Theoretical and Applied Fracture Mechanics 72 (2014) 37–49

Contents lists available at ScienceDirect

Theoretical and Applied Fracture Mechanics

journal homepage: www.elsevier.com/locate/tafmec

Discrete element modeling of shear band in granular materials

Xiaoqiang Gu, Maosong Huang ⇑, Jiangu Qian
Department of Geotechnical Engineering, Key Laboratory of Geotechnical and Underground Engineering of the Ministry of Education, Tongji University, Shanghai, China

a r t i c l e i n f o a b s t r a c t

Article history: Numerical biaxial tests using discrete element method (DEM) were carried out to study the shear band
Available online 30 June 2014 formation in granular materials and the microstructures of the specimen at different strain levels were
carefully examined. The effect of initial density and confining pressure on the characteristics of shear
Keywords: band were investigated. It was found that the progressive formation of the shear band was successfully
Discrete element captured by the DEM simulation with flexible lateral boundaries. During the shearing, several potential
Granular material shear bands are formed at the peak deviatoric shear stress. With further shearing, the potential shear
Biaxial test
bands enlarge and softening behavior happens. Finally, a persistent shear band forms and other potential
Progressive failure
Strain localization
shear bands deteriorate. Generally, loose state and high confining pressure ‘‘delay’’ the onset of shear
Shear band banding. The simulation also showed that the shear band thickness decreases with increasing confining
pressure and density. Meanwhile, the inclination angle of shear band in dense specimen is generally
higher than that in loose specimen. The void ratio of shear band is significantly larger than other parts
of the specimen. At microscopic level, the overall coordination number of the specimen decreases signif-
icantly. The coordination number in the shear band is significantly lower than that outside the shear
band, which is consistent with the large porosity in the shear band. The formation of shear band is
due to the bulking of lateral confining force chains and the shear band is characterized by large void ratio,
low coordination number, and high particle sliding and rotation compared with other part of the
specimen.
Ó 2014 Elsevier Ltd. All rights reserved.

1. Introduction is equivalent to the happening of non-uniqueness of equilibrium

equations for a given constitutive model. Such theories have suc-
It is well recognized that many geotechnical failures are attrib- cessfully predicted the occurrence of shear band based on conven-
uted to the localized failure of soils which is characterized by the tional elasto-plastic constitutive models [30,36,37,15,26,27] and
appearance of shear band. Generally, the soil behaviors within Cosserat continuum theory [38]. Nevertheless, the theoretical
the shear band control the strength and deformation properties method is only applicable to solve simple boundary-value
of the earth structures [9], suggesting the importance of accurately problems, rather than complex problems [18]. Meanwhile, it is
predicting the shear band formation in geotechnical problem. In generally unable to properly predict the thickness and inclination
the past decades, tremendous theoretical, experimental and of the shear band as well as the mechanical behavior after the
numerical studies have been carried out to predict the occurrence occurrence of shear band. For example, Han and Drescher [15] pre-
and explore the fundamental mechanism of localized failure (i.e. dicted that the inclination angle increases with confining pressure
shear band), especially in granular soil. using bifurcation theory, which is opposite to the observation in
Theoretically, the formation of localized failure can be regarded their laboratory tests.
as a type of instability resulted from the bifurcation of uniformly Besides the theoretical study, numerous laboratory tests have
deformed soil [30]. Therefore, the initiation of shear band strongly performed to investigate the shear band of granular soils (e.g.,
depends on the constitutive description of the prelocalization [3,8,33,15,5,20,39,1]. These experimental works have revealed that
homogeneous deformation. The shear band is treated as a ‘‘thin’’ the shear band formation and its characteristics are influenced by a
layer along which two rigid blocks move at inconsistent velocity number of factors, including the density, the confining pressure,
(or strain rate). Mathematically, the initiation of the shear band inherent and stress-induced anisotropy of the material, and the
particle size and shape of the material. With the help of advanced
⇑ Corresponding author. Tel.: +86 21 65983980; fax: +86 21 65985210. technologies, extensive laboratory experiments have also been
E-mail address: [email protected] (M. Huang). performed to look insight into the shear band formation and

http://dx.doi.org/10.1016/j.tafmec.2014.06.008
0167-8442/Ó 2014 Elsevier Ltd. All rights reserved.
38 X. Gu et al. / Theoretical and Applied Fracture Mechanics 72 (2014) 37–49

evolution. These technologies include X-ray Computed Tomogra- 100

phy [14], stereophotogrammetry [8,9], photoelasticity [25], and

Percentage Passing (%)

particle image velocimetry (PIV) [42]. For example, Desrues et al. 80
[8] used stereophotogrammetry to study the shear band in dry
sand. The results indicated that the initiation of shear band took 60
place before the peak in the overall stress–strain curve. Meanwhile,
the shear band initiated at certain points and progressively propa- a
40
gated to a final persistent shear band. Han and Drescher [15] per-
formed drained biaxial tests on poorly-graded Ottawa sand and
20 b
found that the shear strain at shear banding increases with increas-
ing confining pressure. Meanwhile, the shear band thickness and Aspect ratio b/a = 1.5
the shear band inclination angle with respect to the minor princi- 0
0.50 0.60 0.70 0.80 0.90 1.00
ple stress direction decrease as the confining pressure increases.
Particle Size (mm)
Tatsuoka et al. [32] also indicated that the thickness of shear band
decreases as confining pressure and initial density increases. Desr- Fig. 2. Particle size distributions of soil in DEM simulation.
ues and Viggiani [9] summarized the experimental results of strain
localization in sand performed in Grenoble and the influence of soil
initial state (i.e. confining pressure and density), specimen size and
slenderness, grain size and particle size distribution are discussed.
However, these experimental methods are extremely difficult, 400
tedious and generally impossible to continuously provide all the
important aspects of shear band that concerned (e.g. void ratio,
300
contact force and particle rotation). Most importantly, the funda-
mental mechanism related to the shear band formation and how
q (kPa)
the influencing factors affect the characteristics of shear band is 200
not well understood.
From the viewpoint of micromechanics, the granular soil is an
assemblage of discrete particles and the soil behavior depends on 100
the particle interactions through contacts that bear and transfer
the loads. Therefore, the contact-based microstructure of the gran- 0
ular soil, such as the contact number, contact force and their distri- 0 100 200 300 400 500 600
butions, is expected to play an important role in determining the P (kPa)
macroscopic behavior. The discrete element method (DEM) pro- (a)
posed by Cundall and Strack [6] provides a powerful tool for sim-
ulating macroscopic behaviors and analyzing the microstructure 400
C
of granular soil. DEM captures the particulate nature of granular E
soil and it has been extensively used to investigate the macro- G
D
scopic behavior at micromechanical (or particulate) level, includ- 300
F
ing small strain stiffness [7,40,10,11], monotonic shear behaviors B
q (kPa)

I H
[34,12], soil liquefaction [31], and critical state soil mechanics 200
[13,12]. It also has been employed by several researchers to inves-
tigate the strain localization or shear band in granular materials
A
[4,17,41,18,21]. For example, Bardet and Proubet [4] performed 100
DEM simulation to investigate the structure of shear band and
found that the thickness of shear band decreases with axial strain
0
0.0 2.0 4.0 6.0 8.0 10.0
(%)
σ1 (b)

0.5 B
Top cap
(Wall) C

0.0

0.0 A 2.0 4.0 6.0 8.0 10.0

ε v (%)

D
-0.5
σ3 I
H
G
-1.0
Membrane
particle

-1.5
(%)
Soil particle (c)
(Clump)
Fig. 3. Macroscopic behaviors of the dense specimen with porosity n0 = 0.112 at
Fig. 1. Specimen in the DEM simulation. p0 = 200 kPa: (a) stress path; (b) deviatoric stress; (c) volumetric strain;
 X. Gu et al. / Theoretical and Applied Fracture Mechanics 72 (2014) 37–49 39

Table 1 from 9 to 7.5 times of the average particle diameter. They also out-
Parameters in simulation. lined the importance of particle rotation in the shear band. How-
Particle density 2650 kg/m3 No. of clumps 12431 ever, they used circular particles in the simulation which have no
Interparticle friction 0.5 Clump aspect ratio 1.5 rolling resistance at contacts. Iwashita and Oda [17] incorporated
Wall stiffness 1  107 kN/m Wall-particle friction 0 rolling resistance at contact into DEM and found that the develop-
Particle normal Variablea Particle shear kt = kn
stiffness kn stiffness kt
ment of shear bands can be simulated well only when the rolling
Membrane bonding 1.0  10300 N Local damping factor 0.7 resistance at contacts was considered in DEM. They also found gen-
strength eration of large voids and particle rotation resulted from bulking of
a
The particle normal stiffness kn = k0  r, where k0 = 1  106 kN/m2 and r is the
force chains. Jiang et al. [18] performed numerical biaxial tests on
radius of the particle. circular disks and found that the thickness of shear band is 10–14
times of the mean particle diameter and it increases with
increasing axial strain, which is contrary to the results by Bardet

Fig. 4. Distribution of particle translational velocity vectors at different shear stains: (a) 0.2%; (b) 0.6%; (c) 1.2%; (d) 1.8%; (e) 2.2%; (f) 2.4%; (g) 3.2%; (h) 10%.
40 X. Gu et al. / Theoretical and Applied Fracture Mechanics 72 (2014) 37–49

7.0
6.0
5.0
4.0
3.0
2.0
1.0
0.50
0.10

(a) (b) (c) (d)

7.0
6.0
5.0
4.0
3.0
2.0
1.0
0.50
0.10

(e) (f) (g) (h)

Fig. 5. Distribution of particle rotational velocity in the specimen at different shear stains: (a) 0.2%; (b) 0.6%; (c) 1.2%; (d) 1.8%; (e) 2.2%; (f) 2.4%; (g) 3.2%; (h) 10%.

and Proubet [4]. They also found the inclination of shear band are analyzed and attempts are made to link the shear banding with
decreases with increasing axial strain, and its value ranges from the evolutions of microstructure.
54° to 50°, which can be overall predicted by the Mohr–Coulomb
theory. However, the effect of initial state (i.e. effective confining 2. DEM modeling
pressure and soil density) has not been investigated. Moreover,
the progressive formation of shear band has not been clearly The well recognized DEM program PFC2D [16] is used to per-
described. The overall microstructure of the granular soil and its form the simulation. In PFC, the soil grains are modeled by discrete
distribution inside and outside the shear at different strain levels rigid spherical or disk-shaped particles that interact only at con-
have not been carefully examined in the previous DEM tacts. The contact forces are determined by a force–displacement
simulations. Most importantly, how the initial state affects the relationship which is usually termed as contact law. Meanwhile,
microstructure and how the microstructure evolutes inside and the movement of each particle is determined by the Newton’s sec-
outside the shear band have not been well understood. ond law of motion based on the unbalanced forces on each particle.
In this study, drained biaxial tests are simulated by DEM on PFC uses an explicit time-stepping algorithm that requires the
specimens with different initial confining pressures and densities. repeated application of the Newton’s second law of motion for
Non-circular particles are used to account for the rolling resistance updating the position of particle and the contact law for calculating
of real particles. The macroscopic stress–strain response of the contact forces. In PFC, the contact law contains several models
specimen and the progressive formation and evolution of the shear including the stiffness model, the slip model and bonding model
band are carefully investigated. The effects of initial confining pres- to describe the different types of interaction between particle
sure and density on the occurrence of shear band, the thickness contacts. The stiffness model relates the contact forces and the rel-
and inclination of shear band are also examined. Various kinds of ative displacement at the contact in the normal and tangential
microstructure of the specimen inside and outside the shear band directions. Linear model and non-linear Hertz–Mindlin model are
 X. Gu et al. / Theoretical and Applied Fracture Mechanics 72 (2014) 37–49 41

0.130

0.125

0.120

0.115

0.110

0.105

(a) (b) (c) (d)

0.140
0.135
0.130
0.125
0.120
0.115
0.110
0.105

(e) (f) (g) (h)

Fig. 6. Distribution of porosity in the specimen at different shear stains: (a) 0.2%; (b) 0.6%; (c) 1.2%; (d) 1.8%; (e) 2.2%; (f) 2.4%; (g) 3.2%; (h) 10%.

available in PFC. The slip model controls the occurrence of slip at represented by a clump of two spheres with an aspect ratio of
the contact. In PFC, the Coulomb friction law is adopted as the slip 1.5. The clump cannot break throughout the simulation and behav-
model. If the tangential force T is larger than the available maxi- iors as a single grain. Linear contact law is used and the normal
mum friction force lN determined by the friction coefficient l contact stiffness of the particle kn varies according to kn = k0  r,
and normal contact force N, slip occurs at the contact. The bonding where k0 = 105 kN/m2 and r is the radius of the particle [43,12].
model is used to describe the cohesion or the cementation effect at Yan and Dong [43] showed that it is essential to account for the
the contact. Furthermore, artificial damping can be introduced to size-dependent contact stiffness to ensure size-independent mac-
dissipate energy and achieve a steady-state solution. The detail roscopic soil stiffness. The tangential contact stiffness kt is assumed
of DEM theory and the characteristics of PFC can be found in the to be equal to kn. The particle size distribution of the numerical
PFC manual. specimen and the schematic of a clump (i.e. soil grain) are shown
Here, the laboratory biaxial test in plain strain condition by Han in Fig. 2. The nominal mean particle size D50 is 0.72 mm, exactly
and Drescher [15] is simulated by PFC. The specimen is represented the same as the Ottawa sand used in the laboratory test, and the
by a rectangular space of 140 mm height and 40 mm width (Fig. 1). coefficient of uniformity of the soil is 1.13. The top cap and the
To account for the non-spherical shape, the soil grain is pedestal are represented by a pair of frictionless horizontal walls
42 X. Gu et al. / Theoretical and Applied Fracture Mechanics 72 (2014) 37–49

and the membrane is simulated by a flexible boundary with 0.8

bonded particles. The bonded membrane particles can rotate freely
at the contact but cannot break. Table 1 lists the basic parameters
0.6
used in the simulation.
The specimen is prepared by radius expansion method. First,

q/p
12,431 balls with 50% of their final sizes are randomly generated 0.4
in the space and then expanded to their final sizes. Note that the
interparticle friction coefficient l is temporarily set to zero during 0.2 100 kPa 200 kPa
the expansion. The generated specimen is cycled to equilibrium
state and then each ball is replaced by a clump with identical area
400 kPa 800 kPa
(see Fig. 2). Then the specimen is brought to an isotropic stress of 0.0
0.0 2.0 4.0 6.0 8.0 10.0
10 kPa by servo control of the walls. During this stage, different (%)
interparticle friction coefficients are used to generate specimens (a)
with different densities [23,43]. Generally, low l value generates
dense specimen, while high l value generates loose specimen. 1.0
Upon finishing this stage, l is changed to 0.5 and the specimen 100 kPa 200 kPa
is cycled again to equilibrium state. Now, the specimen is at its ini- 0.5 400 kPa 800 kPa
tial state and is ready for isotropic consolidation and monotonic
shearing. 0.0
Conventional drained biaxial tests are simulated in this study.

ε v (%)
0.0 2.0 4.0 6.0 8.0 10.0
In the test, the strain-controlled top and bottom walls move close -0.5
to compress the specimen. The stresses on the lateral boundaries
are kept constant by applying forces on the center of each mem- -1.0
brane particle [41]. Note that the applied force on each membrane
-1.5
particle updates from step to step. It is found that flexible lateral
boundary is essential for DEM simulation to capture the character-
-2.0
istics of shear band. To ensure quasi-static conditions, the nominal (%)
axial strain rate is set to be 0.2/s in the simulation and the axial
(b)
strain applied in one calculation step is around 0.8 * 107 based
on the time step of 4.0 * 107 s. Meanwhile, the ratio of the average Fig. 7. Effect of initial confining pressure on the (a) stress ratio and (b) volumetric
unbalanced force to the average normal contact force in the simu- strain.
lation is monitored during the shearing. If this ratio is larger than
0.01, the top and bottom walls will stop moving temporarily and
are 0.2%, 0.6%, 1.2%, 1.8%, 2.2%, 2.4%, 3.2%, 10% and 4.8%,
additional calculation steps are carried out to reach quasi-static
equilibrium. Andrade et al. [2] showed that the critical state respectively.
To illustrate the progressive nature of shear band formation,
strength of the soil can significantly increase as the strain rate
increases due to the dynamic effect. Parametric studies are per- Fig. 4 compares the distribution of particle translational velocity
vectors at different characteristic points (or shear strain levels).
formed in this study and it is found that the quasi-static equilib-
rium condition is generally satisfied with the used nominal axial As seen in Fig. 4, the distribution of particle velocity vectors at
point A is quite uniform, indicating the specimen deforms uni-
strain rate of 0.2/s.
formly. At point B of maximum volumetric contraction, the distri-
bution of particle velocity vector becomes non-uniform. It seems
3. Results and discussions that two potential shear bands form at the bottom of the specimen.
Note that point B is well before the peak, indicating the specimen is
3.1. Macroscopic behaviors still in the hardening state although it starts to dilate (or becomes
looser). With further increase of shear strain, more potential shear
Followed the conventional biaxial tests, the mean effective bands form and become more obvious. At point D of peak deviator-
stress P and the deviatoric stress q are defined as ic stress, it can be seen in Fig. 4(d) that many potential shear bands
generate, especially in the top part of the specimen. After the peak,
P ¼ ðr1 þ r3 Þ=2 ð1Þ
most of the potential shear bands disappear and two major shear

q ¼ ðr1  r3 Þ=2 ð2Þ

0.8
where r1 and r3 are the axial stress and radial stress, respectively.
Fig. 3 shows the macroscopic shear behaviors of the specimen
with an initial porosity of 0.112 under an initial confining pressure 0.6
p0 of 200 kPa. As expected in a drained biaxial test, the stress path
is a straight line with a slope of 1.0 (Fig. 3a). The deviatoric stress q
q/p

0.4
increases with increasing shear strain c to a peak and then
decreases to a stable value (i.e. critical state). It agrees well with
the softening behavior observed in experimental biaxial test [15], 0.2 200 kPa-loose
which is believed to be caused by the formation of shear band.
200 kPa-dense
The specimen contracts first (positive volumetric strain ev) and
then dilates continuously, which means that the specimen is ini- 0.0
0.0 2.0 4.0 6.0 8.0 10.0
tially dense. For better illustration, nine characteristics points A-I
(%)
are marked on the stress–strain curve to illustrate the progressive
formation of the shear band. The shear strain levels of points A–I Fig. 8. Comparison of the stress ratio curves of the dense and loose specimens.
 X. Gu et al. / Theoretical and Applied Fracture Mechanics 72 (2014) 37–49 43

0.160
0.155
0.150
0.145
0.140
0.135
0.130
0.125
0.120
0.115

100 kPa 200 kPa 400 kPa 800 kPa

(a)

0.220
0.215
0.210
0.200
0.190
0.180
0.170
0.160
0.150
0.140
0.120

100 kPa 200 kPa 400 kPa 800 kPa

(b)
Fig. 9. Distribution of porosity in the specimen at 10% shear strain: (a) dense; (b) loose.

bands forms with an ‘‘X’’ configuration. The major shear bands are attempts to generate potential shear bands to resistant the exter-
unsymmetrical. Moreover, it seems that the right shear band is nal shear load. At the peak deviatoric stress, the specimen gener-
more dominate as its velocity discontinuity is more significant. ates more than 6 potential shear bands. It should be emphasized
As shear strain further increases, however, the left shear band that these potential shear bands are not persistent. As shear strain
becomes dominant and the right shear band disappears at point further increases, the potential shear bands disappear and major
G. After point G, the position of the shear band remains the same shear bands form, while the overall macroscopic stress ratio starts
and the width of shear band seems increase with further shearing. to decrease. It indicates that the potential shear bands at the peak
Fig. 4 convincingly illustrates the progressive evolution of shear are saturated and the specimen cannot generate additional ones to
band in granular materials. resist the eternal shear load. With further shearing, the potential
The macroscopic behavior in Fig. 3 and the progressive shear shear bands changes into major ones and finally a persistent shear
band formation in Fig. 4 suggest that the specimen still shows a band. Meanwhile, more particles are involved in the shear band,
hardening behavior although the particle translational velocity resulting in the increase of shear band thickness. Han and Drescher
vectors indicate that potential shear bands form. The deviatoric [15] reported that the thickness at the strain corresponding to the
stress corresponding to the maximum volumetric strain is well start point of the residual point (e.g. point I in Fig. 3b) is only
below the peak value. With further shearing, the specimen start around 60% of that at very large deformation.
to dilate (i.e. the overall void ratio starts to increases), but the spec- Fig. 4(h) also indicates that the particles in the shear bands
imen still shows a hardening behavior. It seems that the specimen rotate, while the particles outside the shear band behaviors like a
44 X. Gu et al. / Theoretical and Applied Fracture Mechanics 72 (2014) 37–49

rigid block moving along the shear band. It is consistent with the 30
observation that particle rotation plays an important role in the Dense
formation of shear band [24,4]. For better illustration, Fig. 5 shows 25
Loose
the particle rotational velocity at different shear strain levels. Sim-
ilar to the distribution of particle translational velocity, the distri- 20

T SB /D 50
bution of rotational velocity is somewhat uniform at the beginning
of shearing. Then, potential shear bands turn out and they are sat- 15
urated at the peak stress. With further shearing, the distribution of
particle rotational velocity is localized in two major shear band and 10
finally evolutes into the persistent shear band. Fig. 5 clearly illus-
5
trates that the movement of particles is rotational in the shear
band, but translational outside the shear band. Tordesillas and
0
Muthuswamy [35] indicated that particle rotates due to the 0 200 400 600 800 1000
bulking of force chains and results in the formation of shear band. Confining pressure (kPa)
It is consistent with the DEM simulation.
(a)
In laboratory, instead of the particle velocity, the distribution of
void ratio (or porosity) is usually used to monitor the shear band 70
formation by advance technologies. It has been found that the void Dense
ratio in the shear band is significantly larger than that outside of

Inclination angle (deg.)

65 Loose
the shear band and the void ratio in the shear band may be much
larger than the maximum void ratio determined in the laboratory.
60
In this study, the void ratio distribution of the specimen is deter-
mined by 1200 measurement circles with diameter of 10 times
55
of D50. Fig. 6 shows the porosity contour of the specimen at differ-
ent characteristic points. It is clear that the density of the specimen
is generally uniform at the initial state. As the shear strain applies, 50
the particles slide and rotate at the contacts, resulted in particle
arrangement and deformation. It can be seen in Fig. 6(a–d) that 45
0 200 400 600 800 1000
the porosity distribution does not change significantly before the Confining pressure (kPa)
peak deviatoric stress. At the peak, several zones dilate obviously
and seem to form several potential shear bands indicated by lines. (b)
With further shearing, these zones continuously enlarge and con- Fig. 10. Effect of initial state on the (a) thickness and (b) inclination angle of shear
nected to each other. It seems that an ‘‘intact’’ shear band forms band.
at point G. The thickness of the shear band increases as the shear
strain further increases. Meanwhile, it is interesting to note that
the dilation zone outside the intact shear band contracts. The evo-
lution of distribution of particle velocity and porosity indicates that 7.0
the major shear band forms at the peak shear stress, resulted in the
reduction of shear strength. It is consistent with the theoretical A
6.0
prediction. However, the shear band based on the porosity may
be somewhat ‘‘delayed’’ since the large void ratio is resulted from B
the accumulation of particle sliding and rotation.
CN

5.0
It is well recognized that the thickness and the inclination angle E G I H
are two important aspects of a shear band. Note that the thickness C
D
and the inclination angle of a persistent shear band generally may 4.0
evolve throughout a test. Hence, it is important to clearly define at F
which point of the test are they measured [9]. Based on Fig. 6h, the
thickness and the inclination angle to the horizontal direction of 3.0
0.0 2.0 4.0 6.0 8.0 10.0
the shear band at 10% shear strain are determined to be 13.9D50
εs (%)
and 59.2°, respectively. Han and Drescher [15] reported that the
measured thickness of the shear band at large shear strain at Fig. 11. Evolution of the coordination number of the dense specimen confined at
200 kPa is 10 mm, which exactly equals 13.9D50. The inclination 200 kPa.
angle in the simulation is slightly higher than the measurement
value 56° by Han and Drescher [15].
To investigate the effect of initial confining pressure on the Generally, the contraction increases with initial confining pressure
shear banding, biaxial tests are also performed on the specimen as expected by the critical state soil mechanics.
at confining pressures of 100, 400 and 800 kPa, besides at Simulation was also carried out on a loose specimen with initial
200 kPa. Fig. 7 compares the results at different confining pres- porosity of 0.127 under a confining pressure of 200 kPa. Fig. 8 com-
sures. As seen in Fig. 7, the stress ratio increases more sharply at pares the stress ratio-shear strain curves of the loose and dense
low confining pressures. Meanwhile, the shear strain level corre- specimens. It is clear that the peak stress ratio of the dense speci-
sponding to the peak increases as the initial confining pressure men is much higher than that of the loose one, indicating the dense
increases, but the peak and residual stress ratio are almost the specimen is more dilative. However, they reach the same stress
same at different confining pressures. It is consistent with the ratio at very large shear strains as expected by the critical state soil
observation by Desrues and Viggiani [9] that for a given density mechanics. Meanwhile, the shear strain corresponding to the peak
high initial confining pressure ‘‘delays’’ the appearance of shear of the loose specimen is larger than that of the dense one, which
band. Meanwhile, all the specimens contract first and then dilate. means the formation of shear band is later in the loose specimen.
 X. Gu et al. / Theoretical and Applied Fracture Mechanics 72 (2014) 37–49 45

6.00
5.70
5.40
5.00
4.70
4.40
4.25
4.10
3.80
3.50
3.20
3.00
2.80

(a) (b) (c) (d)

6.0
5.7
5.4
5.0
4.7
4.4
4.1
3.8
3.5
3.2
3.0
2.8

(e) (f) (g) (h)

Fig. 12. Distribution of coordination number in the dense specimen at different shear stains: (a) 0.2%; (b) 0.6%; (c) 1.2%; (d) 1.8%; (e) 2.2%; (f) 2.4%; (g) 3.2%; (h) 10%.

The distribution of the porosity in the loose specimen showed that band forms at the bottom of the specimen at 100–400 kPa. The
shear banding does occur at the bottom of the specimen (Fig. 9b). shear band intersects and reflects on the bottom wall. On the con-
It is worth noting that the characteristics of the shear band trary, the shear band forms at the center part of the specimen at
depend on the initial state of soil. Fig. 9 shows the distributions 800 kPa. Compared with the dense specimen, there is no minor
of porosity of the loose and dense specimens under different con- shear band formed in the loose specimen. Moreover, it is interest-
fining pressures at a shear strain of 10%. As seen in Fig. 9(a), for the ing to note that the thickness of the shear band decreases with
dense specimen, the shear band at 100 kPa is a ‘‘conjugate’’ type at increasing confining pressure.
the top part of the specimen. When the initial confining pressure Based on Fig. 9, the thickness and the inclination angle of the
increases to 200–800 kPa, the major shear band occurs at the cen- shear band at different initial states are determined, as shown in
ter part of the specimen. At the confining pressure of 200 kPa, it Fig. 10. Note that the results are based on the shear strain of
seems that there is a minor shear band at the top part of the spec- 10%. As seen in Fig. 10(a), the shear band thickness TSB is normal-
imen and forms an X-shaped pattern with the major one. However, ized by the mean particle diameter D50. It is of interest to note that
the minor shear band is parallel to the major one at the confining TSB/D50 decreases with increasing confining pressures for both
pressure of 400 kPa. Nevertheless, the minor shear band at 800 kPa dense and loose specimens. Meanwhile, the shear band thickness
is similar to that at 200 kPa. Moreover, it seems that the minor of the dense specimen is larger than that of specimen, especially
shear band becomes less obvious as the initial confining pressure at low initial confining pressures. Generally, TSB/D50 decreases from
increases. On the other hand, for the loose specimen, the shear 22.3 to 11.2 for the loose specimen and from 17.0 to 11.0 for the
46 X. Gu et al. / Theoretical and Applied Fracture Mechanics 72 (2014) 37–49

Fmax =0.63 N Fmax = 1.45 N Fmax = 1.55 N Fmax = 1.53 N

(a) (b) (c) (d)

Fig. 13. Distribution of normal contact forces in the dense specimen at different shear stains: (a) 0.2%; (b) 2.4%; (c) 3.2%; (d) 10%.

Desrues and Viggiani [9] showed that the shear band inclination
40
angle continuously decreases as the confining pressure increases.
B
Sliding contact percentage (%)

According to Mohr–Coulomb theory, the theoretical prediction of

30 the shear band inclination angle is p/4 + u/2, where u is the fric-
tion angle. As seen in Figs. 6 and 7, the stress ratio at 10% shear
A strain is around 0.42° for all the specimens, which is corresponding
20 to a friction angle of 24.8°. The predicted inclination angle is 57.4°,
C which is close to the values of the dense specimen but somewhat
E larger than those of the loose specimen.
D G
10
I H
3.2. Evolution of microstructures
F
0
0.0 2.0 4.0 6.0 8.0 10.0 The above results show that the DEM simulations successfully
ε s (%) capture the progressive formation of shear band in granular soils.
It should be emphasized that the macroscopic behavior of the soil
Fig. 14. Evolution of the overall sliding contact percentage in the dense specimen. is a result of particle interactions at contacts such as sliding and
rotation. To explain the macroscopic behavior at grain scale, the
initial microstructure and its evolution during the shearing, includ-
dense specimen when the confining pressure increases from ing the contact number and contact forces (i.e. normal contact
100 kPa to 800 kPa. Desrues and Viggiani [9] showed that for dry force and tangential contact force), are carefully analyzed. Fig. 11
Hostun sand TSB/D50 decreases from 31 to 17 for the loose speci- shows the evolution of coordination number CN (i.e. the average
men and from 22 to 13 for the dense specimen for the same range number of contact per particle) in the undrained and drained tests.
of confining pressure. It is also consistent with the laboratory Note that only the particles with more than two contacts are used
observations that the thickness of shear band is about 10–20 times to calculate CN, while all particles are used to evaluate the porosity
D50 [28,22]. It is clear that the DEM simulation successfully capture [34]. As seen in Fig. 11, CN remains nearly the same before point A
the effect of confining pressure on the shear band thickness. and then continuously decreases with increasing shear strains.
As seen in Fig. 10(b), the inclination angle of the shear band is Compared with Fig. 3c, although the specimen is generally in con-
always larger in a denser specimen at the same confining pressure. traction and the soil density increases before point C, CN decreases
For the dense specimen, the shear band inclination angle increases significantly. It indicates that there is no unique relationship
from 57.3° to 60.5° as the confining pressure increases from between CN and porosity (or density). Rothenburg and Kruyt [29]
100 kPa to 800 kPa. It should be noted that the inclination angle also showed that the relationship between CN and e is affected
at 100 kPa is based on the left shear band. For the loose specimen, by the anisotropy of contact orientation.
however, the shear band inclination angle decreases from 55.4° to After point G, it seems the CN value reaches a constant value (or
51.9° as the confining pressure increases from 100 kPa to 800 kPa. critical value), although the deviatoric stress still decreases with
 X. Gu et al. / Theoretical and Applied Fracture Mechanics 72 (2014) 37–49 47

0.30
0.25
0.20
0.17
0.15
0.13
0.10
0.070
0.050
0.045
0.042
0.040
0.030
0.010

(a) (b) (c) (d)

0.25
0.25
0.20
0.17
0.15
0.10
0.070
0.050
0.045
0.042
0.040
0.030
0.010

(e) (f) (g) (h)

Fig. 15. Distribution of sliding contact percentage in the dense specimen at different shear stains: (a) 0.2%; (b) 0.6%; (c) 1.2%; (d) 1.8%; (e) 2.2%; (f) 2.4%; (g) 3.2%; (h) 10%.

increasing shear strain (see Fig. 3b). Gu et al. [12] indicates that the together with the maximum contact force Fmax. The thickness of
CN value at critical state depends on the mean effective stress at line in Fig. 13 is proportional to the magnitude of contact force.
the critical state. In order to illustrate the difference of CN value It is clear that the normal contact force is generally uniform at
inside and outside shear band, the distributions of CN in the spec- the start of the shearing. As the shear strain increases, the axial
imen at different shear strains are showed in Fig. 12. It is clear that stress increases and column-like force paths (or strong force
CN decreases uniformly in the specimen first and then localizes in chains) form. At large shear strain, several strong force chains col-
several zones. At the peak (point D), several potential shear bands lapse (or bulking occurs) and results in large voids as indicated in
form in the specimen. The evolution of CN is similar to that of the Fig. 13. Similar observations have been made by Oda et al. [24] in
porosity in Fig. 6. It indicates that the CN depends on the soil den- laboratory. Tordesillas and Muthuswamy [35] showed that the
sity. However, the shear band based on the coordination number is formation and the characteristics of shear band is determined by
not as obvious as the porosity. the bulking of lateral confined force chains.
Besides the contact number, contact forces are also important In a DEM simulation, sliding of a contact is governed by the
aspects in micromechanics. Oda et al. [24] performed biaxial test Coulomb friction law. In this study, contact sliding is assumed to
on photoelastic rods and found that the external loads are trans- occur when jTj=ðlNÞ > 0:9999, where T and N are the shear contact
ferred by force chains and the force chains may collapse due to force and normal contact force respectively, and l is the contact
the loss of lateral supporting contacts due to dilatancy. Fig. 13 plots friction coefficient. When contact sliding occurs, particle rear-
the distribution of normal contact force at different shear strains, rangement happens, resulting in permanent deformation. Fig. 14
48 X. Gu et al. / Theoretical and Applied Fracture Mechanics 72 (2014) 37–49

shows the percentage of contacts that sliding happens during the 2012CB719803) and National Natural Science Foundation of China
shearing. The percentage of sliding contacts increases sharply to (Grant Nos. 51308408 and 51238009). The authors are also very
a peak value around point B and then decreases continuously to grateful to the reviewers for their valuable comments and
an ultimate value of around 3%. It indicates that when the speci- suggestions.
men starts to dilate, the sliding contacts concentrate in the poten-
tial shear bands and the overall sliding contact percentage References
decreases. At point F, the sliding contact percentage is close to
the final value, which indicates that major shear band forms. It [1] K.A. Alshibli, S.N. Batiste, S. Sture, Strain localization in sand: plane strain
agrees well with the progressive shear band formation in Fig. 6. versus triaxial compression, J. Geotech. Geoenviron. Eng. 129 (6) (2003) 483–
494.
Gu et al. [12] reported that the ultimate value of sliding contact
[2] J.E. Andrade, Q. Chen, P.H. Le, C.F. Avila, T.M. Evans, On the rheology of dilative
percentage is around 16–20% in the DEM simulation with rigid granular media: bridging solid- and fluid-like behavior, J. Mech. Phys. Solids 60
lateral boundary which cannot yield shear band. Kuhn [19] showed (2012). 1122–113.
[3] J.R.F. Arthur, K.S. Chua, T. Dunstan, Induced anisotropy in a sand, Geotechnique
that the slip deformation was not uniform and it was most intense
27 (1) (1977) 613–622.
in thin ‘‘microbands’’. [4] J.P. Bardet, J. Proubet, A numerical investigation of the structure of persistent
For better illustration, the distribution of sliding contact per- shear bands in granular media, Geotechnique 41 (4) (1991) 599–613.
centage in the specimen at different shear strains is shown in [5] J. Chu, S.C.R. Lo, I.K. Lee, Strain softening and shear band formation of sand in
multi-axial testing, Geotechnique 46 (1) (1996) 63–82.
Fig. 15. As seen in Fig. 15, the distribution of sliding contact per- [6] P.A. Cundall, O.D.L. Strack, A discrete numerical model for granular assemblies,
centage is uniform and the value is high, which indicates the spec- Géotechnique 29 (1) (1979) 47–65.
imen deforms uniformly. With further shearing, the uniformity [7] P.A. Cundall, J.T. Jenkins, I. Ishibashi, Evolution of elastic moduli in a deforming
granular assembly, in: J. Biarez, R. Gourves (Eds.), Powders and Grains,
decreases and the sliding contacts gradually concentrate in several Balkema, Rotterdam, The Netherlands, 1989, pp. 319–322.
zones and the sliding contact percentage in other parts of the spec- [8] J. Desrues, J. Lanier, P. Stutz, Localization of the deformation in tests on sand
imen decreases significantly. At large strains, almost all the sliding sample, Eng. Fract. Mech. 27 (4) (1985) 909–921.
[9] J. Desrues, G. Viggiani, Strain localization in sand an overview of the
contacts are in the shear band. The specimen looks like two rigid experimental results obtained in Grenoble using stereophotogrammetry, Int.
blocks sliding along the shear band. Similarly, the distribution of J. Numer. Anal. Methods Geomech. 28 (2004) 279–321.
particle rotation is most significant in the shear band (see Fig. 5). [10] X.Q. Gu, J. Yang, A discrete element analysis of elastic properties of granular
materials, Granul. Matter 15 (2) (2013) 139–148.
Therefore, the particles in the shear band slide and rotate intensely, [11] X.Q. Gu, J. Yang, M.S. Huang, DEM simulations of the small strain stiffness of
resulting in the deformation localization in the shear band. granular soils: effect of stress ratio, Granul. Matter 15 (3) (2013) 287–298.
[12] X.Q. Gu, M.S. Huang, J.G. Qian, DEM investigation on the evolution of
microstructure in granular soils under shearing, Granul. Matter 16 (1)
4. Summary and conclusions (2014) 91–106.
[13] N. Guo, J.D. Zhao, The signature of shear-induced anisotropy in granular media,
Comput. Geotech. 47 (2013) 1–15.
In this study, the characteristics of shear band in granular [14] S.A. Hall, N. Lenoir, G. Viggiani, J. Desrues, P. Besuelle, Strain localisation in
materials at different initial densities and confining pressures were sand under triaxial loading: characterisation by X-ray micro tomography and
3D digital image correlation, in: Proceeding of First International Symposium
investigated by DEM simulations. The effects of soil density and
on Computational Geomechanics, Juan-les-Pins, France, 2009, pp. 1–9.
confining pressure on the macroscopic response, the pattern of [15] C. Han, A. Drescher, Shear bands in biaxial tests on dry coarse sand, Soils
shear band, and the thickness and inclination angle of shear band Found. 33 (1) (1993) 118–132.
were carefully analyzed. It was found that the DEM simulation [16] Itasca, Particle Flow Code (PFC2D) Manual, Itasca Consulting Group, Inc., Minn,
2003.
with lateral flexible boundaries can successfully capture the shear [17] K. Iwashita, M. Oda, Rolling resistance at contacts in simulation of shear band
band formation in both dense and loose granular materials. During development by DEM, J. Eng. Mech. 124 (1998) 285–292.
the shearing, several potential shear bands were formed at the [18] M.J. Jiang, H.H. Zhu, X.M. Li, Strain localization analyses of idealized sands in
biaxial tests by distinct element method, Front. Archit. Civ. Eng. China 4 (2)
peak deviatoric shear stress. With further shearing, the potential (2010) 208–222.
shear bands evolute into major shear bands and softening behavior [19] M.R. Kuhn, Structured deformation in granular materials, Mech. Mater. 31
happens. Finally, a persistent shear band forms and other shear (1999) 407–419.
[20] P.V. Lade, Q. Wang, Analysis of shear band in true triaxial tests on sand, J. Eng.
bands deteriorate. The void ratio in the shear band is significantly Mech. 127 (8) (2001) 762–768.
larger than other parts of the specimen. For the same specimen, the [21] A. Mohamed, M. Gutierrez, Comprehensive study of the effects of rolling
pattern of shear band may be different at different initial confining resistance on the stress–strain and strain localization behavior of granular
materials, Granul. Matter 12 (2010) 527–541.
pressures. Generally, loose state and high confining pressure [22] H.B. Muhlhaus, I. Vardoulakis, The thickness of shear bands in granular
‘‘delay’’ the occurrence of shear banding. The simulation showed materials, Geotechnique 37 (1987) 271–283.
that the shear band thickness decreases with increasing confining [23] D. Muir Wood, K. Maeda, Changing grading of soil: effect on critical states, Acta
Geotech. 3 (2008) 3–14.
pressure and density. The shear band inclination angle is higher in
[24] M. Oda, J. Konishi, S. Nemat-Nasser, Experimental micromechanical evaluation
dense specimens and its value is close to the prediction by of strength of granular materials: effect of particle rolling, Mech. Mater. 1
Mohr–Coulomb theory in general. At microscopic level, the overall (1982) 269–283.
coordination number of the specimen decreases significantly. The [25] M. Oda, Inherent and induced anisotropy in plasticity theory of granular soils,
Mech. Mater. 16 (1993) 35–45.
coordination number in the shear band is significantly lower than [26] E. Papamichos, I. Vardoulakis, Shear band formation in sand according to
outside the shear band, which is consistent with the large porosity noncoaxial plasticity model, Geotechnique 45 (4) (1995) 649–661.
in the shear band. Most of the contacts do not slide simultaneously [27] J.G. Qian, J. Yang, M.S. Huang, Three-dimensional non-coaxial plasticity
modeling of shear band formation in geomaterials, J. Eng. Mech. 34 (4)
during the shearing. When shear bands forms, almost all the (2008) 322–329.
sliding contacts concentrate in the shear band. At particulate scale, [28] K.H. Roscoe, The influence of strains in soil mechanics, Geotechnique 20 (1970)
the formation of shear band is due to the bulking of lateral confin- 129–170.
[29] L. Rothenburg, N.P. Kruyt, Critical state and evolution of coordination number
ing force chains and the shear band is characterized by high parti- in simulated granular materials, Int. J. Solids Struct. 41 (2004) 5763–5774.
cle sliding and rotation compared with other part of the specimen. [30] J.W. Rudnicki, R. Rice, Conditions for the localization of deformation pressure
sensitive dilatant materials, J. Mech. Phys. Solids 23 (1975) 371–394.
[31] T.G. Sitharam, J.S. Vinod, B.V. Ravishankar, Post-liquefaction undrained
Acknowledgements monotonic behavior of sands: experiments and DEM simulations,
Géotechnique 59 (9) (2009) 739–749.
[32] F. Tatsuoka, M. Sakamoto, T. Kawamura, S. Fukushima, Strength and
The work presented in this paper is supported by the National deformation characteristics of sand in plane strain compression at extremely
Basic Research Program of China (973 Program, Grant No. low confining pressures, Soils Found. 26 (1) (1986) 65–84.
 X. Gu et al. / Theoretical and Applied Fracture Mechanics 72 (2014) 37–49 49

[33] F. Tatsuoka, S. Nakamura, C.-C. Huang, T. Kazuo, Strength anisotropy and shear [39] Q. Wang, P.V. Lade, Shear banding in true triaxial tests and its effect on failure
band direction in plane strain tests of sand, Soils Found. 30 (1) (1990) 35–54. in sand, J. Eng. Mech. 127 (8) (2001) 754–761.
[34] C. Thornton, Numerical simulations of deviatoric shear deformation of [40] Y.H. Wang, C.M.B. Mok, Mechanisms of small-strain shear-modulus anisotropy
granular media, Géotechnique 50 (1) (2000) 43–53. in soils, J. Geotech. Geoenviron. Eng. 134 (10) (2008) 1516–1530.
[35] A. Tordesillas, M. Muthuswamy, On the modeling of confined buckling of force [41] Y.H. Wang, S.C. Leung, A particulate scale investigation of cemented sand
chains, J. Mech. Phys. Solids 57 (4) (2009) 706–727. behavior, Can. Geotech. J. 45 (2008) 29–44.
[36] I. Vadoulakis, Shear band inclination and shear modulus of sand in biaxial test, [42] D.J. White, W.A. Take, M.D. Bolton, Soil deformation measurement using
Int. J. Numer. Anal. Methods Geomech. 4 (1980) 103–119. particle image velocimetry (PIV) and photogrammetry, Geotechnique 53 (7)
[37] I. Vadoulakis, Constitutive properties of dry sand observable in the triaxial test, (2003) 619–631.
Acta Mech. 38 (1981) 219–239. [43] W.M. Yan, J.J. Dong, Effect of particle grading on the response of an idealized
[38] I. Vadoulakis, Shear banding and liquefaction in granular materials on the granular assemblage, Int. J. Geomech. 11 (4) (2011) 276–285.
basis of Cosserat continuum theory, Arch. Appl. Mech. 59 (2) (1989) 106–113.