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

net/publication/256934077

FE simulation of shear localization along granular soil–structure interfaces

using micro-polar elasto-plasticity

Article in Mechanics Research Communications · January 2012

DOI: 10.1016/j.mechrescom.2011.10.001

CITATIONS READS

8 69

3 authors, including:

Babak Ebrahimian Mustafa I. Alsaleh

Shahid Beheshti University Caterpillar Inc.
132 PUBLICATIONS 361 CITATIONS 50 PUBLICATIONS 805 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Babak Ebrahimian on 03 September 2021.

The user has requested enhancement of the downloaded file.

 Mechanics Research Communications 39 (2012) 28–34

Contents lists available at SciVerse ScienceDirect

Mechanics Research Communications

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

FE simulation of shear localization along granular soil–structure interfaces using

micro-polar elasto-plasticity
Babak Ebrahimian a,∗ , Asadollah Noorzad a , Mustafa I. Alsaleh b
a
School of Civil Engineering, Faculty of Engineering, University of Tehran, P.O. Box 11365/4563, 16 Azar Street, Enghelab Avenue, Tehran, Iran
b
R&D, Virtual Product Development Technology, Caterpillar Inc., Mossville, IL 61552, USA

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

Article history: The paper presents FE simulation of shear localization along the interface between cohesionless granular
Received 6 May 2011 soil layer and a rough surface of bounding structure under large shearing movement. The distribution and
Available online 12 October 2011 evolution of static and kinematic quantities are mainly focused regarding infinite layer of micro-polar
material under constant vertical pressure and free dilatancy, particularly with advanced large movement
Keywords: of bounding structure. The FE results demonstrate that the displacement magnitude of bounding structure
Soil–structure interface
has significant effect on the distribution and evolution of state variables and polar quantities in the
Micro-polar (Cosserat) continuum
granular soil layer.
Elasto-plasticity
Shear localization © 2011 Elsevier Ltd. All rights reserved.

1. Introduction interface, on the evolution and location of shear localization are

In most numerical investigations of shear localization, using
micro-polar hypoplastic models, an oscillation is seen in the dis-
tributions of void ratio and stress components across the shear
band (Tejchman and Gudehus, 2001; Huang and Bauer, 2003). This
kind of oscillation is unrealistic and not consistent with the exper-
imental results available in the literature. In such studies, the void
ratio corresponded to the stationary state is not unique. This ratio is
deformation mode dependent and can significantly exceed its crit-
ical value. However, it is experimentally evident that the void ratio
has its maximum quantity in the shear band but still with smooth
distribution (Kishida and Uesugi, 1987; Desrues et al., 1996).
Following the relevant studies already existed in the literature,
the research presented here is the first numerical simulation which
addresses the distribution and evolution of state variables and polar
quantities, particularly under large shear deformation, within infi-
nite micro-polar elasto-plastic granular soil layer. The mechanical
behavior of cohesionless granular soil is described using micro-
polar Lade’s single hardening model. Lade’s model accounts for
the dilatancy effects which is missing in many elasto-plastic mod-
els. The mentioned model can describe the formation of shear
bands, their thickness, spacing and the related scale effect due to
the presence of an internal length in the form of mean grain size
(Alsaleh et al., 2006). In this research, the effects of two different
micro-polar kinematic boundary conditions, prescribed along the
∗ Corresponding author. Tel.: +98 21 88911071; fax: +98 21 66403808.
E-mail addresses: bebrahimian@ut.ac.ir, ebrahimian.babak@gmail.com
(B. Ebrahimian).
studied. Here, the effects of displacement magnitude, applied to the
bounding structure, on the distribution and evolution of state vari-
ables and polar quantities are considered. This effort is done under
large movement of bounding structure parallel to the interface.
Smooth distributions of void ratio and shear stress components are
obtained within the shear band, what the other previous numerical
investigations did not receive.
2. Constitutive relations
The kinematics of a micro-polar continuum is characterized
by the macro-displacement field (ui ) and the micro-rotation (ωic )
where, i = 1, 2, 3. In the continuum mechanics, the so-called velocity
gradient (L), is written as Lij = ∂u̇i /∂xj . L can be also represented as
the sum of symmetric part, ε̇ij = 1/2(∂u̇i /∂xj + ∂u̇j /∂xi ), and skew
symmetric part, Wij = 1/2(∂u̇i /∂xj − ∂u̇j /∂xi ). Furthermore, Wij is
also termed as macro-spin tensor and can be given by macro-spin
vector (ω̇k ) according to Wij = − ∈ ijk ω̇k . Cosserat spin tensor, corre-
sponding to micro-motion, is defined as Wijc = − ∈ ijk ω̇kc . It is noted
that Cosserat spin is not related to macro spin which is a rota-
tion of continuum. Cosserat strain rate (or rate of deformation) is
read as the difference between macro-motion and micro-motion:
ε̇cij = ε̇ij + Wij − Wijc = ∂u̇i /∂xj + ∈ kij ω̇kc . Micro-curvature vector of
deformation or the gradient of particle rotation can be defined as
ij = ∂ωic /∂xj .
In this study, the Lade’s single hardening constitutive model
(Lade and Nelson, 1987; Kim and Lade, 1988; Lade and Kim, 1988),
enhanced with Cosserat rotation and couple stress via internal
length, is adopted to simulate the mechanical behavior of soil. The

0093-6413/$ – see front matter © 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.mechrescom.2011.10.001
 B. Ebrahimian et al. / Mechanics Research Communications 39 (2012) 28–34
enhancement has been conducted through second stress and devi-
atoric stress invariants in order to incorporate the internal length.
This model is an elasto-plastic soil model with a single yield surface,
expressed in terms of stress invariants. Lade’s model has a non-
linear elasticity function where non-associative flow rule and high
non-linear plastic work-based hardening function are assumed.
The model formulations are briefly discussed in the following; the
details are found in (Alsaleh et al., 2006). The model involves 11
parameters that can be determined by four compression tests, one
isotropic and three CD triaxial. The proposed following non-linear
elasticity modulus, used in this work, is defined as
 2  1 +   J  
I I 2
E = ML Pa + 6
Pa 1 − 2 Pa2
where Pa is atmospheric pressure used to normalize stresses;
II is first invariant of the stress tensor, II =  ii (i = 1, 2,
3); ML and  are dimensionless material parameters which
are determined by a series of simple experiments including
loading–unloading–reloading cycles (more details are seen in Lade
and Nelson (1987));  is constant Poisson’s ratio; J2 is second devi-
atoric stress invariant which is given by
 fP =
J2 = h3 [(11 − 22 )2 + (33 − 22 )2 + (11 − 33 )2 ]
 +  2  h4 2
+
12 21
+ (m1 + m22 )
2 l2
where h3 and h4 are balancing factors between couple stresses (mi )
and stresses ( ij ); l is internal length (equal to d50 , where d50 is
mean grain size). The balancing factors depend on the contact dis-
tributions of soil grains and are assumed as one in the current study
(Vardoulakis and Sulem, 1995). Plastic flow occurs when the state
of stress touches the yield criterion (fp ) causing the material to
undergo plastic deformations. The plastic strain increment is cal-
culated by the flow rule as ε̇Pij = ˙ p (∂gp /∂ij ). The plastic curvature
increment of rotation can be given as ˙ P = ˙ p (∂gp /∂mi ).  ij is stress
i stress–strain curve.
components of stress tensor; mi is couple stress components of
stress tensor; ˙ p is proportionality factor (a positive scalar); gp is
plastic potential function given as (Lade and Kim, 1988):
 I 
II3 II2 I
gp = 1 − + 2
IIII III Pa
where 2 and are plastic potential parameters that can be esti-
mated from experiments (Lade and Kim, 1988); the parameter
2 controls the intersection of plastic potential with hydro-
static pressure axis; the exponent determines the curvature
of plastic potential in the principal stress space; the parameter
−1.27 is weighting factor (Kim and Lade, 1988); m
1 = 0.00155m
is failure criterion parameter that can be determined from exper-
iments as shown in (Lade and Kim, 1988); III and IIII are second
and third stress invariants, respectively. The second stress invari-
ant was enhanced to incorporate the effects of couple stresses (m1 ,
m2 ) as
m1 m2
III = h1 (12 21 − 11 22 − 11 33 − 22 33 ) + h2
l2
where hr (r = 1, 2) is weighing coefficient for statics that depends
on the contact distributions of soil grains and here is assumed as
one. Vardoulakis and Sulem (1995) have applied an approach simi-
lar to ours, and modified the second stress invariant. The balancing
factors (h1 –h4 ), selected for this study, are based on the best rec-
ommended values previously tested and proposed by others in
the literature. This selection was based on the micro-mechanical
consideration of particle displacement and particle rotation includ-
ing inter-particle slip and inter-particle shear in granular media
29
(Vardoulakis and Sulem, 1995). The third invariant of stress tensor
is computed as IIII = ( 11  22  33 −  33  12  21 ). Lade and Kim (1988)
developed a yield function based on the work-hardening and work-
softening laws for frictional materials: fp = fp () − fP (WP ) = 0. Wp
is plastic work; and
  I h
II3 II2 I
fP = 1 − eq (5)
IIII III Pa
where h is material yield parameter that can be determined from
experimental data obtained from two arbitrary failure points along
the hydrostatic axis; q = (˛L S)/(1 − (1 − ˛L )S) is model parameter,
with the ranges of zero to one at hydrostatic condition and fail-
(1) ure, respectively (Lade and Kim, 1988); ˛L is material parameter
m
(Lade and Kim, 1988); S = 1/ 1 ((II3 /IIII ) − 27)(II /Pa ) is stress level
with value ranges of zero to one at hydrostatic loading condition
and failure, respectively; 1 is failure criterion parameter (Lade and
Kim, 1988); it is a dimensionless constant and can be determined
upon triaxial compression test results. Lade’s yield surface takes
the shape of an eye drop; each yield surface represents a contour
for the plastic work:
 1 1/ L  W 1/ L
P
(6)
D Pa
where L = P/h and D = C/(27 1 + 3) L are constants; fP increases
only if plastic work (Wp ) increases. Now, plastic work is esti-
(2) mated during isotropic (hydrostatic) loading as WP = CPa (II /Pa )P . The
parameters C and P can be estimated using the results of an isotropic
compression test regarding the best fitting of WP = CPa (II /Pa )P with
the test results. During hardening regime, Eq. (6) is used to describe
the non-linear hardening component. As the materials are failed or
met the instability point, hardening regime is terminated and soft-
ening regime is started. It is an exponential function, expressed
as fP = A exp(−B(WP /Pa )). The constants A and B should be esti-
mated once at failure, where S = q = 1.0, by A = (fP exp(BWP /Pa ))S=1
and B = (b(∂fP /∂(WP /Pa ))/fP )S=1 when 0 ≤ b ≤ 1.0; b is material
constant that describes the concavity of softening part in the
The proposed constitutive relations are not directly dependent
on the initial void ratio while most model parameters such as
modulus of elasticity and other shear strength parameters are sig-
nificantly influenced by the initial void ratio. Therefore, the void
(3) ratio effect is accounted implicitly within the model. In order to
study the effect of void distribution, the model parameters are
obtained at different densities, through which the effect of void
ratio is mapped to the mechanical response of the material. This is
one of the advantages of the proposed constitutive model.
In 2D Cosserat continuum, the stress tensor is expressed as
T
m1 m2
{} = 11 22 33 12 21 (7)
l l
As described earlier, m1 and m2 are couple stress components; l is
length scale (equal to d50 , where d50 is mean grain size). It should
be mentioned that stress tensor is non-symmetric due to the effect
of couple stresses. Strain matrix for 2D Cosserat continuum can be
given in a vector form as
(4)
T
{ }={ 11 22 33 12 21 l1 l2 } (8)
where 1 and 2 are micro-curvatures. Similar to stress tensor, the
objective strain vector is also non-symmetric and deviated from the
classical one. The strain and curvature tensors are decomposed to
an elastic part denoted by a superscript ‘e’ and a plastic part denoted
by a superscript ‘p’: ε̇ij = ε̇eij + ε̇Pij and ˙ ij = ˙ ije + ˙ ijP . As dealing with
stress and strain increments seem more convenient, the follow-
ing constitutive laws can be used: {} ˙ = [D]{ ˙ }; [D] is elasto-plastic
stiffness matrix, derived from Cosserat continuum. This matrix is a
30 B. Ebrahimian et al. / Mechanics Research Communications 39 (2012) 28–34
coupled stiffness matrix including the terms of stresses and couple
stresses. Furthermore, each point in the plane strain condition has
three degrees of freedom including two displacements (u1 and u2 )
T
and one rotation (ω3c ): U = { u1 u2 ω3c } .
3. Finite element modeling
In this research, the present micro-polar Lade’s single harden-
ing model is implemented in a finite element program in order
to investigate the phenomenon of shear localization along granu-
lar soil–structure interfaces during large movement of bounding
structure. The numerical implementation of constitutive model
is not mainly concerned here. However, the numerical aspects
are briefly described and for details (Alsaleh, 2004; Alshibli et al.,
2006) are suggested. In order to describe the displacements and
Cosserat rotation within the element, a 4-noded isoparametric ele-
ment with four integration points is used for plane strain condition
(Alsaleh et al., 2006). Based on this type of element, a bi-linear
shape function will be used. Here, the 4-noded element is chosen
for minimizing any expected locking in the element (Belytschko
et al., 2000). The selective reduced integration technique is imple-
mented to avoid the probable volumetric lock during softening
regime. Newton-Raphson iteration technique is applied to fulfill the
static equilibrium equations; an updated Lagrange formulation is
employed in the calculations to consider large deformation, which
may be important as long as shear localization takes place. Finite
element equations are obtained for discrete system based on the
virtual work equation, where geometric non-linearity is taken into
account by Lagrangian description. The finite element implemen-
tation is performed using the non-linear finite element program
ABAQUS. It should be mentioned that the element with material
rotation is not provided in this program. Therefore, a User Element
Subroutine (UEL) was written and tested thoroughly by the authors
to describe the new element type with both translation and rota-
tion degrees of freedom to solve the system of the finite element
equations within micro-polar framework (Alsaleh, 2004; Alsaleh
et al., 2006). All the internal state variables (stresses, plastic work,
void ratio, etc.) are updated within the UEL using explicit forward
Euler integration scheme. Regarding rectangular co-ordinate sys-
tem and assumed plane strain conditions, the kinematic quantities
are displacements (u1 and u2 ) and micro-rotation (Cosserat rota-
tion, ω3c ), presented in Fig. 1(a). The non-zero static quantities are
stress components ( 11 ,  22 ,  33 ,  12 ,  21 ) and couple stress com-
ponents (m1 , m2 ), shown in Fig. 1(a). It is sufficient to discretize a
small section of lateral infinite layer by finite elements due to the
horizontal and vertical symmetry of the layer, shown in Fig. 1(b).
The Lade’s model parameters, calibrated for a dense silica sand, are
given in Table 1 (Alsaleh, 2004).
Particularly, a section with the initial height of h0 = 4 cm and
the width of b = 10 cm is discretized by 4-noded elements, each of
which is 1.25 mm × 1.25 mm. The size of element is less than 5 × d50
(with d50 = 1 mm) to obtain mesh independent results (Tejchman
and Bauer, 1996). Here, special micro-polar boundary conditions
are introduced across the sections (x1 = 0) and (x1 = b) in the lat-
eral infinite layer (Bauer and Huang, 1999). In this regard, each
node on the left boundary (x1 = 0) and its corresponding on the
right (x1 = b) should have the same displacements and Cosserat
rotation. The prescribed lateral constraints also imply that in the
direction of bounding structure movement, arbitrary field quanti-
ties ˚ are independent of co-ordinates, that is: ∂˚(x1 , x2 )/∂x1 = 0.
In the present study, the top surface of granular layer is assumed to
be fixed not to occur sliding and rotating. The vertical pressure (P0 )
is kept constant at the top surface of layer. However, the height of
layer can be changed as the result of dilatancy or contractancy of
material under shearing.
It is assumed that the granular layer is initially homogeneous
and isotropic. The relevant boundary conditions are:  ij = P0 I, e = e0
and mi = 0; where, P0 = initial pressure (=100 kPa); e0 = initial void
ratio (=0.6). The mentioned boundary conditions also imply that
the shear stresses and couple stresses are assumed as zero in the
initial state. For modeling the interaction between granular soil and
bounding structure, it is assumed that the grains are captured by
the rough surface of bounding structure. It means that at the bot-
tom boundary (x2 = 0), the vertical displacement (u2 ) is zero, and the
prescribed horizontal displacement (u1B ) is the same for bounding
structure and bottom surface of granular layer. This assumption
also implies that the possibility of relative displacement, resulted
from lower skin frictions, is not considered along the surface of
bounding structure. In other words, full shearing of granular layer
is assumed along the rough surface of bounding structure. The bot-
tom surface of granular layer is horizontally moved by applying
a quasi-static displacement. In this research, two different micro-
polar kinematic boundary conditions (Cases I and II) are considered
for studying the effects of rotation resistance of soil particles along
the interface (x2 = 0): Case I: free Cosserat rotation is assumed along
the interface (x2 = 0) and modeled with zero couple stresses; Case
II: constrained Cosserat rotation is assumed along the interface
(x2 = 0) and modeled with zero Cosserat rotations. The boundary
condition with zero couple stress, along the interface (Case I), is in
accordance with the presumption of classical continuum mechan-
ics. In fact, zero couple stress, along the interface, will occur when
grain rotation (micro-rotation) and continuum rotation (macro-
rotation) are equal. It is noted that also in case of zero Cosserat
rotation along the bottom and top surfaces of granular soil layer
(Case II), there is a plane in the middle of localized zone (or in
the middle of layer) where the couple stresses are zero, i.e. the
signs of couple stresses are switched in the plane. In other words,
zero couple stresses, prescribed along the interface, is one reason-
able choice for micro-polar boundary condition. This assumption
is motivated by the fact that zero couple stresses are generally
appeared in the middle of the shear band in cases where the shear
band can freely develop in the granular body (Huang and Bauer,
2003). Here, constrained Cosserat rotation along the interface (Case
II) is corresponded to the behavior of very rough surface of bound-
ing structure. Very rough surfaces can capture even small grains in
their boundary layers without occurring any sliding or rotating. The
condition of constrained Cosserat rotations along the interface was
verified during shearing tests in Couette apparatus performed with
steel rods and silo model tests with coins (Tejchman, 1989, 1997).
The lack of rotation and slip along very rough walls was also con-
firmed by wall friction experiments (Tejchman, 1989; Tejchman
and Wu, 1995). More importantly, in a continuum description, the
assumption of vanishing rotation resistance should not be confused
with the rolling of discrete grains along the interface. The latter
is in the kinematics of a system of discrete grains with specific
sizes (Bauer and Huang, 2004). However, in analyzing the discrete
grains, the rolling of a grain along the contact surface leads to the
relative displacement between grain center and bounding struc-
ture. In the present continuum model it is assumed that no relative
displacement takes place along the interface.
4. Numerical results
4.1. Free Cosserat rotations along the interface (m2 = 0)
In the first case, the rotation resistance of soil particles can
be neglected along the interface and thus the couple stresses
are assumed as zero; that is the soil particles are free to rotate.
Together with zero rotation resistance of soil particles, prescribed
along the top surface of granular layer, the micro-polar boundary
 B. Ebrahimian et al. / Mechanics Research Communications 39 (2012) 28–34 31

Fig. 1. Modeling plane strain infinite granular layer under constant vertical pressure in contact with rigid moving structure: (a) kinematic and static quantities of micro-polar
material under plane strain condition and (b) initially undeformed finite element mesh with assumed boundary conditions.

Table 1
Lade’s model parameters for dense silica sand.

Elastic properties Failure criterion Plastic potential Yield criterion Hardening/softening

ML , ,  m, 1 , a , 2 h, ˛L law C, P, b

292.6, 0.25, 0.13 0.37, 84.1, 0.0 2.2, −3.06 0.95, 0.3 0.00007, 2.6, 1.0

conditions are non-symmetric. In contrast with non-polar (clas-
sical) continuum, the horizontal displacement field (u1 /h0 ) is
non-linear from the beginning of shearing, shown in Fig. 2(a). The
finding is confirmed by the experimental observations (Garga and
Infante Sedano, 2002) as well as other numerical studies (Tejchman
and Bauer, 1996). Void ratios (e) have the largest values in the
interface for different values of bounding structure displacement
1997), while outside the shear band, the values are nearly zero.
These results are in accordance with the experimental observations
(Kishida and Uesugi, 1987; Tejchman, 1997; Oda and Kazama, 1998;
Oda et al., 2004) as well as DEM simulations (Bardet, 1994; Iwashita
and Oda, 1998). The values of Cosserat rotation increase as the nor-
malized horizontal displacement of bounding structure increases.
The distribution of 21 ∗ across the granular layer height is given in

(Fig. 2(b)). The result is also confirmed by the experiments of Fig. 2(d). The profile of 21 ∗ is more non-uniform across the layer

Desrues et al. (1996). Here, a smooth distribution of void ratio with height as bounding structure displacement increases (Fig. 2(d)).
no oscillation is obtained in the shear band for all magnitudes of Here, the magnitude of bounding structure displacement has strong
normalized horizontal displacement of bounding structure, oppos- influence on the distributions of 21 ∗ .

ing to (Tejchman and Gudehus, 2001; Huang and Bauer, 2003). It Fig. 3 shows the evolution of normalized normal stresses
can be concluded that the volume of soil body increases in the shear ∗ ,  ∗ ,  ∗ ), normalized shear stresses ( ∗ ,  ∗ ) and void
(11 22 33 12 21
band as a result of pronounced dilation for shearing under constant ratio (e) presented in different positions across the height.
normal pressure; it is also seen in the experimental observations Normal stress curves are started from isotropic states, that is
(Kishida and Uesugi, 1987). According to Fig. 2(c), the Cosserat rota- ∗ =  ∗ =  ∗ = −1. Furthermore, in the shear band, x /h = 0.00,
11 22 33 2 0
tions have their significant values in the shear band (Tejchman, normal stresses (11 ∗ ,  ∗ ,  ∗ ) go towards the stationary state,
22 33

Fig. 2. Distributions of (a) u1 /h0 , (b) e, (c) ω3c and (d) 21
∗
across x2 /d50 under P0 = 100 kPa for different U1B /h0 (e0 = 0.6, d50 = 1 mm and zero interface couple stresses).
32 B. Ebrahimian et al. / Mechanics Research Communications 39 (2012) 28–34
∗ ∗ ∗ ∗ ∗
Fig. 3. Evolutions of (a), (b) 11 , 22 , 33 , (c) 12 , 21 and (d) e at different x2 /h0 under P0 = 100 kPa (e0 = 0.6, d50 = 1 mm and zero interface couple stresses).
∗ =  ∗ =  ∗ = −1. It is noted that softening is more pronounced
11
22 33
for the elements, located inside the shear band comparing with
those located outside (Fig. 3(a) and (b)). The evolution of 12∗ is the
∗ is different depending on the
same for all elements, while that of 21
position of elements, across the height, shown in Fig. 3(c). Normal-
ized shear stresses, first increase steeply up to their peak values and
then they decrease gradually. For large displacements of bounding
structure, they tend towards a stationary value which depends
on the assumed stress limit condition. This pattern is confirmed
by experimental observations (Kishida and Uesugi, 1987) and
qualitatively fits with other numerical studies (Huang and Bauer,
2003). According to Fig. 3(d), at the beginning of bounding struc-
ture displacement, all void ratio curves decrease slightly and then
increase. It means that the materials first become denser and then
show dilatancy as the deformation increases. Void ratios increase
sharply inside the shear band (x2 /h0 = 0.00) and tend towards a
stationary value of ec = 0.84; however, outside the shear band they
decrease slightly and then become almost constant with continu-
ous shearing. This indicates that the material, after its peak state,
behaves as a rigid body outside the shear band with continuous
shearing (Tejchman and Gudehus, 2001; Huang and Bauer, 2003).
Here, the contour plot of void ratio is given along with deformed
configuration of granular layer (Fig. 4). The brighter zones, in the
plot, are of higher void ratios or where failure may start. The defor-
mations are obviously localized in the soil near bounding structure
surface. The predicted thickness of shear band is about 13 d50 . It
should be mentioned that the magnitude of horizontal displace-
ment has no significant effect on the thickness of shear band. The
shear band thickness is in accordance with the experimental results
(Oda et al., 2004). The boundary conditions of infinite extended
granular layer, assumed here, are completely different from those
of experiments. Therefore, no quantitative comparison would be
possible. Despite using totally different constitutive model, the
numerical results, obtained in this study, are qualitatively similar to
those of Tejchman and Bauer (1996). However, the results, gained
here, are slightly higher than those of experimental measurements
provided by Tatsuoka et al. (1991) and DEM-simulations (Bardet
and Proubet, 1992).
4.2. Constrained Cosserat rotations along the interface (ω3c = 0)
Here, constrained Cosserat rotation along the interface is stud-
ied (Case II). This case is corresponded to the behavior of very
rough surface of bounding structure. The distribution of quanti-
ties across the height of sheared layer is strongly affected by the
interface behavior between granular layer and bounding struc-
ture, confirmed here. According to Tejchman (1997), the interface
behaviors are significantly affected by the boundary conditions of
entire system. The higher void ratio values are obtained within the
shear band; they have extremely lower values outside the shear
band (Fig. 5(a)). These results correspond to experimental obser-
vations (Kishida and Uesugi, 1987; Oda and Kazama, 1998). In this
study, void ratio is distributed smoothly in the shear band which is
more realistic unlike the results already published in the literature
(Tejchman and Gudehus, 2001; Huang and Bauer, 2003). The shear
band can be characterized by the appearance of Cosserat rotation
(Fig. 5(b)). This indicates that polar effects in granular materials
are significant when shear localization appears as also detected in
experiments (Oda and Iwashita, 2000; Garga and Infante Sedano,
2002) and in previous numerical simulations (Tejchman, 1997;
Garga and Infante Sedano, 2002). It is found that the maximum
values of ω3c are lower for ω3c = 0, comparing with m2 = 0.
The 11∗ and 33 ∗ profiles are shown in Fig. 6(a) and (b),
respectively. As the horizontal displacement of bounding structure
increases, the distribution of normalized normal stresses (11∗ , ∗ )
33
become more non-uniform across the height. So far, the distri-
butions of normal and shear stresses have not been obtained by
elasto-plastic Cosserat model, regarding infinite sheared layer in
contact with bounding structure under large shearing movement.
The normalized stresses |11 ∗ | and | ∗ | have their minimum val-
33
ues in the middle of layer excluding u1B /h0 = 0.125, where they
occur at the boundaries. The normalized normal stresses (11 ∗ , ∗ )
33
 B. Ebrahimian et al. / Mechanics Research Communications 39 (2012) 28–34 33

Fig. 4. Deformed shape of granular layer under P0 = 100 kPa along with contour plot of void ratio (e0 = 0.6, d50 = 1 mm, u1B = 4.50 h0 and zero interface couple stresses).

Fig. 5. Distributions of (a) e and (b) ω3c across x2 /d50 under P0 = 100 kPa for different U1B /h0 (e0 = 0.6, d50 = 1 mm and zero interface Cosserat rotations).

∗ ∗ ∗ ∗ ∗
Fig. 6. Distributions of (a), (b) 11 , 22 , 33 , (c) 12 and (d) 21 across x2 /d50 under P0 = 100 kPa for different U1B /h0 (e0 = 0.6, d50 = 1 mm and zero interface Cosserat rotations).

approach to −1 in the middle of shear band under large shearing. At ∗ ) varies (Fig. 6(d)). Therefore, the stress ten-
vertical shear stress (21
∗ =  ∗ =  ∗ = −1 is obtained.
the limit state, the condition of 11 sor also becomes non-symmetric in case of initial isotropic stress
22 33
∗ ) is constant across the
The normalized horizontal shear stress (12 state. Here, the magnitude of bounding structure displacement
height; needed for equilibrium (Fig. 6(c)); however, normalized has strong influence on the distributions of 21 ∗ . The maximum

Fig. 7. Deformed shape of granular layer under P0 = 100 kPa along with contour plot of void ratio (e0 = 0.6, d50 = 1 mm, u1B = 4.50 h0 and zero interface Cosserat rotations).
34 B. Ebrahimian et al. / Mechanics Research Communications 39 (2012) 28–34
values of |21∗ | are slightly higher for m = 0, comparing with
2
c
those for ω3 = 0. In this research, stress components are also uni-
formly distributed in the shear band, opposing to what obtained
in Tejchman and Gudehus (2001) and Huang and Bauer (2003). No
oscillation is observed in the distribution of normalized stresses
within the shear band.
The contour plot of void ratio and deformed configuration of
granular layer are presented in Fig. 7. According to this figure, the
deformation for large shearing is localized in the narrow zone in the
middle of layer. The brighter strip, in the figure, is corresponded to
the higher void ratio, because of dilatancy within the shear band,
and the darker one to the lower void ratio. Comparing the con-
tour plots, presented in Figs. 4 and 7, the location and thickness of
shear band strongly depend on micro-polar boundary conditions,
prescribed along the interface. In this case, the predicted thickness
of shear band is about 23 d50 . Consequently, the thickness of shear
band is narrower for free rotation resistance in comparison with
that of constrained rotation resistance along the interface.
5. Conclusions
The results, obtained for soil behavior under large shearing in
the vicinity of interface, are outlined as follows:
• The location and thickness of shear band in the granular body
are strongly influenced by micro-polar kinematic boundary con-
ditions, prescribed along the interface. For an infinite extended
narrow layer, shear strain localization appears parallel to the
direction of shearing. Shear band is formed at the middle of layer
in the symmetric Cosserat boundary conditions, prescribed at the
top and bottom of layer; otherwise, it is located closer to the
boundary with less restriction of particle rotation.
• Polar quantities are noticeable in the shear band. High values of
Cosserat rotation and increasing void ratios can be used to iden-
tify the shear band. Unlike conventional non-polar continuum,
horizontal displacements are non-linearly distributed across the
height of sheared layer, even if the material is homogeneous
and non-polarized in the initial state. The state quantities tend
towards a stationary state in large monotonic shearing.
• The distribution of state quantities within sheared layer strongly
depends on the prescribed micro-polar boundary conditions,
indicating the rotation resistance of soil particles against the
surface of bounding structure. Furthermore, the distribution of
normal and shear stresses, across the height of granular layer,
depends highly on the magnitudes of horizontal displacement
applied to the bounding structure. As the displacement of bound-
ing structure increases, the distribution of stresses becomes more
non-uniform across the height.
• A smooth distribution of void ratio is obtained across the shear
band; therefore, no oscillation is observed which is consistent
with the experimental observations. It is emphasized that the
stress components in the shear band are also uniformly dis-
tributed opposing to the earlier results already published in the
literature.
• Enhanced elasto-plastic Lade’s model, with embedded Cosserat
rotations and couple stresses, can simulate properly the local-
ization phenomenon in the granular materials during plane
shearing. This conclusion is based on the numerical results,
View publication stats
obtained in this study, conducting numerous finite element anal-
yses. It should be mentioned that the FE results have been
verified qualitatively by experimental observations as well as
other numerical studies.
References
Alsaleh, M., 2004. Numerical modeling for strain localization in granular materi-
als using Cosserat theory enhanced with microfabric properties. Ph.D. Thesis.
Louisiana State University, Baton Rouge, LA.
Alsaleh, M., Voyiadjis, G., Alshibli, K., 2006. Modeling strain localization in granular
materials using micropolar theory: mathematical formulations. Int. J. Numer.
Anal. Methods Geomech. 30 (15), 1501–1524.
Alshibli, K., Alsaleh, M., Voyiadjis, G., 2006. Modeling strain localization in granular
materials using micropolar theory: numerical implementation and verification.
Int. J. Numer. Anal. Methods Geomech. 30 (15), 1525–1544.
Bardet, J.P., Proubet, J., 1992. A numerical investigation of the structure of persistent
shear bands. Géotechnique 41, 599–613.
Bardet, J.P., 1994. Observations on the effects of particle rotations on the failure of
idealized granular materials. Mech. Mater. 18, 159–182.
Bauer, E., Huang, W., 1999. Numerical study of polar effects in shear zone. In: Pande,
G.N., Pietruszczak, S., Schweiger, H. (Eds.), Proc. of 7th Int. Symp. on Numerical
Models in Geomechanics. Balkema, Rotterdam, pp. 133–138.
Bauer, E., Huang, W., 2004. Effect of particle rotations on the interface behaviour
between a granular layer and parallel rough plates. Sci. Eng. Compos. Mater. 11
(4), 239–246.
Belytschko, T., Liu, W.K., Moran, B., 2000. Nonlinear Finite Elements for Continua
and Structure. John Wiley & Sons, LTD, New York, USA.
Desrues, J., Chambon, R., Mokni, M., Mazerolle, F., 1996. Void ratio evolution inside
shear bands in triaxial sand specimens studied by computed tomography.
Géotechnique 46, 529–546.
Garga, V.K., Infante Sedano, J.A., 2002. Steady state strength of sands in a constant
volume ring shear apparatus. Geotech. Test. J. 25 (4), 414–421.
Huang, W., Bauer, E., 2003. Numerical investigations of shear localization in a
micro-polar hypoplastic material. Int. J. Numer. Anal. Methods Geomech. 27,
325–352.
Iwashita, K., Oda, M., 1998. Shear band development in modified DEM: importance
of couple stress. TASK Q. 2, 443–461.
Kim, M.K., Lade, P.V., 1988. Single hardening constitutive model for frictional mate-
rials. Comput. Geotech. 5, 307–324.
Kishida, H., Uesugi, M., 1987. Tests of the interface between sand and steel in the
simple shear apparatus. Géotechnique 37, 45–52.
Lade, P.V., Nelson, R.B., 1987. Modeling the elastic behavior of granular materials.
Int. J. Numer. Anal. Methods Geomech. 11, 521–542.
Lade, P.V., Kim, M.K., 1988. Single hardening plasticity model for frictional materials.
Comput. Geotech. 6, 13–29.
Oda, M., Kazama, H., 1998. Microstructure of shear bands and its relation to the
mechanics of dilatancy and failure of dense granular soils. Géotechnique 48,
465–481.
Oda, M., Iwashita, K., 2000. Study on couple stresses and shear band development
in granular media based on numerical simulation analyses. Int. J. Eng. Sci. 38,
1713–1740.
Oda, M., Takemura, T., Takahashi, M., 2004. Microstructure in shear band observed
by microfocus X-ray computed tomography. Géotechnique 54 (8), 539–542.
Tatsuoka, F., Okahara, M., Tanaka, T., Tani, K., Morimoto, T., Siddiquee, M.S.A., 1991.
Progressive failure and particle size effect in bearing capacity of footing on sand.
Proc. ASCE Geotech. Eng. Congr. 27 (2), 788–802.
Tejchman, J., 1989. Scherzonenbildung und verspannungseffekte in granulaten
unter berükcksichtigung on korndrehungen. In: Gudehus, G., Natau, O. (Eds.),
Publication Series of the Institute of Soil and Rock Mechanics, vol. 117. University
Karlsruhe, Karlsruhe, pp. 1–236.
Tejchman, J., Wu, W., 1995. Experimental and numerical study of sand–steel inter-
faces. Int. J. Numer. Anal. Methods Geomech. 19, 513–536.
Tejchman, J., Bauer, E., 1996. Numerical simulation of shear band formation with a
polar hypoplastic model. Comput. Geotech. 19 (3), 221–244.
Tejchman, J., 1997. Modelling of shear localisation and autogeneous dynamic effects
in granular bodies. In: Gudehus, G., Natau, O. (Eds.), Publication Series of the
Institute of Soil and Rock Mechanics, vol. 140. University Karlsruhe, Karlsruhe,
pp. 1–353.
Tejchman, J., Gudehus, G., 2001. Shearing of a narrow granular layer with polar
quantities. Int. J. Numer. Anal. Methods Geomech. 76 (2), 513–536.
Vardoulakis, I., Sulem, J., 1995. Bifurcation Analysis in Geomechanics. Blackie Aca-
demic & Professional, Glasgow, UK.