A Review of Two Different

Approaches to Hypoplasticity
Claudio Tamagnini
Universita di Perugia, Italy

Gioacchino Viggiani, Rene Chambon

Laboratoire 3S, Grenoble, France
June 16, 1999

Contents
1 Introduction 2
2 Mathematical structure 4
2.1 Preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2 K{hypoplasticity . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.3 CLoE hypoplasticity . . . . . . . . . . . . . . . . . . . . . . . 9
3 Invertibility, consistency and limit states 14
3.1 General results . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.2 Amorphous hypoplastic models . . . . . . . . . . . . . . . . . 16
3.3 Endomorphous hypoplastic models . . . . . . . . . . . . . . . 20
4 Strain localization and bifurcation analysis 23
4.1 Introduction and historical background . . . . . . . . . . . . . 23
4.2 General results . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.3 The question of the shear modulus . . . . . . . . . . . . . . . 29
5 Conclusions 34
A Gudehus/Bauer K{hypoplastic model 43
B von Wol ersdor K{hypoplastic model 43

1
1 Introduction
Non{linearity and irreversibility are striking features of soil behavior, af-
fecting the response of any geotechnical \structure", be it, for example, a
foundation, an excavation, an earth dam, or a natural slope. From a math-
ematical viewpoint (i.e., at the constitutive level), di erent strategies have
been proposed to deal with such features of soil behavior, including:
 classical perfect or hardening plasticity with strain{dependent elastic
moduli (non{linear elastic, or \variable moduli" models, see e.g. [47,
65];
 generalized, or bounding surface plasticity, e.g. [24, 64];
 incrementally non{linear theories, including incrementally non{linear
elastoplasticity [24, 85], endochronic plasticity, e.g. [79], and incremen-
tally non{linear rate{type models [11, 52, 26]).
Although very popular in the geotechnical community (especially in the
U.K.), variable moduli models su er from a number of important shortcom-
ings, which have been thoroughly discussed in the literature (e.g.[60, 66]).
In particular, it has been shown by Nelson already back in 1977 [60], that
serious numerical instabilities may result from their use in FE computations
when the direction of the stress increment approaches the boundary between
the loading and unloading tensorial zones. As demonstrated by Gudehus
[34], this is due to the lack of continuity of the relevant stress response enve-
lope. Moreover, the use of pseudo{Hookean tangent sti ness tensors in the
hypoelastic equations leads to coaxiality between stress and strain rates and
to a complete volumetric{deviatoric uncoupling. Both features appear to be
questionable on the basis of currently available experimental data.
The origins of bounding surface plasticity as applied to soils can be traced
back to the late `70s. Their development was mainly motivated by the need
to accurately describe the irreversible and hysteretic soil behavior observed
along cyclic paths. Di erently from classical elastoplasticity (with or without
elastic non{linearity), in bounding surface plasticity, the constitutive equa-
tion is always characterized, for any particular loading mechanism, by more
than one possible responses | plastic loading and elastic unloading | i.e.,
plastic ow can occur at any point in the domain of admissible stress states.
This makes the material behavior not only non{linear (for nite loading
paths), but also incrementally multi{linear.
The particular way in which the di erent loading-unloading conditions
are de ned guarantees the continuity of the incremental response at the

2
boundary between the tensorial zones (neutral loading conditions). More-
over, the derivation of plastic strain rates from a suitable ow rule allows for
deviatoric{volumetric coupling as well as non{coaxiality of stress and strain
rates.
In spite of its merits, bounding surface plasticity has not found widespread
application in geotechnical practice, and remains mainly con ned to research
applications, mostly due to the relative complexity of its mathematical struc-
ture and the diculties associated to the calibration of the relevant param-
eters.
In recent years, a completely di erent approach to soil non{linearity has
been proposed, leading to a class of constitutive models generally referred
to as incrementally non{linear models [25]. The distinctive features of such
theories are:
i) the absence of any kinematic decomposition of strain rates into re-
versible and irreversible parts;
ii) the continuously non{linear dependence of the tangent sti ness tensor
on strain rate direction.
That soils are thoroughly non{linear, rather than bilinear or multilinear
(multi{mechanisms plasticity), is in fact suggested by qualitative consid-
erations on the mechanisms responsible for inelastic deformations at the
microstructural level [75], as well as by the limitations shown by classical
elastoplastic models with a single smooth yield surface in predicting the on-
set of shear banding, see, e.g. [70, 67, 49, 81], and by limited experimental
evidence (Royis & Doanh 1998). Note that feature (ii), but not feature (i), is
also present in some bounding surface models [24, 85], and in the endochronic
models developed by Bazant [6] and by Valanis and coworkers (e.g., [79]).
Although originally used for the rst time by Dafalias in the context of
incrementally non{linear hardening plasticity [24], the term hypoplasticity
has subsequently been adopted by Kolymbas & Wu [88] to refer to a class
of incrementally non{linear, rate{type models developed as an extension of
the classical theory of hypoelasticity [78]. A general outline of the theory
was laid down by Kolymbas in 1991 [52], and several review papers followed
thereafter (e.g., [56, 54, 57]).
The objective of this paper is neither to give an outline of hypoplasticity
(which can be found in the aforemetioned papers) nor to present one of the
various approaches in detail. Rather, attention is focused on two particular
classes of constitutive equations developed in Karlsruhe by Kolymbas and
coworkers [56, 54, 57] | referred to in the following as K{hypoplastic models

3
| and those developed in Grenoble by Chambon and coworkers [11, 12, 16]
under the general name of CLoE models.
The problem of whether the latter can be termed \hypoplastic" | as
they in fact should, according to the de nition given in [88] | is virtually
irrelevant. What is more interesting, in the authors' opinion, is that although
these two approaches share a number of similarities, the motivations for their
| independent | development were di erent in various respects. As a result,
major di erences are apparent in their original formulation, as well as in their
respective subsequent developments.
This paper is an attempt of critically reviewing the two classes of models
with respect to a number of features which are of importance in the light
of their application to both the fundamental understanding of granular soil
behavior, and the analysis of practical problems in geotechnical engineering.
It is the authors' hope that through such a comparative analysis, a better
understanding can be obtained of both possibilities and limitations of the
theory of hypoplasticity.
In following, the usual sign convention of solid mechanics is adopted through-
out. In line with the principle of e ective stress as stated by Terzaghi [74],
all stresses are e ective stresses, unless otherwise stated. Both direct and
indicial notations will be used to represent vector and tensor quantities ac-
cording to convenience. In the description of stress states, use will sometimes
be made of the three invariant quantities p (mean stress), q (deviator stress)
and  (Lode angle), de ned as follows:
r
p
p := ; 3 tr() q := 32 tr(s ) cos(3) := ; 6 tr(s ) =
1 3
2
(1)
[tr(s )]
2 3 2

where s :=  + p1 is the deviatoric part of the stress tensor, 1 is the unit

tensor ([1]ij = ij ), and tr(x) := x
1 for any 2nd order tensor x.

2 Mathematical structure
2.1 Preliminaries
A rigorous mathematical de nition of a hypoplastic material has been laid
out by Wu & Kolymbas [88] and Kolymbas [52]. Herein, the attention will
be focused on a particular subclass of hypoplastic constitutive equations,
characterized by the following fundamental properties:
1. Let x = (X ; t), (X ; t) : B [0; T ] 7! St , denote the motion of a given
body from its reference con guration B to the current con guration
4
 St at time t, and F (X ; t) := @=@ X the corresponding deformation
gradient. Denoting with F Xt (s), s  0, the history of the deformation
( )

gradient up to time t at material point X , the response functional G

entering in the general form of the constitutive equation for a simple
material [78]:
1 
(t) = sG F =0
t
X (s )
( )
(2)
satis es the identity:
1  1 
(t) = sG F Xt (s)
=0
( )
= G
s=0
F Xt [(s)]
( )
(3)
for every monotonically increasing function  (s) such that  (0) = 0 and
lim  (s) = 1.
s!1
2. There exist suitable tensor{valued functions h : S2  S  Q 7! S ,
2 2

A : S  Q 7! L and b : S  Q 7! S such that:

2 4 2 2

r = h (; d; q) = A (; q) d + b (; q) kdk (4)

where S is the space of symmetric second{order tensors; L is the space
2 4

of fourth{order tensors; Q is the space of the additional state variables

q which de ne the previous history of the material, and whose evolution
is governed by suitable evolution equations in rate form;
r := _ + ! ; ! (5)
is the Jaumann{Zaremba stress rate;
l := FF
_ ; 1
d := 21 ;l + lT
! := 12 ;l ; lT
(6)
are the spatial velocity gradient,pthe rate of deformation and the spin
tensor, respectively; and kdk := d
d is the Euclidean norm of d.
REMARK 1. Property (1) states that the behavior of the material is not in-
uenced by any change in the time scale, i.e., the material is rate{independent.
As shown by Truesdell & Noll [78], this property implies that the function h
is positively homogeneous of degree 1 in d:
h (; d; q) = h (; d; q) 8 > 0 (7)
This hypothesis rules out some recent versions of argotropic hypoplastic
models, which introduce in the basic formulation (4) some form of rate{
dependence, see, e.g., [52, 2, 38, 61].
5
REMARK 2. According to the principle of material frame indi erence, the
function h must be an isotropic tensor{valued function of all its arguments,
see [78]. For example, if q 2 R 3 as in [87], this means that for any orthogonal
tensor Q:
h ;QQT ; QdQT ; Qq
= Qh (; d; q) QT (8)
REMARK 3. A fundamental di erence between hypoplasticity and hypoe-
lasticity [78] is represented by the term b kdk on the RHS of eq. (4), which
renders the function h non{linear in d. This feature is necessary in order
to correctly describe irreversible behavior upon stress reversal along a given
loading path. However, as pointed out by Chambon [12], a possible limi-
tation | to be evaluated through experimental veri cation | is that, due
to the particular structure of the rate equation (4), for any given state and
strain rate d:

r := h (d) ; h (;d) = 2b kdk (9)
i.e., the di erence between the stress rates upon reversal of the rate of defor-
mation is independent on the direction of d.
REMARK 4. When Q = ;, i.e., when Cauchy stress is the only state vari-
able, eq. (4) de nes an amorphous hypoplastic material according to the ter-
minology laid out by Kolymbas [52]. A well{known limitation of amorphous
hypoplasticity is related to cyclic loading, where strong ratcheting e ects are
usually predicted for non{symmetric cyclic paths, see, e.g., [12, 3, 16, 62].
The opposite case of endomorphous hypoplastic materials is obtained when
other additional state variables are present in the formulation. In particu-
lar, when one of the additional state variables is the porosity (or void ratio)
eq. (4) de nes a pycnotropic hypoplastic model [52].
REMARK 5. The experimental observation that a proportional strain path
starting from a nearly stress{free and undistorted state yields a proportional
stress path [32, 40, 33] poses some additional restrictions on the function h.
In the particular case of amorphous hypoplastic materials, it can be shown
that a sucient condition to satisfy this requirement is that h be positively
homogeneous with respect to :
h (; d; q) = mh (; d; q) 8 > 0 (10)
where m denotes the degree of homogeneity. This property, de ned barotropy
in [52], implies that the behavior of the material can be normalized with
respect to pm .

6
2.2 K{hypoplasticity
The development of K{hypoplasticity can be traced back to the pioneering
work of Kolymbas on incrementally non{linear constitutive equations of the
rate{type [48, 41, 50]. Although none of the aforementioned theories can
be classi ed as K{hypoplastic, according to the de nitions laid out in the
previous section, the approach followed in these works for the de nition of the
constitutive function h is the same which lead to the subsequent development
of K{hypoplasticity.
The starting point in Kolymbas' approach is given by the condition of
isotropy (8) for the constitutive functions to be de ned. As | in the partic-
ular case of amorphous materials | a representation theorem due to Wang
[84] provides the most general form of isotropic tensor{valued function of two
tensorial arguments, Kolymbas suggested to derive particular expressions for
the function h by appropriately selecting some terms in Wang's general ex-
pression by trial and error (see, e.g., [53] for details).
Employing this procedure, several particular amorphous K{hypoplastic
models have been proposed and subsequently re ned, see e.g., [88, 86, 92].
As an example, the constitutive model developed by Wu & Bauer [92] is
characterized by the following functional form for the tensors A and b:
A := tr()L b := tr()N (11)
L := C I + C ^ ^
1 2 N := C ^ + C ^
3
2
4
2
(12)
in which: I is the (symmetric) fourth{order identity tensor (2Iijkl = ik jl +
il jk ); ^ := = tr() is a normalized stress tensor; ^ := ^ ; (1=3)1 its devi-
atoric component, and Ck , k = 1; : : : ; 4, are material constants. The model
is characterized by a barotropy of degree one, and its calibration requires
the determination of only 4 material constants, which can be obtained from
standard laboratory tests.
Subsequent developments of K{hypoplasticity have been aimed at improv-
ing the original formulation in the following aspects:
i) Description of barotropy and pycnotropy e ects, in order to model the
dependence of material response on density and e ective stress level,
and to incorporate the concept of critical state [91, 1, 38, 95, 83];
ii) De nition of suitable calibration procedures, associated to a more straight-
forward physical interpretation of the various constants entering the
model, linked to some fundamental properties of the solid skeleton as
grain properties and grain size distribution [1, 45];

7
 iii) Extension of the space of state variables to some suitable structure
tensor in order to: a) allow the modeling of cohesive powders [4]; b)
improve the performance of standard formulations for cyclic loading
[3, 62]; or, c) introduce inherent anisotropy [87];
iv) Extension of the range of application of the theory to multiphase gran-
ular materials [36, 39], and micropolar media [72, 73].
The aspects covered in the developments referenced in points (ii) to (iv) are
beyond the scope of the present work, and thus will not be discussed further.
However, some theoretical formulations referenced in point (i) deserve some
more detailed comment, as they represent a natural evolution of the original
amorphous hypoplastic models.
In particular, two recent, advanced K{hypoplastic models proposed by
Gudehus and Bauer [38, 1] and von Wol ersdor [83] are worth mentioning,
as they can be considered the syntesis of the research work carried out in
Karlsruhe on this subject over the last 25 years. For both models, the general
constitutive equation (4) can be written as follows:
r = fbfeL (^) d + fbfefdN (^) kdk (13)
where:
L := a I + ^ ^
2
1 N := a (^ + ^ )
1 (14)
for the Gudehus/Bauer model, and:
L := ^ 1
^ ;F I + a ^ ^

2 2
N := ^aF
^ (^ + ^ ) (15)
for the von Wol ersdor model. The functions a , F , a, fe, fd and fb entering
1
in eqs. (14){(15) are given for completeness in appendices A and B.
As it is apparent from the comparison between eqs. (11) and (13), the fun-
damental progress from the older generation of amorphous K{hypoplastic
models is represented by the introduction of barotropy and pycnotropy ef-
fects via the scalar factors fb, fe and fd, which allow a uni ed description of
the behavior of granular materials for a wide range of pressures and densi-
ties. For any given rate of deformation, the functions fb and fe control the
\directional sti ness" khk = kdk of the material. In particular, the barotropy
factor fb accounts for the e ect of mean stress while the rst pycnotropy
factor fe accounts for the e ect of densi cation (pycnotropy). The second
pycnotropy factor fd , which operates only on the tensor N , controls the bal-
ance between the linear and the non{linear parts of the function h as the
8
density of the material is changed. A detailed explanation of the role played
by the three factors is given in [1, 38]. An important consequence of the
particular structure of eq. (13) is the possibility of an independent calibra-
tion of the di erent material constants entering in the constitutive functions
(separability, see [38, 1]).

2.3 CLoE hypoplasticity

The origins of CLoE hypoplasticity | where the acronym CLoE stands for
Consistance et Localisation Explicite | can be traced back to the pioneering
work of Chambon and Desrues on strain localization in incrementally non{
linear materials [27, 14, 28].
In analyzing the problem of shear band analysis for incrementally non{
linear materials, Chambon and Desrues observed that a completely general
non{linear bifurcation analysis could be carried out for the following \heuris-
tic" model:
r
s= d ; s kdk (16)
in which  and  are material constants. Based on this observation, the
basic structure of the heuristic model has been subsequently assumed as
the starting point for the development of a class of incrementally non{linear
constitutive equations sharing with eq. (16) a formally identical structure |
i.e., eq. (4) | which are both capable of representing the actual behavior of
real geomaterials, and allow a complete mathematical analysis of the non{
linear shear band localization problem. This particular feature of CLoE, as
well as K{hypoplastic models, is discussed in detail in sect. 4.
In the development of CLoE models, the following basic assumption have
been introduced to derive speci c functional forms for the tensors A and b
in eq. (4):
1. To keep the formulation as simple as possible, the set of state variables
is limited to the Cauchy stress tensor. Therefore the general eq. (4)
reduces to:
r = A () d + b () kdk (17)
As already discussed in sect. 2.1, this prevents an appropriate descrip-
tion of cyclic behavior along some particular stress paths;
2. According to the experimentally observed behavior of granular material
and the previous assumption (1), the domain of admissible states B =

9
 B [ @ B is bounded by a surface @ B , formally de ned through a isotropic
tensor function () as:

@ B := 2S 2
() = 0 (18)
and referred to as limit surface.
3. The constitutive equation resulting from the de nition of the tensor
functions A and b in eq. (17) is invertible in all points of B except on
the limit surface, i.e.:
8 (; r ) 2 B  S 9 d 2 S 2 2
r = A () d + b () kdk
REMARK 6. Due to assumption (1), the tensors A and b admit the following
representation as a 6  6 matrix and a six{components column vector in the
reference frame of the principal stress directions [9]:
2 3 2 3
A 1111 A 1122 A1133 0 0 0 b 11
6A A A 0 0 0 7 6b 7
6 2211 2222 2233 7 6 722

A = 66A
6
3311
0
A 3322
0
A3333
0 A
0 0
0
0
0
7
7
7 b = 666b0 777
33
(19)
6 1212 7 6 7
4 0 0 0 0 A 2323 0 5 405
0 0 0 0 0 A 3131 0
Note that this results holds also for K{hypoplastic models for which the
stress tensor is the only tensor{valued state variable. However, while all K{
hypoplastic models previously discussed are de ned in such a way that A
presents minor as well as major symmetries, this is in general not true for
CLoE models, for which Aijkl 6= Aklij .
REMARK 7. Assumptions (1) and (2) imply that for any stress state  2
@ B the stress rate predicted by eq. (4) is directed inside the limit surface,
regardless of the direction of the rate of deformation, i.e.:
@
r  0 (20)
@
The above condition has a direct geometrical interpretation in terms of Gude-
hus' response envelopes [34]: the response envelope corresponding to a stress
state on the limit surface must be tangent to () = 0 [11, 12, 16].
REMARK 8. As rst demonstrated in [11, 12], assumption (3) holds if and
only if:

10
 i) the tensor A is invertible, thus the constitutive equation (17) can be
rewritten as:
r = A [d + B kdk] where: B := A; b 1
(21)
ii) the tensor B is such that:
kB k ; 1 < 0 (22)
An alternative proof is given by Chambon in a contribution to this conference
[13]. As the condition:
^() = kB k ; 1 = 0 (23)
de nes a surface | referred to in the following as invertibility surface |
which must include all admissible stress states in order for the constitutive
equation to be invertible at all points, the assumption (3) implies that limit
surface and invertibility surface must coincide:
()  ^() ) () = kB k ; 1 = 0 (24)
Eq. (24) and (20) represent a fundamental constraint in the development of
functional forms for the tensors A and b, known as consistency condition
[11, 12, 16]. This point is detailed further in the following sect. 3.
REMARK 9. A second important consequence of assumption (3) is more
directly related to experimentally observed behavior of granular material.
According to the de nition given by Chambon [12, 13], as loss of invert-
ibility can be associated to the occurrence of material softening, all CLoE
models are formulated in order to describe a non{softening behavior under
homogeneous deformation, i.e. for such loading conditions as, e.g., drained
TX compression and extension, the predicted stress{strain response is mono-
tonic, and the maximum value of deviatoric stress is obtained asymptotically
for large (theoretically in nite) deformations. The basic idea underlying this
| admittedly strong | hypothesis is that the commonly observed response
on such materials as, e.g., dense sands | showing a well de ned peak stress
at a relatively small strain, and a subsequent reduction of deviatoric stress as
axial deformation is increased | is a consequence of geometric, rather than
material softening, due to strain localization into shear bands, or to other dif-
fuse heterogeneous deformation modes (i.e., barrelling or bulging). Although
true material softening can actually occur in geomaterials due to some kind
of damaging processes at the microscopic level, some experimental observa-
tions | see e.g., [30] | suggest that this can be considered as a convenient
11
working hypothesis for such materials as uncemented sands. In this respect,
it is also worth noting that this assumption is consistent with the main mo-
tivation for the development of CLoE hypoplasticity, i.e., the possibility of a
complete non{linear analysis of shear band bifurcation processes.
Di erently from K{hypoplastic models, the strategy adopted to develop
speci c functional forms for the constitutive tensors A and b in the rst
generation of CLoE models, i.e. CLoE v1.00{v1.02 and subsequent minor
modi cations [43, 29, 16], is based on the following interpolation procedure.
A number of special loading paths, known as basic paths, are de ned along
which the material response to particular loading conditions is described via
suitable response functions interpolating experimentally observed data (e.g.,
a ; r = f (a) for TX compression). For each stress state, a set of image
points are then de ned along the basic paths, for which A and b can be easily
determined by di erentiating the response functions. The actual values of
A and b are then evaluated by interpolating the corresponding tensors at
the image points according to the actual values of the Lode angle and the
normalized deviatoric stress q=ql , where ql is the deviator stress on the limit
surface corresponding to current values of p and .
In performing the interpolation, the consistency condition at the limit sur-
face, eq. (20), is enforced through a suitable rotation of tensor A. To ease the
calibration procedure and to link material constants to commonly observed
features of soil behavior, the basic paths are selected among those which are
experimentally accessible by means of standard laboratory equipments |
i.e., triaxial compression and extension; isotropic compression starting from
isotropic and anisotropic stress states. These conventional laboratory tests
can provide informations on all the unknown components of the constitutive
tensors except for the \shear moduli" A , A and A . The question
1212 2323 3131
1

of shear moduli identi cation is discussed in more detail in the following

sect. 4.3, in connection with shear band analysis.
This approach presents the advantage of directly linking the constitutive
functions to the observed material behavior along an as large as possible
set of loading paths, and should be contrasted with common practice in the
development of elastoplastic constitutive models, where the mathematical
formulation of the constitutive equations is generally based on experimental
investigation on a much more limited number of stress{paths. However, this
advantage is paid in terms of a relatively large number of parameters required
to describe the response functions, and of the complexity of the interpolation
procedure, which does not allow to de ne an explicit form for the functions
A() and b().
1 Here, A components are de ned in the principal stress space, see remark 6.

12
 While recent developments of K{hypoplasticity aimed at extending the
eld of application of older versions of the theory by introducing the modi -
cations already discussed in the previous section, the development of CLoE
hypoplasticity followed a completely di erent path. In fact, starting from
the basic requirements (1){(3), new versions of the theory have been devel-
oped by Chambon and his coworkers [17, 23, 19, 20] which represent drastic
simpli cations of the original CLoE v1.00, and which, for this reason, will be
referred to in the following as MiniCLoE models.
In MiniCLoE models the concept of interpolation of the constitutive ten-
sors from the image points along the basic paths is abandoned, and | in
analogy to K{hypoplasticity | the tensors A and B in eq. (21) are de ned
explicitly, taking into account the fundamental constraints of invertibility
and consistency to the limit surface.
An example of MiniCLoE model is given by the so{called \von Mises
CLoE". As it is clear from its name, the limit surface is given by the von
Mises yield function of classical perfect plasticity:
r
() := q ; k = 3 ksk ; k = 0 (25)
2
where k is a material constant, representing the deviator stress on the limit
surface. The expression for the tensor A is directly derived from the sti ness
tensor for isotropic linear elasticity, except for the \shear moduli" A ,
A and A , while the tensor B is directly obtained by imposing the
1212

2323 3131
consistency and invertibility conditions, eqs. (20), (22). In the vector/matrix
representation of eq. (19), the corresponding expressions for A and b are
given by [23]:
2 3
K + 4G=3 K ; 2G=3 K ; 2G=3 0 0 0
6K ; 2G=3 K + 4G=3 K ; 2G=3 0 0 0 7
6 7
6K ; 2G=3 K ; 2G=3 K + 4G=3 0 0 0 7
A := 66 0 0 0 j (q) 0 0
7
7 (26)
6 7
4 0 0 0 0 j (q) 0 5
0 0 0 0 0 j (q )
with:
 q
j (q) := G 1 ; ! k

(27)

13
and:
2p
p3=2 (s11 =k )(2 ; q=k)3
6 3=2 (s22 =k )(2
6p ; q=k)77
B := 66 3=2 (s33 =k)(2
6 ; q=k)77 (28)
6 0 7
7
4 0 5
0
where K and G represent the bulk and shear sti ness parameters, respec-
tively, and ! is a material constant controlling the evolution of shear moduli
with the normalized deviator stress q=k.
Although MiniCLoE models do not allow an accurate description of the
mechanical behavior of real geomaterials due to their relatively simple math-
ematical form, they have been speci cally developed in order to to develop a
consistent interface constitutive equation to model post{localization behav-
ior (Daphnis model, [18, 23]), and to test some fundamental results on: i)
shear band localization, and, ii) existence and uniqueness of the general BVP
in rate form [17, 7, 21, 19, 20].

3 Invertibility, consistency and limit states

3.1 General results
As already discussed in the preceding section, the assumption of invertibil-
ity of the rate constitutive equation for all admissible stress states  2 B
represents one of the major di erences between CLoE and K{hypoplastic
models.
In order to discuss this aspect in more detail, it is useful to recall some
fundamental results rst obtained by Chambon in [11, 12]. Let S be a unit
tensor representing the direction of the stress rate such that:
r = S (29)
where   0 is the stress rate norm. Provided that the tensor A is invertible,
the constitutive equation (21) is invertible if, for any possible S , there exist:
i) a unit rate of deformation tensor d, with kdk = 1, and, ii) a stress rate
norm  > 0, such that:
d = H ; B with: H := A; S 1
(30)

14
From eq. (30) and the condition kd k = 1, the following quadratic equation
is obtained for the scalar unknown :
 kH k ; 2H
B + kB k ; 1 = 0
2 2
(31) 2

When eq. (31) admits only a single real and positive solution, then the rate
of deformation corresponding to the assumed stress rate direction is unique,
and is given by eq. (30) . It is trivial to show that this condition is met if
1
and only if the zero{order term in the quadratic equation (31) is negative,
i.e., when (22) is satis ed.
According to eqs. (11){(15), for the K{hypoplastic models discussed in
sect. 2.2, the two second{order tensors appearing in eqs. (31) can be written
as follows:
H = fe1fb L; S B = fdL; N
1
(32) 1

where it is assumed that for amorphous constitutive equations as (11), the

barotropy and pycnotropy functions fe, fd and fb are given by:
fe = fd = 1 fb = tr() (33)
Following von Wol ersdor [83], a general expression for the tensor L; 1

appearing in (32) is given by:

L; = ;1 + 1 k^ k
;1 + k^ k
I ; ^ ^ 
1
2
(34) 2
2
2
1 2

in which:
1 =C 1 2 = C =C 2 1 (Wu & Bauer model) (35)
1 =a 2
1 2 = 1=a 2
1 (Gudehus/Bauer model) (36)
1 = F = k^ k
2 2
2 = a =F 2 2
(von Wol ersdor model) (37)
Eq. (31) then specializes to K{hypoplasticity as follows:
 
 2
L;1S 2
; 2 fffd
; ;1
; ;1
L S
L N

(fefb ) 2
e b
+ fd 2
L; N ; 1 = 0
1 2
(38)
and the second invertibility condition, eq. (22) reads:
jfdj L; N < 11
(39)

15
3.2 Amorphous hypoplastic models
For amorphous hypoplastic materials with barotropy of the rst order |
as all CLoE models of the rst generation, or Wu & Bauer model | the
invertibility conditions (22) or (39) are met in the interior of the invertibility
surface:
(
B = AL; Nb (CLoE)
;
^ () = kB k ; 1 = 0 with:
1

(40)
(K{hypo) 1

Due to the positive homogeneity of A, b, L and N with respect to , see

remark 5, the invertibility surface de ned by eqs. (40) is a cone with the
vertex at the origin.
A direct inspection of eqs. (30) reveals that for stress states on the invert-
ibility surface:
 = 0 , r = 0 , d = ;B (41)
that is, there exists a particular rate of deformation direction for which the
stress rate vanishes. For this reason, and motivated by an analogous de ni-
tion introduced in the theory of hypoelasticity [77, 76], the surface ^ () = 0
has been also referred to in several works on K{hypoplasticity as yield sur-
face, e.g. [52, 93, 54], or failure surface, e.g. [92, 95, 94, 96].
REMARK 10. Due to the fundamental assumptions (2) and (3), in all CLoE
models the invertibility surface is assumed coincident with the limit surface
(), and the constitutive tensors A and b are de ned in such a way to
satisfy the consistency condition (20), see sect. 2.3.
On the contrary, in amorphous K{hypoplastic models the invertibility sur-
face does not represent a boundary for admissible stress states [52, 93, 96],
due to the di erent strategy employed in the formulation of the relevant con-
stitutive tensors. A geometrical explanation for this result can be obtained
by considering in more detail the structure of eq. (38) which, for a state on
the limit surface, reduces to:
 L; S ; 2f ;L; S

;L; N
= 0
2
1 2 1 1
(42)
d
fefb
Apart from the solution  = 0, the other possible solution of eq. (42) is
subject to the condition that  must be positive, as it represents the norm
of the stress rate. Therefore, as fb , fe and fd are positive , the stress rate
2

A limit condition of vanishing stress rate is possible for a cryptoplastic state | i.e.,
2
when e = ed and fd = 0 | only if A is non{invertible.

16
direction is constrained by the following inequality:
; ;1
; ;1
L S
L N
= S
;L;T L;
N > 0 1
(43)
which means that the stress rate direction must remain con ned in the half{
space de ned by the plane:
; ;T ;1

(S ) = S
L L N =0 (44)
tangent to the six{dimensional stress{response envelope at the origin of the
stress rate hyperspace. A schematic 2{d representation for the case of ax-
isymmetric loading is shown in g. 1a. Now, when the stress rate space is
superimposed to the actual stress space with the origin at the current stress
state, the plane (S ) = 0 does not coincide, in general, with the tangent to
the invertibility surface ^() = 0. As in this case the invertibility surface
cuts the response envelope, there exist some particular rate of deformation
directions for which the stress rate is directed outside the invertibility sur-
face ( g. 1b). By appropriately selecting the loading path [93, 96], it is thus
possible to reach stress states for which invertibility is lost and a homoge-
neous bifurcation condition occurs, as two di erent stress rates are generally
associated to a given rate of deformation directions. This particular behavior
requires a special consideration when solving practical boundary value prob-
lems, as in this case | as well as when shear band bifurcation occurs, see
the following sect. 4 | special numerical techniques are required in order to
obtain objective numerical solutions for those BVPs for which uniqueness of
solution is lost.
REMARK 11. The consistency condition for CLoE models, discussed in the
previous sect. 2.3 (remark 7), can be interpreted geometrically in the same
way, i.e., as a restriction imposed on the possible forms of the tensors A and
b in order to have the plane (S) = 0 coincident to the tangent plane to the
invertibility/limit surface ^() = () = 0. In mathematical terms this can
be expressed as:
@ = @ ^ = ; ;A;T A;
b ( > 0) (45)
@ @
1

Eq. (45) has been rst derived by Chambon in [11, 12].

In [96] and [57], it has been argued that a pathological consequence of
the enforcement of consistency condition (45) in CLoE v1.00{v1.02 models
is that the directional sti ness of the material tends to zero for any possible
stress rate direction as the stress state approaches the limit surface, i.e., the
response envelope shrinks to a point. This would imply that limit states
17
 ŷ(s ) = 0

s11 S11
S11
p(S ) = 0
- L-T L-1 N
Ñŷ
p(S ) = 0 s 2S33
- L-T L-1 N

lS (d ) 2S33
lS (d )

2s33

(a) (b)

Figure 1: Schematic representation of a stress response envelope for kB k = 1.

could not be left, once they are reached. In fact, this is not the case, as it
can be easily shown by considering again eq. (31), written for a point on the
limit surface:
 kH k ; 2H
B = 0
2 2
(46)
The response envelope can be considered shrinking to a point if, for any
admissible stress rate directions, the solution  = 0 is a double root of
eq. (46), that is, if:
H
B = S
;A;T A;
b = 0 8S 2 S 1
(47) 2

Given the assumed regularity conditions for A, this can happen only for
stress rate directions tangent to the plane (S ) = 0, but not for all stress
rate directions.
A heuristic proof of the above result is given in g. 2, in which a number
of response envelopes for axisymmetric loading have been drawn for CLoE
v1.00, starting from di erent axisymmetric stress states with r = ;100:0
kPa, and assuming kdk = 1:0  10; s; . In deriving the response evelopes
3 1

the set of material constants suggested in [71] for dense Hostun sand has been
adopted. As it is clear from g. 2a, the envelopes reduce in size and rotate as
the current stress state approaches the limit surface, but do not shrink to a
point. From the enlargement in g. 2b it can be appreciated that the response
envelope for the stress state on the limit surface (point F) is characterized

18
sa (kPa) sa (kPa)
limit surface
-600 F
-625

F
-575 directional stiffness
in unloading
-400
(b)
-525
isotropic axis
-100 -150 -200
-200 2s r (kPa)

sa

(a)
0
sr sr
0 -200 -400
2s r (kPa)
sa

Figure 2: Stress response envelopes for CLoE v1.00 under axisymmetric load-
ing conditions.

by a relatively high but nite aspect ratio, and that the directional sti ness
for unloading at constant radial stress is, in fact, relatively large due to the
rotation imposed by the consistency condition.
REMARK 12. Although in amorphous K{hypoplastic models the invertibil-
ity surface does not represent, in general, a boundary to all the admissible
stress states, the existence of a limit surface | or bound surface in the termi-
nology adopted in [52, 93, 96] | must be predicted by the model to comply
with the well{established experimental evidence that there exist some stress
states which cannot be reached by any possible strain path. In other words,
the region of the stress space accessible via the integration of the constitutive
equation along any conceivable strain path must be bounded. Moreover, this
region should be checked against experimental evidence in order to prevent
the occurrence of some unreasonable stress conditions, e.g., with one or more
positive (i.e. tensile) principal values.
Unfortunately, the above requirements cannot be considered satis ed a
19
priori for amorphous K{hypoplastic constitutive equations, and have to be
checked on a case{by{case basis. As a matter of fact, Wu & Niemunis [93]
report that \[ : : : ] only a suitable choice of the tensorial expressions for
L and N leads to the existence of the bound surface", while Gudehus &
Kolymbas [42] observe that in some cases, the domain of admissible states
for incrementally non{linear constitutive equation developed along the lines
discussed in sect. 2.2 might include a non negligible region of tensile stresses.
An analytical method for locating the bound surface has been given by
Kolymbas in [52]. However, Wu & Niemunis [93, 96] observed that Kolymbas
approach fails for non{axisymmetric stress states. To overcome this prob-
lem, they propose an alternative procedure, based on a expression formally
equivalent to eq. (45). However, as in K{hypoplasticity the gradient @ =@ 
is not known in advance, the determination of the bound surface has to be
carried out numerically.
It is the writers' opinion that, in this respect, the explicit de nition of a
limit surface in the mathematical formulation represents a de nite advan-
tage of CLoE models as compared to equivalent, amorphous K{hypoplastic
formulations. In fact, although this assumption is by no means necessary,
it is certainly convenient from a practical point of view, expecially when
the solution of boundary value problems with the nite element method is
concerned, as:
i) it allows a straightforward check on the occurrence of physically impos-
sible states due to inaccurate numerical integration at the stress{point
level; and,
ii) in conjunction with the consistency condition (24), it guarantees the
invertibility of the rate equation, without which the well{posedness of
some particular (rate) boundary value problem | and thus the con -
dence on the numerical results obtained | could be lost.

3.3 Endomorphous hypoplastic models

For endomorphous hypoplastic materials of the type discussed in sect. 2.2,
the invertibility condition for the constitutive equation in rate form is given
by eq. (39). Due to the pycnotropy function fd , the invertibility surface is
de ned, in this case, in a state variable space including stress and void ratio:
^ (; e) = [fd (e)] L; N ; 1 = 0
2 1 2
(48)

20
Similar considerations apply for the limit (or bound) surface, which | if
existing | could be described by an analogous equation:
(; e) = 0 (49)
However, in this case no analytical or numerical procedures are available for
its determination.
As for amorphous hypoplastic models, when loss of invertibility occurs,
a limit state is reached characterized by a vanishing stress rate, and by an
associated rate of deformation direction:
=0 , r = 0 , d = ;fdL; N 1
(50)
In addiction, the inclusion of void ratio in the set of state variables allows to
embed the concept of critical states introduced by Casagrande [8].
According to Casagrande's de nition, critical states are those states (c; ec)
for which the stress rate vanishes under continued isochoric deformation:
e = ec = const. dc
1 = 0 L (^c) dc + fd;cN (^c) kdck = 0 (51)
Eqs. (51) and (51) pose an additional restriction to the possible choices for
the tensors L and N , since at critical state:
2 3

dc = ;fd;cL; (^c) N (^c) ) L; (^c) N (^c)
1 = 0 (52)

kdck
1 1

The consequences of this restriction on the functional form of the second{

order tensor N are detailed in [83].
The projection of the critical state locus onto the (e; p) space, describing
the evolution of critical void ratio with mean stress level, is given by the
empirical relation (93). In order for the projection of the critical state locus
onto the stress space to be represented as a single critical stress surface,
the pycnotropy function fd;c at critical states must be independent of e. In
particular, in both Gudehus/Bauer [38, 1] and von Wol ersdor [83] models
the function fd has been selected in order to have fd;c = 1, see eq. (91).
By taking the squared norm of eq. (52) , the equation of the critical stress
1
surface is obtained as:
 ) = L; N ; 1 = 0
c(
1 2
(53)
As eq. (53) is homogeneous of zero order in , it represents a cone in the
stress space with the vertex at the origin.

21
REMARK 13. According to eqs. (48), (53) and (91), when the material is
sheared along a loading path with a prevalent deviatoric component (i.e.,
in drained or constant p strain controlled TX compression tests), the locus
of limit states | in the sense de ned by eq. (50) | is found above the
critical state cone when the material is denser than critical (dense sands,
ep < ec and fd;p > 1), while peak and critical states generally coincide when
the initial density of the material is lower than critical (loose sands, fd;p =
fd;c), see [1]. This pattern of behavior appears in substantial agreement
with experimental observations, as demonstrated by the comparison between
measured and predicted peak friction angles reported in [1] for a medium
quartz sand with various initial densities.
REMARK 14. An interesting consequence of the mathematical structure of
endomorphous K{hypoplastic models described in sect. 2.2 can be deduced
by specializing eq. (50) to a particular constitutive equation. As an example,
for the Gudehus/Bauer model [38, 1] eq. (50)3 reads:
h i
d = ;fd a ;a2 +1 k^ k2
;a21 ; k^ k2
^ + ;a21 + k^ k2
^ (54)
1 1

from which:
d
1 = ;fd a ;a +1 k^ k
;a ; k^ k

2 2
2
1
2
(55)
1 1

since ^
1 = 1 and ^
1 = 0. To correctly recover a condition of isochoric

ow, at critical states k^ ck = a . Therefore, at peak states located outside
1
the critical state cone; (i.e., when
failure occurs at void ratios lower than
critical) the quantity a ; k^ k is negative, and, according to eq. (55) the
2 2

associated rate of deformation is dilatant (d
1 > 0). As the material dilates,
1

the void ratio is progressively increased and fd reduces accordingly, until a

condition of critical state is reached at a lower stress ratio and higher void
ratio.
This behavior is not limited to a particular deviatoric path in stress space.
According to Gudehus [37], for K{hypoplastic models of this type, critical
states can be considered as asymptotic states (attractors) for all deviatoric
paths leading to steady state ow. However, as the introduction of the con-
cept of critical state in the mathematical structure of the constitutive equa-
tion implies the possibility of developing material softening for some partic-
ular initial states and loading paths, this development of K{hypoplasticity
is not compatible with the assumption of invertibility on which CLoE hy-
poplasticity is based.

22
REMARK 15. Although for the aforementioned reason no direct comparison
is possible between CLoE and endomorphous K{hypoplasticity, it is interest-
ing to note that | similarly to all CLoE formulations | in K{hypoplastic
models of this type, the shape of the critical state cone is assumed a pri-
ori. For example, in von Wol ersdor model [83], the various functions
entering the constitutive tensors (see App. B) are de ned in order for the
function c = 0 to be coincident with the Matsuoka/Nakai failure condition
[59]. This is to be contrasted with earlier K{hypoplastic formulations, see
e.g. [93, 92, 95], for which the invertibility (failure) surface is derived as by{
products of the particular assumptions for the constitutive tensors L and
N.
4 Strain localization and bifurcation analysis
4.1 Introduction and historical background
Cino's notes

The theoretical problem of the prediction of the onset of strain localiza-

tion in non{standard, pressure dependent materials such as granular soils,
considered as a assumed mode (shear band) bifurcation process in the sense
of Rice [67], has received much attention since the early works of Rice and
Rudnicki [70, 67, 68].
As a matter of fact, the inadequacy of classical ow theories of hardening
plasticity with a single mechanism to provide a satisfactory description of ex-
perimentally observed shear banding processes was certainly one of the main
motivations for the development of incrementally non{linear and hypoplastic
constitutive theories, see, e.g. [49, 14, 10, 16].
The rst shear band analysis for incrementally non{linear materials |
although not hypoplastic according to the de nition given in sect. 2 | dates
back to the pioneering work of Kolymbas in 1981 [49]. A key point in Kolym-
bas' approach is the assumption of a vanishing rate of deformation outside
the band.
A few years later, in 1985, Chambon & Desrues [14] presented a general
bifurcation analysis for the heuristic model de ned by eq. (16), which, as
dicussed in sect. 2.3, can be considered the originator of the subsequent
CLoE hypoplastic models. A key result in Chambon & Desrues [14] analysis
is the fact that when bifurcation occurs, the ratio between the strain rate
inside and outside the band goes to in nity, i.e., the material outside the

23
band stops deforming. Similar results have been subsequently obtained by
Loret [58] for a slightly di erent incrementally non{linear model.
A general shear band bifurcation criterion for hypoplastic materials has
been subsequently given by Chambon and his coworkers [12, 22, 43, 15],
starting from the assumption of a continuous variation of the tangent map
between rate of deformation and stress rate spaces with the state of the ma-
terial during the loading process (continuity in the large, according to the
de nition given in [44]). See [15] for details. The extension of the afore-
mentioned works to include the case of a discontinuous variation of material
\sti ness" during the loading process has been recently discussed in [19].
Non{linear shear band analyses for homomorphic K{hypoplastic models
have been presented by Kolymbas & Rombach in 1989 [55] (assuming kdk ' 0
outside the band), and by Wu & Sikora in 1991{92 [89, 90]. Recently, Bauer
& Huang [5] have extended the aforementioned studies to K{hypoplastic
formulations with barotropy and pyknotropy, discussing the dependence of
stress ratio at bifurcation and shear band inclination on void ratio and mean
stress.

4.2 General results

In the following, some general results of the non{linear shear band analysis
for general hypoplastic materials discussed in [19] are brie y summarized.
Note that, as the analysis is completely general, these results can be directly
applied to both CLoE and K{hypoplastic models.
The development of a shear band bifurcation in a homogeneous deforming
body is completely de ned in terms of the following kinematic condition
among spatial velocity gradients l := grad v:
l =l +g n
1 0
(56)
in which the indexes 0 and 1 denote the relevant quantities inside and outside
the shear band, respectively; n is the unit vector normal to the shear band;
and g is a vector de ning the velocity gradient jump in the direction n.
The requirement of continuing equilibrium at the onset of localization also
implies the following static condition:
_ n ; _ n = 0
1 0
(57)
in which _ and _ are the Cauchy stress rates inside and outside the band,
0 1

respectively.
The bifurcation condition is obtained by specifying in the static condition
the hypoplastic constitutive equation (5), which for our purposes can be more
24
conveniently rewritten in terms of the material time derivative of Cauchy
stress in the following general form:
_ = Ml + b kdk (58)
where:
M := A + C ; b := AB (59)
for CLoE models, or:
M := fsL + C ; b := fsfdN (60)
for K{hypoplastic models, and the fourth order tensor C in eqs. (59), (60),
de ned as:
Cijkl := 21 (ik lj ; il ki ; ik jl + il jk ) (61)
accounts for corotational terms in appearing in the de nition of the stress
rate, eq. (5).
From eqs. (57), (58) it follows [12, 15, 19]:
[M (g n)] n + ; d ; d
bn = 0
1 0
(62)
As eqs. (56)-(58) are homogeneous in d , without loss of generality it can
0

be assumed that d = 1, and:

0

d = (1 + r) d = 1 + r ; r  ;1
1 0
(63)
Eq. (62) can therefore be written as:
[M (g n)] n + rbn = 0 (64)
subject to the condition:
1 l +l T +g n+n g = 1+r
0 0
(65)
2
Bifurcation is possible if and only if the above system of nonlinear equations
(64), (65) has at least one solution for g and r such that r  ;1.
De ne the second order acoustic tensor P associated with M as:
Pik := Mijklnj nl (66)
the following results hold, depending on the invertibility of P [12, 15, 19].
25
CASE 1. If P is invertible, then eq. (64) can be solved for g , with:
g = ;rP ; bn 1
(67)
Eq. (65) thus reads:
d ; r sym P ; b (n n)
0 1
=1+r (68)
Shear band bifurcation is possible if and only if there exists some n for which
eq. (68) has a solution r  ;1. Eq. (68) can be written in the following
compact form:
;

N ; 1 r + 2 (S ; 1) r = 0
2 2
(69)
where:
 ;1
N := sym P b (n n) ; S := d
P ; b (n
0 1
n) (70)
The only non{trivial solution of eq. (69) is:

r = 2(1 ; S )  ;1 (71)
N ;1 2

since for r = 0 the only possible solution of eq. (64) is g = 0 (no localization).
According to eq. (71), it can be shown [19] that shear band bifurcation is
possible if and only if:
 ;1
N = sym
2
P b (n n)  1 2
(72)
CASE 2. If P is not invertible, then the bifurcation condition reads [15, 19]:
det (P ) = 0 (73)
This second condition is apparently similar to the classical bifurcation condi-
tion for incrementally linear materials (see, e.g., [67]); however, in this case,
the tensor P is derived from the linear part only of the constitutive equa-
tion. According to Chambon et al. [15], the bifurcation condition de ned
by eq. (73) is practically never met before the \norm" criterion given by
eq. (72) for CLoE models, for which the tensor A has the structure of an
elastic tensor. The same conclusion is reached by Bauer & Huang [5] for the
Gudehus{Bauer model [38, 1].

26
REMARK 16. As proven in [19], if and only if non{linear shear band bifur-
cation condition (72) holds, there exists a spatial velocity gradient l such
that:
  
det [Q (dir l )] = 0 with: Qik = Mijkl + bij dkl nj nl (74)
kd k
in which the tensor:
 := M + b kddk

(75)
represents the directional linearization of the non{linear rate equation (58).
However, this results has only a theoretical interest, since dir l is not known
a priori.
REMARK 17. The limiting case N = 1 in the bifurcation condition (72)
occurs if at the beginning of the loading process bifurcation is not possible
| which seems a fairly reasonable assumption | and the response of the
material is continuous in the large (see [15]). In this case, which is usually
the rule for hypoplastic models discussed in this work, eq. (71) yields the
two solutions r = 1 and r = ;1, according to which the ratio between
the norms of the rate of deformation inside and outside the band is in nite.
Note that the two cases of zero rate of deformation inside the band (r = ;1)
or outside the band (r = 1) are perfectly symmetric from a mathematical
point of view.
From a physical standpoint, this means that at the onset of localization,
strain rate concentrates inside the band, i.e., l is negligible with respect to l .
0 1

This theoretical result appears to be con rmed by a number of experimental

observations in both cohesive and granular soils [27, 82]. As observed in [15],
the static condition (57) implies in this case:
_
(n g) = 0 ) _
sym (g n) = _
d = 0 1
(76)
This means that, under the assumption of small strains (r ' _ ), localization
implies the loss of positiveness of the second order work.
REMARK 18. When material response changes discontinuously with the
loading parameter, abrupt bifurcation is possible with N > 1. In this case
the ratio between the norms of the rate of deformation inside and outside the
band remains nite, and shear banding becomes possible for a fan of shear
band orientations. As a direct consequence, in this case various possible bi-
furcation modes can be obtained from numerical simulation of the loading
process for the same initial conditions, depending on the details of the nu-
merical procedure employed, see [19]. This represents a clear drawback for
this kind of constitutive models.
27
REMARK 19. The shear band analyses for K{hypoplasticity mentioned in
the previous section [55, 89, 90, 5] di er from the one outlined above in that
they all rely on a particular kinematic assumption, concerning:
1. the rate of deformation outside the band (in particular, d0 = 0) [55];
 p 
2. the dilatancy  := tan; g =( 2g ) inside the band [89];
1
2 1

3. the rate of deformation norm jump [90, 5]:

[ kdk] := d ; d
0 1
(77)
Interestingly, the result outlined in remark 17 provides an a posteriori theo-
retical argument in favor of the hypothesis made in the bifurcation analyses
presented by Kolymbas and Kolymbas & Rombach [49, 55]. The hypotesis
put forward by Wu & Sikora (1991) [89], on the contrary, does not appear
to be necessary, as it represents an additional constitutive assumption for
the material inside the band. Following a di erent approach, Wu & Sikora
(1992) [90] and Bauer & Huang [5] solve the bifurcation problem posed by
eq. (64) by assuming the absolute value of the deformation norm jump:
 := d ; d 0 1
(78)
In particular, they observe that the earliest bifurcation point (lower bound
solution in [90]) is reached when:
[ kdk] = k[ d] k (79)
Eq. (79) implies the following:
d = d ;   0
1 0
(80)
which represents a condition of continuous bifurcation ( =  , see [28])
1 0

and includes the condition of vanishing rate of deformation outside (or inside)
the band as a particular case (; = 0 or  = 0). As a matter of fact, by
1

comparing eqs. (63) and (80) it follows that:

=r+1 (81)
and since for a continuous variation of the response with the loading param-
eter the only possible solutions for r at bifurcation are r = 1 and r = ;1,
the lower bound solution of Wu & Sikora [90] and Bauer & Huang [5] co-
incides with the bifurcation condition obtained by Chambon and coworkers
[14, 28, 43, 15, 19].

28
4.3 The question of the shear modulus
The principal and most obvious use of the above results in practical applica-
tions is to de ne the conditions upon which uniqueness of the solution can
be lost due to shear banding, either at the scale of laboratory tests (e.g.,
[49, 43, 15, 5] or in the analysis of full scale boundary value problems (e.g.,
[55, 31]).
However, another possible application | motivated by the availability of a
closed form solution for the nonlinear shear band problem | is the the use of
experimental localization data to obtain informations on such details of the
constitutive equations which cannot be directly de ned using experimental
data from tests with xed principal stress and strain directions.
In particular, since the pioneering works of Hill & Hutchinson [46], Rud-
nicki & Rice [70] and Rice [67], it is well known that the onset of the bifur-
cation and the orientation of the shear band can be strongly a ected by the
way in which the constitutive model describes the response of the material for
deformation paths which, upon bifurcation, change abruptly their direction.
In the particular case of rectilinear plane strain compression or extension,
this means that a major role in de ning bifurcation conditions is played by
the shear modulus in the plane of deformation (see, e.g., [80, 81, 63] for a
discussion on this topic in the framework of hardening plasticity).
Following an initial idea by Vardoulakis [80], the possibility of using exper-
imental shear band observations for the calibration of shear moduli, which
are not accessible experimentally in standard triaxial or plane strain tests
with xed principal stress and strain directions, has been rst suggested for
CLoE models in [12], and then successfully exploited in [22, 43, 15, 31] for the
particular CLoE model discussed in [16], and its subsequent modi cations.
In this respect, the particular structure assumed for the fourth order ten-
sor A represents an important di erence between CLoE and K{hypoplastic
models. In fact, as the stress tensor is the only tensor{valued state variable
in both CLoE and the K{hypoplastic models considered in sect. 2.2, the rep-
resentation of the tensors A and b given by eq. (19) (see remark 6) in the
reference frame of the pricipal stress directions holds in both cases. From
eq. (19) is immediately apparent that the three components A , A and
A of the tensor A represent the shear moduli relevant to the onset of
1212 2323

3131
localization | i.e., to a so{called loading to the side process.
If, without loss of generality, only plane strain processes in the plane x : x
1 2
corresponding to the principal directions of  and  are considered, then
1 2
attention can be focused on the only relevant shear modulus A . Note 1212
that if the loading process is such that both principal axes of stress and
strain remain xed up to the bifurcation point | as for example biaxial

29
compression tests | this shear modulus plays no role in determining the
material response before the onset of localization.
In all CLoE hypoplastic models, the shear modulus A is considered a
1212
function of the state variables according to the following general equation:
A = j (1 ; !R)
1212 0 (82)
In eq. (82), j is the shear modulus at the isotropic image point (q = 0),
0
either assumed constant (von Mises and Mohr{Coulomb MiniCLoE models,
see [17, 23]), or function of the mean stress p, to be determined starting from
the response in triaxial compression and extension (see, e.g., [43] for details);
! is a material constant, and R is a normalized deviatoric stress, de ned as:
(
R := q=ql (CLoE v1.02; von Mises MiniCLoE) (83)
tan m= tan  (Mohr{Coulomb MiniCLoE)
in which ql is the deviator stress at the image point on the limit surface,  is
the friction angle, and m represent a measure of the stress ratio in terms of
mobilized friction angle. The term (1;!R) in eq. (82) introduces a deviatoric
stress dependent stress{induced anisotropy in the shear modulus which is
consistently recovered (A = j ) as the isotropic state is approached (R !
1212 0
0). See [9] for further details.
The corresponding expressions for the shear modulus A for the three
1212
K{hypoplastic models discussed in sect. 2.2 are given by:
8
<c tr 
> 1 (Wu & Bauer)
2A = >fsa
1212
2
1 (Gudehus/Bauer) (84)
:
fsF = (
) (von Wol ersdor )
2

The quantities appearing in the above equations have been de ned in sect. 2.
From eqs. (82) and (84) it is apparent that, while in K{hypoplastic formu-
lations the shear modulus is strongly related to the response of the material
as observed in rectilinear compression or extension paths | as all the pa-
rameters entering in the di erent expressions in eq. (85) a ect the overall
predicted behavior in such kind of loading | in CLoE models the use of
the parameter ! introduces an additional degree of freedom which allows an
independent calibration of the di erent components of material sti ness in
rectilinear compression and shear.
This feature of CLoE models is exploited in the analysis of bifurcation
problems in the sense that experimental localization data can then be used to
calibrate the shear modulus | through the parameter ! | without a ecting
the response of the material in rectilinear compression or extension.
30
 In the following, a more detailed analysis of the e ects of the di erent
constitutive assumptions for the shear moduli is presented, by comparing
the predicted responses of the di erent hypoplastic models considered upon
some suitable loading condition. Such a comparison has been carried out for
the CLoE v1.03 [15] model, and for the K{hypoplastic models proposed by
Wu & Bauer (WB, [92]), Gudehus/Bauer (GB, [38, 1]) and von Wol ersdor
(vW, [83]). A series of isochoric plane strain compression tests has been
simulated, starting from an isotropic initial stress p = 100:0 kPa. Note that
the kinematic restraint imposed on testing conditions allows to rule out the
e ect of pyknotropy, which is included in vW and GB models, but is not
present in CLoE and WB formulations.
In all CloE simulations, the values of material parameters suggested in [71]
for dense Hostun sand have been adopted, except for the parameter !, which
has been varied in the range [0:2; 1:0]. The material parameters and (where
required) the initial void ratios entering in the three K{hypoplastic models
have been determined in order to match the response predicted by CLoE.
The resulting values are given in tab. 1.
In the calibration of vW and GB models, the values suggested by Herle &
Gudehus [45] for dense Hostun sand have been used as a starting point. It is
worth noting that the only notable di erence between the parameters used
in this work and those suggested in [45] is in the friction angle at critical
state, c, for which a much larger value is needed to compensate the high
friction angles assumed for CLoE in both TX compression (C = 45) and
extension (E = 46).
The results of the various simulations in terms of stress{strain curves on
q:s plane, and of stress paths on q:p plane are shown in g. 3. As no attempt
has been made in re ning the calibration of the di erent material parameters
in order to obtain the best possible match, some di erences are observed
in both stress{strain curves and stress paths. Nonetheless, the substantial
overall agreement among all predictions still allows a meaningful comparison
between the performance of the four di erent hypoplastic models, at least
from a qualitative point of view.
The evolution of shear sti ness with deviatoric strains for the four hy-
poplastic models considered is shown in g. 4 in terms of the non{dimensional
normalized shear modulus:
; := A;n n 1212
(85)
p pa
(1 )

in which pa = 100:0 kPa is the athmospheric pressure, and the barotropy

constant n is either given in tab. 1 (vW and GB models), or equal to zero
(CLoE and WB models, which assume a linear dependence of tangent sti -
31
parameter Wu & Bauer [92] Gudehus [38] v.Wol ersdor [83]
C 100.0 |
1 |
C 950.0 |
2 |
C 870.0 |
3 |
C -1450.0 |
4 |
e | 0.67
0 0.60
c | 40.0 40.0
hs (kPa) | 1.0e6 1.0e6
n | 0.3 0.3
| 0.13 0.13
| 2.0 1.0
ec | 0.96
0 0.96
ed | 0.61
0 0.61
ei | 1.09
0 1.09

Table 1: Material parameters adopted for K{hypoplastic models.

300 300
CLoE
vW
deviatoric stress, q (kPa)

deviatoric stress, q (kPa)

Gd
200 200
WB

CLoE
100 100
vW
Gd
WB
0 0
0.00 0.01 0.02 0.03 0 100 200 300

deviatoric strain, es mean stress, p (kPa)

Figure 3: Plane strain rectilinear compression tests results.

32
 600 300
(a) CLoE model (b) K-hypo models
w= 0.2 vW
norm. shear modulus, G

norm. shear modulus, G

w= 0.6 Gd
400 200
w= 1.0 WB

200 100

0 0
0.00 0.01 0.02 0.03 0.00 0.01 0.02 0.03

deviatoric strain, es deviatoric strain, es

Figure 4: Plane strain rectilinear compression tests: evolution of normalized

shear modulus with deviatoric strain.
ness with mean stress p). The adopted normalization allows to account for
the e ect of barotropy, as described by the di erent formulations.
The three curves relative to CLoE model show that the normalized shear
modulus reduces with increasing stress ratio q=p until, at deviatoric strains
s larger than about 0.02, the stress ratio remains almost constant. The
reduction of the normalized shear modulus is larger for larger values of !, as
implied by eq. (82). A similar behavior is reproduced by the two advanced K{
hypoplastic models vW and GB, although in these two cases, the evolution
of normalized sti ness with deviatoric strains is completely controlled by
the parameters which de ne material behavior under the assumed testing
conditions of rectilinear biaxial compression. On the contrary, a constant
normalized shear sti ness | which appears a rather unrealistic response |
is predicted by the WB model due to the relatively simple structure assumed
for the tensor A. Note that the same response is given by CLoE when ! = 0.
An indication of the level of stress{induced anisotropy in the shear sti ness
developed along the loading process is given by the ratio:
 := A ^ (86) 1212

G
between the current shear sti ness A and the corresponding equivalent 1212
isotropic shear sti ness
G^ := 41 [(A ; A ) + (A ; A )] 1111 1122 (87) 2222 2211

de ned in terms of the components of the 3  3 submatrix of A containing

the normal components of tangent sti ness.
33
 3 3
anisotropy ratio, Q

anisotropy ratio, Q
2 2
CLoE model
w = 0.0
K-hypo models
1 w = 0.2 1
vW
w = 0.6
Gd
w = 1.0
WB
0 0
0.00 0.01 0.02 0.03 0.00 0.01 0.02 0.03
deviatoric strain, es deviatoric strain, es

Figure 5: Plane strain rectilinear compression tests: evolution of anisotropy

ratio with deviatoric strain.
The evolution of  with deviatoric strain is shown in g. 5. As expected,
since the behavior of the material is isotropic under isotropic stress condi-
tions, all the curves start with  = 1. CLoE simulations are shown in g. 5a.
For ! = 0:2, a rapid increase of  is observed in the initial stage of the
loading process, followed by a peak and by a subsequent reduction to a fairly
stable value, as a condition of almost proportional loading with q=p ' const.
is reached. However, as ! is increased a rather di erent behavior is observed,
with much lower values of . In particular, for ! = 1:0 the anisotropy ratio
is always less than unity, and reduces steadily as deviatoric strain increase.
This rather complex behavior is due to the interplay between the di erent
e ects of stress ratio on the three components of A entering in the de nition
(86), which are controlled by ! according to eq. (82).
The corresponding curves for the K{hypoplastic models are shown in g. 5b.
For the WB model, the anisotropy ratio increases steadily up to a nal value
 ' 2:4, re ecting the decrease of the normal components A , A and 1111 2222
A = A with increasing stress ratio. The behavior predicted by the two
1122 2211
advanced GB and vW models is rather similar to that predicted by CLoE for
small ! values. However, it is worth noting that, in this case, the anisotropy
ratio remains always greater than unity.

5 Conclusions
To be written.

34
References
[1] Bauer E. (1996). Calibration of a comprehensive hypoplastic equation for
granular materials. Soils and Foundations, 36(1), 13-26.
[2] Bauer E., Niemunis A., Herle I. (1993). Visco{hypoplastic model for co-
hesive soils. In: Modern Approaches to Plasticity, Kolymbas ed., Elsevier,
365-384.
[3] Bauer E., Wu W. (1993). A hypoplastic model for granular soils under
cyclic loading. In: Modern Approaches to Plasticity, Kolymbas ed., Else-
vier, 247-258.
[4] Bauer E., Wu W. (1994). Extension of hypoplastic constitutive model
with respect to cohesive powders. Proc. Computer Methods and Advances
in Geomechanics, Siriwardane & Zaman eds., 1, Balkema, 531-536.
[5] Bauer E., Huang W. (1998). The dependence of shear banding on pressure
and density in hypoplasticity. Proc. Localization and Bifurcation Theory
for Soils and Rocks, Gifu, Adachi et al. eds., Balkema, 81-90.
[6] Bazant Z.P. (1978) Endochronic inelasticity and incremental plasticity.
Int. Journal of Solids and Structures, 14, 691-714.
[7] Caillerie D., Chambon R. (1994). Existence and uniqueness for B.V. prob-
lems involving CLoE models. Proc. Localization and Bifurcation Theory
for Soils and Rocks, Aussois, Chambon et al. eds., Balkema, 35-40.
[8] Casagrande A. (1936). Characteristics of cohesionless soils a ecting the
stability of earth lls. J. Boston Soc. Civil Engineers, 23(1), 13-32.
[9] Chambon R. (1981). Contribution a la modelisation numerique non{
lineaire des sols. These de Doctorat d'etat, Universite de Grenoble.
[10] Chambon R. (1986). Bifurcation par localisation en bande de cisalle-
ment, une approche avec des lois incrementalement non lineaires. Journal
de Mechanique Theorique et Appliquee, 5(2), 277-298.
[11] Chambon R. (1989). Une classe de lois de comportement
incrementalement nonlineaires pour les sols non visqueux, resolution de
quelques problemes de coherence. C. R. Acad. Sci., 308(II), 1571-1576.
[12] Chambon R. (1989). Bases theoriques d'une loi de comportement incre-
mentale consistante pour les sols. Groupe C.O.S.M. Rapport de Recherche.

35
[13] Chambon R. (1999). Uniqueness, second order work and bifurcation in
hypoplasticity. This Conference.
[14] Chambon R., Desrues J. (1985). Bifurcation par localisation et non
linearite incrementale: un example heuristique d'analyse complete. Proc.
Plastic Instability, Paris, Presses ENPC, 101-113.
[15] Chambon R., Desrues J., Tillard D. (1994). Shear moduli identi cation
versus experimental localization data. Proc. Localization and Bifurcation
Theory for Soils and Rocks, Aussois, Chambon et al. eds., Balkema, 101-
111.
[16] Chambon R., Desrues J., Hammad W., Charlier R. (1994). CLoE, a
new rate{type constitutive model for geomaterials. Theoretical basis and
implementation. Int. J. Num. Anal. Meth. Geomech., 18, 253-278.
[17] Chambon R., Crochepeyre S. (1995). Loss of uniqueness of the solution
in a boundary value problem. Proc. NUMOG V, Pande & Pietrusczczak
eds., 165-171.
[18] Chambon R., Crochepeyre S. (1997). Incremental non linearity, bifur-
cation and shear banding. Proc. Localization and Bifurcation Theory for
Soils and Rocks, Gifu, Adachi et al. eds., Balkema.
[19] Chambon R., Crochepeyre S., Desrues J. (1999). Localization criteria
for non{linear constitutive equations of geomaterials. Mech. Cohesive-
Frictional Materials, in print.
[20] Chambon R., Caillerie D., Desrues J., Crochepeyre S. (1999). A com-
parison of incremental behavior of elastoplastic and CLoE models. Int.
J. Num. Anal. Meth. Geomech., in print.
[21] Chambon R., Caillerie D. (1999). Existence and uniqueness theorems
for boundary value problems involving incrementally non linear models.
Int. Journal of Solids and Structures, in print.
[22] Charlier R., Chambon R., Desrues J., Hammad W. (1991). Shear band
bifurcation and soil modelling: a rate type constitutive law for explicit
localization analysis. Proc. 3 Int. Conf. on Constitutive Laws for Engi-
nd

neering Materials, Tucson, Desai & Krempl eds., ASME Press.

[23] Crochepeyre S. (1998) Contribution a la modelisation numerique
et theorique de la localisation et de la post localisation dans les
geomateriaux. These de Doctorat d'etat, Universite de Grenoble.
36
[24] Dafalias Y.F. (1986). Bounding surface plasticity. I: Mathematical foun-
dation and hypoplasticity. J. of Engng. Mech., ASCE, 112(9), 966-987.
[25] Darve F. (1991a). The expression of rheological laws in incremental form
and the main classes of constitutive equations. In: Geomaterials: Con-
stitutive Equations and Modelling, Darve ed., Elsevier, 123-148.
[26] Darve F. (1991b). Incrementally non{linear constitutive relationships.
In: Geomaterials: Constitutive Equations and Modelling, Darve ed., El-
sevier, 213-237.
[27] Desrues J. (1984). La localisation de la deformation dans le materiaux
granulaires. These de Doctorat d'etat, Universite de Grenoble.
[28] Desrues J., Chambon R. (1989). Shear band analysis for granular ma-
terials: the question of incremental non{linearity. Ingenieur{Archiv, 59,
187-196.
[29] Desrues J., Chambon R. (1993). A new rate{type constitutive model for
geomaterials: CLoE. In: Modern Approaches to Plasticity, Kolymbas ed.,
Elsevier, 309-324.
[30] Drescher A., Vardoulakis I. (1982). Geometric softening in triaxial tests
on granular material. Geotechnique, 32(4), 291-303.
[31] El Hassan N., Desrues J., Chambon R. (1997). Numerical modelling of
borehole instability using a non linear incremental model with bifurcation
analysis. Proc. Int. Symp. on Deformation and Progressive Failure in
Geomechanics, Oka et al. eds., Nagoya, 677-682.
[32] Goldscheider M. (1976). Grenzbedingung und Flieregel von Sand.
Mech. Res. Comm., 3, 463-468.
[33] Goldscheider M. (1984). True triaxial tests on dense sand. In: Constitu-
tive Relations for Soils, Gudehus et al. eds., Grenoble, 11-54.
[34] Gudehus G. (1979). A comparison of some constitutive laws for soils
under radially symmetric loading and unloading. Proc. 3 Int. Conf.
rd

Num. Meth. Geomech., Aachen, Wittke ed., Balkema, 4, 1309-1324.

[35] Gudehus G. (1985). Requirements for constitutive relations for soils. In:
Mechanics of Geomaterials, Bazant ed., John Wiley & Sons., 47-63.

37
[36] Gudehus G. (1995). A comprehensive concept for non{staurated gran-
ular bodies. Proc. Unsaturated Soils, Alonso & Delage eds., Balkema, 2,
725-737.
[37] Gudehus G. (1995). Attractors for granular storage and ow. Proc.
Partec `95, 333-345.
[38] Gudehus G. (1996). A comprehensive constitutive equation for granular
materials. Soils and Foundations, 36(1), 1-12.
[39] Gudehus G. (1996). Constitutive relations for granulate{liquid mixtures
with a pectic constituent. Mechanics of Materials, 22, 93-103.
[40] Gudehus G., Goldscheider M., Winter H. (1977). Mechanical properties
of sand and clay and numerical integration methods: some sources of
errors and bounds of accuracy. In: Finite Elements in Geomechanics,
Gudehus ed., John Wiley & Sons, 121-150.
[41] Gudehus G., Kolymbas D. (1979). A constitutive law of the rate{type
for soils. Proc. 3 Int. Conf. Num. Meth. Geomech., Aachen, Wittke ed.,
rd

Balkema, 1, 319-329.
[42] Gudehus G., Kolymbas D. (1985). Numerical testing of constitutive rela-
tions for soils. Proc. 5 Int. Conf. Num. Meth. Geomech., Nagoya, Adachi
th

et al. eds., Balkema, 1, 63-81.

[43] Hammad W.I. (1991). Modelisation non lineaire et etude experimentale
des bandes des cisallement dans les sables. These de Doctorat d'etat,
Universite de Grenoble.
[44] Hashiguchi K. (1993). A basic formulation of elastoplastic constitutive
equations. In: Modern Approaches to Plasticity, Kolymbas ed., Elsevier,
39-58.
[45] Herle I., Gudehus G. (1999). Determination of parameters of a hypoplas-
tic constitutive model from grain properties. Mech. Cohesive-Frictional
Materials, in print.
[46] Hutchinson J.W., Hill R. (1975) Bifurcation phenomena in the plane
tension test. Journal of the Mechanics and Physics of Solids, 23, 239-
264.
[47] Jardine R.J., Potts D.M., Fourie A.B., Burland J.B. (1986). Studies of
the in uence of non-linear stress-strain characteristics in soil-structure
interaction. Geotechnique, 36, 377-396.
38
[48] Kolymbas D. (1977). A rate{dependent constitutive equation for soils.
Mechanical Research Communications, 4, 367-372.
[49] Kolymbas D. (1981). Bifurcation analysis for sand samples with a non{
linear constitutive equation. Ingenieur{Archiv, 50, 131-140.
[50] Kolymbas D. (1984). A constitutive law of the rate type for soils |
Position, calibration and prediction. In: Constitutive Relations for Soils,
Gudehus et al. eds., Grenoble, 419-437.
[51] Kolymbas D. (1987). A novel constitutive law for soils. Proc. 2 Int.
nd

Conf. on Constitutive Laws for Engineering Materials, Tucson, Elsevier,

1, 319-326.
[52] Kolymbas D. (1991). An outline of hypoplasticity. Arch. Appl. Mech.,
61, 143-151.
[53] Kolymbas D. (1991). Computer-aided design of constitutive laws. Int.
J. Num. Anal. Meth. Geomech., 15, 593-604.
[54] Kolymbas D. (1994). Hypoplasticity as a constitutive framework for
granular materials. Proc. Computer Methods and Advances in Geome-
chanics, Siriwardane & Zaman eds., 1, Balkema, 197-208.
[55] Kolymbas D., Rombach G. (1989). Shear band formation in generalized
hypoelasticity. Ingenieur{Archiv, 59, 177-186.
[56] Kolymbas D., Wu W. (1993). Introduction to hypoplasticity. In: Modern
Approaches to Plasticity, Kolymbas ed., Elsevier, 213-224.
[57] Kolymbas D., Herle I. (1997). Hypoplasticity: a framework to model
granular materials. In: Behaviour of Granular Materials, Cambou ed.,
Springer, Berlin, 239-268.
[58] Loret B. (1987). Non{linearite incrementale et localisation des
deformations: quelques remarques. Journal de Mecanique Theorique et
Appliquee, 6(3), 423-459.
[59] Matsuoka H., Nakai T. (1977). Stress{strain relationship of soil based
on the `SMP'. In: Constitutive Equations of Soils, Specialty Session 9, IX
ICSMFE, Tokio, 153-162.
[60] Nelson I. (1977). Constitutive models for use in numerical computations.
Proc. Plastic and Long Term E ects, DMSR `77, Karlsruhe, 2, 45-97.

39
[61] Niemunis A. (1996). A visco{plastic model for clay and its FE imple-
mentation. Proc. XI Coll. Franco{Polonais en Mechanique des Sols et
des Roches Appliquee, Gdansk, 151-162.
[62] Niemunis A., Herle I. (1997). Hypoplastic model for cohesionless soils
with elastic strain range. Mech. Cohesive-Frictional Materials, 2, 279-299.
[63] Papamichos E., Vardoulakis I., Han C. (1993). Noncoaxial ow theory
of plasticity: shear failure predictions in sand. In: Modern Approaches to
Plasticity, Kolymbas ed., Elsevier, 585-598.
[64] Pastor M., Zienkiewicz O.C., Chan A.H.C. (1990). Generalized plasticity
and the modelling of soil behaviour. Int. J. Num. Anal. Meth. Geomech.,
14, 151-190.
[65] Puzrin A.M., Burland J.B. (1998). Non{linear model of small{strain
behaviour of soils. Geotechnique, 48(2), 217-233.
[66] Pyke R. (1986). The use of linear elastic and piecewise linear models
in nite element analyses. In: Geomech. Modeling in Engng. Practice,
Dungar & Studer eds., Balkema, 167-188.
[67] Rice J.R. (1976). The localization of plastic deformations. Theoretical
and Applied Mechanics, Koiter ed., North{Holland, 207-220.
[68] Rice J.R., Rudnicki J.W. (1980). A note on some features of the theory
of localization of deformation. Int. Journal of Solids and Structures, 16,
597-605.
[69] Royis P., Doanh T. (1998). Theoretical analysis of strain response en-
velopes using incrementally non{linear constitutive equations. Int. J.
Num. Anal. Meth. Geomech., 22, 97-132.
[70] Rudnicki J.W., Rice J.R. (1975). Conditions for the localization of defor-
mation in pressure{sensitive dilatant material. Journal of the Mechanics
and Physics of Solids, 23, 371-394.
[71] Tamagnini C., Viggiani G., Chambon R., Desrues J. (1998). Evalua-
tion of di erent strategies for the integration of hypoplastic constitutive
equations. Application to the CLoE model. Mech. Cohesive{Frictional
Materials, in print.
[72] Tejchman J. (1994). Numerical study of localized deformation in a
Cosserat continuum. Proc. Localization and Bifurcation Theory for Soils
and Rocks, Aussois, Chambon et al. eds., Balkema, 257-274.
40
[73] Tejchman J. (1997). Modelling od shear localization and autogeneous
dynamic e ects in granular bodies. Pubbl. Series of the Inst. of Soil Mech.
and Rock Mech., University of Karlsruhe, vol. 140.
[74] Terzaghi K. (1948). Theoretical Soil Mechanics. Wiley, New York.
[75] Tobita Y. (1997). Importance of incremental nonlinearity in the defor-
mation of granular materials. Proc. IUTAM Symp. on Mech. of Granular
and Porous Materials, Fleck & Cocks eds., Kluwer, 139-150.
[76] Tokuoka T. (1971). Yield conditions and ow rules derived from hypo{
elasticity. Arch. Rat. Mechanics and Analysis, 42, 239-252.
[77] Truesdell C.A. (1956). Hypo{elastic shear. J. Appl. Physics, 27, 441-447.
[78] Truesdell C.A., Noll W. (1965). The non{linear eld theories of me-
chanics. In: Encyclopedia of Physics, S. Flugge ed., Vol. III/3, Springer,
Berlin.
[79] Valanis K.C., Lee C.F. (1984). Endochronic theory of cyclic plasticity
with applications. J. Appl. Mech., ASME, 51, 789-794.
[80] Vardoulakis I. (1980). Shear band inclination and shear modulus of sands
in biaxial tests. Int. Journal for Numerical and Analytical Methods in
Geomechanics, 4, 103-119.
[81] Vardoulakis I., Graf B. (1985). Calibration of constitutive models for
granular materials using data from biaxial experiments. Geotechnique,
35(3), 299-317.
[82] Viggiani G., Finno R.J., Harris W.W. (1994). Experimental observations
of strain localisation in plane strain compression of a sti clay. Proc. Lo-
calization and Bifurcation Theory for Soils and Rocks, Aussois, Chambon
et al. eds., Balkema, 189-198.
[83] von Wol ersdor P.{A. (1996). A hypoplastic relation for granular ma-
terials with a prede ned limit state surface. Mech. Cohesive{Frictional
Materials, 1, 251-271.
[84] Wang C.{C. (1970). A new representation theorem for isotropic tensor
functions, parts I and II. Archive for Rational Mechanics and Analysis,
36, 166-223.
[85] Wang Z.{L., Dafalias Y.F., Shen C.K. (1990). Bounding surface hy-
poplasticity model for sand. J. Engng. Mech., ASCE, 116(5), 983-1002.
41
[86] Wu W. (1992). Hypoplasticity as a mathematical model for the mechan-
ical behavior of granular materials. Publ. Series of the Institute of Soil
and Rock Mechanics, University of Karlsruhe, n. 129.
[87] Wu W. (1998). Rational approach to anisotropy of sand. Int. J. Num.
Anal. Meth. Geomech., 22, 921-940.
[88] Wu W., Kolymbas D. (1990). Numerical testing of the stability criterion
for hypoplastic constitutive equations. Mechanics of Materials, 9, 243-
253.
[89] Wu W., Sikora Z. (1991). Localized bifurcation in hypoplasticity. Int.
Journal of Engineering Science, 29(2), 195-201.
[90] Wu W., Sikora Z. (1992). Localized bifurcation of pressure sensitive di-
latant granular materials. Mechanical Research Communications, 19(4),
289-299.
[91] Wu W., Bauer E. (1993). A hypoplastic model for barotropy an py-
knotropy of granular soils. In: Modern Approaches to Plasticity, Kolym-
bas ed., Elsevier, 225-245.
[92] Wu W., Bauer E. (1994). A simple hypoplastic constitutive model for
sand. Int. J. Num. Anal. Meth. Geomech., 18, 833-862.
[93] Wu W., Niemunis A. (1994). Beyond invertibility surface in granular
materials. Proc. Localization and Bifurcation Theory for Soils and Rocks,
Aussois, Chambon et al. eds., Balkema, 113-126.
[94] Wu W., Niemunis A. (1996). Failure criterion, ow rule and dissipation
function derived from hypoplasticity. Mech. Cohesive{Frictional Materi-
als, 1, 143-163.
[95] Wu W., Bauer E., Kolymbas D. (1996). Hypoplastic constitutive model
with critical state for granular materials. Mechanics of Materials, 23,
45-69.
[96] Wu W., Niemunis A. (1997). Beyond failure in granular materials. Int.
J. Num. Anal. Meth. Geomech., 21, 153-174.

42
A Gudehus/Bauer K{hypoplastic model
The constitutive functions entering in the Gudehus/Bauer [38, 1] K{hypoplastic
model, eqs. (13){(14), are given by:
a := c + c k^ k 1[1 + cos(3)]
1 (88)
1 2

and:
 (1    
hs 3p ;n ei
fb(p) := nh
)
1 + ei 0
(89)
i hs ec ei 0
e 
fe(e; p) := ec (90)
 
e
fd (e; p) := e ; e ; e d
(91)
c d
with:
n    
3
ei := ei exp ; h p (92)
0
s
  n 
ec := ec exp ; 3hp
0 (93)
s
  n 
ed := ed exp ; 3hp
0 (94)
s
and:
   
hi := c1 + 13 ; eei ; ed p1 0 0
(95)
2
1 c ; ed 3c 0 0 1
r    
3 3 ;
c := 8 sin  sin c 3 3 +
c := 8 sin  sin  c
(96)
1 2
c c
The quantities hs (granulate hardness), c (critical friction angle), ei , ei , 0 0
ec , ed , , and n appearing in eqs. (89){(96) are material constant, to be
0 0
determined from standard laboratory tests, or modi ed classi cation tests.

B von Wol ersdor K{hypoplastic model

The functions de ning barotropy and pycnotropy factors (fb, fe and fd),
as well as those de ning the characteristic void ratios (ei , ec and ed) which

43
enter in the von Wol ersdor [83] K{hypoplastic model are still given by
eqs. (89){(91) and eqs. (92){(94), respectively.
The two functions a and F appearing in eq. (15) are de ned as follows:
p
a := p3 (3 ;sinsin
c
c)
(97)
2 2
 =
1
F := 8 tan + p
2 2 ; tan 2
; p1 tan
1 2

(98)
2 + 2 tan cos(3) 2 2
with:
p
tan := 3 k^ k (99)
As in Gudehus/Bauer model, c represents the critical friction angle of the
material.

44