The 12th International Conference of

International Association for Computer Methods and Advances in Geomechanics (IACMAG)

1-6 October, 2008
Goa, India

A Lode Angle Dependent Formulation of the Hardening Soil Model

T. Benz, M. Wehnert
Wechselwirkung – Numerische Geotechnik, Stuttgart, Germany

P.A. Vermeer
Institute for Geotechnical Engineering, University of Stuttgart, Germany

Keywords: Double Hardening, Hardening Soil, Matsuoka-Nakai, constitutive soil model

ABSTRACT: This paper presents a new formulation of a well known double hardening model often referred to as
the Hardening Soil model. As an alternative to the existing Hardening Soil model’s Mohr-Coulomb failure surface,
the new formulation allows for the incorporation of smooth failure surfaces, such as the failure surface proposed
by Matsuoka & Nakai or that proposed by Lade & Duncan. Although the incorporation of all other failure criteria
that can be formulated as a function of the Lode angle is also possible, the examples presented at the end of this
paper concentrate on the Hardening Soil model with Matsuoka-Nakai failure criterion. The Lode angle dependent
formulation with Matsuoka-Nakai yield surface is validated in element tests and in an excavation example. A
comparison between results obtained from the Hardening Soil model with the Mohr-Coulomb type yield surface
and that with the Mastuoka-Nakai type yield surface reveals the failure criterion’s actual influence on material
strength and stiffness when applied in the Hardening Soil Model in plane strain conditions.

1 Introduction
Constitutive soil models that incorporate shear hardening and volumetric hardening mechanisms, in short referred
to as double hardening models, were first proposed in the 1970’s. Since then, the double hardening concept has
proved to be very useful in the numerical analysis of geotechnical problems. A particularly well known double
hardening model is the Hardening Soil model that was developed by Schanz (1988) and Schanz et al. (1999) on
the basis of the double hardening model by Vermeer (1978). This paper revisits the formulation of the original
Hardening Soil double hardening model and extends it towards a new Lode angle dependent formulation. Within
the new formulation it is now possible to use other alternative failure criteria than the Mohr-Coulomb criterion. In
the examples presented at the end of this paper, the Mohr-Coulomb failure criterion is replaced by the smooth
Matsuoka-Nakai failure criterion as shown in Figure 1.
The Mohr-Coulomb failure criterion (Mohr 1900) for soils, which is implemented in the original Hardening Soil
model, is one of the earliest and most trusted failure criterion. It has been experimentally verified in triaxial com-
pression and extension and is of striking simplicity. However, the Mohr-Coulomb criterion is very conservative for
intermediate principal stress states between triaxial compression and extension. Matsuoka and Nakai (1974,
1982) proposed a failure criterion that is in better agreement with experimental data. They propose the concept of
a Spatial Mobilized Plane (SMP), which defines the plane of maximum spatial, averaged particle mobilization in
principal stress space. Replacing the Mohr-Coulomb yield surface and plastic potential by a smoother surface is
also an advantage from a numerical point of view.

2 Model formulation

2.1 Definitions
Within this paper, compressive stress is taken as positive. Tensile stress is taken as negative. Stresses are
always taken to be effective values without any special indication by a prime. Infinitesimal deformation theory is
applied. Cauchy stress is related to linearized infinitesimal strain. Tensorial quantities are generally expressed in
indicial notation. The order of a tensor is indicated by the number of unrepeated (free) subscripts. Whenever a
subscript appears exactly twice in a product, that subscript will take on the values 1, 2, 3 successively, and the
resulting terms are summed (Einstein’s summation convention). Eigenvalues of stress and strain tensors (prin-
cipal stresses and strains) are denoted by one subscript only, e.g. σ i and ε i with i = 1, 2, 3. The order of principal
stresses is σ 1 ≥ σ 2 ≥ σ 3 .
The Roscoe stress invariants p (mean stress) and q (deviatoric stress), are defined as:

653
 Fig. 1. Two double hardening models with different failure criteria in principle stress space (left) and in deviatoric
plane (right) – (a) Mohr-Coulomb failure criterion – (b) Matsuoka-Nakai failure criterion.

σ ii 3 1 1
p= and q = (σ ij − δ ijσ kk )(σ ij − δ ijσ kk ), (1)
3 2 3 3
In triaxial compression with σ 1 ≥ σ 2 = σ 3 , the Roscoe invariants simplify to:
1
p = (σ axial + 2σ lateral ) and q = (σ axial − σ lateral ). (2)
3
In analogy to the stress invariants, volumetric strain ε v and shear strain γ s are defined as:
3 1 1
ε v = ε ii and γ s = (ε ij − δ ij ε kk )(ε ij − δ ij ε kk ), (3)
2 3 3
which simplify to
ε v = ε axial + 2ε lateral and γ s = (ε axial − ε lateral ), (4)
in triaxial compression. Shear strain relates to the deviatoric strain invariant ε q as follows:
3 3 2 1 1
γs = εq = (ε ij − δ ij ε kk )(ε ij − δ ij ε kk ) . (5)
2 2 3 3 3

2.2 Shear hardening

In drained triaxial primary loading, the experimentally observed relationship between axial strain and deviatoric
stress in soils can be well approximated by a hyperbolic function. Kondner and Zelasko (1963) described the
hyperbolic stress-strain relationship for drained triaxial loading as follows:
q 2 sin ϕ q
ε 1 = ε 50 with qa = (σ 3 + c cot ϕ ) and ε 50 = a . (6)
qa − q 1 − sin ϕ 2E50
where E50 gives the secant stiffness in primary triaxial loading at a deviatoric stress level equal to half the failure
stress as illustrated in Figure 2. Duncan and Chang (1970) based their hypoelastic model on the above
formulation by Kondner and Zelasko, additionally introducing the deviatoric measure qf in the form:
q 2 sin ϕ q
ε 1 = ε 50 for q < qf = (σ 3 + c cot ϕ ) and qa = f . (7)
qa − q 1 − sin ϕ Rf
where Rf is a failure ratio that modifies the hyperbolic stress-strain curve defined by Kondner and Zelasko when
chosen lower than 1. The conceptual difference in the formulations by Kondner and Zelasko and that by Duncan
and Chang is illustrated in Figure 2. Extending the hypoelastic Duncan-Chang model to an elastoplastic
formulation, Schanz (1988) proposed the following yield function:
q q 2q
fs = a − − γ ps . (8)
E50 qa − q Eur
where γ ps is an internal material variable for the accumulated plastic deviatoric strain, q = σ 1 − σ 3 is defined for
triaxial loading, and qa is the asymptotic deviatoric stress as defined in the original Duncan-Chang model (Equa-
tion 7). As the stress-strain relation of soils in unloading and reloading can be roughly approximated by a linear
function, the HS model assumes isotropic elasticity inside the yield function: The elastic unloading-reloading
stiffness Eur relates elastic stress to elastic strain (see above).

654
 q q asymptote
qa
qa asymptote
qf
1 1
E50 E50
½qa ½qf Eur
1
e1 e1
Fig. 2. Hyperbolic stress-strain law by Kondner & Zelasko (left) and its modification after Duncan & Chang (right).

For constant volumetric strain, the equivalent of the original Hardening Soil model (Equation 7) with the approach
by Duncan and Chang is given by defining:
γ ps = ε1p − ε 2p − ε 3p and thus γ ps = 2ε1p , (9)
as then the following relation holds:
q 1 q q q q q
ε 1 = ε 1e + ε 1p = − γ ps = a = 50 = ε 50 . (10)
Eur 2 2E50 qa − q E50 qa − q qa − q
Unfortunately, the definition of the strain measure used in the original Hardening Soil model is not compatible with
the shear strain γ s defined in equation 3 and, as a consequence, does not vanish during volumetric straining.
Therefore a representation of deviatoric stress q against the deviatoric strain γ will show an offset when volumetric
loading is applied prior to deviatoric loading and hence, is not objective. The following revised formulation of the
Hardening Soil yield function therefore uses the shear strain measure γ sps , which goes back to the second
deviatoric invariant as defined in Equation 3. In triaxial conditions, the shear strain γ sps is defined as:
3
γ sps = ε1p − ε 3p so that γ sps = ε1p for ε vp = 0. (11)
2
It can now be easily proven that in order to keep the new yield function equivalent to the approach by Duncan and
Chang and thus equivalent to the original Hardening Soil model, one has to write:
3 qa q 3 2q
fs = − − γ sps (12)
4 E50 qa − q 2 Eur
Next, the new yield function can be written in terms of mobilized friction ϕm , which is in triaxial compression by
means of the Mohr-Coulomb criterion, defined as:
σ1 − σ 3
sin ϕm = (13)
σ 1 + σ 3 + 2c cot ϕ
so that
q = sin ϕm (σ 1 + σ 3 + 2c cot ϕ ) (14)
The Mohr-Coulomb criterion assumes failure in triaxial conditions whenever:
2 sin ϕ
qf = (σ 3 + c cot ϕ ) (15)
1 − sin ϕ
and thus, the ratio q /qa can be expressed as:
q q 1 − sin ϕ sin ϕm σ 1 + σ 3 + 2c cot ϕ
= Rf = Rf
qa qf sin ϕ 1 2σ 3 + 2c cot ϕ
(16)
q ⎛ 1 − sin ϕ ⎞ ⎛ sin ϕm ⎞
Rf ⎜ = ⎟⎜ ⎟.
qa ⎝ sin ϕ ⎠ ⎝ 1 − sin ϕm ⎠
Applying Equation 16 to Equation 12 results in the final form of the shear hardening yield function:

fs =
3 q ( 1− sin ϕm
sin ϕm ) −
3 q
− γ sps (17)
2 Ei ( 1− sin ϕm
sin ϕm ) − Rf ( 1− sin ϕ
sin ϕ ) 2 Eur

where sin ϕm is the mobilized friction angle in triaxial compression. The transition from E50 to Ei ≈ 2E50 is made
because of the double hardening model’s second yield surface that will be introduced later. The second yield
surface will affect material stiffness such that, the meaning of Ei in the final model is not as closely related to the
hyperbolic model by Kondner and Zelasko as the one of E50 .
The shear hardening function given in Equation 17 is not limited to the Mohr-Coulomb failure criterion as for
example the yield function of the original Hardening Soil model is. In mobilized friction, the Matsuoka-Nakai yield
criterion for example, can be written as:

655
 9−
I1I2

sin ϕm ≡ I3
. (18)
1−
I1I2
I3

This definition yields deviatoric isolines of mobilized friction that are similar to the shape of the Matsuoka-Nakai
yield criterion. Alternatively, mobilized friction can be expressed in the Lode angle dependent formulation:
3q
sin ϕm ≡ . (19)
6L(θ )p ƒ + q
where L is varying between 1 and δ = Me Mc for triaxial compression and extension respectively and p ƒ = p + c cot ϕ
accounts for a cohesion related apex shift along the hydrostatic axis. In this way, many Lode dependent yield
functions can be assigned to the yield function given in Equation 17. Lode dependent formulations of the
Matsuoka-Nakai and the Lade criterion in the form of functional relationships of L(θ ) are for example introduced
by Bardet (1990).
The stiffness moduli Ei and Eur are scaled for their stress dependency with a power law (Janbu 1963, Ohde 1951)
m m
⎛ σ + c cot ϕ ⎞ ⎛ σ + c cot ϕ ⎞
Ei = Eiref ⎜ ref3 ⎟ and Eur = Eurref ⎜ ref3 ⎟ (20)
⎝ p + c cot ϕ ⎠ ⎝ p + c cot ϕ ⎠
where Eiref and Eurref are the material stiffness moduli at the reference pressure pref and m is the exponent of the
power law. In Equation 20 the minor principal stress σ 3 is used as an indicator of the actual stress state in the
σ
material instead of the mean stress p = 3ii .

2.3 Volumetric hardening

Similar to the above defined yield function that resembles lines of constant plastic shear strain, a yield function
that resembles lines of constant void ratio or constant volumetric strain is introduced next. Within the double har-
dening model, this second cap-type yield function accounts for volumetric hardening, this corrects for overly stiff
primary oedometric or isotropic loading, obtained in pure shear hardening models. From experimental evidence,
loci of constant void ratio are usually defined as an ellipse in p-q space, (e.g. Modified Cam Clay model):
q2
f MCC = 2 − p 2 − ppp (21)
Μ
where Μ is the slope of the critical state line in the p-q plane and pp is an internal material hardening variable for
pre-consolidation stress. The original HS model’s cap-type yield surface is defined slightly differently:
q
2
f c = 2 − p 2 − pp2 (22)
α
where α is an internal material constant, controlling the steepness of the cap in p-q space and q is a special stress
measure, defined as:
3 − sin ϕ
q = σ 1 + (δ −1 − 1)σ 2 − δ −1σ 3 with δ = . (23)
3 + sin ϕ
The definition of the special stress measure q is necessary to adopt the cap-type yield surface’s deviatoric shape
to the Mohr-Coulomb shape of the original Hardening Soil model’s cone-type yield surface. For compatibility with
other failure criteria, e.g. Matsuoka-Nakai or Lade, the stress measure q is avoided and the cap type yield surface
is rewritten as:
q2
fc = − p 2 − pp2 . (24)
( L(θ )α )
2

2.4 Plastic potentials

The original Hardening Soil model uses a non-associated potential for the cone type yield surface and an
associated potential for the cape type yield surface. Likewise does the Lode angle dependent formulation. As the
plastic flow directions for low mobilized friction are almost radial and the error made for higher mobilization levels
is also tolerable, the Lode angle dependent formulation uses potentials with radial deviatoric flow. The non-
associated potential to the cone-type yield surface is defined as:
6 sinψ m
g s = Μψ m p ƒ where Μψ m = . (25)
3 − sinψ m
The associated potential to the cap-type yield surface is written as:
q2
gc = − p 2 − pp2 . (26)
( L(θ )α )
2

656
 a) q b) q n+1 s n+1
n msij
ij
corner region sTijrial
f
t corner region
n+1 s
m ij n+1 c n+1 c
n+1 n =m
m tij sTijrial s n+1
ij ij
f c
f =g
c sij
s s
n g n g
apex gray sij s sij
n+1 f c c
region sij f =g
p p
apex c cotj st elastic region pp c cotj elastic region pp

Fig. 3. (a) The two vector cone-cap return strategy. (b) Apex gray region and tension cut-off (with σ 2 = 0 ).

The angle of mobilized dilatancy ψ m in Equation 25 can for example be calculated according to Rowe’s stress dila-
tancy theory (Rowe 1962). A main drawback of Rowe’s approach, is the highly contractive behavior at low mobi-
lized friction angles. In the original Hardening Soil model the mobilized dilatancy angle, sinψ m is therefore set to
be greater than or equal to zero overriding Rowe’s original equation:
sin ϕm − sin ϕcs
sinψ m = ≥ 0. (27)
1 − sin ϕm sin ϕcs

2.5 Hardening rules

The evolution laws of the original Hardening Soil model are defined as:
d γ ps = d λ s hγ ps with hγ ps = 1

⎛ σ + c cot ϕ ⎞
m (28)
dpp =
= 2H ⎜ ref3d λ c hpp ⎟ p with hpp
⎝ p + c cot ϕ ⎠
where m represents the power law exponent, and H relates plastic volumetric strain ε vp to pre-consolidation stress
pp as follows:
m
⎛ σ + c cot ϕ ⎞
dpp = H ⎜ ref3 ⎟ dεv .
p
(29)
⎝ p + c cot ϕ ⎠
In decomposing volumetric strain into elastic and plastic contributions, H can be rewritten as a function of the bulk
stiffness in unloading-reloading K s and the bulk stiffness in primary loading K c :
KsKc 1
H= = Ks . (30)
K s − K c KKs − 1
c

where due to the assumption of isotropic elasticity, the elastic bulk stiffness K s relates to Eurref as follows:
Eurref
Ks = . (31)
3(1 − 2ν )
The model parameter H can therefore be determined by the bulk stiffness ratio K s /K c . As the physical meaning of
the latter is more evident, it is often used to quantify H. The hardening rules of the Lode angle dependent
formulation are the same except hγ ps = 3 2. This change reflects the new shear strain measure used in the Lode
s
angle dependent formulation and assure that both formulations yield the same result in triaxial compression.

2.6 Implementation aspects

For the integration of the constitutive equations, an implicit closest point projection algorithm was chosen.
Although the Matsuoka-Nakai yield criterion is smooth in a deviatoric section, corner problems still arise at the
non-smooth intersections of the cone-type yield surface and the cap-type yield surface as well as at the apex.
Koiter (1960) additively decomposes plastic strain rates in such corner problems as follows:
d ε ijp = d λ cone mij (σ ij , q ∗ ) + d λ cap mij (σ ij , q ∗ ). (33)
where mij = ∂g /∂σ ij is the derivative of the plastic potential. The two vector cone-cap return strategy is illustrated in
Figure 3a. As the size of the gray corner region in hardening situations is not known beforehand, the scheme
proposed by Bonnier (2000) is used to determine to which surface the trial stress is to be returned.
The gray apex region is defined by the gradient to the cone-type potential surface as shown in Figure 3b. If the
apex corresponds to an admissible tensile stress, the trial stress is returned to the apex. If the apex point violates
the user defined maximum allowable tensile stress, a return mapping scheme to the respective tension cut-off
planes is evoked. The tension cut-off criterion is based on minimum principal stress, which implies three (fixed)
orthogonal tension cut-off planes in principal stress space:

657
 Table 1. Parameters for the double hardening model used in the analyses.

Parameter Unit Hostun RF Sand (L1) Sand (L2) Sand (L3)

1) User defined parameters
Unsaturated/saturated weight γ unsat / γ sat [kN/m³] - 19 / 20 19 / 20 19 / 20
ref
Triaxial secant stiffness E50 [kN/m²] 30000 45000 75000 105000
ref
Oedometric tangent stiffness Eoed [kN/m²] 30000 45000 75000 105000
Unloading/reloading stiffness Eurref [kN/m²] 90000 180000 300000 315000
Power of stress dependency m [–] 0.55 0.55 0.55 0.55
Cohesion (effective) c [kN/m²] 0.0 1.0 1.0 1.0
Friction angle (effective) ϕ [°] 42.0 35.0 38.0 38.0
Dilatancy angle ψ [°] 16.0 5.0 6.0 6.0
Poisson’s ratio ν [–] 0.25 0.2 0.2 0.2
Reference stress for stiffness pref [kN/m²] 100 100 100 100
K0-value (normal consolidation) K0nc [–] 0.40 0.43 0.38 0.38
Failure ratio Rf = qf qa [–] 0.9 0.9 0.9 0.9
Tensile strength σTension [kN/m²] 0.0 0.0 0.0 0.0
2) Internal parameters
Initial secant stiffness Eiref [kN/m²] 65488 96662 154447 208642
Cap parameter (steepness) α [–] 1.47 1.48 1.87 1.88
Cap parameter (stiffness ratio) KS KC [–] 1.84 2.15 2.07 1.59
3) State parameters
Plastic shear strain γ sps [–] - - - -
Pre-consolidation pressure pp [kN/m²] - - - -

fi t = σ Tension − σ i , (34)
where σ Tension is the user defined maximum allowable tensile stress.

2.7 Material parameters

A summary of the material parameters and state parameters introduced above is presented in Table 1. A diffe-
rentiation is made between user input and internal parameters because the latter cannot be quantified as results
of standard triaxial and oedometer tests directly and hence, are not expected to be entered by the user.
Internal model parameters are the stiffness measures Eiref and H, and the cap-type yield surface’s steepness α .
ref ref
These internal parameters mainly relate to the user input parameters E50 , Eoed and K 0nc respectively, where Eoed
ref
is
the tangent stiffness at σ 1 = p in K 0 (oedometer) loading, and K 0 is the stress ratio of horizontal effective stress
ref nc

ref ref
to vertical effective stress in a normally consolidated state. Note that E50 , and Eoed are not elastic stiffness
constants.
In double hardening situations, i.e. both yield loci are hardened simultaneously, analytical back-calculation of
internal model parameters is impossible. Therefore, the internal parameters are solved for in an iterative scheme
ref ref
so that the double hardening model simulates the user input E50 in a triaxial element test and both, Eoed and K 0nc in
an oedometer element test, to within a tolerated error.

3 Model validation
The Lode angle dependent formulation of the Hardening Soil model is validated by element tests that are cal-
culated on the Gauss-Point level and by analyses of a boundary value problem using the finite element code
PLAXIS V 8. The original Hardening Soil model that is implemented in the calculation kernel of this code serves
as a reference. The Matsuoka-Nakai yield surface is considered in the Lode angle dependent formulation of the
double hardening model. In the following, the abbreviation HSMC refers to the original Hardening Soil model with
Mohr-Coulomb yield surface and HSMN refers to the Lode angle dependent formulation of the double hardening
model with Matsuoka-Nakai yield surface.

3.1 Element tests

In triaxial compression, triaxial extension, and K0-loading, the HSMC and the HSMN yield identical results, which
is a first validation of the HSMN formulation and its numerical implementation. In plane strain biaxial tests,

658
however, the material strength estimate of the Matsuoka-Nakai criterion increases compared to that of the Mohr-
Coulomb criterion. Figure 4 shows biaxial test data on dense Hostun sand compiled by Desrues et al. (2000) and
their numerical back calculation using the HSMC and the HSMN models. The material data set used in the calcu-
lation (Table 1) was calibrated for triaxial and oedometer tests before. The test data validates the use of
Matsuoka-Nakai failure criterion for dense Hostun sand.

3.2 Excavation in Berlin sand

The working group 1.6 „Numerical methods in Geotechnics“ of the German Geotechnical Society (DGGT) has
organized several comparative finite element studies (benchmarks). One of these benchmark examples is the
installation of a triple anchored deep excavation wall in Berlin sand. The reference solution by Schweiger (2002)
is used here as the starting point: Both, the mesh shown in Figure 5, and the soil parameters given in Table 1 are
taken from this reference solution. The calculation assumes plane strain conditions. Note that in the reference
solution, the stiffness of the lower sand layer is artificially increased as the models investigated do not account for
small-strain stiffness. In using a small-strain stiffness model, the extra definition of sand layer 3 could be omitted
as shown by Benz (2007).

Fig. 4. Element tests on dense Hostun RF sand.

Fig. 5. Results of the analyses on the excavation in Berlin sand.

659
A considerable quantitative difference between the results of the HSMC and those of the HSMN model is found
whenever plastic hardening is evoked (see Figure 5 - settlement trough and wall displacement). Although the
HSMN model hardens with the same rate as the HSMC model, its overall stiffness is higher due to its higher peak
value in plane strain conditions. Consequently, less plastic straining occurs in the HSMN loading problem.

4 Summary and conclusion

The original Hardening Soil model comprises ideas by Kondner (1963), Duncan & Chang (1970), Ohde (1951) or
Janbu (1963), Rowe (1962) and Vermeer (1978). Standard lab tests, such as triaxial and oedometer tests provide
the model’s basic characteristics. The new Lode angle dependent formulation now adds the possibility of incor-
porating different failure criteria, as for example the one proposed by Matsuoka & Nakai, which is not only smooth
but also considers the influence of the intermediate principal stress on material strength. It should be noted that
same as the original Hardening Soil model, the new model formulation does not account for initial anisotropy,
stress induced anisotropy, or rotation of principle stresses. However, the formulation may be subsequently exten-
ded towards incorporation of these features. A small-strain stiffness extension of the Hardening Soil model is
already available (Benz 2007). The Hardening Soil model with Matsuoka-Nakai failure criterion predicts additional
material strength and stiffness in plane strain conditions. In the examples presented in this paper, the new model
performs better than its Mohr-Coulomb counterpart.

5 Acknowledgements
The Lode angle dependent formulation of the Hardening Soil model emanated from a research project that was
initiated and funded by the Federal Waterways Engineering and Research Institute (BAW), in Karlsruhe, Ger-
many. The financial support and technical advice provided by the BAW is gratefully acknowledged. Further, the
authors thank Paul Bonnier from PLAXIS B.V. for his valuable assistance with the numerical implementation.

6 References
Bardet J.P. 1990. Lode dependences for isotropic pressure-sensitive elastoplastic materials. Transactions of the ASME. 57(9),
498–506.

Benz T. 2007. Small-strain stiffness of soils and its numerical consequences. Universität Stuttgart. PhD Thesis.

Bonnier PG. 2000. Implementational aspects of constitutive modelling. SCMEP Workshop No 1 at NTNU.

Desrues J., Vermeer P.A., Zweschper B. 2000. Database for Tests on Hostun RF Sand. Universität Stuttgart. Tech Report 13.

Drucker D.C. 1956. On uniqueness in the theory of plasticity. Quart Appl Math. XIV, 35–42.

Duncan J.M., Chang C.-Y. 1970. Nonlinear Analysis of Stress and Strain in Soil. Proc. ASCE: Journal of the Soil Mechanics
and Foundation Division. 96,1629–1653.

Janbu N. 1963. Soil Compressibility as Determined by Oedometer and Triaxial Tests. Proc. 3rd ECSMFE, Wiesbaden. 19–25.

Koiter W.T. 1960. General Theorems for Elastic-Plastic Solids. Progress in Solid Mechanics. Sneddon, Hill eds. 165–221.

Kondner R.L., Zelasko J.S. 1963. A hyperbolic stress-strain formulation for sand. 2nd Pan. Am. CSMFE. Brazil, 289–324.

Matsuoka H. 1974. Stress-strain relationships of sands based on the mobilized plane. Soils and Foundations. 14(2), 47–61.

Matsuoka H., Nakai T. 1982. A new failure criterion for soils in three dimensional stresses. IUTAM Conference on Deformation
and Failure of Granular Materials. Delft, 253–263.

Mohr O. 1900. Welche Umstände bedingen die Plastizitätsgrenze und den Bruch eines Materials? VDI-Zeitschrift 44, 1524.

Ohde J. 1951. Grundbaumechanik. Hütte.

Rowe P.W. 1962. The Stress-Dilatancy Relation for Static Equilibrium of an Assembly of Particles in Contact. Proc. of the Royal
Society of London. Series A, Mathematical and Physical Sciences. 269, 500–527.

Schanz T. 1988. Zur Modellierung des mechanischen Verhaltens von Reibungsmaterialien. Universität Stuttgart: Habilitation.

Schanz T., Vermeer P.A., Bonnier P.G. 1999. Formulation and verification of the Hardening-Soil Model. Beyond 2000 in
Computational Geotechnics. Brinkgreve, ed. Rotterdam: Balkema, 281–290.

Schweiger H.F. 2002. Results from numerical benchmark exercises in geotechnics. Numge 2002. 5th European Conference
Numerical Methods in Geotechnical Engineering, Mestat, ed. Paris, 305–314.

Vermeer P.A. 1978. A double hardening model for sand. Géotechnique 28(4), 413–433.

660