are turned to be purely frictional, and apart from the For cases with μ ≤ 0.2, the samples behave in a contractive
inter-particle frictions, particle rearrangement and repack- manner and both the mean normal stress and deviatoric stress
ing are also readily formed to counter-resist the shear load reduce during shearing, even down to a zero stress state for
[36,37]. the cases with μ = 0.05 and 0.1, which is very similar to the
Figure 5c illustrates the change of the void ratio against liquefaction state defined in saturated soils. For those with
the deviatoric stress. Again, the inter-particle friction depen- higher frictions, the specimen more likely tends to dilate to a
dence of peak shear strength for higher friction cases can higher stress level, and there is much less contractive behav-
be observed, although the critical shear strength is not very ior observed on the stress paths. Figure 6b demonstrates the
much affected. It is also noted that the higher friction may stress ratios, q/ p  varies with axial strain εa for various fric-
bring the specimen to the critical state with a larger void tion coefficients. At εa ≈ 2 %, q/ p  reaches the peak values

123
438 Z. X. Yang et al.

(a) 6 μ=0.05 (b)1.5 μ=0.05

μ=0.1 μ=0.1
3 μ=0.2 1.2 μ=0.2
μ=0.5 μ=0.5
μ=1.0 μ=1.0
0 0.9

q /p
εv (%)

-3 0.6

-6 0.3

-9 0.0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
εa (%) εa (%)

(c) 4000 (d) 50

μ=0.05 μ=0.05
μ=0.1 μ=0.1
40 μ=0.2
μ=0.2
3000
μ=0.5 μ=0.5
q (kPa)

μ=1.0 μ=1.0
30

φ mob (°)
2000
20

1000
10

0 0
0.16 0.20 0.24 0.28 0 5 10 15 20 25 30

e εa (%)

Fig. 5 Drained biaxial shear with various inter-particle frictions. a εv ∼ εa , b q/ p  ∼ εa , c q ∼ e, d φmob ∼ εa

and then drops to residual value ∼0.8 for μ ≥ 0.5 while flat- ior with higher inter-particle frictions is much more inhibited,
tens off for μ ≤ 0.2. An alternative view on the variation of and the dilative response is becoming dominant after passing
the mean normal stress with the axial strain is presented in the phase transformation state.
Fig. 6c. Presented in Fig. 6d is the mobilized friction angle, To better illustrate the significance of inter-particle fric-
φmob . For cases with μ = 0.5 and 1.0, φmob increases to tion on the shear failure and dilatancy, Fig. 8 shows three
the peak values of about 34◦ and 40◦ at εa ≈ 2 %, respec- typical undrained stress paths A, B and C, for different inter-
tively, and then reaches the residual value of about 24◦ at particle frictions with μa > μb > μc . It is seen curve
εa ≈ 20 % for both. However for μ < 0.2, no clear dila- A exhibits highly dilative shear failure, while curve C
tation is observed. The findings are quite similar to those of shows completely contractive shear failure, with curve B in
drained conditions. between. It is noted that for a given stress ratio q/ p  , dilatan-
cies D of points a, b and c on the curves are different, with
5 Influence of overall behavior Da < Db < Dc . Recalling the famous Rowe’s stress-strain
dilatancy equation [11], dεv /dε 1 = 1−(σ1 /σ2 )/ tan2 (45◦ +
p p

ϕμ /2), where D = dε v /dε 1 is a measure of dilatancy; σ1 /σ2

p p
To further examine the effect of inter-particle friction on the
dilatancy response, Fig. 7 presents the relationship of the is the ratio between the major and minor principal stresses;
stress ratio versus the dilatancy, where the stress ratio is nor- ϕμ is the inter-particle friction angle, an increase in the inter-
malized by the corresponding stress ratio at the critical state, particle friction will result in a more contractive response (D
and the dilatancy is expressed as dεv /dε 1 . As the elastic increases). It is apparent that the Rowe’s dilatancy equation
part is relatively small, no attempt was made to differentiate fails to predict the dilative behavior of granular materials
the plastic from the total strain. It is clearly shown that with with higher inter-particle friction and the influence of the
smaller inter-particle frictions, the granular assembly tends inter-particle friction on the dilatancy is not appropriately
to contract till the critical state, while the contractive behav- reflected.

123
On the influence of inter-particle friction and dilatancy 439

(a) 2000 (b) 1.5

1600 1.2
μ=0.05
μ=0.1
1200 0.9
μ=0.2
q (kPa)

μ=0.5

q/p
800 μ=1.0 0.6 μ=0.05
p’ and q → 0 μ=0.1
for tests μ=0.2
0.3
400
μ=0.05 and 0.1 μ=0.5
μ=1.0
0 0.0
0 400 800 1200 1600 2000 0 5 10 15 20 25 30

p (kPa) εa (%)

(c) 5000 (d) 50

μ=0.05
μ=0.1
4000 40 μ=0.2
μ=0.5
30
μ=1.0

φ mob (°)
3000
p (kPa)

μ=0.05
2000 μ=0.1 20
μ=0.2
μ=0.5
1000 μ=1.0 10
p’ → 0 for tests
μ=0.05 and 0.1
0 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
εa (%) εa (%)

Fig. 6 Undrained biaxial shear with various inter-particle frictions. a p  ∼ q, b q/ p  ∼ εa , c q ∼ εa , d φmob ∼ εa

1.0

P.T.
0.5 contractive
μ
0.0 μ μ μ
D D D
dεv/dε1

dilative
-0.5
μ=0.05 C.S.
μ
μ=0.1
-1.0
μ=0.2 μ
μ=0.5
-1.5
μ=1.0

-2.0
0.0 0.5 1.0 1.5 2.0 Fig. 8 Illustration of the inter-particle friction on the shear failure and
η /η res dilatancy

Fig. 7 Stress-dilatancy paths with various inter-particle frictions

DEM simulations conducted by Thornton [23] using state stress ratio is fairly not affected by the friction for both
P FC 3D with sphere elements suggest that both the peak the drained and undrained conditions with μ ≥ 0.2. This is
stress ratio and critical state stress ratio depended on the in agreement with the experimental results by Skinner [12],
inter-particle friction, but this effect is limited for μ ≥ 0.5 in which shearing on wet and dry assemblies (with distinct
or so. The results shown in this study with clump particles inter-particle frictions) of glass ballotini gives almost identi-
are generally consistent with his conclusion, but the critical cal stress ratio at critical state.

123
440 Z. X. Yang et al.

80

C.S. friction angle of Clumped particles

Peak friction angle of Clumped particles
60 (present study)
Horne's trend (1969)
Peak friction angle of Oval I
φ mob Peak friction angle of Oval II
40 (Oda et al 1982)
Skinner's results with error bars (1969)
C.S. friction angle of Sphere
Peak friction angle of Sphere
20 (Thornton 2000)
C.S. friction angle of Polygonal particle
Peak friction angle of Polygonal particle
(Pena et al 2008)
0
0.0 0.3 0.6 0.9 1.2 1.5
inter-particle friction μ

Fig. 9 Comparison of mobilized friction angles

Oda et al. [42] conducted biaxial compression testing on the total, and the accumulated mean particle rotations of all
the assemblies of oval cross-sectional rods with two differ- particles, are introduced to quantify the particle sliding and
ent inter-particle frictions and found that the peak stress ratio rolling, respectively.
depended on the inter-particle friction. However, this con- Slipping is assumed to be occurred at a contact if the tan-
clusion is only valid for Oval II (with more elongated cross gential contact force is equal to the ultimate tangential force,
section), and there is a poor dependence for Oval I parti- which is determined by the normal contact force and inter-
cles (with more rounded cross section). A detailed com- particle friction angle following the Coulomb’s friction law
parison is given by Fig. 9, including experimental data, [26]. Figure 10a–d show the relationship between the sliding
simulation results and theoretical predictions. The DEM fraction and the axial strain, and sliding fraction and stress
results include the simulations using various particle shapes: ratio, against the inter-particle friction coefficients, for both
spherical, oval, polygonal and clumped particles. It is noted drained and undrained tests respectively. It can be observed
that the theoretical curve given by Horne [43] only consid- that the sliding fraction is highly dependent on the inter-parti-
ers the mobilized friction angle at the critical state, and the cle friction angle, and the lower the friction angle, the higher
particle rotation effect is not considered. It is shown that is the sliding fraction. For example, under drained condi-
his prediction only agrees well with the data at the low fric- tion, the mean sliding fraction is almost 40 % for cases of
tion, while for high friction, the prediction significantly devi- μ = 0.05 and 0.1, and decreases to 30 % when μ = 0.2
ates from both experimental data and numerical simulation, and drops further to 10 % for μ = 0.5 and a few percents
which has been also discussed by Thornton [23]. The appar- for μ = 1.0. This relationship between the sliding fraction
ent discrepancy is largely attributed to the particle rolling and inter-particle friction angle emphasizes that the micro-
effect, which can not be simply ignored, which is consis- scale particle sliding that is governed by the inter-particle
tent with Peña et al. [22] and Alonso-Marroquín et al. [44], friction angle is closely relevant to the macroscopic behavior
in which they used circular and polygonal particles respec- of granular soils shown previously. It is also seen that for both
tively for the discrete element simulations. The independence drained and undrained tests, the development of the sliding
of macroscopic shear resistance on inter-particle friction was fraction is closely related to the volumetric behavior of the
explained by formation of rotational patterns in the particu- samples, see Figs. 5, 6: the decrease of sliding fraction with
late system [44]. Extremely, if inter-particle friction μ = 0, εa or q  / p is co-occurred with the volume contraction (cases
the shear resistance is still available, which suggests that the μ = 0.05 and 0.1), while the increase of sliding fraction is
inter-particle friction is not only the cause of macroscopic echoed when the volume is dilatant but decrease when it is
friction, but also the interlocking and particle structures, see contracted (cases μ = 0.2, 0.5 and 1.0).
[12,22,28,45]. For granular materials, the externally applied loading is
transmitted by the force network via contacts of particles.
The evolution of the microstructure engendered by loading
6 Microscale observations can be characterized by the contact unit normal, branch vec-
tor as shown in Fig. 16 in the Appendix and Coordination
To examine potential microscopic mechanism underlying the Number (CN), defined as the mean contact numbers per par-
mechanical response of granular soils described above, the ticle for a granular assembly. Evaluation of the microstruc-
‘sliding fraction’, defined as the ratio of slipping contacts to ture quantities and how they evolve under loading are given

123
On the influence of inter-particle friction and dilatancy 441

(a) 60 (b)1.5
μ=0.05
μ=0.1 μ=0.05
μ=0.2 1.2 μ=0.1
45 μ=0.2
μ=0.5
μ=1.0 μ=0.5
sliding fraction (%)

0.9 μ=1.0
30

'
q/p
0.6

15
0.3

0.0
0
0 5 10 15 20 25 30 0 15 30 45 60
ε a (%) sliding fraction (%)

(c) 60 (d)1.5
μ=0.05
μ=0.05
μ=0.1
1.2 μ=0.1
45 μ=0.2
μ=0.2
sliding fraction (%)

μ=0.5
μ=0.5
μ=1.0 0.9 μ=1.0
30
'
q/p

0.6

15
0.3

0 0.0
0 5 10 15 20 25 30
0 15 30 45 60
ε a (%) sliding fraction (%)

Fig. 10 Sliding contact fraction over various inter-particle frictions. a, b Drained tests; c, d undrained tests

in the Appendix. The internal force chains can be depicted by gests that the lower friction may produce specimen with a
contact forces (both normal and tangential), mobilized fric- larger CN, while the higher friction tends to give smaller CN
tion coefficient, defined as the ratio of the local contact tan- by rearranging the particles via sliding and rolling. In gen-
gential force to the normal counterpart, and mean mobilized eral, CN is increased with decreasing inter-particle friction,
friction coefficient by averaging over all contacts among the as particle sliding is eased and particles packed more densely
assembly of particles. (see Fig. 10). The evolution of the CN values is closely related
Figure 11a presents the mean mobilized friction coeffi- to the volumetric behavior of the specimens: CN increases
cient for various values of inter-particle friction under drained initially with the axial strain up to a certain axial strain (gen-
condition. It can be seen that the mean mobilized friction erally less than 2 %) for all μ values, which is corresponding
coefficients increase initially with the axial strain and then to the volume contraction of the specimens, and then flat-
flatten off up to a certain axial strain (generally less than ten off for μ ≤ 0.2 but drop off nearly to a constant for
2 %) for all μ values. However, the magnitudes of the mean μ ≥ 0.5 when the specimens dilate, although the specimens
mobilized friction coefficient are dependent on the inter-par- have similar initial void ratio of 0.15–0.17.
ticle frictions. It shows that the maximum mean mobilized Figure 12 shows the evolution of anisotropy for these
friction coefficients are 0.043 for μ = 0.05 and 0.083 for contact parameters during the shear loading. In general, the
μ = 0.1, which are more than 80 % of the assigned inter- contact property evolves in a different manner as the inter-
particle frictions, but they reduce to only 62 % for μ = 0.5 particle friction varies. As shown in Fig. 12a, in general,
(maxima = 0.309) and 44 % for μ = 1.0 (maxima = 0.443), for cases with higher μ show greater anisotropy intensity
respectively. d , and the increase of the shear strain will enhance the
Shown in Fig. 11b is the evolution of CN for various values anisotropy intensity . The anisotropy intensity may tend
of inter-particle friction under drained condition, and sug- to flatten off at large strain for cases with higher μ, except

123
442 Z. X. Yang et al.

(a) 0.8 At initial state, both the contact normal and contact nor-
μ=0.05 mal force show circular distributions while tangential force
μ=0.1
mean moblized friction coefficient

and mobilized friction coefficient are very small under iso-

0.6 μ=0.2
μ=0.5 tropic stress state. The contact force chain formed by small
μ=1.0 and nearly circular grids has no apparent orientation, indi-
0.4 cating that the sample is isotropic, which is also justified by
the distributions of above four micro-scale variables.
At peak state, the principal directions of the contact param-
0.2 eters unanimously orient vertically, in response to the vertical
deviatoric loading. There are a few sparse strong force chains
in vertical direction along with dense vertical oriented grids.
0.0
0 5 10 15 20 25 30 By comparison with those of initial state, at the peak state,
εa (%) more contacts are formed in the direction of major principal
stress (vertical), while less contacts are in the minor principal
(b) 6 stress direction (horizontal), making the contact unit normal
distribution from being circular to elliptical vertically.
At ultimate state, the general distributions of contract
parameters are similar to those at peak state. However, as
5 shown in Figs. 5b, 12a, for μ = 0.5, the behavior is strain
softening after the peak state, while the evolution of con-
CN

μ=0.05 μ=0.5 tact normal anisotropy intensity d is increasing with the

μ=0.1 μ=1.0
shear strain until ultimate state, such that the distribution of
4 μ=0.2
contact normal at the ultimate state is even more anisotropic
than that at the peak state. Generally the stress ratio q/ p 
depends on the contact normal and contact force distribu-
3 tions in the micro-scale. If comparing the stress ratio distri-
0 5 10 15 20 25 30
bution (Fig. 5b) with the contact force anisotropy (Fig. 12b,
εa (%)
c), it is found that the stress ratio is more related to the con-
Fig. 11 Mobilized friction coefficient and coordination number with tact force rather than contact unit normal. This observation
various inter-particle frictions for drained tests. a Mean mobilized fric- is also supported by the DEM simulation results of Li [34].
tion coefficient; b CN It is also noted that the force chain formation is similar to
the experimental results, in which the growth of force net-
for case μ = 1.0, which drops off and fluctuates after the work by column-like structures are observed in the direction
peak. The numerical results are in accordance with exist- paralleled to the major principal stress direction, carrying and
ing experimental observations [46,47]: the concentration of transmitting the deviatoric load [42,48,49].
the contact unit normals towards the major principal stress For different inter-particle frictions, the contact unit nor-
direction. Figure 12b, c give the evolutions of anisotropy mal distributions at various states are quite similar, but sig-
intensities nd and td for the contact normal and tangen- nificant discrepancies are noted in the distributions of the
tial forces respectively. As shown in the figure, except for contact forces, especially for the tangential force. Figure 14
μ = 1.0, higher μ will give rise to higher nd and td , which presents the tangential force distribution at ultimate stage for
are also increasing with shear strain εa before flattening off μ = 0.05, μ = 0.2 and μ = 1.0. It is seen that the magni-
in the large strain. It is also evident to see that higher μ tends tude of the contact tangential force is highly dependent on the
to augment the anisotropic distributions of both the contact inter-particle frictions, although their distribution patterns are
normal and tangential forces, such that results in a higher similar. The major differences in the tangential forces may be
degree of induced structural anisotropy. A similar observa- the source of the different macroscopic behaviors including
tion is also found in Fig. 12d, where the mobilized friction the dilatancy characteristics for different inter-particle fric-
coefficient is computed as the ratio of local tangential force tion stated previously. Figure 15 shows the corresponding
to normal force. contact force chains. The finer force network is developed
As an example, Fig. 13 illustrates the evolutions of the for μ = 0.05, while much coarser networks accompanied by
microstructures including the contact unit normal, contact larger pores can be observed for cases with larger frictions
normal and tangential forces, mobilized friction coefficient, μ = 0.2 and 1.0. This is consistent with the observations
and also contact force network at three featuring stages, i.e. aforementioned that higher CN is for smaller inter-particle
initial, peak and ultimate states, for drained test with μ = 0.5. friction and vise versa.

123
On the influence of inter-particle friction and dilatancy 443

(a) 0.60 (b) 0.60

n
contact normal force anisotropy Δd
contact normal anisotropy Δd

0.45 0.45

0.30 0.30
μ=0.05
μ=0.1 μ=0.05
μ=0.2 μ=0.1
0.15 0.15
μ=0.5 μ=0.2
μ=1.0 μ=0.5
μ=1.0
0.00 0.00
0 5 10 15 20 25 30 0 5 10 15 20 25 30
ε a (%) εa (%)

(c) 0.60 (d) 0.60

μ=0.05
contact tangential force anisotropy Δd
t

μ=0.05

mobilized friction anisotropy Δd

μ=0.1

ϕ
μ=0.1
0.45 μ=0.2 0.45
μ=0.2
μ=0.5
μ=0.5
μ=1.0
μ=1.0
0.30 0.30

0.15 0.15

0.00 0.00
0 5 10 15 20 25 30 0 5 10 15 20 25 30
ε a (%) ε a (%)

Fig. 12 Evolution of contact parameters’ anisotropy with various inter-particle frictions for drained tests. a εa ∼ d , b εa ∼ nd , c εa ∼ td ,
φ
d εa ∼ d

7 Conclusions the famous Rowe’s stress-dilatancy formulation fails to

capture.
This paper presents a systematic DEM analysis of the gran- The striking overall behavior with varying inter-particle
ular assembly sheared biaxially under both drained and friction is interpreted by the interplay between the two forms
undrained conditions. The samples all have similar initial of kinematics of the particles, sliding and rolling in the micro-
isotropic fabric and density, and the same confining pressure, scale. The strengths of samples with various inter-particle
but with different inter-particle friction coefficient. The focus frictions appear to be influenced largely by the partitioning of
of this paper is placed on the interpretation of the mechani- deformation between inter-particle sliding and rolling, which
cal behavior observed in the macro-scale by the microscopic is also helpful to explain the distinct distributions of CN and
analysis. inter-particle forces. The anisotropic distribution of a num-
It is noted that the inter-particle friction significantly influ- ber of contact parameters including contact unit normal, con-
ences the mechanical behavior of the granular assembly, tact forces, and mobilized friction coefficient are identified
under both drained and undrained conditions. The increase of and analyzed. The result shows that the different evolution
inter-particle friction brings the stress-strain response of the patterns of contact anisotropy due to different inter-particle
samples from hardening to softening, and a peak stress-ratio frictions can be linked with and interpret the observed mac-
and strength are observed with higher frictions. However, roscopic behavior.
the ultimate state stress ratio and strength are not appreciably Results presented in this paper lead to a better under-
affected by the inter-particle friction, which is consistent with standing of the role of inter-particle friction on the macro-
previous investigations reported in the literature. Remark- and micro-mechanical behavior of the granular materials.
ably, it is noted that for a given stress ratio, the decrease of Noting that the limitation of the two-dimensional analy-
the inter-particle friction reduces the rate of dilation, which sis and ideal particle shape employed, validations with a

123
444 Z. X. Yang et al.

Initial state Peak state Ultimate state

90 90 90
30 30 30
120 60 120 60 120 60

Contact unit normal distribution (%)

Contact unit normal distribution (%)

Contact unit normal distribution (%)

20 20 20
150 30 150 30 150 30

10 10 10

Contact 0 180 0 0 180 0 0 180 0

normal 10 10 10

210 330 210 330 210 330

20 20 20

240 300 240 300 240 300

30 30 30
270 270 270

90
1600 90 90
120 60
1600 1600
120 60 120 60
1200
1200 1200
150 30
800
Contact normal force (N)

150 30 150 30
800 800

Contact normal force (N)

Contact normal force (N)

400
400 400
Contact 0 180 0
0 180 0 0 180 0

normal 400
400 400
800
210 330
800 800
force 1200
210 330 210 330

1200 1200
240 300
1600
240 300 240 300
270 1600 1600
270 270

90 90 90
300 300 300
120 60 120 60 120 60

225 225 225

Contact tangential force (N)

Contact tangential force (N)

Contact tangential force (N)

150 30 150 30 150 30

150 150 150

- + - +
Contact
75 75 75 - +
0 180 0 0 180 0 0 180 0

tangentia 75
+ - 75 75 + -
150 150
+ - 150
l force 210 330 210 330 210 330

225 225 225

240 300 240 300 240 300

300 300 300
270 270 270

90 90 90
0.30 0.30 0.30
120 60 120 60 120 60
0.25 0.25 0.25

0.20 0.20 0.20

Mobilized friction coefficient
Mobilized friction coefficient

Mobilized friction coefficient

150 30 150 30 150 30

0.15 0.15 0.15

Mobilize 0.10
- + 0.10
- + 0.10

0.05 0.05 0.05

- +
d friction 0.00

0.05
180 0 0.00

0.05
180 0 0.00

0.05
180 0

+ - + -
coefficie
0.10

0.15
0.10

0.15
+ - 0.10

0.15
210 330 210 330 210 330
0.20 0.20 0.20
nt 0.25 0.25 0.25
240 300 240 300 240 300
0.30 0.30 0.30
270 270 270

Contact
force
chain

Fig. 13 Contact parameters’ distribution for drained test with μ = 0.5

123
On the influence of inter-particle friction and dilatancy 445

90
(a) 8
120 60

4
150 30 (a)
Contact tangential force (N)

2 - +
0 180 0

2
+ -
4
210 330

240 300
8
270

(b)
90
(b) 80 120 60

60

150 30
40
Contact tangential force (N)

20
- +
0 180 0

20 + -
(c)
40
210 330

60

240 300
80
270

90
(c) 400 120 60

300
Fig. 15 Contact force network developed at ultimate state for drained
150 30
200 test. a μ = 0.05, b μ = 0.2, c μ = 1.0
Contact tangential force (N)

100 - +
0 180 0

100 + - three-dimensional model with more realistic particles would

be required, to the development of the constitutive models
200
210 330 addressing the influence of inter-particle friction in the light
300
of the microstructure findings.

240 300 Acknowledgments The financial support provided by the Research

400
270 Grants Council of Hong Kong (HKU 7191/05E) is acknowledged. The
first author is also grateful for support by Natural Science Foundation
Fig. 14 Contact tangential force developed at ultimate state for drained of China (No. 50808159 & No. 51178421) and Fundamental Research
test. a μ = 0.05, b μ = 0.2, c μ = 1.0 Funds for the Central Universities (Program No. 2010QNA4028).

123
446 Z. X. Yang et al.

Appendix: Interpretation of micro-quantities spatial directional distribution of quantity r can be expressed

into
Anisotropy is an important feature for granular materials,   
and can be characterized mathematically by fabric tensors. r (ϕ) = r0 1 + rd cos 2 ϕ − ϕdr (A3)
The fabric tensors can be quantified by using the data reg-
where r0 represents an isotropic distribution over the direc-
istered during the numerical simulations in the micro-scale.
tions and the pair of parameters rd and ϕdr are denoting the
Presently, the investigations are focused on the two-dimen-
fabric anisotropy intensity and principal direction, and can
sional conditions. Figure 16 illustrates the representations
be also obtained through the components of di j by using
for micro-quantities of two neighboring particles in contact.
Eq. (A2).
Referring to the work by Bathurst and Rothenburg [50], one
It is aware that the local tangential force could be posi-
may use the second-order form of Fourier series expansion to
tive or negative, reflecting the sense of rotation imparted on
approximate the directional distribution of those micro-quan-
the particle. The sign convention adopted here follows that
tities. For quantities only having information in orientation
positive tangential force tends to rotate a particle clockwise,
(no information in magnitude), such as particle orientation,
and negative one means counter-clockwise. In this regard, the
contact unit normal and branch vector orientation etc., the
contact normal and tangential force directional distributions
directional distribution in terms of the density function can
can be written as follows:
be written in the form:
  
E(ϕ) = E 0 (1 + di j n i n j ) = E 0 [1 + d cos 2(ϕ − ϕd )] f n (ϕ) = f 0 1 + nd cos 2 ϕ − ϕdn (A4)
 
(A1) f t (ϕ) = − f 0 d sin 2 ϕ − ϕd
t t
(A5)

where ϕ measures the orientation angle of an interested quan- where f 0 represents an isotropic contact normal force and
tity with respect to the horizontal axis, ranging from 0◦ to serves as a scaling factor for the tangential force distribution
360◦ , E 0 = 1/2π designates an isotropic distribution part, in Eq. (A5), and the subscripts/superscripts n and t denote
and d and ϕd are two parameters characterizing the inten- normal and tangential respectively. In the meanwhile, the
sity and principal direction of the anisotropic distribution of a mobilized inter-particle friction could be defined as the ratio
particular quantity, and can be calculated through the compo- between the local contact tangential and normal forces. Its
nents of di j , d11 and d12 (only two is sufficient in expressing sign convention follows that of the tangential force, and may
the tensor di j because of its deviatoric and symmetric prop- have positive or negative value as well, and approximately
erty), using the following relations: takes the same form as Eq. (A5) with subscript/superscript
 μ differentiated:
2 + d 2 and ϕ = 1/2 tan−1 (d /d ) 
d = d11 d 12 11 (A2) μ μ
12
f μ (ϕ) = −d sin 2 ϕ − ϕd (A6)
It is noted that Eq. (A1) gives the directional distribution of
the quantities only having orientational information, while As an effective numerical tool, DEM has the advantage of
Eq. (A2) designates the anisotropy intensity and principal providing the direct information of the internal variables or
direction of such a distribution. It is noted that for an isotro- quantities in the micro-scale, other than the conventional lab-
pic distribution, d vanishes. oratory testing. Eqs. (A1) and (A4)–(A6) can visually present
However, a quantity r may likely possess the information the spatial distributions of micro-quantities, and explicitly
of both the direction and magnitude at the same time, such give the deviations from the corresponding isotropic distri-
as contact force, branch vector, contact vector, etc. Similar to butions. Whenever a certain micro-quantity is isotropic, the
Eq. (A1), a Fourier series approximation characterizing the components of its deviatoric fabric tensor should vanish.

References
Contact point contact unit normal 1. Arthur, J.R.F., Menzies, B.K.: Inherent anisotropy in a sand.
contact normal force Géotechnique 22(1), 115–128 (1972)
Centroid CII contact tangential force
Centroid CI 2. Oda, M.: Initial fabrics and their relations to mechanical properties
contact vector
of granular materials. Soils Found. 12(1), 17–36 (1972)
branch vector
3. Bolton, M.D.: The strength and dilatancy of sands. Géotech-
nique 36(1), 65–78 (1986)
Particle I Particle II 4. Ishihara, K.: Liquefaction and follow failure during earth-
quakes. Géotechnique 43(3), 351–415 (1993)
Fig. 16 Micro-quantities representation for two neighboring particles 5. Yang, Z.X., Li, X.S., Yang, J.: Undrained anisotropy and rotational
in contact shear in granular soil. Géotechnique 57(4), 371–384 (2007)

123
On the influence of inter-particle friction and dilatancy 447

6. Yang, Z.X., Li, X.S., Yang, J.: Quantifying and modelling fab- 29. Cundall, P.A., Strack, O.D.L.: A discrete numerical model for gran-
ric anisotropy of granular soils. Géotechnique 58(4), 237–248 ular assemblies. Géotechnique 29(1), 47–65 (1979)
(2008) 30. Jiang, M.J., Konrad, J.M., Leroueil, S.: An efficient technique for
7. Cabalar, A.F., Clayton, C.R.I.: Some observations of the effects generating homogeneous specimens for DEM studies. Comput.
of pore fluids on the triaxial behaviour of a sand. Granul. Mat- Geotech. 30(7), 579–597 (2003)
ter 12(1), 87–95 (2010) 31. Feng, Y.T., Han, K., Owen, D.R.J.: Filling domains with disks: an
8. Santamarina, J.C., Cho, G.C.: Soil behaviour: the role of particle advancing front approach. Int. J. Numer. Methods Eng. 56(5), 699–
shape. In: Jardine, R.J., Potts, D.M., Higgins, K.G. (eds.) Advances 713 (2003)
in Geotechnical Engineering: The Skempton Conference, vol. 1, 32. Bagi, K.: An algorithm to generate random dense arrangements
pp. 604–617. Thomas Telford, London (2004) for discrete element simulations of granular assemblies. Granul.
9. Mitchell, J.K., Soga, K.: Fundamentals of soil behaviour, 3rd Matter 7(1), 31–43 (2005)
edn. Wiley, NJ (2005) 33. Cui, L., O’Sullivan, C.: Analysis of a triangulation based approach
10. Cavarretta, I.: The influence of particle characteristics on the engi- for specimen generation for discrete element simulations. Granul.
neering behaviour of granular materials. PhD Thesis, The Univer- Matter 5(3), 135–145 (2003)
sity of London (2009) 34. Li, X.: Micro-scale investigation on the quasi-static behavior of
11. Rowe, P.W.: The stress-dilatancy relation for static equilibrium granular material. PhD Thesis, The Hong Kong University of Sci-
of an assembly of particles in contact. Proc. R. Soc. Lond. Ser. ence and Technology (2006)
A 269(1339), 500–527 (1962) 35. Huang, Z.Y., Yang, Z.X., Wang, Z.Y.: Discrete element model-
12. Skinner, A.E.: A note on the influence of interparticle friction ing of sand behavior in a biaxial shear test. J. Zhejiang Univ. Sci.
on the shearing strength of a random assembly of spherical par- A 9(9), 1176–1183 (2008)
ticles. Géotechnique 19(1), 150–157 (1969) 36. Iwashita, K., Oda, M.: Rotational resistance at contacts in sim-
13. Haruyama, M.: Effects of surface roughness on the shear charac- ulation of shear band development by DEM. J. Eng. Mech.
teristics of granular materials. Soils Found. 9(4), 48–67 (1969) ASCE 124, 285–292 (1998)
14. Procter, D.C., Barton, R.R.: Measurements of the angle of inter- 37. Mahmood, Z., Iwashita, K.: Influence of inherent anisotropy on
particle friction. Géotechnique 24(4), 581–604 (1974) mechanical behavior of granular materials based on DEM sim-
15. Cavarretta, I., Coop, M., O’Sullivan, C.: The influence of particle ulations. Int. J. Numer. Anal. Methods Geomech. 34(8), 795–
characteristics on the behaviour of coarse grained soils. Géotech- 819 (2010)
nique 60(6), 413–423 (2010) 38. Fu, P., Dafalias, Y.F.: Study of anisotropy shear strength of granu-
16. Chan, L.C.Y., Page, N.W.: Particle fractal and load effects on inter- lar materials using DEM simulation. Int. J. Numer. Anal. Methods
nal friction in powders. Powder Technol. 90(3), 259–266 (1997) Geomech. 35(10), 1098–1126 (2011)
17. Cho, G.C., Dodds, J., Santamarina, J.C.: Particle shape effects 39. PFC2D User’s manual, Itasca Consulting group, Inc., Minneapolis,
on packing density, stiffness, and strength: natural and crushed USA (2005)
sands. J. Geotech. Geoenviron. Eng. 132(5), 591–602 (2006) 40. Ng, T.T.: Fabric evolution of ellipsoidal arrays with different par-
18. Guo, P., Su, X.: Shear strength, interparticle locking, and dilatancy ticle shapes. ASCE J. Eng. Mech. 127, 994–999 (2001)
of granular materials. Can. Geotech. J. 44, 579–591 (2007) 41. Shodja, H.M., Nezami, E.G.: A micromechanical study of rolling
19. Jensen, R.P., Bosscher, P.J., Plesha, M.E., Edil, T.B.: DEM sim- and sliding contacts in assemblies of oval granules. Int. J. Numer.
ulation of granular media-structure interface: effects of surface Anal. Methods Geomech. 27, 403–424 (2003)
roughness and particle shape. Int. J. Numer. Anal. Methods Geo- 42. Oda, M., Konishi, J., Nemat-Nasser, S.: Experimental microme-
mech. 23, 531–547 (1999) chanical evolution of strength of granular materials: effects of par-
20. Nouguier-Lehon, C., Cambou, B., Vincens, E.: Influnece of pa- ticle rolling. Mech. Mater. 1(4), 269–283 (1982)
ricle shape and angularity on the behavior of granular materi- 43. Horne, M.R.: The behaviour of an assembly of rotund, rigid, cohe-
als : a numerical analysis. Int. J. Numer. Anal. Methods Geo- sionless particles III. Proc. R. Soc. Lond. Ser. A 310(1500), 21–
mech. 27, 1207–1226 (2003) 34 (1969)
21. Peña, A.A., García-Rojo, G., Herrmann, H.J.: Influence of particle 44. Alonso-Marroquín, F., Vardoulakis, I., Herrmann, H.J., Weath-
shape on sheared dense granular media. Granul. Matter 9, 279– erley, D., Mora, P.: Effects of rolling on dissipation in fault
291 (2007) gouges. Phys. Rev. E 74, 031306 (2006)
22. Peña, A.A., Lizcano, A., Alonso-Marroquín, F., Herrmann, 45. Cambou, B.: From Global to Local Variables in Granular Materials,
H.J.: Biaxial test simulations using a packing of polygonal par- Thorton, Editor, Powders & Grains. pp. 73–86. Balkema, Rotter-
ticles. Int. J. Numer. Anal. Methods Geomech. 32, 143–160 (2008) dam (1993)
23. Thornton, C.: Numerical simulations of deviatoric shear deforma- 46. Drescher, A.: An experimental investigation of flow rule for gran-
tion of granular media. Géotechnique 50(1), 43–53 (2000) ular assemblies. Géotechnique 26(4), 49–65 (1976)
24. Powrie, W., Ni, Q., Harkness, R.M., Zhang, X.: Numerical 47. Oda, M., Konishi, J., Nemat-Nasser, S.: Some experimentally based
modelling of plane strain tests on sands using a particulate fundamental results on the mechanical behaviour of granular mate-
approach. Géotechnique 55(4), 297–306 (2005) rials. Géotechnique 30(4), 479–495 (1980)
25. Alonso-Marroquín, F., Luding, S., Herrmann, H.J., Vardoilakis, 48. Oda, M., Kazama, H.: Microstructure of shear bands and its rela-
I.: Role of anisotropy in the elastoplastic response of a polygonal tion to the mechanisms of dilatancy and failure of dense granular
packing. Phys. Rev. E 71, 051304 (2005) soils. Géotechnique 48(4), 465–481 (1998)
26. Suiker, A.S.J., Fleck, N.A.: Frictional collapse of granular assem- 49. Oda, M.: The mechanism of fabric changes during compressional
blies. J. Appl. Mech. Trans. ASME 71(3), 350–358 (2004) deformation of sand. Soils Found. 12(2), 1–18 (1972)
27. Liu, S., Matsuoka, H.: Microscopic interpretation on a stress-dilat- 50. Bathurst, R.J., Rothenburg, L.: Investigation of micromechanical
ancy relationship of granular materials. Soils Found. 43(3), 73– features of idealized granular assemblies using DEM. Eng. Com-
84 (2003) put. 9, 199–210 (1992)
28. Kruyt, N.P., Rothenburg, L.: Strength, dilatancy, energy and dissi-
pation in quasi-static deformation of granular materials. In: García-
Rojo, R., Herrmann, H.J., McNamara, S. (eds.) Powders and
Grains, pp. 251. A.A. Balkema, Rotterdam (2005)

123