Wulfsohn2002 PDF

Wulfsohn, D., and B. A. Adams. 2002. Elastoplastic soil mechanics.
In Advances in Soil Dynamics Volume 2, 1-

116. St. Joseph, Mich.: ASAE.
Chapter 1
Elastoplastic Soil Mechanics
Part I. Constitutive Modeling of Soils

Dvoralai Wulfsohn
The complete design of agricultural and forestry field machinery requires a description
of the response of soil to forces applied by the equipment. Soil structural/mechanical
response during, and as a consequence of, machine-soil interactions is of concern with
respect to soil fertility as well as for optimal mechanical design of traction devices and
implements. Soil, however, is a highly complex particulate multiphase material whose
stress-strain behavior is not completely understood. This is particularly true for
unsaturated soils, which engineers and scientists interested in agriculture frequently
encounter.
Traditional approaches for the evaluation of agricultural machine-soil interactions
have included empirical, semiempirical, and analytical techniques. Empirical methods
involve extensive testing, and the results are only applicable to conditions similar to those
under which the test data were obtained. Analytical methods based on generalized models
of soil have the potential to overcome the limitations of empirical methods.
Semiempirical techniques combining simple “analog” representations of machine-soil
behavior with empirical parameters have had some success, but they are limited to
evaluating greatly simplified geometries.
There has been a concerted effort by scientists to develop predictive models for
agricultural soil behavior based on fundamental mechanics. The capabilities of modern
computer technology combined with numerical computational methods, such as the finite
element method, make it possible to analyze complex machine-soil configurations.
However, there has been only limited success in predicting vehicle traction, soil
compaction, and the performance of tillage tools using theoretically based models. The
main reason for this has been the inability to determine the basic properties of field soils.
In a theme report presented at a conference on Predictive Soil Mechanics, Graham (1993)
emphasized the need to identify and measure real properties of soils based on a physical
appreciation of their relationships, preferably set against a theoretical background, instead
of specifying them based on raw empiricism. In 1968, Wiendieck made a similar
argument with respect to analog representations of running gear-soil interaction, noting
that empirical relationships observed between stresses and displacements obtained under
very specific test conditions were sometimes misconstrued by researchers as being stress-
strain “laws.”
Classical soil mechanics has been associated largely with the solution of problems by
geotechnical engineers dealing with saturated soils subjected to high loads for long
2 Elastoplastic Soil Mechanics
durations. Agricultural engineers are generally interested in the mechanical behavior of

surficial soils subjected to short-duration loads due to vehicle and implement loading
(Hettiaratchi, 1988). As a result, several features distinguish agricultural soils problems:
(1) Soils are generally unsaturated and under relatively low geostatic stresses; (2) Soils
undergo large strains resulting in significant irreversible deformations; and (3) Soils are
subjected to loading at high (dynamic) rates. Further progress in predicting soil response
under these conditions requires valid and realistic soil constitutive relations.
Most of the soil mechanics models developed for agricultural applications have been
derived from elastoplastic constitutive models. At the same time there has been the
formal development of the science of “unsaturated soil mechanics” (Fredlund and
Rahardjo, 1993a). The first part of this chapter will be devoted to a review of stress and
strain variables and their utilization in theoretical models of soil as an elastoplastic
material. Applications of these models in the agricultural engineering literature have been
presented in Volume 1 of this Monograph in the section “Soil Mass Dynamic Load-
Deformation Properties Definable within a System of Mechanics (Stress-Strain)”
(Chancellor, 1994: 149-238). Theoretical aspects necessary for understanding the
concepts discussed in this chapter will be presented here.
This part will be followed by a presentation of critical-state soil mechanics. The
theories of critical-state soil mechanics were developed from the application of plasticity
theory to observed soil behavior. In the first instance, the additional complexities
introduced by considering the unsaturated, multiphase, nature of agricultural soils will
not be considered (these will be discussed in some depth in the last part of this chapter).
The intention is to explain some of the more difficult concepts of the critical-state
framework in such a manner that they can be understood conceptually, even if the reader
is not greatly interested in the mathematical details of the theory.
Mechanics
The approach to soil mechanics described in this chapter is a macroscopic,
phenomenological approach based upon continuum mechanics (fig. 1.1). The
phenomenological approach neglects the physical processes that take place in a body and
provides mathematical descriptions, based on experiments, of the way these processes
manifest themselves externally (Vyalov, 1986). Soil is to some extent a discontinuous
medium, and could be modeled using a particulate mechanics approach. Nevertheless, at
some scale of aggregation the view of soil as a continuum is valid and models of soil
using continuum mechanics has had a large measure of success in the case of saturated
soils. Their extension to unsaturated soils is more difficult and will form the focus of
some discussion in the second part this chapter.
Fundamental to continuum mechanics is the concept of state variables. State variables
are non-material parameters required to characterize the system (e.g., stress and
deformation, temperature). To be compatible with the terminology of continuum
mechanics, the state variables must be independent of the physical properties (Fredlund
and Rahardjo, 1993a). Measurable physical soil properties are incorporated in the
constitutive relations: single-valued expressions which relate one state variable to one or
more other state variables (Fung, 1969). Examples of constitutive relations are a stress
versus strain relationship, and, for unsaturated soils, the soil-water characteristic curve
Advances in Soil Dynamics, Vol. 2 3
(Fredlund and Rahardjo, 1993a).1 Once the constitutive relations have been formulated,
the analysis of practical engineering problems requires knowledge of the soil-device
boundary conditions (external loading and interface geometry) and an adequate
computational method. Moreover, for the practical implementation of the derived
relationships to highly spatially variable agricultural soils, there is a need for simple
experimental techniques to measure or estimate soil physical properties.
CHARACTER OF CLASSICAL SOIL

MECHANICS
Description of State Variables
Verification of Constitutive Relationship
Formulation of Solutions to Practical Problems
Solution of Formulations
(e.g., Numerical methods)
Figure 1.1—The phenomenological approach to soil mechanics based upon continuum mechanics
(courtesy of Dr. D. G. Fredlund, Civil Engineering Department, University of Saskatchewan).
1
This concept and some of its implications will be presented in Part II of this chapter, Critical-State Soil
Mechanics for Unsaturated Agricultural Soils.
Concept of Stress
When a solid body is subjected to external forces it experiences deformation and
stress. Through rigid-body mechanics the external force system can be resolved into a
resultant force F and a resultant couple M. The stress at a point P in a solid body B can
be obtained by considering a small plane area δA at random orientation with a unit
normal vector ni originating at P (fig. 1.2).
There will be a resultant force δFi acting on δA with a couple δMi. The stress vector
at P is defined as the force intensity acting on the surfaces of the element obtained by
letting δA go to zero:
δFi
Ti = lim (1.1)
δA → 0 δA
There are an infinite number of possible ni through the point P depending on the
orientation of δA, and therefore an infinite number of values of Ti. It turns out that the
state of stress at P can be characterized if the stress vectors are known for three mutually
perpendicular area elements having unit normals in the directions of an arbitrary x, y, z
orthogonal coordinate system (Chen and Baladi, 1985). Each stress vector Ti can be
decomposed into three orthogonal components of stress along the coordinate axes (fig.
1.3), i.e., one component normal to the face (normal stress) and two components tangent
to the face (shear stresses):
Ti = σi1 n1 + σi2 n2 + σi3 n3 (1.2)
x3 ni
δM i
δF i
P
δA x2
x1
Figure 1.2—Equilibrium conditions in the continuum under external loading. Decomposing internal
forces acting at an elementary area δA with a unit normal vector ni at point P in a body B produce a
resultant force δFi and resultant couple δMi.
Figure 1.3—Decomposition of element stress vector into normal and shearing stress components.
where n1, n2, n3 are unit vectors along the coordinate axes x, y, and z respectively.2 The
complete state of stress at point P may then be completely defined by three stress vectors
(i.e., i = 1, 2, 3) acting on three mutually perpendicular faces of a vanishingly small cube
shaped volume element aligned with the x, y, z coordinate axes respectively (fig. 1.4a).
The nine components of the three stress vectors form the elements of a second order
tensor [σ], which can be written as:3
 T1   σ 11 σ 12 σ 13 
   
[σ] =  T2  =  σ 21 σ 22 σ 23  (1.3a)
T  σ σ σ 
 3   31 32 33 
or equivalently, in the familiar engineering notation as:
 σx τ xy τ xz 
 
[σ] =  τ σy τ yz  (1.3b)
yx
 
τ
 zx τ zy σz 
2
In tensor analysis the notation x, y, z for the coordinates are replaced by x1, x2, and x3, so that they are
specified by the indices 1, 2, and 3 respectively.
3
A second-order tensor is defined completely by three vectors in the same way as a vector is defined
completely by three scalars. Thus a vector is a first-order tensor (Chen and Baladi, 1985).
Figure 1.4—Stress components on a soil element: (a) Stresses relative to arbitrary orthogonal coordinate
axes x, y, z; (b) Principal stress cube; (c) Stress octahedron (from Hettiaratchi and O’Callaghan, 1980).
The components of stress are shown in the positive directions with compressive stresses defined as
positive as is customary in soil mechanics. (Compressive stresses are defined as positive because
uncemented, saturated soils, cannot sustain tensile stresses.)
The diagonal terms of the stress tensor are the normal stresses and the off-diagonal terms
are the shear stresses. The components of the stress tensor are scalar quantities and have
dimensions of force per unit area. By convention, the first suffix denotes the direction of
the normal of the surface δA and the second suffix the direction along which the stress
component is directed. Thus σ13 (= τxz) refers to the stress component that acts on the
surface whose normal is along the 1- (x-) axis and it is directed in the 3- (z-) direction.
(Tensor and engineering notation will both be used interchangeably.)
As for the couple δMi, as δA→0 any continuous force distribution approaches that of
a parallel unidirectional distribution. This can be represented as a single resultant force
with a line of action positioned at the incremental area dA and a zero couple moment.
The force loading has already been accounted for in terms of stresses, so in the limit the
couple moment can be dispensed with (Shames, 1989). Thus, there is no contribution
from the couple δMi in this process. In the absence of body couples, the stress tensor can
be shown to be symmetric (σij = σji) so that the description of the state of stress requires
only six independent stress components.
Principal Stresses. The stress tensor [σ] given in equation 1.3 is associated with the
orthogonal coordinate system shown in figure 1.4a. It is always possible to locate a set of
mutually perpendicular planes along which shear stresses are zero leaving only normal
components of the stress tensor presenting maximum and minimum values (see fig. 1.4b).
These normal components of the stress tensor are known as principal stresses, the planes
are called principal planes, and the unit normal vectors associated with these planes are
known as the principal axes or directions. The state of stress at a point can therefore be
described by the three magnitudes of the principal stresses, σ1, σ2, and σ3, and three
components describing the orientations in space of the principal axes. If a material is
isotropic then the direction of stresses at a point are unimportant and the state of stress
can be completely described in terms of the three principal stresses as:
 σ1 0 0
 
[ σ] =  0 σ2 0 (1.4)
0 0 σ 3 

The largest principal value is known as the major principal stress, the smallest the
minor principal stress, and the remaining value is called the intermediate principal
stress.4
Decomposition of the Stress Tensor. The stress tensor [σ] (eq. 1.3) can be
decomposed into two symmetric stress tensors in the following manner:
 σx τ xy τ xz   σ x − p τ xy τ xz   p 0 0 
     
 τ yx σy τ yz  =  τ yx σy − p τ yz  +  0 p 0  (1.5)
   
 τ zx τ zy σ z   τ zx τ zy σ z − p   0 0 p 
where p is the mean or hydrostatic normal stress:

4
Note that equation 1.4 does not necessarily imply that σ1 > σ2 > σ3; however, the latter is a notation that is
often adopted for convenience.
p = (σx + σy + σz)/3 = (σ1 + σ2 + σ3)/3 (1.6)
which can be considered as a uniform pressure acting in all directions. The first tensor on
the right hand side of equation 1.5 is known as the deviatoric stress tensor [s],
representing pure shear, and the second tensor is known as hydrostatic or spherical stress
tensor.
Stress Invariants. From the above discussion it is seen that the magnitude of stress
components at a point will depend on the orientation of the coordinate system chosen. An
important property of tensors is the existence of invariants: single-valued functions of the
components of the stress tensor which are independent of the coordinate system selected.
Three independent invariants are associated with a second order tensor. The first invariant
of the stress tensor [σ], denoted J1, is given by the sum of the diagonal terms (i.e., the
trace) of the stress tensor:
J1 = tr [σ] = σ11 + σ22 + σ33 = σx + σy + σz (1.7)
Clearly J1 is directly related to hydrostatic stress p (eq. 1.6). The second and third
invariants of the stress tensor are not commonly utilized in stress-strain modeling soils
and will not be given here. Instead, two additional invariants are obtained by
consideration of the deviatoric stress tensor (note that there are still three independent
invariants).
The first invariant of the deviatoric stress tensor [s] is zero, so that this tensor has only
two independent non-zero invariants. The second invariant of the deviatoric stress tensor
is given by:
1
[ ]
J 2 D = (σ x − σ y ) 2 + (σ y − σ z ) 2 + (σ z − σ x ) 2 + τ 2xy + τ 2yz + τ 2zx
6
(1.8a)
or, in terms of the principal stresses, by:
J 2D =
1
6
[
(σ 1 − σ 2 ) 2 + (σ 2 − σ 3 ) 2 + (σ 3 − σ 1 ) 2 ] (1.8b)
The invariant J2D plays an important role in describing yield behavior in soil mechanics.
In some cases the third invariant of deviatoric stress tensor is also used to represent the
yield behavior (see table 1.1).
An alternative representation of the physical meaning of the stress invariants is shown
in fig. 1.4c. The principal stress cube can be replaced by a regular octahedron constructed
of eight triangular planes normal to the hydrostatic axis (or space diagonal, defined by σ1
= σ2 = σ3) called octahedral planes. The stresses acting on these surfaces will consist of
eight identical pairs given by a normal stress, called the octahedral normal stress σoct,
plus a shear stress called the octahedral shear stress τoct. The octahedral normal stress is
equal to the hydrostatic stress p, and the octahedral shear stress is related to the square
root of J2D (table 1.1).
Stress Space. Once the principal orientation is known, it is convenient to use the
principal axes as the reference coordinate axes. The relation between the principal
stresses and the stress invariants can be represented geometrically by a cylindrical
coordinate system in stress space (the space bounded by the principal axes), as shown in
figure 1.5. The angle θ is called the angle of similarity and is related to the third invariant
of the deviatoric stress tensor (see table 1.1). It measures the relative value of the
intermediate principal stress. If isotropy is assumed, stress functions are completely
π π
defined if described over the 60° sector AOB ( − ≤ θ ≤ ). An octahedral plane
6 6
passing through the origin, i.e., σ1 + σ2 + σ3 = 0, is known as the octahedral plane (or Π-
plane). Hydrostatic stress is zero on the Π-plane.
Strain State of a Soil
In a similar manner to that of the stress, it is possible to completely describe the state
of strain at a point in a body by a second order symmetric tensor:
 ε11 ε12 ε13   ε x ε xy ε xz   ε 1 γ xy 1 γ xz 

 
[ε] =  ε 21 ε 22 ε 23  =  ε yx ε y ε yz  =  1 2 γ yx 
x 2 2
εy (1.9)
2 γ yz 
1
 ε ε ε     
 31 32 33   ε zx ε zy ε z   1 2 γ zx 1 γ
2 zy εz 
σ1
3τ oct hydrostatic axis
or space diagonal
plane of constant p B
(octahedral plane) 3σ oct
σ3
O
σ2
Figure 1.5—Principal stress space.
Table 1.1. Commonly used stress and strain invariants for elastoplastic modeling of soils
and relationships between them.
System Stress Invariants Strain Invariants

General 3-dim I1σ = σx + σy + σz = σ1 + σ2 + σ3 I1ε = εx + εy + εz
(derived from = ε1 + ε2 + ε3
I2σ = σxσy + σyσz + σzσx - τxy2 - τxz2 - τzy2
characteristic equation)
= -(σ1σ2 + σ2σ3 + σ1σ3) = εv
I3σ = σxσy σz - σxτzy2 - σyτxz2 - σzτxy2 I2ε = -(ε1ε2 + ε2ε3 + ε1ε3)
- 2τxyτyzτzx I3ε = ε1 ε2 ε3
= σ1 σ2 σ3
General 3-dim J1 = tr σ = I1σ I1 = tr ε = I1ε

(derived from tensor[a]) 2
J2D =
1
2
tr (s) = (-s1s2 + s2s3 + s3s1) [b]
I2D =
1
2
tr (e) 2 [c]
6
[
= 1 (σ x − σ y ) 2 + ( σ y − σ z ) 2 + ( σ z − σ x ) 2 ] I3D =
1
3
tr (e)3 [c]
+ τ2xy + τ2yz + τ2zx
[
= 1 (σ1 − σ2 )2 + (σ2 − σ3 )2 + (σ3 − σ1 )2
6
]
J 3D = 13 tr (s)3 = s1s 2s3
= (σ1 − σm )(σ2 − σm )(σ3 − σm ) [b]
Octahedral plane σoct = J1/3 εoct = I1/3 = εv/3

2J γoct = 2
τoct = 3 2D
I
3 2D
Critical-state models p = σoct = (σ1 + σ2 + σ3)/3 εv = ε1 + ε2 + ε3

3J 2 D = 3 τoct
q = εs = 2γ oct
2
Cap models J1 I1p / 3

J 2D
2 I p2D
1  3 3 J 
θ = − sin −1 − 3D 
3  2 J 3/ 2 
 2D 
[a] Subscript D denotes invariant obtained from deviatoric tensor.

[b] s1, s2, and s3 denote principal values of the deviatoric stress tensor s; σm is the mean stress, (σ1 + σ2 +
σ3)/3.
[c] Expressions for strain invariants are analogous to those for corresponding stress invariants; e is the
deviatoric strain tensor.
Fn
area A Fs
σ
τ
l0
ε
π+γ
2
dl Fs
Fn
Figure 1.6—Stress and strain. A consequence of defining compressive strains as positive for soils is that a
positive shear strain corresponds to an increase in the angle between two initially perpendicular line
segments.
Normal strains (ε11, ε22, ε33 in tensorial notation, or, εx, εy, εz in the engineering
convention) are defined as the change in length per unit length of an infinitesimally small
line segment. Engineering shear strains (γxy, γyz, γzx) are defined as the change in angle
between two initially perpendicular line segments (fig. 1.6). They are related to the
tensorial shear strains by 2ε12 = γ12 or 2εxy = γxy. If strain levels are small (i.e., [ε] << 1),
the strain components of a three-dimensional body can be expressed as:
∂u 1 ∂u ∂u
ε11 = ; ε 22 = 2 ; ε 33 = 3 ;
∂x 1 ∂x 2 ∂x 3
 ∂u ∂u   ∂u ∂u   ∂u ∂u 
ε12 = 12  1 + 2 ; ε13 = 12  1 + 3 ; ε 23 = 12  3 + 2  (1.10)
 ∂x   
 ∂x 2 ∂x 1   3 ∂x 1   ∂x 2 ∂x 3 
where u1 , u2 , and u3 are deformations of the body along three mutually orthogonal
rectangular coordinate system axes such as the one in fig. 1.4a. In the absence of body
moments, the tensor is symmetric (εij = εji) and six independent strain components are
required to describe the state of strain.
Principal Strains and Strain Invariants. The properties of the strain tensor are much
the same as those of the stress tensor. It is possible to demonstrate principal strains ε1, ε2,
ε3 and their directions, the principal strain axes, and strain invariants (table 1.1). In terms
of the principal strains the strain tensor reduces to
ε 1 0 0
[ε] =  0 ε2 0 
(1.11)
 0 0 ε 3 
The strain tensor can be decomposed into two parts:
 εx ε xy ε xz   ε x − 13 ε v ε xy ε xz   13 ε v 0 0 
     
 ε yx εy ε yz  =  ε yx ε y − 13 ε v ε yz + 0 1
3
εv 0  (1.12)
   
 ε zx ε zy ε z   ε zx ε zy ε z − 13 ε v   0 0 1
ε 
3 v
The first tensor on the right is the deviatoric strain tensor [e] and is associated with
distortion (i.e., change in shape without volume change). The second tensor on the right
is the volumetric strain tensor and is associated with volumetric change (i.e., change of
size without change of shape):
For small strains, the first invariant of the strain tensor:
I1 = tr [ε] = εv = εx + εy + εz = ε1 + ε2 + ε3 (1.13)
denotes the volumetric strain (εv = -∆V/V). The second invariant of the deviatoric strain
tensor I2D, which is of particular interest, is given by:
I 2D =
1
6
[ 2
]
(ε 11 − ε 22 ) 2 + (ε 22 − ε 33 ) 2 + (ε 33 − ε 11 ) 2 + ε 12 + ε 223 + ε 13
2
(1.14)
In engineering practice, a closely related term known as the octahedral shear strain γoct is
used:
2
γ oct = I 2D (1.15)
3
In critical-state models (e.g., Cam-clay), the deviatoric strain εs, is commonly used (table
1.1).
A geometrical representation of the state of strain in principal stress space can be
made along the same lines as that shown in figure 1.5, provided the axes σ1, σ2, and σ3
are replaced by the axes ε1, ε2, and ε3 (Vyalov, 1986).
Soil Volumetric Variables. Soil volumetric response is usually analyzed in terms of
volumetric strains. The void ratio e (the total volume of pores divided by the total volume
of solids) or the specific volume v = 1 + e (the volume of an element of soil containing
unit volume of soil minerals) are commonly used as volume change variables for soils. It
can be shown that:
de dv (1.16)
dε v = − =−
+
1 e v
where the negative sign is required in soil mechanics, since compressive strains (and
stresses) are defined as positive. In the agricultural engineering literature the volume of
an unsaturated soil is sometimes expressed instead as the bulk weight volume BWV
(Bailey and VandenBerg, 1968; Chancellor, 1994):
1
BWV = dry bulk density (m3/kg) (1.17)
Note that BWV is proportional to the specific volume, BWV = v/(ρwGs), where Gs is the
specific gravity of soil solids and ρw is the density of water. The basic volume-mass
relationship can be used to relate volume and mass variables for the various phases of an
unsaturated soil:
Sr e = w G s (1.18)
where w = gravimetric water content, and Sr = degree of saturation. The relationship

between bulk density, dry bulk density, void ratio and water content for an unsaturated
soil is presented graphically in figure 1.7.
Axisymmetric Conditions
One of the commonly used tests in soil mechanics is the triaxial test as illustrated in
figure 1.8. A cylindrical soil specimen is subjected to axially symmetric loading
conditions so that the axial and radial stresses are principal stresses, i.e., σa = σ1 and σr =
σ2 = σ3. Similarly, the corresponding strains are principal strains (i.e., εa = ε1 and εr = ε2 =
ε3). The stress and strain invariants are determined as:
J1 = σoct = p = (σ1 + 2σ3)/3 (1.19)
τ oct = ( 32 J 2 D )1 / 2 = 2 / 3 (σ1 − σ 3 )
(1.20)
dεoct = dεv = dε1 + 2dε3 (1.21)
γ oct = 2
3
I 2D = 3
2
( dε 1 − dε 3 ) (1.22)
Axisymmetric conditions are sometimes also assumed in analyses of the response of soil
to a circular “footing” on the surface of the soil (e.g., an idealized soil-tire contact area)
which is subjected to a pressure (or sinkage) that is symmetric about the center of the
footing (wheel slippage, which produces tangential shearing, cannot be analyzed as
axisymmetric, nor can non-circular contact areas).
Plane Strain Conditions
Various three-dimensional problems can be approximated as two-dimensional plane
strain problems. Such conditions are relevant for applications that involve footings whose
length is large compared with their width and loading is uniform in the longitudinal (2-)
direction (fig. 1.9c). For example, a slipping wheel may be idealized as a rectangular
plate that applies a uniform vertical pressure across its width while simultaneously
shearing the soil surface (fig. 1.9b). The direct shear apparatus and simple shear
apparatus test soil samples under conditions of plane strain (fig. 1.9a).
Because of the loading and geometry, the strain (and displacement) along the length of
the contact zone are zero:
ε22 = 0 (1.23)
Plane strain problems can be expressed in terms of the stresses in the plane of shearing
(fig. 1.9c), in which case the stress in the 2-direction, σ22 (from symmetry this must be a
principal stress) is a dependent variable that is forced to adopt whatever value is
necessary to comply with the zero strain constraint in that direction (Wood, 1990).
Figure 1.7—Volume-mass relations for unsaturated soils (from Fredlund and Rahardjo, 1993a).
Figure 1.8—Axisymmetric state of stress in triaxial soil sample: (a) Application of stresses;
(b) Representation of principal stresses.
σ
3
3
2 σ
1
1
σ
1 3
Figure 1.9—Examples of plane strain conditions: (a) Direct shear test; (b) Rectangular plate applying
uniform normal load and tangential shearing to soil surface; (c) Stress element.
Large Deformations
In the preceding development, small strains have been assumed to be applicable, i.e.,
strains are given by the conventional engineering strain, defined as the ratio of the
change in length to the original length:
l − l 0 dl
ε= = (1.24)
l0 l0
where ε = strain, l0 = original length of the material without load, l = length of material
after load is applied, and dl = change in length (fig. 1.6). Most stress-strain models for
soil are developed on the basis of engineering strains; however, in some cases, models for
soils are expressed in terms of the natural strain, defined as the ratio of change in length
to current length l:
dl (1.25)
dε =
l
Engineering strain and natural strain are related by:
l
dl  l 
ε= ∫ = ln  = ln(1 + ε)
 (1.26)
l0 l  l0 
At low strain levels, the natural and engineering strains are essentially equivalent;
however, at larger levels of strain they will differ considerably. Agricultural soil-machine
interactions are characterized by large soil deformations and strains. Bailey et al. (1984)
found that their experimental compaction data were better described using natural strains
instead of engineering strains. Consequently, their stress-strain models have been
developed on the basis of natural strains (e.g., Grisso et al., 1984; Bailey et al., 1986;
Bailey and Johnson, 1989). It is generally acceptable to express constitutive models in
terms of engineering strains even for large deformations if incremental strains (dε) are
used in model formulations.5 This is usually the case when solving using numerical
methods such as the finite element method.
Constitutive Models
The description of states of stress and strain at a point in a loaded, deforming material
has been presented. Stress and strain are state variables and are independent of material
properties. Thus, the analyses presented are applicable for all materials, e.g., whether
steel, timber, concrete, plastic, biomaterials, or soil. However, to simulate the physical
behavior of a structure, it is necessary to have models or relationships between stresses
and strains that incorporate the properties of the material in question. These are the
constitutive laws.
Constitutive Laws
The constitutive law is a functional relationship between stress and strain, i.e.:
ƒ (σ, σ& , ε, ε& ) = 0 (1.27)
5
Strain increments (changes in strain) are also preferred for analyses since there is no true zero for strain.
where σ& and ε& respectively represent stress and strain rates6 (Desai and Siriwardane,
1984). The most general linear relationship between the stress and strain tensors takes the
following form (Reismann and Pawlik, 1980):
[σ] = [C] [ε] (1.28)
where [C], called the material property or stiffness matrix, is a fourth order tensor since it
relates two second order tensors.
Alternatively, the constitutive relationship can be written in the inverse form:
[ε] = [D] [σ] (1.29)
where [D] is called the compliance matrix (or tensor).

For soils both shearing and volumetric effects are coupled, i.e., shearing stresses cause
volumetric strains and normal stresses cause shear strains (Atkinson, 1993). As will be
seen, if the constitutive law is expressed in terms of shearing and volumetric invariants,
then for materials that are isotropic and elastic and perfectly-plastic, the stiffness matrix
is symmetric, while for materials that are isotropic and elastic, the off-diagonal terms are
zero so that shearing and volumetric effects are decoupled. Soil stress-strain behavior is
also largely nonlinear and the stiffness and compliance parameters contained in eqs. 1.28
and 1.29, respectively, will not be constant but depend on the strain and the current state
of stress. Stated in another way, the parameters that make up the material property matrix
experience continuous changes as the loading changes (Desai and Siriwardane, 1984).
They are thus dependent on any past loading experienced by the soil, i.e., the stress
history of the medium. Because of the complexity of soil behavior, most useful
constitutive laws for soil are formulated in incremental form, i.e., relating increments of
stress and strain, dσ and dε:
[dσ] = [C] [dε] (1.30)
A major aim of constitutive modeling is to define [C] (or [D]) in such a way that it can
be used in numerical solution techniques such as a finite element method (Desai and
Siriwardane, 1984). However, model predictions using finite element or other numerical
solutions will still be only as good as the underlying model. Thus, a major thrust of
research in theoretical soil mechanics has been to identify forms of stress-strain
relationship that adequately model the observed physical behavior of soils.
Material Behavior Under Load
The first step in predicting the behavior of materials is to choose an appropriate
idealization of material behavior to assist mathematical analysis. The mechanical
behavior of soils fall into two basic categories: (1) brittle behavior and (2) ductile
behavior (fig. 1.10). The type of behavior an agricultural soil will exhibit depends on
various factors including its bulk density, moisture status, and microstructure (i.e., its
stress history), as well as the manner of loading. Materials that exhibit brittle behavior
exhibit very little strain as stress increases, until at some point the material fractures or
6
In conventional soil mechanics true time effects are often ignored, in which case the rates are treated as
increments of stress and strain (Naylor, 1978b).
Figure 1.10—Typical stress-strain and volumetric response of soils when sheared in a shear box or
triaxial apparatus: (a) Ductile response exhibited by normally consolidated or underconsolidated
(“loose”) samples; (b) Brittle failure exhibited by over-consolidated (“dense”) specimens.
cracks. When undergoing ductile behavior, soil deforms in such a way that can be
regarded as elastic, plastic, viscous, or a combination of these. No mathematical model
can completely describe all aspects of behavior of a material as complicated as soil.
Rather, the main features of the behavior are identified and included in the model
description, while aspects considered of minor importance are ignored.
Figure 1.11 shows graphical representations of certain mechanical analog idealizations
of stress-strain behavior.
Soil rarely exhibits linear-elastic behavior over the range of loading of practical
interest (fig. 1.11a). Despite its shortcomings, the linear elastic theory has been the most
commonly used constitutive model for soils (Chen and Baladi, 1985). For example, soil
has been traditionally modeled as an elastic medium for the prediction of compaction
(volume change) induced by surface loads. Non-linear elastic constitutive relationships
can be used to describe soil loading along a particular loading or unloading path with
reasonable accuracy, and in fact can be very accurate for soils sustaining monotonic
loading. These relationships do not provide good representation, however, of soil
behavior under a broad range of loading, unloading, and reloading conditions (i.e.,
general loading conditions). Some form of incremental elasticity features in most soil
models, i.e., there is assumed to exist some region of stress space within which changes
in stress are associated with fully recoverable deformations.
Soil may exhibit behavior similar in some respects to that idealized in figure 1.11e,
although it seldom shows linear-elastic behavior even when loaded from a stress-free,
undeformed state. If loading continues, at some point the soil microstructure (fabric)
“yields” and there is a distinct change in the stress-strain relationship (i.e., soil stiffness).
Yielding and plastic straining may cause hardening (i.e., an increase in yield stress) or
(a)
(b)
(c)
(d)
(e)
Figure 1.11—Idealized stress-strain curves for materials (from Desai and Siriwardane, 1984).
softening (i.e., a decrease in the yield stress). In the latter case the state has reached, and
passed, a peak in the stress-strain curve, a feature commonly found in over-consolidated
soils (Atkinson, 1993) (fig. 1.10b). During unloading-reloading soil usually exhibits non-
linear elastic behavior until reaching the yielding point. Plasticity analysis applied to soils
is often limited due to the occurrence of slip surfaces7 caused by instability, particularly
for strain softening conditions.
7
Intense shearing may occur in a very thin zone (possibly on the order of 10 grains thick for soils) leading to
discontinuity of the soil fabric (Atkinson, 1993). These zones are called slip or shear planes or surfaces.
Rigid, perfectly-plastic (fig. 1.11b) and elastic, perfectly-plastic (fig. 1.11d) models
are sometimes used in limit equilibrium analysis. In these idealizations no hardening or
softening occurs, as indicated by the horizontal portion of the stress-strain curves
(constant yield stress). The soil could undergo plastic deformations while strain is
increased, but at some point will “fail.” For the study of applications involving the
maximum strength of soils (e.g., traction, tillage) soil is frequently modeled as a rigid
perfectly-plastic material where the stresses at failure are related by the linear Mohr-
Coulomb failure criterion. No account is taken of strains or deformations within the soil
before or after “failure.” The state of stress that exists when the Mohr’s circle of applied
stresses becomes tangent to the failure surface defines the soil strength and the plane of
these stresses defines a failure surface. A plane of sliding is commonly observed in
vehicle traction and tillage tool-induced failure, justifying the failure concept physically
(Freitag, 1985). The limit equilibrium approach has had significant success for prediction
of draft of simple tillage tools using “passive earth pressure theory” (Hettiaratchi et al.,
1966; Hettiaratchi and Reece, 1967; McKyes and Ali, 1977; Perumpral et al., 1983).
Rate Effects. Rate or viscous effects are commonly ignored in conventional soil
mechanics. This has not necessarily been the case in agricultural soil mechanics, or Soil
Dynamics, since most tillage and traction operations involve rapid mobilization of the
soil. Within the context of a Continuum Mechanics framework there have been
comparatively few studies that have used rate-dependent models to analyze soil-machine
interactions. This has largely been because of the additional complexities involved, both
from a theoretical and a computational perspective. Furthermore, since the static and
quasi-static mechanical behavior of unsaturated soils is still little understood compared to
that of saturated soils, efforts have focused on establishing a better understanding of static
agricultural soil behavior. Furthermore, many dynamic constitutive formulations add rate-
dependent terms to a static failure locus. In recent years, there has been an increase in
efforts to model the dynamic behavior of agricultural soils. A brief review of
developments is given in the section on Dynamic Soil Models.
Elasticity
A simple definition of elastic material behavior is that there is a unique relationship
between stress and strain (Desai and Siriwardane, 1984). An elastic medium retains no
permanent deformation in a closed stress cycle; in other words, it returns to its initial state
after a cycle of loading and unloading. An elastic stress-strain relationship may be linear
or non-linear.
One of the most widely recognized applications of elasticity in agricultural soil
mechanics is the use of the Boussinesq formulas to estimate the distribution of vertical
stress beneath a surface load (Chancellor, 1994, pp. 234-238). The inclusion of a stress-
concentration factor is intended to correct for some of the limitations arising from the
assumption of a linear elastic, homogeneous, isotropic material. The integration
(summation) of Boussinesq equations for point loads over a finite region (e.g.,
representing the soil-tire contact area) is an application of the principle of superposition,
another property of linear elastic behavior.
The formal theory of elasticity requires that there exist unique strain energy and
complimentary energy functions (Desai and Siriwardane, 1984), putting thermodynamic
restrictions on the way the model is created. For a homogeneous elastic material, [C] (eq.
1.26) is made up of 81 elements. Symmetry of stress and strain tensors, and strain energy
considerations, reduce the number of independent constants required to 21.
In the special case of linear elasticity the material obeys Hooke’s Law. If the material
is isotropic, then there are only two independent material constants. The pair of constants
E (Young’s modulus, defined as the ratio of stress to strain under direct load) and ν
(Poisson’s ratio, the ratio of lateral strain to longitudinal strain under direct load) are
probably the most well known. When the elasticity of soils is considered, it is more
useful to use an alternative pair of elastic constants: the bulk modulus K, and the shear
modulus G (fig. 1.12). These parameters divide the elastic deformation into volumetric
(change of size at constant shape) and distortional (change of shape at constant volume)
parts, respectively. Thus, the bulk modulus is defined as the ratio of stress to strain under
uniform hydrostatic stress conditions:
p
K= e (1.31)
εv
Similarly, the shear modulus is defined as the ratio of stress to strain under shear. The two
sets of coefficients are related by:
E
G= (1.32)
2(1 + ν)
E
K= (1.33)
3(1 − 2 v)
The elastic constitutive relation (eq. 1.28) for a linear elastic material in terms of the
material parameters K and G is given by:
Hydrostatic Deviatoric
stress, p stress, q
K 3G
Volumetric strain, ε ev Deviatoric strain, εse
(a) (b)
Figure 1.12—Stress-strain behavior of an ideal linear-elastic, isotropic material: (a) Compression and
swelling – change in size at constant shape; (b) Shearing – change in shape at constant volume.
 4G 2G 2G 
K + 3 K−
3
K−
3
0 0 0
 
 σ11  K − 2G K+
4G
K−
2G
0 0 0   ε11 
σ   3 3 3  ε 
 22   2G 2G 4G   22 
σ 33   K − K− K+ 0 0 0  ε 
3 3 3  33 
 =  
σ
 12   0 0 0 2G 0 0  ε12 
σ 23    ε 23 
    
 σ13   0 0 0 0 2G 0   ε13 
 
 0 0 0 0 0 2G  (1.34)
 
The inverse relationship (cf. eq. 1.29) is given by:
 1 1 1 1 1 1 
 3G + − − 0 0 0 
9K 9 K 6G 9K 6G
 
 ε11   1 −
1 1
+
1 1
−
1
0 0 0   σ11 
ε   9K 6G 3G 9K 9K 6G  σ 
 22   1 1 1 1 1 1   22 
− − + 0 0 0  σ 
ε 33   9K 6G 9 K 6G 3G 9K  33 
 = 1  
ε12   0 0 0 0 0  σ12 
ε 23   2G  σ 23 
   1  
 ε13   0 0 0 0 0   σ13 
2G
 1 
 0 0 0 0 0  (1.35)
 2G 
Working in terms of Young’s modulus and Poisson’s ratio, the strain of a soil specimen
under a cylindrical triaxial stress state is described by:
ε e1  1  1 − 2ν  σ1 
 e=    (1.36)
ε 3  E − ν 1 − ν  σ 3 
Using K and G to separate volumetric and distortional effects, the elastic response of
an isotropic material can be written as:
ε ev  1 / K 0  p 
 e =  0 1   (1.37)
ε s   3G  q 
where ε se = 23 (ε1 − ε 3 ) is the elastic deviatoric (shear) strain, and q = (σ1 - σ3) is referred
to as the deviatoric stress (table 1.1). The off-diagonal zeros in equation 1.37 indicate the
absence of coupling between volumetric and distortional effects.
Non-Linear Elastic Models. The stress-strain behavior of soil may be accurately
described using non-linear elastic models for specific loading paths and conditions
(Upadhyaya, 1994). Non-linearity implies that the elastic parameters are not constant, but
depend on the state of stress and/or strain of the soil. There are two general approaches to
formulating non-linear stress-strain laws for implementation in numerical solutions such
as a finite element method: those in which the load is applied all at once, and those in
which it is applied in a series of small increments that approximate a differential (Naylor,
1978a, 1978b). The former method can be applied to virtually all forms of stress-strain
law but tends to be inefficient for strongly non-linear problems. The second category
requires the stress-strain law to be in differential form, or at least be differentiable
(Naylor, 1978a). The tangent modulus method is an example of this approach in which an
incremental analysis is usually conducted using tangent moduli values, such as the
tangent bulk modulus, Kt, and the tangent shear modulus, Gt. The non-linear form of
equation 1.32 can then be written as (cf. eq. 1.30):
 1 1 1 1 1 1 
 3G + − − 0 0 0 
9K t 9K t 6G t 9K t 6G t
 t 
 1 −
1 1
+
1 1
−
1
0 0 0   dσ11 
 dε11   9K 6G t 3G t 9K t 9K t 6G t 
dε   t  dσ 22 
 22   1 −
1 1
−
1 1
+
1
0 0 0  
dε 33   9K 6G t 9K t 6G t 3G t 9K t  dσ 33 
=
t
 1  
d ε
 12   0 0 0 0 0  dσ12 
dε 23   2G t  dσ 23 
   
 dε13  
1
0 0 0 0 0   dσ13 
2G t
 
 1 
0 0 0 0 0
 2G t  (1.38)
For triaxial, cylindrical loading conditions, equation 1.38 reduces to:
dε v  1 / K t 0  dp 
 = 0 1   (1.39)
 dε s   3G t  dq 
The variable moduli Kt and Gt can be represented as explicit functions of stress and
strain, or of stress and strain invariants (fig. 1.13b). Often models are expressed in terms
of tangent Young’s modulus and Poisson’s ratio to represent stress-strain curves for soils
(fig. 1.13c). These are known as variable moduli or variable parameter models and many
Figure 1.13—Variable parameter models: (a) Linear-elastic; (b) Variable G and K; (c) Variable E and ν
(from Desai and Siriwardane, 1984). Stress and strain state variables are defined in table 1.1.
such models have been proposed for soil. In general, soil stiffness (bulk modulus)
increases with hydrostatic stress, and shear modulus reduces as shear stress (or the second
invariant of the deviatoric stress tensor) is increased with little or no change in mean
stress. Often K and G are assumed to vary linearly with stress invariants (Naylor, 1978b;
Desai and Siriwardane, 1984; Upadhyaya, 1994). One possibility is:
Kt = K0 + αK p (1.40)
Gt = G0 + αG q (1.41)
where K0 and G0 are the initial bulk and shear modulus, respectively, q = 3J 2 D = 3
2
τ oct
and αK > 0 and αG < 0 are stress related material parameters.
From equations 1.39, 1.40, and 1.41 the following differential equations are obtained
for axisymmetric cylindrical conditions:
dp
dε v = (1.42)
K0 + αKp
dq
dε s = (1.43)
3(G 0 + α G q )
Integrating equations 1.42 and 1.43, respectively, gives:
1  K0 + αKp 
εv = ln  (1.44)
α K  K0 

1  G + αGq 
εs = ln 0 
 (1.45)
3G 0  G0 
A modification of equation 1.42 is often used to describe the non-linear elastic behavior
of soil before it yields. If K0 is assumed to be equal to zero, integration of equation 1.42
yields:
1
εv = ln p + constant (1.46a)
αK
or, using the relationship between volumetric strain and the void ratio (eq. 1.16):
e = eK - κ ln p (1.46b)
where eK is the soil void ratio at unit hydrostatic stress, and κ is the logarithmic bulk
modulus and is related to -1/αK. Equation 1.46b is often used to describe the elastic
Figure 1.14—A hyperbolic representation of stress-strain curve for soils.
unloading-reloading response of soil. For normally consolidated soil, isotropic

compression is described using a similar expression:
e = eN - λ ln p (1.47)
where eN and λ are soil parameters. Equations 1.46b and 1.47 are used in describing soil
behavior in the Cam-clay critical-state formulation.
The use of hyperbola to represent stress-strain curves for soils was first proposed by
Kondner and coworkers (Kondner, 1963; Kondner and Zelasko, 1963; Kondner and
Horner, 1965) in the form:
ε
(σ1 − σ3) = a + bε (1.48)
where 1/a is the initial tangential Young’s modulus and 1/b is the asymptotic value of
failure stress, i.e., (σ1 −σ3)ult (fig. 1.14). The compressive strength at failure (σ1 − σ3)f is
slightly smaller than this asymptotic value, so the curve is cut off at the failure value. By
relating the asymptotic value to compressive failure stress using a factor, Rf, i.e.:
(σ1 − σ3)f = Rφ (σ1 − σ3)ult (1.49)
Equation 1.48 can be rewritten as:

ε
q = σ1 − σ 3 =
1 εR f 
 +  (1.50)
 E i (σ 1 − σ 3 ) f 
The factor Rf < 1 is called the failure ratio, and is independent of confining pressure
(Duncan and Chang, 1970a). The response of many soils depends on the confining
pressure for a given stress path. Duncan and Chang (1970a, 1970b) used a relationship
proposed by Janbu (1963) to express initial tangent modules as a function of confining
stress as:
n
 σ 
E i = K ' p atm  3 
 (1.51)
 p atm 
where K´ and n are empirical constants, patm is the atmospheric pressure, and σ3 is the
minor principal stress. The deviatoric stress at failure was expressed in terms of the
Mohr-Coulomb criterion as:
2c cos φ + 2σ 3 sin φ
(σ 1 − σ 3 ) f = (1.52)
1 − sin φ
where c is the soil cohesion and φ is the soil internal angle of friction. The tangent
Young’s modulus was obtained by differentiating equation (1.50) with respect to axial
strain ε, and after some manipulation the following expression was obtained for the
tangent modulus value at any stress condition (Duncan and Chang, 1970a):
2 n
 R (1 − sin φ)(σ1 − σ 3 )   σ3 
E t = 1 − f  K ' p atm  
 (1.53)
 2 c cos φ + 2σ 3 sin φ   p atm 
This expression involves five parameters (Rf, c, φ, K´, n) that have direct physical
meanings and can be obtained from triaxial tests on cylindrical specimens. It is also very
easy to implement the model into incremental stress analysis techniques such as a finite
element method.
The model of Bailey et al. (1984) is another form of hyperbolic model that has proved
successful in describing anisotropic consolidation of agricultural soils for various soil
textures over a wide range of consolidation pressures (e.g., Chi et al., 1993c).
Naylor (1978b) warns that when differential models that assume constant ν and
varying E (or G) are used, if Et disappears as a failure stress state is reached, both Gt and
the tangential bulk modulus Kt disappear as well. Such behavior is illustrated in figure
1.15 in which such models exhibit a progressive decrease in volume as volumetric strain
increases with axial strain without leveling off. If shearing continues long enough, the
material will literally “disappear.” When both variable Et and νt are used, the behavior
near failure cannot be described adequately, leading to numerical instabilities (Chen,
1984). In practice, ν increases to 0.5 and K to infinity as failure is approached, while the
shear modulus is left unchanged so that the failed soil is modeled as having the capacity
to sustain further shear stresses. Duncan and Chang (1970a, 1970b) and Chi and
Kushwaha (1990) used a constant value of Poisson’s ratio in their studies. Desai and
Siriwardane (1984) and Chi et al. (1993b) reported hyperbolic models for variable
Poisson’s ratio similar to equation 1.53. Chi et al. (1993a) proposed a simplified form of
the hyperbolic relationship to represent variable Poisson’s ratio. These relationships along
with experimental data are presented in Volume 1 of this monograph (Chancellor, 1994).
Figure 1.15 —Comparison of four stress-strain laws for a drained triaxial test on normally consolidated
clay. (1) E-ν model: E = 100p - 130q, ν = 0.33; (2) K-G model: K = 100p, G = 38p - 49q; (3) Hyperbolic
model: a = 0.234, b = 0.320, ν = 0.33; (4) Critical-state model: E = 250,000 kPa, ν = 0.33, p0 = 150 kPa, χ =
(λ - κ)/(1 + e) = 0.01 (from Naylor, 1978b). With a suitable choice of parameters, all models can be made
to fit experimental q-ε1 data with sufficient accuracy. Predictions of volumetric strain εv may be
significantly different and are of more significance.
Although variable moduli and hyperbolic models are easy to incorporate into
numerical analysis, and are capable of producing quite accurate results for monotonic
loading conditions, the path dependency of the parameters limits their applicability
(Upadhyaya, 1994). Abrupt increases in soil stiffness on unloading cannot be auto-
matically brought about by variable moduli models (Naylor, 1978b) thus these models are
not capable of accurately predicting the response of soils volume under unloading-
reloading paths. A further limitation is that because volumetric and deviatoric
components are decoupled, these models cannot be used to determine change behavior
under shear (Desai and Siriwardane, 1984). Plasticity based models can overcome some
of the limitations of nonlinear elastic formulations. Since variable parameter models
represent soil behavior much better than linear isotropic elastic formulations, they are
often used in elastoplastic models to describe soil behavior up to the point of yielding.
Elastoplastic Behavior
Figure 1.16 represents a typical stress-strain curve for a metal bar under uniaxial
tension. Soil often exhibits similar behavior under isotropic compression as well as
triaxial compression. In the case of metals, when the load is increased gradually, it
usually exhibits linear elastic behavior up to point A, so that if the load is removed the
body regains its original state. Beyond point A subsequent loading is still reversible,
although the stress-strain relation is non-linear. However, there is a point B, termed the
yield point, beyond which unloading is not fully reversible. Following yield the bar will
further deform if loaded, say, to point C. Between the first yield at B and failure there are
simultaneous elastic (reversible) and plastic (irreversible) components of strain and the
material behavior is termed elastoplastic. If the load is removed at some point on this
path, the specimen will usually follow an unloading path such as CDE.
There will be some permanent strain (i.e., irrecoverable deformation) in the body, and
the material is said to have undergone plastic deformation (plastic flow). If this specimen
is reloaded from point E, it may follow a path such as EFC during which the material
usually behaves elastically until it reaches point C. The material is said to have undergone
work hardening since the material’s yield point has been effectively raised (i.e., point C,
which represents a “new” yield point for the material, corresponds to a higher stress level
σ G
K
C
J
B
σy
A
F H
D
E I ε
Plastic Elastic
Figure 1.16—Elastoplastic stress-strain behavior under uniaxial tension. The shaded area represents
energy dissipated during the elastic hysteresis loop.
than point B). Beyond point C the material will exhibit elastoplastic behavior again and
follow path CG. If the specimen is unloaded and reloaded at point G it may trace an
unloading-reloading curve GHIJ similar to the curve CDEF. Because the reloading paths
do not follow the original loading path, strains will be dependent on the history of stress
states when plastic deformations occur. For instance, points J and K represent different
states of stress for the same state of strain. Unloading-reloading curves are generally
close to each other and the shaded area under the curve CDEF represents elastic
hysteresis (anelastic) energy loss. It is the reversibility that is the important feature
distinguishing between elastic and plastic straining of a material. For an elastic material
the state of stress is a function of the current state of deformation only. The deformations
do not depend on how or along what stress path the state of stress was reached (Desai and
Siriwardane, 1984). In contrast, plastic behavior is characterized by the history-dependent
information.
The theory of plasticity provides a mathematical framework to generalize this idea of
one-dimensional yielding to more general conditions, in which any suitable combination
of hydrostatic and deviatoric stresses may lead to yield and plastic flow. In classical
plasticity theory five types of statement are employed to describe the stress-strain
relationships for an elastoplastic material under general loading conditions. These are:
1. Elastic properties describing recoverable deformations of the soil.
2. A yield surface defining the boundary of the elastic domain and thus determining
when plastic deformation occurs.
3. A hardening law describing the rate at which the yield surface grows or shrinks.
The hardening law generalizes the concept of uniaxial yield stress being increased
by strain hardening or decreased by strain softening to more general stress states.
4. A plastic potential surface in stress space that specifies the mode of plastic
deformation in stress space by specifying relative magnitudes of incremental plastic
strains when the material is yielding.
5. A flow rule that relates plastic strain increments to stress increments.
Given the above ingredients, the complete constitutive relationship for an elastoplastic
material can be formulated. As an example, the stress-strain relationship for the modified
Cam-clay model will be derived later in this chapter, after concepts of critical-state soil
mechanics have been introduced.
Elastic behavior has already been presented. The remaining features of an elastoplastic
formulation (2 - 5) will be described in the following sections.
Yield Surface. The yield surface (or yield locus) generalizes the concept of the yield
stress, which defines the elastic domain in uniaxial plasticity, to two- and three-
dimensional stress states. In general a yield criterion is written:
ƒ(σij) = 0 (1.54)
describing a surface in three-dimensional stress space. If the material is assumed to be

isotropic, the yield criterion can be written in terms of the principal stresses, or more
conveniently, in terms of the invariants of the stress tensor. The general shape of a failure
surface in three-dimensional space can be described by its loci in the hydrostatic planes
and in the Π-plane. (Specific forms of yield surface encountered in soil mechanics will be
discussed further on.)
The yield surface may change in size, position or shape in stress space as the soil is
loaded (i.e., due to work hardening or softening) (fig. 1.17). Geometrically, the point
representing the current state of stress must lie on or within the yield surface. In classical
plasticity theory any state of stress located within the surface, i.e., ƒ(σij) < 0, is
considered to be under elastic state. If during loading the state of stress reaches the yield
criterion the material undergoes plastic flow (ƒ = 0 produces perfectly-plastic behavior).
Stress states outside the surface are impossible to attain. Thus, once the soil is in a plastic
state, loading leads to another plastic state and the stress state must continue to satisfy the
yield criterion. This requirement is called the consistency condition (Prager, 1949). Yield
surfaces must also be convex and must contain the stress origin (Hill, 1950). The
requirement of convexity arises from the condition of irreversibility in plasticity theory,
which states that because of the irreversible nature of plastic strains, the work done on the
plastic deformations will be positive (Prager, 1949; Drucker, 1950; Desai and
Siriwardane, 1984). The plastic strains during flow are governed by a hardening law and
by a flow rule.
C
B
Figure 1.17—Successive yield surfaces in stress space: A→B stress state increment inside current yield
locus, and B→C→D stress increments expanding the yield locus (strain-hardening).
Hardening Law. The hardening law is a function describing the change in size and
position of the yield surface in stress space. The hardening law is incorporated into the
yield surface equation by writing:
ƒ(σij, h) = 0, (1.55)
where h is the hardening parameter.

Two hypotheses have been proposed to define the degree of hardening (Desai and
Siriwardane, 1984). The work hardening hypothesis assumes that hardening depends only
on the plastic work and is independent of the strain path. According to this hypothesis,
the yield criterion can be written as:
ƒ = ƒ(σij, Wp). (1.56)
The second hypothesis, referred to as the strain hardening hypothesis, assumes that
plastic strain is a measure of hardening so that the yield criterion can be written as:
ƒ = ƒ(σij, εp). (1.57)
In the critical-state model, hardening is assumed to be governed by the accumulated

plastic volumetric strain. In soils where post-peak strength loss is important, h may be a
measure of the accumulated plastic deviatoric strain (Naylor, 1978b).8 In the most general
formulations, hardening can be expressed as a function of both volumetric and deviatoric
strains.
Hardening which results in the yield surface expanding or shrinking about the origin is
referred to as isotropic hardening. If the yield surface translates in space, hardening is
referred to as kinematic hardening. Mixed hardening refers to a combination of the two.
Flow Rules and the Plastic Potential. The magnitude of plastic strain is associated
with the change in size of the yield locus. The various components of total incremental
plastic strains that occur are determined through knowledge of the plastic potential and a
flow rule. The plastic potential function g = 0 defines a surface in stress space that passes
through the point representing the current state of stress and to which the plastic strain
increments are normal. The plastic strain increments are related to the stress increments
by the flow rule. The flow rule is given in terms of the ratios of the plastic strain
increments and so describes the relative sizes of individual strain increments, but not their
absolute magnitudes. It is generally described by the following equation:
∂g
dε ijp = L (1.58)
∂σ ij
where L is a non-negative scalar called the plastic multiplier or loading index. Its value
depends on the assumed hardening law for the soil, or if there is no hardening, on the
boundary conditions. The soil is said to follow an associative flow rule (or normality
since in this situation vector of dεp is normal to the yield locus) if the yield function and
the plastic potential are the same, g = ƒ, otherwise the flow rule is non-associative (fig.
1.18).
8
The yielding of sands is associated largely with shearing, i.e., major particle rearrangement.
σj , σj ,
p
dε j dε j
p
p dε
dε
plastic potential plastic potential
p yield stress
dε j
yield surface
p
dε i
yield surface
p
σi , dε i σi , dε i
p
(a) (b)
Figure 1.18—Plastic potential surface and flow rules: (a) Associated flow, (b) Non-associated flow.
Stability and Work Hardening. Drucker (1950) postulated that a stable system in
equilibrium when loaded by an external agency absorbs work from the external agency,
whereas an unstable system releases work. Thus, if the external agency is incapable of
extracting work from the system, the system will collapse. Stable deformation is
equivalent to strain hardening behavior, whereas unstable deformation corresponds to
strain softening in a uniaxial test (fig. 1.19). Drucker showed that this definition of
stability means that work done on the plastic deformations will be positive:
dWp = σ d εp > 0 (1.59)
From this concept, Drucker deduced that the yield surface must be convex and that
normality, i.e., the associative flow rule, applies (Drucker, 1950; Kurtay and Reece,
1970).
(a) (b)
Figure 1.19—Stable (a) and unstable (b) responses in a tension test, corresponding to strain-hardening
and strain-softening behavior, respectively (from Britto and Gunn, 1987).
Evidence for Yielding in Soils

One-dimensional consolidation provides an excellent illustration of the elements of
soil elastoplastic behavior (fig. 1.20). Initially, over-consolidated soil behavior is fairly
stiff until it reaches the break, or “yield,” at the preconsolidation stress. Post-yield
normally consolidated behavior is more compressible and distinctly non-reversible. If
unloaded and reloaded, soil will exhibit nonlinear elastic (i.e., recoverable) behavior until
it reaches the new yield point on the normal consolidation line (i.e., the preconsolidation
stress for the loop). Some hysteresis is evident in the unloading-reloading loop. Similar
patterns can be observed in isotropic consolidation tests and triaxial compression tests.
Wood (1990) reviewed experimental evidence for yielding in saturated soils. Yielding
is often readily observed in natural, undisturbed clays. The point of yielding is often
much less clear in insensitive soils and if the soil structure has been previously disturbed
(see fig. 1.20). Yielding is also less easy to see in tests that do not subject the soil to pure
tension. Often the best approach is to plot experimental data in many ways (e.g., q versus
p, q versus εs, q versus ε1, e versus p, p versus εv) and make a number of independent
estimates of the yield point. Moulin (1988) presented a graphical approach that allows
various combinations of stress and strain to be examined in a single diagram. Early
evidence for yielding in clays in one-dimensional oedometer tests was presented by
Bjerrum and Kenney (1967) who suggested then that the preconsolidation pressure be
Figure 1.20—One-dimensional consolidation of soils (from Holtz and Kovacs, 1981). The rounded
curvature of the consolidation curve near the “yield” point as defined by the preconsolidation stress,
shows that this soil has been disturbed. The point on the normal consolidation curve just prior to
unloading is the preconsolidation stress for the unloading-reloading curve.
considered a yielding point. Further evidence was provided by subsequent work for
triaxial tests by Bjerrum and his colleagues at the Norwegian Geotechnical Institute.
Mitchell (1970) and Crooks and Graham (1976) identified yield stresses in tests in lightly
over-consolidated clays following a range of general stress paths. They established yield
surfaces in stress space from the locus of points of set yield stresses (fig. 1.21). Similar
behavior is evident in tests of unsaturated soils; however, failure is more difficult to
define.9
Anisotropy is an important feature of many field or natural soils, even for elastic
behavior. The effect of anisotropy is often indicated by significant non-symmetry of the
yield locus about the p-axis when plotted as a function of the stress invariants. Soil that
has experienced only isotropic stresses and deformations in its history would be expected
to show yield loci which cross the p-axis at right angles and would have identical
stiffness for small changes in q whether loaded in extension or compression (Wood,
1990).
The experimental data thus support the application of concepts of metal plasticity to
yielding and deformation of soils, although there are some significant departures in
details. For example, for both sands and clay soils, unlike metals, the deviatoric stress q is
strongly dependent on hydrostatic pressure p. There is support for normality in clays,
although there is experimental evidence that plastic strain increments are not strictly
normal to yield surfaces even for plastic clays. However, the deviations may be small
enough that the assumption of associative flow is reasonable (fig. 1.22) (Wong and
Mitchell, 1975; Graham et al., 1983a). Poorooshasb et al. (1966, 1967), Lade and Duncan
(1973, 1975, 1976), and others have shown that cohesionless soils (i.e., sands) do not
exhibit associated flow behavior. In such soils the assumption of normality results in
excessive dilatancy (i.e., prediction of volume increase during plastic flow) (Drucker and
Prager, 1952). Non-associative flow rules are commonly used for soils in which strain
softening can occur.
Figure 1.21—Yield surfaces for lightly over-consolidated clay in triaxial compression (Mitchell, 1970).
9
Several failure criteria have been proposed for unsaturated soils. All depict some maximum combination of
stresses that an unsaturated soil can resist (see Fredlund and Rahardjo, 1993a: 225-227).
Figure 1.22—Evaluation of associative flow rule for data obtained from triaxial tests on undisturbed
Winnipeg clay: (a) Plastic strain increment vectors plotted at yield points; (b) Departure from normality
(data from Graham et al., 1983, replotted by Wood, 1990). Notation: η = q/p, δeqp = de pp = de pv
Specific Yield Surfaces for Soils

Naylor (1987b) classified yield surfaces for soils into two categories: those that can be
represented by open-ended cones in stress space; and those that consist of capped cones
(cap models). The former category of yield surfaces are generally assumed not to strain
harden, as evident by a fixed cone angle, although they may strain soften. They define
conventional “failure” criteria, or the locus of ultimate or maximum stress states at
failure. Cap models allow for strain hardening of soil prior to failure; the overall yield
surfaces consist of a fixed failure envelope that intersects with a series of “moving” yield
surfaces that form the hardening caps. Yield criteria are generally established based on
experimental data. The initial experimental data were obtained for yielding of metals and
as a result the development of soil plasticity theory has been strongly influenced by the
theory of metal plasticity (Chen and Baladi, 1985). For ductile metals two commonly
applied yield criteria are those due to Tresca and to von Mises. According to Tresca’s
criterion (or the maximum shear stress theory), plastic yielding starts when the maximum
shear stress reaches a value equal to σy, the yield stress in uniaxial tension. The yield
stress σy must be determined experimentally. In principal stress space this criterion can be
represented as a surface of a prism with a regular hexagonal cross-section, centered on
the hydrostatic axis (figs 1.23 and 1.25). The von Mises yield criterion (or distortion
energy theory) states that yielding starts when the distortional energy reaches a value
equal to the elastic distortional energy at yield in uniaxial tension. This is equivalent to
the second invariant J2D reaching a value equal to k = σy/ 3 :
1
6
[ ]
( σ 1 − σ 2 ) 2 + (σ 2 − σ 3 ) 2 + ( σ 3 − σ 1 ) 2 − k 2 = 0 (1.60)
In principal stress space the von Mises criterion is equivalent to a cylindrical surface
which circumscribes the Tresca surface (see fig. 1.25). Both the Tresca and von Mises
criteria assume that the yield strength in tension and compression are identical (the size of
the hexagon or circle is independent of the values of p), clearly a limitation for soils
where shear strength increases with hydrostatic stress. The only case in which these
models can be used is for undrained saturated soils when the analysis is performed in
terms of total stresses (Chen and Baladi, 1985). The Tresca criterion is implicitly invoked
in calculations of bearing capacity of soils (Wood, 1990). Yong and Windisch (1970)
used the Tresca criterion and an associated flow rule to estimate contact stresses at the
wheel-soil interface in a frictionless clay soil (i.e., for which soil shear strength will not
increase with hydrostatic stress) from measured instantaneous deformation patterns.
Because of the very short soil loading period by the moving wheel they assumed there
was no dissipation of pore pressures so that a total stress analysis was appropriate.
A common failure model for soils is the Mohr-Coulomb criterion:
|τf| = c + σn tan φ (1.61)
where τf = shear strength, c = soil cohesion, σn = normal stress on the plane of failure,
and φ = soil internal angle of friction. This equation is normally interpreted in terms of a
failure envelope of Mohr’s circle plots (Chancellor, 1994). By rewriting the equation as:
σ1 - σ3 = sin φ (σ1 + σ3 + 2c cot φ) (1.62)
it can be represented in principal stress space as an irregular hexagonal pyramid (figs

1.23 and1.24). Equation 1.62 shows that the Mohr-Coulomb criterion ignores the effects
Figure 1.23—Mohr-Coulomb and Tresca yield criteria (from Desai and Siriwardane, 1984).
Figure 1.24—Soil failure criteria (from Chen and Baladi, 1985). Note that 3 σoct = 3 p is the
hydrostatic axis (compression to the left of the origin), and 3 τoct is the deviatoric axis. All these yield
criteria assume isotropy (resulting in six-fold symmetry over 60° segments) and convexity in principal
stress space. The failure criteria shown will always predict dilation. This can be seen by imagining an
outward pointing plastic strain increment vector at any point on the surface and assuming that the
principal strain axes are coincident with the principal stress axes. The direction of the plastic strain
vector will always be such that the plastic volumetric strain component is negative (directed to the right),
indicating a plastic volume increase.
of the intermediate principal stress. It does account for the increase in strength with
increasing hydrostatic stress. In terms of stress invariants J1, √J2D, and θ, the Mohr-
Coulomb criterion can be expressed as:
J 1 sin φ + J 2 D cos θ − 1
3
J 2 D sin φ sin θ − c cos φ = 0 (1.63)
For the special case of frictionless materials for which φ = 0, Mohr-Coulomb reduces
to the maximum-shear-stress criterion of Tresca (fig. 1.23). The general validity of
equation 1.61 has been well established for agricultural soils. Figure 1.25 presents
experimental strength data plotted in the Π-plane for a variety of soil textures, showing
that Mohr-Coulomb presents a good, slightly conservative soil failure criterion.
Unfortunately, the Mohr-Coulomb criterion is not convenient to use in three-dimensional
applications due to the presence of corners (mathematical singularities). A further
limitation of the Mohr-Coulomb criterion is that it does not represent soil strength
correctly in the tensile zone.
Figure 1.25—Comparison of some experimental results of measurement of strength of disturbed samples

of soils with yield criteria on the Π-plane (from Feda, 1982). The Drucker-Prager locus coincides with the
von Mises criterion in this projection.
Fredlund et al. (1978) proposed a Mohr-Coulomb type failure criterion for unsaturated
soil:
τf = c´ + (σn - ua) tan φ´ + (ua - uw) tan φb (1.64)
where c´ and φ´ are the effective cohesion and effective internal friction angle,
respectively (i.e., for saturated soil), (σn - ua) is called the net stress, ua is the pore-air
pressure, uw is the pore-water pressure, and φb is the friction angle with respect to
changes in matric suction (ua - uw). Equation 1.64 describes a three-dimensional surface
as shown in figure 1.26. For many agricultural applications, the pore space can be
assumed to be interconnected up to the surface so that the pore-air pressure ua is equal to
zero (gauge) and the net stress is also the total (applied) normal stress. Equation 1.64 can
then be equated to equation 1.61 if the total cohesion c is equal to {c´ + (ua - uw) tan φb},
and if φ is equal to φ´. The parameter φb decreases with increasing suction (Escario and
Sáez, 1986; Gan et al., 1988) describing a nonlinear variation of shear strength with
matric suction, and φ´ may be a function of suction for some soils (Escario et al., 1989).
Fredlund et al. (1995) presented a model to estimate the shear strength function (eq. 1.64)
using c´, φ´, and the soil-water characteristic function. Wulfsohn et al. (1996) verified the
Fredlund et al. (1995) model for a sandy clay loam soil and demonstrated its use to
estimate the variation of tool draft and maximum traction with soil water content.
Drucker and Prager (1952) presented a smooth generalization of Mohr-Coulomb that
includes the effects of all the principal stresses on soil shearing resistance. In terms of
stress invariants the Drucker-Prager criterion can be expressed as:
J 2 D − αJ 1 − k = 0 (1.65)
Figure 1.26—Extended Mohr-Coulomb failure criterion for unsaturated soil (from Wulfsohn et al.,
1996).
where α > 0 and k > 0 are material parameters that can be related to the Mohr-Coulomb
coefficients c and φ, respectively.10 In stress space the criterion plots as a conical surface
(fig. 1.24). For purely cohesive soil, α = 0 and the criterion reduces to the von Mises
condition. The Drucker-Prager criterion was proposed as a classical plasticity theory
yield criterion.11 For stress states within the yield surface, the material undergoes elastic
strain only. When equation 1.65 is satisfied the soil flows plastically, and undergoes both
elastic and plastic strains. This criterion has the drawbacks of the Mohr-Coulomb yield
surface described above and, furthermore, the circular yield locus does not give as good a
fit to soil failure data. It has been established that there is a dependence of q on the
intermediate principal stress and the locus of failure points on the Π-plane is not circular
(Hvorslev, 1960; Bishop, 1972) (see fig. 1.25). The primary advantages of the model are
its computational simplicity and the fact that the coefficients can be matched with Mohr-
Coulomb. Desai and Siriwardane (1984) describe how to obtain Drucker-Prager
parameters from laboratory test data. Further details as well as limitations of the Drucker-
Prager criterion are given by Chen and Baladi (1985), Desai and Siriwardane (1984), and
Christian and Desai (1977).
The Lade-Duncan surface (Lade, 1977, Lade and Nelson, 1984) (fig. 1.24) has been
found suitable for a wider range of pressures than the other criteria for both sands and
normally consolidated clays (Chen and Baladi, 1985). The failure criterion overcomes
some of the shortcomings of the Mohr-Coulomb and Drucker-Prager criteria. It accounts
for decreasing friction angle with increasing confining pressures (as indicated by
curvature of the locus in the hydrostatic plane) so that the weakness of soils at low
confinement pressures and in tension can be better represented. It also incorporates the
dependence of deviatoric stress on the intermediate principal stress as seen by the θ-
dependence of the locus in the Π-plane. Zienkiewicz (1976) presented another
generalization of the Mohr-Coulomb criterion that eliminates the angularity of the yield
surface and incorporates a tension cut-off. Zienkiewicz (1976) showed that the criterion
provided better agreement with some experimental data (Lade, 1972; Lade and Duncan,
1975). Ohmaki (1979) presented a generalizing equation that can produce various failure
cones depending on the values adopted for the equation parameters.
Cap Models. Many soils experience plastic deformations almost from the very start of
loading (even during hydrostatic loading), undergoing a process of continuous yielding
until the soil finally fails (Desai and Siriwardane, 1984). Drucker et al. (1957) proposed
that this type of behavior could be described by modeling the soil as an elastoplastic
work-hardening material. They reproduced the foregoing behavior by defining a series of
geometrically similar “moving” yield surfaces (applied along with a conventional
normality rule) that intersect a fixed failure cone centered on the hydrostatic axis (figs
1.27 and 1.28). For example, Drucker et al. (1957) proposed that the series of yield
surfaces might resemble a series of spherical end-caps sitting on a Drucker-Prager cone.
The yield surfaces that define the hardening behavior are often called hardening caps
10
The relationships between k and c and between α and φ are different when determined using triaxial
compression tests (i.e., axisymmetric conditions) than when determined using shear box tests (i.e., plane strain
conditions) (Desai and Siriwardane, 1984, p. 245).
11
Chen and Baladi (1985) claim that the Drucker-Prager extension of the von Mises (Coulomb) yield criterion
for metals to a three-dimensional soil mechanics problem (Drucker and Prager, 1952) was the first major
advance made towards the development of soil plasticity theory.
(fig. 1.27). To properly define plastic strains under hydrostatic loading, the caps must
intersect the hydrostatic axis. During the successive yielding the material exhibits strain-
or work-hardening behavior. Thus, when the stress point moves beyond the current yield
surface, a new yield surface, or cap, is established. Drucker et al. (1957) also suggested
that soil bulk density be used to determine the degree of hardening, and thus the size of
the cap. The fixed and moving yield surfaces are assumed to intersect in such a way that
at the intersection the tangents to the caps are parallel to the hydrostatic axis. Thus, when
the state of stress reaches a point on the yield cap that is locally parallel to the hydrostatic
axis (i.e., it also reaches the fixed yield surface), the incremental plastic strain vector is
normal to the hydrostatic axis as well as the yield surface, implying that no further
volume changes and no further hardening will take place (fig. 1.27).
Generalized cap models have been proposed by DiMaggio and Sandler (1971) (fig.
1.29d), Sandler et al. (1976), Prevost and Höeg (1975), and Lade (1977, 1979). The
model of Lade (1977, 1979), developed for sands and normally-consolidated clays, uses a
modified version of the Lade and Duncan (1975) cone together with spherical caps. Both
the conical yield surface and the cap are allowed to harden isotropically. An associated
flow rule is used with the cap, while a non-associated flow rule is used with the conical
portion of the yield surface.
The two significant innovations in soil mechanics introduced by Drucker, i.e., the idea
of hardening caps fitted to a cone, and the use of soil current bulk density to define
hardening, led in turn to the development of several different cap models. The ones
J2D dεp
Fixed yield
surface
dεp
J1
Hardening
caps
Figure 1.27—Cap models to represent soil strain-hardening behavior.

probably best known are the “original” and “modified” Cam clay models12 developed by
researchers at the University of Cambridge (Roscoe et al., 1963; Roscoe and Burland,
1968) (fig. 1.29a,b). These form a major part of what is known as critical-state soil
mechanics. Several critical-state models have been proposed that combine, in different
ways, elliptical caps similar to the modified Cam clay model, with the Mohr-Coulomb
criterion (Zienkiewicz et al., 1975; Naylor, 1978b), (figs 1.28, 1.29c). Dafalias (1987)
presented an extension of the modified Cam clay critical-state model from isotropy to
anisotropy. The resulting equations were only slightly more complex than the ones for the
isotropic case.
Several differences between the generalized cap and the Cam clay models are
summarized by Desai and Siriwardane (1984, pp. 295-296). Perhaps the most important13
is that in the critical-state models the moving cap plays the main role in defining yielding,
while the fixed yield state is used essentially to define the “critical-state” (at which there
are no further changes in soil stresses and volume during continuous shearing). In
contrast, the cap models use both the fixed and moving yield surfaces to define the
yielding process. In these models the fixed yield surface is treated as the “ultimate”
yielding surface at which no further volume changes occur.
Several studies have been conducted to compare the performance of constitutive
models for soil. With an appropriate choice of parameters, many of these models can
reproduce certain features of laboratory test strength data equally well; however,
predictions of volumetric strain may be very different (fig. 1.15). In a simulation of
compaction of a wet clay by tires, Chi and Tessier (1995) found that a Drucker-Prager
cap model generally predicted higher volumetric strains than the modified Cam clay
model. Desai et al. (1981) compared six different constitutive relations for soils: Von
Mises, Mohr-Coulomb, Drucker-Prager, a critical-state model (Cam clay), a cap model,
and a viscoplastic formulation. The models were incorporated in a finite element
computer code that also incorporated analysis for geometric nonlinearities. Model
parameters were obtained from laboratory tests. The cap model best reproduced features
of soil behavior from the laboratory tests. Araya and Gao (1995) used a Drucker-Prager
cap model in a three-dimensional finite element analysis of subsoiler cutting with
pressurized air injection. Good agreement was obtained between predicted and
experimental forward soil rupture distances. Side rupture distances were overpredicted
and vertical soil motion underpredicted. Draft also tended to be overpredicted.
Zienkiewicz (1976) concluded that the cap model of DiMaggio and Sandler (1971)
reproduces deformation behavior of soils most realistically.
12
The original Cam clay model was formulated for normally and lightly overconsolidated (“loose”) clays
undergoing triaxial extension and compression only. Various explanations for the discontinuity of the flow rule
on the p-axis have been proposed. Roscoe et al. (1963), for example, assumed the existence of heterogeneous
plastic distortion of localized regions of triaxial specimens. For numerical applications this discontinuity is
problematic and in the CRISP finite element program (Britto and Gunn, 1987) it has effectively been smoothed
out (Wood, 1990). The modified Cam clay was developed as an extension of the original Cam clay model for
more general stress states. The yield locus crosses the p-axis at right angles so that under isotropic compression
and for low stress ratios (q/p) the model predicts insignificant plastic shear strains.
13
Some of the other differences noted by Desai and Sirwardane (1984) apply to the Cam clay critical-state
formulations specifically, but not necessarily to more sophisticated models based upon the critical-state concept,
i.e., the choice of stress and strain invariants for describing generalized loading states, and the portion of the cap
in stress space used to define yielding.
Figure 1.28—Cap model that consists of Mohr-Coulomb failure cone with Cam clay ellipses as caps.
Drawn for M = 0.9, φ = 23º (from Wood, 1990). The size of the cap is defined by the current value of soil
bulk density (or the corresponding preconsolidation stress).
Critical-State Soil Mechanics

Partly for historical reasons, classical soil mechanics separates problems into three
primary areas, the choice of model depending on the type of problem considered: (1)
seepage analysis for the analysis of water through soil; (2) volume change analysis in
which the soil is modeled as an elastic medium for the estimation of stresses induced by
applied loads and the prediction of compaction; and (3) limit equilibrium methods in
which the soil is modeled as a rigid perfectly-plastic material for the study of applications
involving the maximum strength of soils.
Critical state
Critical state line (fixed
q, Critical
line (fixed) q, yield surface)
dεSp point
dεS
p
dεp
dε
p
p, dεVp
p0 p0 p0 p, dεVp
Moving yield
surface (cap)
Moving cap
(a) (b)
(c) (d)
Figure 1.29—Cap models: (a) Original Cam clay critical-state model with log-spiral cap; (b) Modified
Cam clay yield surfaces; (c) A Mohr-Coulomb critical-state model with tension cut-off (from Naylor,
1978b); (d) Generalized cap model of DiMaggio and Sandler (from Desai and Siriwardane, 1984).
The critical-state theory of soil mechanics was originally developed by researchers at

Cambridge University, England, for saturated soils (Roscoe et al., 1958; Schofield and
Wroth, 1968). The important contribution of critical-state soil mechanics is that both
volume (i.e., structural) changes as well as stress changes are considered along with the
loading path when describing soil behavior. The theory unified concepts of consolidation,
compression, yielding, and failure of soils into a single framework (Atkinson and
Bransby, 1978). Accounts of the theory for saturated soils can be found in Schofield and
Wroth (1968), Atkinson and Bransby (1978), Wood (1990), and Atkinson (1993).
In this section some of the important features of critical-state theory will be presented.
One very simple elastoplastic cap model based on the critical-state concept, called Cam
clay,14 will be described. The Cam clay model was developed based on soil response
observed during tests using reconstituted soils and can be considered to represent
“idealized” soil behavior. Some modifications to this simple model intended to predict
more realistic soil behavior will also be discussed.
14
The model developed here is more correctly called “modified” Cam clay to distinguish it from the earlier
model called Cam clay which is generally out of use (often referred to as the “original” Cam clay model).
State Variables
One of the basic assumptions of critical-state theory is that the soil is an homogeneous,
isotropic medium. The assumption of isotropy reduces the problem description to three
independent stress variables and three independent strain variables. A further
simplification made is that the yield surface is symmetric about the hydrostatic (σ1 = σ2 =
σ3) axis. This results in the need for only two independent stress invariants which in turn
lead to two corresponding strains. It is customary to define two stress components, the
hydrostatic stress p and the deviatoric stress q (table 1.1), which for axisymmetric triaxial
conditions (σ2 = σ3 and ε2 = ε3) are given by:
σ1 + 2σ3
p= (1.66)
3
q = σ1 - σ3 (1.67)
These parameters are increasingly used in soil mechanics because of their association
with changes in volume and in shape, respectively. The corresponding (i.e., “work
conjugate”15) strain invariants are defined as follows:
Volumetric strain εv = ε1 + 2ε3 (1.68)
Deviatoric (shear) strain εs = 2/3 (ε1 - ε3) (1.69)

The strain invariants εv and εs describe the strains at the start of a test or loading
sequence. The symbols dεv and dεs will be used for strain increments.
The void ratio e or the specific volume v is commonly used as the volume change
variable in critical-state formulations. The use of specific volume allows volume changes
of different sized soil samples to be compared. It is also mathematically more convenient
to use than void ratio because of its somewhat simpler relationship to volumetric strain
(eq. 1.16).
Soil States, State Space, and State Boundaries
The soil volume strain and stress condition can thus be depicted in a three-dimensional
space defined by p, q, and v, termed state space. Critical-state theory postulates that the
behavior of a soil may be entirely defined if its state in state space is known. In other
words, the response of a soil will depend on its mean stress, the imposed shear stress, and
the physical arrangement of its soil particles (as measured by the specific volume or void
ratio).16 In practice not all combinations of p, q, and v are possible and certain boundaries
limiting the values of the state variables have to be identified. Yield loci may be traced
15
The correct form of strain invariants are chosen by noting that as an element deforms under load, the work
done under external loads is invariant, and the corresponding product of stress and incremental strain invariants
must always equal the work done by external loads (Schofield and Wroth, 1968).
16
This applies for a saturated soil, in which the water content of the sample is uniquely related to the void ratio.
For an unsaturated soil, the relationship between void ratio and water content is dependent on the degree of
saturation (fig. 1.7 and eq. 1.18) so that an additional “volumetric” variable (e.g., water content) is required to
describe the “state” of a soil. The specific volume (or void ratio) is considered to be a more fundamental
volumetric variable for saturated soils than water content (although the latter is easier to measure) because it
represents the relative disposition of particles and voids for any soil whether saturated or not (Wood, 1990).
Various approaches to describing the state of an unsaturated soil within the framework of critical-state soil
mechanics will be discussed in the last part of this chapter.
out in v-p space as well as in q-p space to define a unique three-dimensional state
boundary surface in p-q-v space for any given soil (fig. 1.30). The components of the
state boundary surface identified in figure 1.30 play important roles in defining the
shearing and compression behavior of the soil.
The loci of p, q, and v values traced out by a soil element undergoing deformation are
referred to as state paths. State paths must at all times lie within or traverse the state
boundaries, but can never cross them. For states of stress below the state boundary only
elastic deformations are assumed to occur. On the other hand, if the state of stress lies on
the state boundary, both elastic and plastic deformations may occur (fig. 1.31).
The three-dimensional state boundaries are often presented by their projections on the
v-p and q-p planes. There are no shear stresses or deformations on the v-p plane while no
volume changes occur on the q-p plane. Note that the projections of the state boundary
surface on the q-p plane for constant values of v form the hardening caps (fig. 1.31).
State boundary critical- state line

surface
p
critical state
wall
normal
consolidation
line
sub-critical zone super-critical zone

(ductile failure) v (brittle failure)
Figure 1.30—Three-dimensional view of normal consolidation line, critical-state line, and series of yield
loci making up the state boundary in p-q-v state space. For elastic states of stress, state paths lie within
the state boundary; for plastic states the state paths traverse the state boundary. The end points of state
paths lie on the critical-state line. The critical-state wall divides state space into zones associated with
ductile and brittle failure.
q Critical
Fixed yield surface or
Critical-State Line
point
D M
C Moving yield
B surface (cap)
A
p
p0A p0
Figure 1.31—Yield locus in q-p space. If the material is over-consolidated at A with preconsolidation
stress p0A and loaded along stress path AB it will respond elastically until it reaches the current yield
locus at B. Upon further loading, the material hardens through successive states of yielding during which
the yield surface expands, until the material reaches an ultimate state on the critical-state line at D. The
size of the yield locus p0 is used to define the hardening behavior.
The Critical-State Line

An important distinguishing feature of all “critical-state” models is the concept of
critical-state proposed independently by Roscoe et al. (1958) and Parry (1958). The early
work of Taylor (1948) with sands and Henkel (1959) with clays showed that soil
specimens reach a “steady state” condition when continuously sheared. When a loose soil
sample is sheared, it reaches a critical void ratio, after which the void ratio remains
constant during subsequent deformations (i.e., no further volume change takes place).
Deformation is in such a manner as to decrease the volume of the sample. When a dense
soil sample is sheared, it reaches a peak stress as shown in figure 1.10, and then reaches a
lower residual stress. Initially the soil compacts, and then it dilates until the specific
volume (volumetric strain) reaches a constant value that corresponds to its critical value
(Desai and Siriwardane, 1984).
The critical-state theory states that soil undergoing shear deformation passes through
successive states of yielding as it traverses the state boundary, during which the material
hardens or softens. The soil ultimately reaches a critical (or ultimate) yield state at which
large shear deformations could occur indefinitely with no change in stress or plastic
volumetric strain. In terms of the state variables, a specimen is at critical-state when:
∂p ∂q ∂v
= = =0 (1.70)
∂ε s ∂ε s ∂ε s
It is not sufficient for only one or two of these ratios to be zero; they must all be
simultaneously satisfied for a specimen to have reached critical-state.17 Clearly,
specimens with different initial consolidation pressures and specific volumes will have
different final values of p, q and v when they reach their respective critical-states.
However, the theory postulates that the locus of all such critical-state points in p-q-v
space, termed the critical-state line (CSL), is unique for a given soil, so that under
continuous shearing the soil will reach a state located on the CSL and “fail,” irrespective
of the stress path followed on its way to the CSL (Atkinson and Bransby, 1978) (fig.
1.31).
The critical-state line when projected on the q-p plane may be expressed by:
q=Μp (1.71)
and, when projected on the v-p plane, by:
v = Γ - λ ln p (1.72)
where λ, Γ, and Μ18 are soil constants (fig. 1.33). The parameter Γ is defined as the value
of v corresponding to unit p (i.e., ln p = 0) on the critical-state line. The value of Γ will
depend on the units used for hydrostatic stress.
The CSL is thus a curved line in p-q-v space (fig. 1.30) given by the intersection of the
two planes described by equations 1.71 and 1.72. The CSL rises (i.e., q increases) as p
increases and v decreases.
Comparison of equation (1.71) with the Mohr-Coulomb failure criterion suggests that
soils are failing in a purely frictional manner at critical-state (there is no “cohesion”
intercept). As Wood (1990) expresses it, the deformations have become so large that the
soil has been thoroughly remolded and churned up, and any bonding between particles
that might have led to some cohesive strength has broken down.19 The slope of the CSL in
the q-p plane, M, is related to the internal angle of friction, φ, by:
6 sin φ
Μ= (1.73)
3 − sin φ
for triaxial compression (Schofield and Wroth, 1968).

Domains within State Space
The mechanical behavior of a soil is a combination of the geometry of the state
boundary surface, the initial position of the soil in state space, and the stress path. The
state space for a given soil is conceptually divided into two zones separated by an
imaginary vertical plane containing the critical-state line, called the critical-state wall
(fig. 1.30). The state space on the side of the critical-state wall containing the origin of
the p-q-v axes, is called the super-critical domain (or dry of critical). The state space on
the side of the critical-state wall far from the origin is designated the sub-critical domain
17
Pore-water pressure changes also decrease to negligible values in saturated soils. The situation is not as clear
for unsaturated soils in which both pore-air and pore-water pressures may vary.
18
Capital µ (mu).
19
As will be seen, this does not necessarily occur with unsaturated, cemented soils.
(or wet of critical). As a general rule, state paths traversing the state boundary wet of
critical are associated with compaction (decrease in v) and ductile failure, whereas, state
paths traversing the state boundary dry of critical are associated with dilation (increase in
v) and brittle failure. In either case, the state of the element will eventually lie on the
critical-state line. These types of behavior are illustrated in figure 1.32. The stress-strain
responses depicted in figure 1.32 compare qualitatively with the type of behavior
associated with loose (normally consolidated and lightly over-consolidated) and dense
(heavily over-consolidated) soils (fig. 1.10).
Volume-Pressure Relations
In critical-state soil mechanics the normal (isotropic) consolidation, unloading
(swelling) and reloading lines are idealized as straight lines in v-ln(p) space with slopes
-λ and -κ (fig. 1.33b).
The normal consolidation line (NCL, also called the λ-line) is defined in state space
by:
q=0 (1.74)
v = Ν - λ ln p (1.75)
where Ν 20 is a soil constant defined as the value of v when p = 1. Note that the NCL is
parallel to the CSL on the v-ln p plot. The normal consolidation line is the part of the
state boundary surface in the sub-critical domain lying in the q = 0 plane, i.e., on the floor
of p-q-v space (fig. 1.30).
The equation of the unloading-reloading line (url, often referred to as a κ-line) is
given by:
v = vκ - κ ln p (1.76)
where vκ is the specific volume of an over-consolidated soil at unit p.

The Cam clay equations for the CSL, NCL and url are summarized in table 1.2. As
shown, a somewhat different notation is commonly used when state boundaries are
defined in p-q-e space as opposed to p-q-v space.
Table 1.2. Cam clay equations in the p-q-e and p-q-v domains.
Name p-q-e p-q-v (v = 1 + e)

CSL (strength envelope) q = Mp q = Mp
CSL e = ecs - λ ln p v = Γ - λ ln p
NCL q = 0, e = eN - λ ln p q = 0, v = Ν - λ ln p
unloading-reloading line URL e = eκ - κ ln p v = vκ - κ ln p
20
Capital ν (nu).
q
p
1 super-critical
2 (brittle rupture)
sub-critical
(ductile yielding)
q, q,
dε p dε sp
s CSL
1
1
2
2
p, dε vp εs
p
c
v v
sub-
critical
2
NCL
2 1
1
CSL
super-critical
p, dε vp εs
Figure 1.32—Soil behavior predicted by the Cam clay critical-state model: (1) Compaction in the sub-
critical domain, stable strain-hardening characteristic; (2) Dilation in the super-critical domain, unstable
strain-softening characteristic ( – – – – – elastic states,  plastic flow).
q, Critical-State Line
εsp
dεsp
dεvp
Cam clay
M Yield Locus
p0 p0 p, εvp
2
(a)
N
NCL
Γ CSL
λ
λ
κ
url
p=1 ln p0 ln p
(b)
Figure 1.33—Modified Cam clay state boundaries: (a) Elliptical yield locus for Cam clay model and
critical-state line in p-q plane; (b) Critical-state line (CSL), normal (isotropic) consolidation (NCL), and
unloading-reloading (url) lines in v-ln(p) plane.
Derivation of Cam Clay Constitutive Law

Modified Cam clay is a classical plasticity model. Its main features are a particular
form of nonlinear elasticity, which models the increasing bulk stiffness as the material
undergoes compression, and the assumption of normality (associated flow), together with
a particular form of hardening law that allows the yield surface to grow or shrink.
Elastic Response. For stress states within the current yield surface, deformations are
assumed to be completely recoverable. For an isotropic elastic medium the stress-strain
relationship can be written:
dε ev  1 / K 0  dp 
 e =    (1.77)
dε s   0 1 / 3G  dq 
Elastic strains correspond to movement on an unloading-reloading (url) line.

Differentiating equation 1.76 gives:
dp
dv e = − κ (1.78)
p
where the superscript “e” is used to indicate that this describes an elastic volume change.
From equation 1.16, which relates an increment of specific volume to volumetric strain,
equation 1.79 can be rewritten as:
dp
dε ev = κ (1.79)
vp
By comparison with equation 1.77,

vp (1.80)
K=
κ
so that a constant κ in the v-ln(p) plane implies a bulk modulus K that increases with
hydrostatic stress p.
Changes in deviatoric stress q within the current yield locus produce elastic shear
strains dε s , which can be calculated from equation 1.77 with an appropriate value of
e
shear modulus G. Assumptions made in the choice of G are fairly arbitrary. If a constant
value of Poisson’s ratio is assumed, the shear modulus G is deduced from K and
Poisson’s ratio ν by:
3(1 − 2ν)K
G=
2(1 + ν)
vp 3(1 − 2ν)
= ⋅
κ 2(1 + ν) (1.81)
An alternative choice might be to assume a constant G. In this case, the variation of K

with p implies a variation of ν. In the original critical-state formulations, it is assumed
e
that there is no recoverable energy associated with shear distortion so that dε s = 0.
Equation of the Yield Locus. Yielding of the soil occurs whenever the stresses satisfy
the equation for the elliptical Cam clay yield surface:
ƒ(p, q, p0) = M2 {p(p - p0)} + q2 = 0 (1.82a)
or
M 2 + η2 p 0
= (1.82b)
M2 p
where η = q/p is the stress ratio.21 This simple shape introduces just one more soil
parameter, p0, the value of p where the current yield surface intersects the p-axis (fig.
1.33a). It gives a measure of the “size” of the current yield surface and is equal to the
maximum hydrostatic stress to which the soil has been subjected (i.e., the
preconsolidation pressure). The top of the Cam clay locus where η is equal to Μ has:
p
p cs = 0 (1.83)
2
When the state of stress reaches the yield criterion the material undergoes plastic
deformation. Once the soil is in a plastic state it will continue to satisfy the yield criterion
(unless unloading begins) and any further loading will cause the size of the yield surface,
as given by p0, to change. The tip stress, p0, is considered as the hardening parameter
(compare eqs. 1.82a and 1.55).
Hardening Law. In the Cam clay model the yield locus is assumed to expand at
constant shape. The hardening/softening is assumed to depend only on the volumetric
plastic strain component p0 = p0( dε pv ), and is such that expansion of the yield surface is
associated with normal compression of the soil, while increases in the volume of the soil
skeleton (dilation) cause the yield surface to shrink.
Consider figure 1.34. The unloading-reloading line url 1 is a line joining a set of
elastically attainable combinations of p and v for a soil having a current yield locus with
tip stress p01. After some yielding has occurred there is a new unloading-reloading line,
url 2 associated with a new value of tip stress p02 = p01+ dp0. During plastic flow the total
volume change is assumed to be divisible into elastic and plastic components, so that
p e
dv = dv - dv (1.84)
p
where dv is the plastic component of the incremental specific volume. The irrecoverable
change in volume is simply the volume change remaining when the value of hydrostatic
stress is increased from p01 to p02 and then reduced to p01 again (fig. 1.34). From equations
1.75 and 1.76 this is:
dv p = −[λ ⋅ d(ln p 0 ) − κ ⋅ d(ln p 0 )]
dp 0
= −( λ − κ ) (1.85)
p0
The plastic volumetric strain is given by:
dv p (λ − κ) dp 0
dε pv = − = (1.86)
v v p0
where v is the initial specific volume before the increment.
21
The Cam clay yield surface equations were derived by making certain assumptions about the nature of energy
dissipated in a soil specimen supporting stress (p, q) as it strains. In modified Cam clay the energy dissipated is
assumed as dW = p (dε pv ) 2 + M 2 (dε sp ) 2 (Schofield and Wroth, 1968). This leads to equation 1.91, which
combined with an associative flow rule, describes a series of elliptical yield surfaces in q-p space.
Figure 1.34—Change in unloading-reloading line from url 1 to url 2 as yield locus expands from Y1 to Y2.
During plastic flow the total volume change is the sum of elastic and plastic components.
The hardening law can be written as:
∂p 0 vp 0
= (1.87)
∂ε pv λ − κ
Plastic Potential and Flow Rule Equation 1.87 describes the change in size of the
yield surface in terms of the magnitude of the plastic volumetric strain. It is still
necessary to calculate the magnitude of any plastic deviatoric strains. Writing the flow
rule (eq. 1.58) in terms of volumetric and deviatoric plastic strain increments yields:
∂g
dε pv = L (1.88a)
∂p
∂g
dε sp = L . (1.88b)
∂q
In the Cam clay formulation it is assumed that the soil obeys the normality condition,
thus:
g = ƒ = q2 - M2 [p(p0 - p)] = 0 (1.89)
The vector of plastic strain increments is then in the direction of the outward normal to
the yield locus so that (from eqs. 1.88 and 1.89) during plastic flow:
dε sp ∂g ∂q 2q 2q p
= = =
dε pv ∂g ∂p M 2 (2p − p 0 ) M 2 (2 − p 0 p )
(1.90)
M 2 − η2
= .
2η
In the above description, plastic strain increments are governed by the particular stress
state at which yielding occurs rather than on the stress path followed to reach the yield
surface, a distinguishing feature of plastic response.
Figure 1.35 shows the Cam clay yield locus with superimposed incremental strain
vectors normal to the yield locus. It should become clear how this single simple model is
able to produce the different types of stress-strain behavior associated with loose and
dense specimens (fig. 1.32). When yielding takes place with η < Μ (sub-critical) then
p
there are compressive volumetric strains (+ dε v ) and the yield surface expands
(hardening) as the stress state approaches critical. When yielding takes place with η > Μ
(super-critical) then there are dilative volumetric strains (- dε v ) and the yield surface
p
contracts (softening) as the sample approaches critical (The sharp peak proposed by the
model is smoothed off in real soils). As η approaches M the ratio dε v dε s , and
p p
q, dε sp
CSL η = Μ
dε vp = 0
-dε vp dilation +dε vp
compaction
η>Μ dε sp
η<Μ dε vp
p, dε vp
Figure 1.35—Plastic strain increment vectors normal to the Cam clay yield locus under an associative
flow rule.
therefore dv/dεs, approaches zero in a smooth manner. At critical-state, η = Μ, the stress

state is at the top of the current yield locus, so indefinite plastic shearing can occur
without further expansion or contraction of the yield locus ( dε pv = 0), an ultimate limiting
condition. Note that it is not sufficient for the stress ratio to be equal to Μ for a critical-
state to develop; the specific volume has to be at critical too (i.e., both equations 1.71 and
1.72 for the CSL must be satisfied). Thus, the stress ratio η = Μ produces a critical-state
provided plastic deformations are occurring (Wood, 1990).
Plastic Stress-Strain Relationship. The yield surface, flow rule, and hardening law
relationships can be assembled into an incremental constitutive relationship for use in
numerical solutions of boundary value problems. The differential form of the yield loci
ƒ(σij, h) = ƒ{p, q, p0( dε v )} = 0 is:
p
∂ƒ ∂ƒ ∂ƒ
dƒ = dp + dq + p dε pv = 0
∂p ∂q ∂ε v
∂ƒ ∂ƒ ∂ ƒ ∂p 0 p (1.91)
= dp + dq + dε v = 0
∂p ∂q ∂p 0 ∂ε pv
where the chain rule of differentiation has been used. Substituting equation 1.88 for dε v
p
into equation 1.91, gives:
∂ƒ ∂ƒ ∂ƒ  ∂p 0 ∂g 
dp + dq + L  p =0 (1.92)
∂p ∂q ∂p 0  ∂ε v ∂p 
which yields an expression for the plastic multiplier L:
 ∂ƒ ∂ƒ 
−  dp + dq 
 ∂p ∂q 
L= (1.93)
∂ ƒ  ∂p 0 ∂g ∂p 0 ∂g 
 + 
∂p 0  ∂ε pv ∂p ∂εsp ∂q 
Substituting the expression for L back into the flow rule (eq. 1.88) yields the general Cam
clay plastic stress-strain relationship:
 ∂f ∂g ∂f ∂g 
dε pv  −1  ∂p ∂p ∂q ∂p  dp 
 p =    (1.94)
dε s  ∂f  ∂p 0 ∂g   ∂f ∂g ∂f ∂g  dq 
 
∂p 0  ∂ε pv ∂p   ∂p ∂q ∂q ∂q 
Cam Clay Elastoplastic Constitutive Relation. The complete description of the Cam
clay elastoplastic model can now be written. The total incremental strains are given by
the sum of elastic and plastic incremental strains:
dε = dε e + dε p (1.95)
The elastic stress-strain relationship is given by:
dε ev   κ / vp 0 
 e=  (1.96)
dε s   0 1 / 3G 
Substituting the appropriate terms into equation 1.94, after some manipulation the plastic
stress-strain response can be expressed as:
dε pv  λ−κ Μ 2 − η 2 2η 
 p = 2  2 
(1.97)
dε s  vp(Μ + η )  2η 4η /(Μ − η )
2 22
Equation 1.97 applies only if plastic deformation is occurring (i.e., ƒ = 0), otherwise
dεp = 0. Note that the symmetry of the plastic compliance matrix arises because of the
assumption of normality. A non-associated flow rule would produce a non-symmetric
compliance or stiffness matrix.
The above model expressed soil behavior using quantities relevant to conventional
triaxial tests and consolidation tests, i.e., p, q and v (or e). The Cam clay model can just
as well be formulated for general three-dimensional states of stress using the appropriate
stress invariants (table 1.1). A Cam clay elastoplastic constitutive relation that can be
applied for solution of three-dimensional boundary value problems is derived by Desai
and Siriwardane (1984).
Lade and Nelson (1984) present a general mathematical procedure for formulating
incremental elastoplastic constitutive relations for materials modeled using multiple,
intersecting yield surfaces. Each yield surface has its own associated or non-associated
flow rule. They describe a procedure for determining the material stiffness matrix for use
in finite element or finite difference analyses. Scott (1985) also describes a general
procedure for implementing incremental soil plasticity models from the yield criterion,
flow rule and hardening law.
Cam Clay Soil Parameters

One of the merits of the Cam clay model is its ease of use. The model requires the
specification of five soil parameters (see table 1.2) that can be obtained using standard
laboratory triaxial and consolidation tests:
λ, M, N, κ, ν (or G)
The soil parameter Γ (eq. 1.72) is a function of these parameters, given by:
p0
Γ = Ν - (λ - κ) ln ( ) (1.98)
pC
where pC is the value of p at which the CSL crosses the yield surface, and p0/pc is a
constant that depends on the assumed yield locus. For modified Cam clay, p0/pc = 2 (for
the original Cam clay model, ln (p0/pc) = 1). Each yield locus has associated with it its
own unloading-reloading line. The parameter vκ (eq. 1.76) is a function of stress history,
and is related to p0 by:
vκ = Ν - (λ - κ) ln (p0) (1.99)
The preconsolidation stress, p0, is obtained from isotropic consolidation tests or from
oedometer tests. The determination of parameters and typical values for agricultural soils
are summarized in Part 2 of this chapter.
The Overall State Boundary

According to the Cam clay model, all state paths will eventually reach the CSL. The
stress conditions that apply at the CSL represent the ultimate strength of the soil (i.e., its
critical-state strength), and this is the lowest strength the soil will reach provided that the
strains within it are reasonably uniform and not excessively large (Smith, 1990).
Extensive experimental data from shear box tests (Hvorslev, 1937) and triaxial
compression and extension tests (Parry, 1960) suggest that although the Cam clay model
correctly predicts a peak strength for over-consolidated soils, the magnitude of the peak
strength is over-predicted. Instead, the locus of failure points from shear tests on over-
consolidated soils can be idealized by a straight line called the Hvorslev surface with the
equation:
Γ−v
q = (M − h c ) exp  + hcp (1.100)
 λ 
where the slope of the Hvorslev surface, hc, is a soil constant (Atkinson and Bransby,
1978). An important feature of the Hvorslev surface is that it illustrates that the shear
strength of a soil at failure depends on the specific volume (i.e., the water content) as well
as on the hydrostatic stress. The left-hand strength component in equation 1.100 depends
on the current specific volume and the value of certain soil constants and is such that it
increases as specific volume v decreases (it is indirectly related to soil “cohesion”). The
right-hand term is proportional to the hydrostatic stress, and so may be thought of as
frictional in nature. In an ideal situation the stress paths of over-consolidated soils will
reach the Hvorslev surface and then traverse this surface to reach critical-state. In
practice, failure of undrained soils can occur before the critical-state line is reached if the
irregularities in the soil are significant (Smith, 1990). This situation is fairly common for
structured agricultural field soils.
Shear box tests conducted by Hvorslev (1937) and the peak strength data obtained by
Parry (1960) in a large number of triaxial tests as well as other data, show that samples
fail and reach their greatest strength at what could be called a critical-state (Atkinson and
Bransby, 1978; Wood, 1990). Thus the Hvorslev surface extends only so far as the CSL.
There will also be a limit to the extent of the Hvorslev failure line at low p values. If it is
supposed that the soil is unable to withstand tensile stresses (i.e., the minor principal
stress σ3 cannot be less than zero) then tensile failure must occur for triaxial compression
conditions if ever the axial stress σa is less than the radial stress σr. The lowest possible
value of q is zero which means the tensile failure boundary must pass through the origin.
The highest possible value for q will occur when the radial stress σr = 0 and q, therefore,
equals σa. At this stage p = σa/3, which means that η = q/p = 3. This traces another state
boundary rising from the origin with a slope of 3:1 called the no-tension (or tension cut-
off) surface.22 Stress paths that reach the no-tension surface will fracture without
necessarily reaching critical-state.
Normally compressed and lightly over-consolidated samples do proceed in a stable
fashion to a critical-state without a premature peak, and the Cam clay yield surface is
retained for sub-critical soil states. This portion of the state boundary is known as the
Roscoe surface. The Hvorslev surface is adopted as the state boundary surface for heavily
over-consolidated soils. The Hvorslev and Roscoe state boundary surfaces intersect and
the line of intersection is the critical-state line.
The complete state boundary surface is then a composite surface made up of surfaces
corresponding to ductile yielding (Roscoe surface), Hvorslev rupture, and tensile fracture.
The curved Roscoe surface joins the normal consolidation line to the critical-state line.
The Hvorslev surface links up with the Roscoe surface at the critical-state line. Its locus
is limited to its left-hand side by the no-tension surface. Figure 1.36 shows the complete
state boundary plotted in p-q space. Any constant v section will have the shape shown in
figure 1.36, but the size of the section will be such that the point A always lies on the
NCL. In three-dimensional p-q-v space, the shape of the complete boundary surface is
that shown in figure 1.37. Allowing for the change of view, the shape of any constant v
section of the surface can be seen to be the same as that shown in figure 1.36.
q
Impossible
states Critica l-s tate line
Hvors lev
s urface Rosc oe
surface
Pos sible
s tates Normal cons olidation
No-tens ion line
3 line
p
1 A
Figure 1.36—The complete state boundary in p-q space.
22
If the soil could sustain tensile stresses, the line corresponding to tensile failure would lie to the left of the no-
tension line and might be curved. This possibility is relevant to many cohesive powders, and to unsaturated and
cemented soils.
Figure 1.37—The complete state boundary plotted in three-dimensional state space p-q-v (from Stafford,
1981).
Advances
The Cam clay model that was described in some detail in this chapter is an illustrative,
largely qualitative model that has had some success in predicting soil behavior in
practical cases. Many of the assumptions made in the derivation of Cam clay were
convenient, yet arbitrary. In this section, some developments toward more realistic
models suitable for machine-soil interactions in agricultural soils are introduced. An in-
depth treatment of advanced elastoplastic and viscous models is beyond the scope of this
chapter. The cited references may be referred to for additional details.
Dynamic Soil Models

Time and Rate Effects. There are several categories of time effects that influence soil
response. It is useful to distinguish between effects due to high rate-short duration
loading, those due to (viscous) flow and creep, and those due to aging and other time-
hardening phenomena.
Considerable strain rate and inertial effects may occur in agricultural and forestry
applications. In cohesive soils, shear strength may increase substantially with increasing
strain rates (Casagrande and Shannon, 1948; Payne, 1956; Turnage, 1975; Stafford and
Tanner, 1983; McKyes, 1985). Inertial effects are particularly important in frictional soils,
whereas strain rate tends not to affect shearing resistance (Atakol and Larew, 1970;
Koolen and Kuipers, 1983; McKyes, 1985). Even when large masses are not involved,
inertial forces may still be significant if accelerations of soil particles are high enough.
Examples include the movement of soil particles near the edges of wheels of vehicles
operating at high speeds (Freitag, 1985) and soil cutting at high speeds (McKyes, 1985).
Soil deformations associated with cultivation practices are also extremely large. Thus, the
special problems associated with large deformation analysis are highly relevant and an
important area for further research. Finite element formulations that incorporate large
strains and deformations along with material nonlinearities are described by Zienkiewicz
and Taylor (1988), Desai and Phan (1980), and Bathe (1982).
Every material demonstrates viscous properties, i.e., a time dependence of the state of
stress and strain (or deformation). These become more pronounced during plastic
deformations. Evidence of the viscoelastic behavior of many soils is shown by the
relaxation (lessening) of stresses with time under a constant load, and the recovery of
deformation with time after the load is removed. Elasticity, viscosity, and plasticity can
be all considered as cardinal rheological properties of a continuum (Vyalov, 1986). Due
to the short duration of loading involved true viscous effects may often be ignored in
analyses of machine-soil interactions. Short loading durations are incorporated in
agricultural soil models through the use of undrained pore-water (i.e., constant water
content) testing procedures and analyses. Viscous effects are often considered in the
analysis of paddy soils using rheological models (Chancellor, 1994: 207-217). The
theories of infinitesimal viscoelasticity and viscoplasticity combine small strain elasticity
and plasticity, respectively, with rheology.
There is some evidence that at high loading rates, viscous effects may be important
even for relatively dry agricultural soils. Using triaxial compression tests with loading
rates from 0.35 to 6.2 m/s, Niyamapa and Salokhe (1992) observed that the nature of
failure of specimens changed from brittle shear to plastic flow at 4.4 m/s for a sandy loam
and at 5.7 m/s for a silty loam. Under high speeds, small cracks and fluidization were also
observed in the samples. Olson and Weber (1965) and Stafford (1979, 1981) reported that
the mode of failure caused by narrow tines changed from brittle fracture with well-
defined shear planes to continuous flow as the speed of the tine was increased (fig. 1.38).
The change in failure mode was accompanied by a distinct change in the characteristics
of the draft variation. Elijah and Weber (1971) observed four types of failure caused by
cutting blades: brittle shear, bending, tensile, and plastic flow. The mode of failure
depended on the cutting speed and the blade rake angle. Rosa (1997) concluded that
viscous effects were significant during high-speed tillage (2.8 to 8.4 m/s) in a remolded
sandy clay loam soil.
Time effects due to aging, which are not readily linked to rates of deformation, can
produce significant changes in soil properties over time due to physico-chemical
reactions and diffusional processes. Some “sensitive” soils may exhibit thixotropic
hardening over time to produce a highly compacted soil structure even after moderate
disturbance (such as moderate compaction, or even slight soil loosening), as soil particles
and particle groups tend to flocculate (Mitchell, 1993). The cementation effects of
organic matter and other soil minerals may be destroyed on disturbance. Pore-water
tension changes with time after disturbance or due to climatic factors, can cause
significant changes in shear and tensile stresses, and extreme swelling/expansion of some
soils.
Dynamic Soil Strength. A logarithmic relationship between shearing resistance and
rate of strain has been established for saturated clays (Yong and Japp, 1968; Mitchell,
1993). Various logarithmic expressions have also been proposed for strength of
agricultural soils (Qun and Shen, 1988; Zeng and Yao, 1992). Linear relationships
between shearing resistance and shear rate have been proposed by some researchers
(Hassan and Chancellor, 1970; Awadhwal and Singh, 1992). Soil structure was found by
Awadhwal and Singh (1992) to significantly influence the effect of shear rate on shear
strength. The rate of shear strength increase with shear rate was four times higher for an
unpuddled wet loam soil than for puddled soil in which the structure had been destroyed
by rototilling (fig. 2.182, Vol. 1). The authors also established a log-log decreasing
relationship between the secant shear modulus and shear rate for shearing speeds between
0.0045 and 1.33 m/s.
Figure 1.38—Failure of clay (w = 18.2%) by a rigid narrow tine at different speeds (from Stafford, 1981).
As speed increased, the nature of failure changed from brittle shear to compression and flow between 2
and 3 m/s.
Strain Rate Invariants. In viscous formulations, appropriate and consistent

descriptions of the strain rates are required. The expressions for rates of strain are
analogous to those for the strain invariants. Therefore, the first invariant of the rate of
strain tensor will be equal to:
&I = &I = ε& = 3ε& = ε& + ε& + ε& = ε& + ε& + ε& (1.101)
1ε 1 v oct x y z 1 2 3
where ε& i are rates of linear strain. The second deviatoric strain invariants are related by:
γ& oct
2
= 23 &I 2 D = 1
9
[(ε& 11 ] [
− ε& 22 ) 2 + (ε& 22 − ε& 33 ) 2 + (ε& 33 − ε& 11 ) 2 + 23 (ε& 12 ) 2 + (ε& 23 ) 2 + (ε& 13 ) 2 ] (1.102)
where ε& ij = 12 γ& ij . Expressions analogous to those for the other forms of invariants (table
1.1) also apply for strain rates. The notations ε& and γ& are adopted for the sake of symbol
identification. Strictly speaking, the shear rate is the derivative of the shear strain
increments rather than the total derivative of shear strain (Vyalov, 1986).
Limit Equilibrium Models. Several researchers have added inertial and shear rate
terms to established models based on passive earth pressure theory for tillage tool draft.
Such models have had some success in predicting tool draft as a function of speed.
McKyes (1985) incorporated inertial terms in a wedge model of soil cutting. Predictions
were compared with draft data reported by Luth and Wismer (1971) for tools cutting
through a sandy soil at speeds up to 3 m/s. A prediction error of about 25% was reported
for a speed of 2.5 m/s. Swick and Perumpral (1988) incorporated dynamic effects due to
soil shearing rate and inertial forces in the limit equilibrium model of Perumpral et al.
(1983). The effects of shear rate on both soil strength and on soil-metal friction were
incorporated in the model. Reasonable predictions were obtained for prediction of draft
of narrow tools operating at speeds up to 1.2 m/s. Inertial forces were found to be an
important component of draft in the soils tested.
Zeng and Yao (1992) developed a model for prediction of draft during soil cutting by
tillage tools. Soil shear strength was related to shear rate and applied normal stress using
a relationship of the form:
ln τ = C1 + C 2 ln γ& + C 3 ln(1 + C 4 σ n ) (1.103)
where C1, C2, C3, and C4 are experimentally determined soil parameters, γ& is the shear
rate, and σn is the normal stress. Equation 1.103 was used in a manner similar to Mohr-
Coulomb in limit equilibrium analysis, i.e., soil failed without achieving plastic flow. The
model also incorporated inertial effects and a relationship between soil-blade sliding
speed and soil-metal friction.
Viscoelastic Models. Linear viscoelastic formulations conceptualize the material as
being composed, at the microscopic level, of linear elastic elements (“springs”)
connected together with linear viscous elements (“dashpots”) to form a network. The
infinitesimal stress tensor at any given time is related to the infinitesimal strain tensor and
the history of infinitesimal strains through the elastic stress-strain law and a stress
relaxation function, which is a function of time. Reviews of viscoelasticity can be found
in Coleman and Noll (1961) and Lee (1962). The Burgers model, commonly used as a
viscoelastic model for paddy soils (Chancellor, 1994; Pan 1986), is an example of a
rheological model that is composed of two linear elastic elements and two linear viscous
elements. Lu and Pan (1986) used the Burgers model in a finite element analysis of paddy
soils. Oida and Tanaka (1981) and Oida (1984) used a three-element rheological model to
derive a constitutive relationship for paddy soils. An application of viscoelasticity for a
two-dimensional finite element analysis of wheel sinkage was presented by Oida (1984).
Shen and Kushwaha (1995) proposed a constitutive relationship that added a
logarithmic strain rate term to the Duncan and Chang (1970) hyperbolic stress-strain law
to represent the increase in soil strength with loading rate. In this approach, strain rate
effects were imbedded in the tangential modulus of elasticity, Et, and therefore in the
stiffness matrix of the finite element formulation. Rosa (1997) investigated a similar
model, with variable Poisson’s ratio, in a finite element analysis of dynamic soil cutting
with narrow tools. However, similar problems were encountered to those that occur with
static hyperbolic models with variable ν and E, i.e., the onset of numerical instabilities
once soil failure is initiated. The addition of a strain rate term in the hyperbolic
constitutive law, and thus the finite element stiffness matrix, magnified the problem, as
did the inclusion of inertial terms in a lumped mass matrix. Rosa and Wulfsohn (1999)
reported some success with a viscoelastic formulation. Inertial terms were assumed
negligible compared to viscous terms and therefore removed from the model, and viscous
strain rate terms were incorporated in a lumped damping matrix instead of the stiffness
matrix. Acceptable predictions were obtained for a rigid, narrow tool, with draft
overpredicted by less than 1% at a speed of 2.8 m/s, but by 25% at 8.4 m/s. Numerical
instabilities were still encountered as the number of “failed” elements increased. The
authors suggested that viscoplastic formulations, in which failure is better handled, might
prove more suitable for analysis of high-speed soil-tool interaction.
Viscoplastic Models. Viscoplastic models are the most general constitutive
representation of material behavior. The combination of rheological effects with
plasticity-based constitutive models make the stress and strain states functions of time as
well as load history. Zienkiewicz (1976) concluded that viscoplastic models, rather than
the elastoplastic models that neglect viscous effects, are needed to describe dynamic soil
behavior.
The effect of strain rate on strength of cohesive soils is typically modeled by assuming
that the size of the whole yield surface depends on the rate at which the soil is being
deformed. It is often assumed that dynamic effects can be superimposed on static soil
behavior. Thus, there will be a yield locus corresponding to static soil response (infinitely
slow loading), but as the deformation rate increases the size of the yield locus is increased
by an amount related to the strain rate. One such way to incorporate strain rate effects in a
critical-state-based constitutive model is shown in figure 1.39. Alternatively, the rate at
which plastic deformations occur (i.e., viscoplastic deformations) can be considered to be
dependent on the extent by which the current stress state lies outside the static yield
surface. The static yield surface itself may still expand or contract due to strain-hardening
or softening. Experiments by Akai et al. (1977) showed that the vectors of rates of
steady-state creep (viscous flow) for a soft sedimentary rock were approximately normal
to moving (dynamic) surfaces based on the Lade-Duncan failure criterion.
The mathematical framework for elasto-viscoplastic models was established by
Perzyna (1963, 1966) and Olszak and Perzyna (1966). General viscoplastic formulations
for small strains are reported by Zienkiewicz (1976), Zienkiewicz and Taylor (1988),
Akai et al. (1977), Baladi and Rohani (1984), and Desai and Zhang (1987). To simplify
model formulations, the assumption of no viscous strains in the elastic range is often
Figure 1.39—Incorporation of strain rate effects in critical-state based model for saturated clay (from
Graham et al., 1983b).
made. One of the drawbacks of such models is the large number of soil parameters
required; for example, the model of Baladi and Rohani (1984) requires 16 parameters.
Critical-state-based elasto-viscoplastic models have been proposed by Davis and
Mullenger (1978), Adachi (1976), Adachi and Oka (1982), and Kutter and Sathialingam
(1992). It should be noted, however, that most of these models were developed for
analysis of creep in soils, i.e., the response of soils under sustained long-term loading, or
for the long-term effects of dynamic loading due to earthquakes, rather than for the
analysis of short-term rate effects on soil strength and soil-structure interactions. Chung
and Lee (1975) described the formulation of a visco-elastoplastic finite element soil
model, based on a critical-state yield surface, for the study of rigid wheel-soil interaction.
Xie and Zhang (1985) used an elasto-viscoplastic cap model to simulate dynamic soil-
tool interaction in a finite element code. Large deformations were incorporated using an
updated Lagrangian approach.
Kinematic and Mixed Hardening Models
One of the basic assumptions made in classical plasticity analysis is that within the
yield surface soil deformations are entirely elastic and without hysteresis. This
assumption leads to poor predictions when loading is largely confined to within the
current yield surface, for example under cyclic loading (e.g., multiple passage of tires).
Under cyclic loading there may be a continual build up of pore pressures and
deformations so that failure is reached at a deviatoric stress well below the strength
expected for static loading (Wood, 1982; Wood, 1990). More realistic models that allow
for some plastic behavior upon reloading within the yield surface have been proposed.
There are various ways to incorporate such behavior in models. Nested surface models
define a family of subloading surfaces nested within each other. Each inner surface
represents a specific region of constant strain-hardening (i.e., constant tangent stiffness,
Et) and translates (kinematic hardening) independently of the others (Mroz, 1967; Iwan,
1967; Zienkiewicz et al., 1977). The bounding (outer) surface does not translate, but it
may expand or contract (isotropic hardening) and it controls deformations at large strains.
q q
M
p p
q q
p εs
Figure 1.40—Illustration of a nested surface model combined with Cam clay: (a) Position of yield loci for
initial stress state A; (b) Position of yield loci for stress state B after loading stress path AB; (c) Position of
yield loci for stress state C after unloading stress path BC; (d) Stress-strain response.
As the soil stress state moves through stress space, it carries around the current load
surface with it. The load surface translates until it touches the next outer surface, at which
point they both move together (with the stress state) until they meet the next outer
surface. The larger the load surfaces that are moving, the lower the corresponding tangent
to stiffness. A change in direction of stress path at any stage requires the stress state to
first traverse the inner yield locus, with consequent high stiffness and low strains. The
result is a piecewise linear stress-strain response (fig. 1.40). This type of model can be
shown be equivalent to a series of parallel spring and slider elements (Wood, 1990). Such
models provide a convenient way of storing information about the history of the soil
(Wood, 1990). Mixed hardening-nested models (Prevost, 1977, 1978, 1981) allow both
kinematic and isotropic hardening of the inner loading surfaces.
Al-Tabbaa and Wood (1989) use a two-surface variation of the above concept
combined with the Cam clay model. A small inner yield locus, used to define truly elastic
states, is carried around by the current stress state. Changes in stress to states within the
yield locus are associated with high stiffness (small strains). Changes in stress states
within the region between the inner yield locus and the outer surface are associated with a
lower stiffness, while states beyond the outer yield locus are associated with large plastic
deformations. Instead of defining a series of inner load surfaces, bounding surface
models establish a single bounding or limiting surface while soil stiffness is made a
function of the “distance” of the current state of stress to the bounding surface (Dafalias
and Popov, 1975; Krieg, 1975). The result is a smooth transition from elastic to plastic
behavior, and a much-simplified model as compared to the nesting surfaces. The model
proposed by Mroz et al. (1979) establishes two surfaces: a yield locus enclosing
elastically attainable states and a bounding surface representing the consolidation history
of the soil. The bounding surface is assumed to isotropically harden/soften, while the
yield surface is allowed to translate and change in size within the region enclosed by the
bounding surface. The hardening modulus is taken as a function of the distance between
the current stress state on the yield surface and its conjugate point on the bounding
surface. The matching point is located at the point where the two surfaces have the same
unit normal direction. In the models of Dafalias and Popov (1975, 1977), both surfaces
are allowed to translate and expand or contract simultaneously. Dafalias and Herrmann
(1980, 1982, 1986) use a bounding surface formulation that did not explicitly define a
yield surface. An image of the current stress state is projected on the bounding surface by
a radial line that extends from a point located within the bounding surface (the
“projection center,” a material property), through the current stress point (fig. 1.41).
Kinematic and mixed hardening models can be easily combined with various
elastoplastic models. Typically, while reproducing cyclic soil behavior very well, the
models require a large number of soil parameters. For example, the bounding surface
model described by Dafalias and Herrmann (1982) requires 16 parameters. Naylor (1985)
and Kaliakin and Dafalias (1989) present simplified bounding surface models that can be
implemented with other models while requiring few additional parameters (fig. 1.41).
Kutter and Sathialingam (1992) incorporated a simple bounding surface in a viscoplastic
critical-state model for dynamic behavior of clays.
q
M
B
δ
A
p
yield locus
bounding surface
Figure 1.41—A simple Kaliakin-Dafalias type bounding surface model combined with the Cam clay
model. The bounding surface is an ellipse with the same aspect ratio as the yield locus. Hardening is
related to the distance δ between the current stress state A and its image on the bounding surface at B, as
established by projecting a line from the projection center located at the origin, through the current
stress state.
Part II. Critical-State Soil Mechanics for

Unsaturated Agricultural Soils
Dvoralai Wulfsohn Bankole A. Adams
Machine-soil interactions involve the simultaneous occurrence of shear strength and

volume changes. For example, the passage of a tire on agricultural or forested land
compacts or compresses the soil, which in turn increases its strength. During tillage
operations the tool reduces the soil strength while it loosens the soil and increases its void
volume. The coupling of volume change and shear strength behavior of soils in a
relatively simple framework is a major attraction for the use of critical-state theories for
various soil engineering problems. Several examples illustrating these concepts are
presented here. The primary purpose of this section is to report on advances made in the
application of critical-state soil mechanics to agricultural soils. To do so, it will first be
necessary to discuss the major differences between the behavior of agricultural soils,
which are generally unsaturated, and saturated soils, for which the critical-state theory
was originally developed. An understanding of the mechanics of unsaturated soils will
facilitate a better interpretation of the challenges associated with the applications of
critical-state theory to unsaturated soils, and provide a wider range of tools for
formulating solutions to problems associated with agricultural and forest soils. The focus
will be on literature covering agricultural soils; however, because the emergence of the
science for an “unsaturated soil mechanics” has been slow, there is only a limited
literature applying these theories to modify critical-state theory for unsaturated soils.
Much of the literature dealing with this approach is found in the geotechnical engineering
area. Therefore, the state-of-the-art in application of critical-state theory to unsaturated
soils in general will also be presented.
Mechanics of Unsaturated Soils

To this point the nature of the stresses used in the formulation of elastoplastic soil
models such as the critical-state theory for agricultural soils has not been described. To
do so it is necessary to review the concepts of effective stress and stress state variables
for unsaturated soils.
Effective Stress and Stress State Variables
When dealing with a saturated (two-phase) soil consisting of a soil skeleton and pore
fluid (air or water), the normal stress components must be separated into two parts: the
stress components carried by the soil structure (effective stresses), and the stress
component carried by the pore fluid (pore pressure):
σ = σ´ + u (1.104)
where σ = total applied stress, σ´ = effective stress, and u = pore pressure (Terzaghi,
1936; Terzaghi and Peck, 1948). The effective stress is considered to be the quantity that
entirely governs the mechanical effects of changes in stress for a saturated soil. Thus, the
critical-state description of saturated soil behavior is developed in terms of effective
stresses; the hydrostatic stress variable is the effective hydrostatic stress p´:
p´ = p - uw (1.105)
where uw is the pore-water pressure. The deviatoric effective stress variable q´ is equal to
q (eq. 1.67) because, as a difference in stresses, it is independent of pore pressure.
Agricultural (i.e., surficial) soils are generally unsaturated; the voids are occupied by
both air and water. Since different pressures can exist in the air phase and the water
phase, this simple idea of effective stress is not valid for unsaturated soils.
Several attempts have been made to define a single-valued “effective stress” for
unsaturated soils. Croney et al. (1958) and Bishop (1959) hypothesized that the pore
pressure can be separated into two parts: the stress carried by the water and the stress
carried by the air. In 1959, Bishop proposed an effective stress relationship for
unsaturated soils that included a soil parameter χ as a stress-partitioning factor. At
saturation, χ will be equal to 1 and the expression will reduce to Terzaghi’s effective
stress equation (eq. 1.104). Similar expressions have been proposed for the “effective
stress” of an unsaturated soil by Aitchinson (1961, 1973), Jennings (1961), and Richards
(1966).
Jennings and Burland (1962) and Burland (1965) showed that the mechanical behavior
of unsaturated soils could not be explained fully using the effective stress concept. In
particular, the dependency of soil compressibility on suction is not properly reproduced;
the isotropic yield stress is incorrectly predicted as decreasing with increasing suction,
which is in contrast with experimental observations (Fredlund et al., 1978); and the
effective stress concept is not capable of correctly predicting the collapse of loose soils
upon wetting (Josa et al., 1992). The soil parameter χ could not be determined easily and
it was different when determined for shear strength and for volume change (Blight, 1967;
Fredlund and Rahardjo, 1993a). Various studies have found that χ is a function of the
degree of saturation, the geometry of the air-water-interface, soil structure, cyclic drying
and wetting, and other factors (Feda, 1982). These attempts at proposing a single-valued
effective stress equation for unsaturated soils were therefore problematic from a
continuum mechanics perspective in that they made the proposed stress variable a
function of material properties. Karube (1988) proposed a single-valued effective stress
relationship for unsaturated soil where at constant suction the effective stress was defined
by (σ - ua) and the suction was considered a factor contributing to the soil constants
(similar to the total stress approach), so that the failure equations reduced to a form
similar to those for saturated soil. Data for elastic volumetric strain and changes in water
content from a series of triaxial consolidation tests on an unsaturated kaolin clay did not
support this approach.
Fredlund and Morgenstern (1977) analyzed unsaturated soil as a four-phase system. In
addition to the three phases of water, air, and soil minerals, they identified a fourth phase
representing the air-water interphase called the “contractile skin” (fig. 1.42). On the basis
of multiphase continuum mechanics they argued that the total stress σ, pore-air pressure
ua, and pore-water pressure uw must be combined in two independent stress parameters in
order to describe the mechanical behavior of unsaturated soils. For practical engineering
(a) (b)
Figure 1.42—Unsaturated soil as a four phase medium: (a) Element of unsaturated soil showing the four
phases: solid, water, air and contractile skin (from Fredlund and Morgenstern, 1977); (b) The stress state
variables for an unsaturated soil (from Fredlund and Rahardjo, 1993a).
analysis, the stress state variables most commonly selected are the “net stress”: (σ - ua),
and the “matric suction” (ua - uw).23 The stress state variables separate the effect due to
changes in normal stress from the effect due to change in pore pressure. The important
distinction between this approach and previous ones is that net stress and matric suction
are considered as independent stress variables, eliminating the need for a single-valued
effective stress equation for unsaturated soil (Fredlund and Rahardjo, 1993a).
Furthermore, unlike the various single-valued effective stress equations proposed for
unsaturated soils, they are independent of material properties and are thus true state
variables compatible with continuum mechanics. Using this approach, Fredlund and
Morgenstern provided a rigorous, scientific basis for an “unsaturated soil mechanics,”
supported by over 25 years of experimental evidence. Similarly, the effective stress (eq.
1.104) was reinterpreted as the stress state variable for saturated soils.24
The use of the term “effective” stress to describe stresses in unsaturated soils is
strongly discouraged by those who advocate the use of the independent stress state
variable approach, because of the tendency for confusion with a single-valued stress state
variable as is appropriate for saturated soils. Accordingly, in this chapter the term
“effective stress” will be used exclusively to refer to the stress state variable for saturated
soils.
Effect of Suction on Mechanical Behavior
With the introduction of the matric suction as an independent stress variable, it is
appropriate here to describe the effect of soil suction on unsaturated soil behavior. The
total suction of a soil is made up of two components: the matric suction (due to surface
tension at air-water interfaces) and the osmotic suction (due to the dissolved salts in pore-
water). Both suction components have an effect on the mechanical behavior of a soil.
23
Note that matric suction is a stress invariant.
24
It is important to reiterate that the continuum mechanics approach is a phenomenological one; the state
variables are not derived based on consideration of intergranular stresses or a capillary model. Such attempts to
derive the effective stress for saturated soils based on a micromechanical (as opposed to continuum) approach
have not been successful (see Bishop, 1959).
Since the osmotic suction is related to the solute concentrations in the pore-water, the role
of osmotic suction is equally applicable to both saturated and unsaturated soils (Fredlund
and Rahardjo, 1993a). Matric suction is associated only with unsaturated soils and varies
with time due to environmental changes; as a result, the matric suction of a field soil at
any location may vary substantially over the growing season. At higher initial water
contents, when water content of the soil is varied the matric suction component of soil
suction is primarily affected. Often, the influence of osmotic suction is quantitatively
negligible compared to the matric suction. The effect of osmotic suction changes on soil
behavior may be significant in cases where the salt content of the soil is altered, due to
chemical contamination, soil salinization, or precipitation on a very dry soil. In cases
where both matric and osmotic suctions have similar quantitative effects on soil behavior,
the osmotic suction can be considered as an independent isotropic stress state variable, or
there is some evidence that it may be possible to combine the matric and osmotic
potentials into a single total suction stress tensor (Fredlund and Rahardjo, 1993a). To
date, most unsaturated soil mechanics research has been concerned with the effect of
matric suction on soil mechanical behavior.
Studies have shown that increases in matric suction generally increase soil strength,
both shear and tensile (e.g., Greacen, 1960; Chancellor and Vomocil, 1970; Williams and
Shaykewich, 1970; Fredlund et al. 1978; Gulhati and Satija, 1981; Escario and Sáez,
1986; Braunack and Malafant, 1988; Drumright, 1989; Mullins et al., 1990). Escario and
Sáez (1986) and Fredlund et al. (1987) found a non-linear relationship between shear
strength and matric suction. At large values of matric suction, strength may drop for some
soils (Williams and Shaykewich, 1970; Mullins and Panayiotopoulos, 1984).
Suction or water content changes have significant implications for volume change
behavior. Matric suction contributes to soil stiffening so that soil compressibility
decreases while the preconsolidation pressure increases (Matyas and Radhakrishna, 1968;
Alonso, 1993; Cui and Delage, 1993; Kohgo et al., 1993a). Cui and Delage (1993)
observed a transition from ductile to brittle failure as a result of increase in matric suction
(fig. 1.43). Other researchers too have observed that increasing matric suction has a
Figure 1.43—Effect of matric suction, s, on stress-strain behavior of an unsaturated soil. Post-peak

softening behavior is shown for suctions at 400 kPa or higher (from Cui and Delage, 1993).
similar effect on stress-strain behavior as increasing bulk density. Unsaturated soils

(especially clays) may also experience swelling and shrinking as matric suction varies
due to changing environmental conditions (Matyas and Radhakrishna, 1968; Fredlund
and Rahardjo, 1993a). During drying of a field soil, shrinkage of the clay may cause
cracking, so that many agricultural soils behave as homogeneous materials at low matric
suctions, but fail by crack propagation at higher suctions (Utomo and Dexter, 1981).
Wetting (i.e., reducing suction) of medium- to low-density unsaturated soils also induces
collapse (i.e., irrecoverable volumetric compression). So called “hardsetting soils” are
characterized by an inability to establish water-stable aggregates. During drying
(increasing matric suction), hardsetting soils sharply increase in strength and set to a
hard, structureless mass; however, upon rewetting they become very soft and the soil
mass will slump (Mullins et al., 1994).
The Soil-Water Characteristic Curve. Matric suction is closely related to the pore
size distribution and the water content of the soil. Matric suction can be viewed as a
stress variable that influences soil properties that are dependent on pore size, pore
distribution, and the pore fluids. The extent of the influence of matric suction on these
parameters will be affected by the structure produced due to the application of external
stresses. The relationship between water content (or degree of saturation) and matric
suction for a given soil structure is referred to as the soil-water characteristic, water
potential, or moisture retention curve. This relationship is affected by the soil texture,
type of minerals in the soil, and the microstructure (Warkentin, 1971). Soil-water
characteristic relationships have been used traditionally in agronomy to monitor the
availability of free water for plant roots and to estimate pore size distribution. The soil-
water characteristic is also emerging as being of great importance in quantifying
unsaturated soil mechanical behavior. It has been suggested that the soil-water
characteristic is an important constitutive relation25 for unsaturated soils (Fredlund and
Rahardjo, 1993b). It is possible to estimate a wide range of unsaturated soil property
functions, including permeability, thermal conductivity, coefficient of diffusitivity, and
shear strength parameters (see fig. 1.44) from the soil-water characteristic relationship
(Johansen, 1975; Reicosky et al., 1981; Fredlund et al., 1995; Fredlund, 1996, 1997). Use
of the soil-water characteristic is particularly attractive for agricultural applications, since
it provides a link between soil mechanical properties and conditions for plant growth.
Stress State Variables for Agricultural Conditions
Total or applied stresses have been used by most researchers in the study of
agricultural soil mechanics. The two common reasons given for the use of a total stress
approach are the difficulty in quantifying internal stresses at air-water interfaces (i.e.,
matric suction), and that changes in pore-water pressures are minimal and of little effect
in agricultural soils. The influence of soil suction is then assumed to be embodied in the
magnitudes of the appropriate critical-state parameters expressed in terms of total stress
(Hettiaratchi, 1987). The reasons for the use of total stresses in the study of unsaturated
soil behavior cannot be always justified in view of recent advances in the study of
unsaturated soil behavior. It is now possible to measure and control the pore-air and pore-
water pressures during laboratory testing (Fredlund and Rahardjo, 1993a; Wulfsohn et al.,
25
The soil-water characteristic curve relates a stress state variable (i.e., the matric suction) to a strain state (or
volumetric) variable (i.e., soil water content). Thus, by definition, the soil-water characteristic is a constitutive
relationship.
Figure 1.44—Relationship between the soil-water characteristic and shear strength for a sand and a
clayey silt. The non-linear variation of shear strength with suction can be directly related to the soil-
water characteristic relationship and effective strength parameters (from Fredlund and Rahardjo,
1993b).
1998). The measurement of matric suction in field conditions remains a major challenge,
particularly for fine-textured soils; however, developments in time-domain reflectometry
and other technologies are beginning to overcome these constraints (e.g., Whalley et al.,
1994; Woodburn et al., 1993; Ridley and Burland, 1994; Ridley and Wray, 1995). The
assumption that pore-water pressure changes are not significant for constant water
content conditions is not valid when the applied external stress exceeds some critical
value (Wulfsohn et al., 1998).
Towner (1983) stressed the need to appropriately define external as well as internal
stresses in the application of the critical-state theory to agricultural soils. Towner argued
out that the uniqueness of the critical-state line, a basic assumption in critical-state soil
mechanics, will depend on the use of appropriate stress state variables. Kirby (1989)
pointed out that in order to formulate an unsaturated critical-state framework, issues that
must be addressed include how stresses are quantified in unsaturated soils, the
nonhomogeneous deformation characteristic of the Hvorslev surface, and the volume
change that occurs in unsaturated soil in “undrained” (constant gravimetric water content)
tests. Nevertheless, Kirby (1989, 1991b) concluded that for practical convenience the
yield surface, critical-state line, normal compression line, and rebound lines could be
described in terms of total or applied stresses.
An alternative to the use of the total stress approach is the use of independent stress
state variables in which the effect of total applied stress is separated from the effects due
to the internal stress at pore air-water interfaces. This approach has been used by a
number of researchers who retained net stress, suction, void ratio and water content as
stress and strain state variables, respectively (e.g., Bishop and Blight, 1963; Aitchinson,
1965; Burland, 1965; Matyas and Radhakishna, 1967; Fredlund, 1973; Alonso et al.,
1990; Wheeler and Sivakumar, 1993; Adams and Wulfsohn, 1996, 1997, 1998).
In summary, two approaches are used in the unsaturated and agricultural soil
mechanics literature to classify stresses and the pore volume. In the total stress approach,
hydrostatic and deviatoric stresses for unsaturated soils, p and q, are defined by equations
1.66 and 1.67, respectively (for axisymmetric stress states), and the specific volume v =
1 + e is used to define the volume state of the soil. In the independent stress state variable
approach matric suction s = (ua - uw) is included as an independent stress variable, and the
mean net stress is used to define the hydrostatic stress state as:
p´´ = (σ1+ σ2+ σ3)/3 - ua (1.106)
where p´´ = mean net stress and ua = pore-air pressure. The expression for deviatoric
stress, q, (eq. 1.67) remains the same, while the specific volume, v, now includes changes
in the pore-air and pore-water volumes. Water content, w, degree of saturation, Sr,
moisture ratio,26 ϑ = Sre, or specific water volume,27 vw = 1 + Sre = 1 + wGs, may be used
as a deformation state variable in addition to specific volume or void ratio. The specific
water volume can be shown to be the work conjugate to matric suction, i.e., the strain
increment dvw/v is correctly associated with suction as to give a component of work input
per unit volume of soil (just as the volumetric strain increment and the deviatoric strain
increment are associated with p´´ and q, respectively).
26
The proportion of pore volume occupied by water (Klausner, 1991; Philip and Smiles, 1969); equal to the
volume of water in a volume of soil containing a unit volume of solids.
27
The volume of water plus solids in a volume of soil containing a unit volume of solids (Wheeler, 1991).
The Work Input to an Unsaturated Soil

Care is needed in the choice of stresses and strains used in constitutive models so that
they are work conjugate, i.e., the product σ dε must produce the correct incremental work
per unit volume performed by stresses on a specimen. Wheeler and Sivakumar (1995)
stated without proof that the specific water volume, defined as the volume of water in a
unit volume of soil, vw = 1 + Sre, is the work conjugate of the matric suction, s = ua - uw.
Houlsby (1997) verified this by considering the energy balance on an element of
unsaturated soil for general loading conditions.28 The analysis was fairly involved and
will not be repeated here. We will present an analysis similar to that of Schofield and
Wroth (1968) for a saturated soil.
Consider an unsaturated triaxial specimen loaded as shown in figure 1.45. The
specimen is initially in equilibrium under static loads. The cylindrical specimen has
initial volume v (of which e = v - 1 consists of voids) and length l, so that its cross-
sectional area is a = v/l. We can express the sample volume as:
v = 1 + e = 1 + Sre + (1 - Sr)e = 1 + ϑw + ϑa (1.107)
where ϑw = Sre is the proportion of pore volume occupied by water (the “moisture ratio”)
and ϑa = (1 - Sre) is the proportion of pore volume occupied by air.
The triaxial specimen is enclosed in an air- and water-impermeable sheath of
negligible thickness and strength. The specimen is sealed at its base on a pedestal
containing a saturated high-air entry disk that allows passage of water but not of air,
thereby ensuring continuity between pore-water within the specimen and a water
compartment in the pedestal. A coarse low-air entry disk, with low attraction to water,
placed on top of the specimen allows free flow of air between the specimen and an air-
filled compartment. The cell, pore-air, and pore-water compartments are all connected by
rigid pipes to cylinders. The pressures in the cylinders are controlled by weightless
pistons of unit cross-sectional area. The test system is contained within an imaginary
boundary penetrated by stiff weightless rods that carry upper platforms to which small
load-increments can be applied and removed. The displacement of any load-increment is
the same as that of its associated load within the system.
We probe the equilibrium of the specimen by slowly applying a set of load-increments
[duw, dua, a(dσ1 - dσ3), dσ3, any of which may be zero] simultaneously to the upper
platforms. We will consider mechanical work only, and neglect the (small) work input to
the sample across the system boundary needed to provoke a stress change. We will also
assume that the solids phase is incompressible. The external agency provokes an
incremental axial displacement dl and incremental volumetric deformation dv. The
overall volume change (i.e., of the soil structure) is equal the sum of air and water
volume changes (Fredlund and Rahardjo, 1993a).29 Differentiating equation 1.107:
dv = de = dϑw + dϑa = Srde + e·dSr + (1 - Sr)de - e·dSr (1.108)
28
In his analysis, Houlsby (1997) used as stress variables the matric suction and a second stress equation
involving total stress and degree of saturation, obtained by substituting Sr for χ in Bishop’s proposed effective
stress equation for unsaturated soils. This is the same combination of variables used by Bolzon et al. (1996).
The strain conjugates derived by Houlsby (1997) can be shown to be equivalent to those obtained here.
29
The volume change of a fluid (air or water) will be largely due to compression under undrained conditions,
and to flow of the fluid through the soil structure under drained conditions.
Load duw d σ3 dua a(dσ 1 - dσ3)

increments
dϑ a Displacements
dϑ w
Equilibrium
dv dl
ua
System
boundary
a
(σ 1-σ 3)
σ3
uw σ3
σ3
l v
(b)
(a)
Figure 1.45—Triaxial probing test system and unsaturated soil specimen: (a) Probing load increments
and observed displacements; (b) Representation of principal stresses σ1 and σ2 = σ3 acting on a soil
element in the test system due to the all-round confining pressure (σ3) and axial loading by a ram
(σ1 - σ3).
The incremental work done (within the system) on the specimen during displacements
is:
dW = −u a dϑ a − u w dϑ w + σ 3 de + (σ1 − σ 3 )adl . (1.109)
Substituting equation (1.108) into (1.109) and rearranging we obtain:

dW = − u a [(1 − S r )de − e ⋅ dS r ] − u w [S r de + edS r ] + σ 3 de + (σ 1 − σ 3 ) ⋅ a ⋅ dl

(1.110)
= (u a − u w )[S r de + e ⋅ dS r ] + (σ 3 − u a ) ⋅ de + (σ 1 − σ 3 ) ⋅ a ⋅ dl
where
[Srde + e·dSr] = d[Sre] = dϑw. (1.111)
To obtain the incremental work input per unit volume of soil, we divide both sides of
the resulting expression by v = 1 + e:
dW  dϑ 
= (u a − u w ) w  + (σ 3 − u a ) ⋅ dε V + (σ1 − σ 3 ) ⋅ dε 3
v  1+ e 
(1.112)
 dϑ 
= (u a − u w ) w  + (σ 3 − u a ) ⋅ (dε1 + 2dε 3 ) + (σ1 − σ 3 ) ⋅ dε 3
 1+ e 
where we have used the relations εV = –de/(1 + e) and ε1 = dl/l. Rearranging, we can
write:
dW  dϑ 
= (u a − u w ) w  + 2(σ 3 − u a ) ⋅ dε 3 + (σ1 − u a ) ⋅ dε1 (1.113)
1+ e  1+ e 
Further manipulation yields:
dW  dϑ  σ + 2σ 3
= (u a − u w ) w  + ( 1 − u a ) ⋅ (dε 1 + 2dε 3 ) + (σ1 − σ 3 ) ⋅ 23 (dε 1 − dε 3 )
1+ e 1+ e  3
(1.114)
which is (using the critical-state soil mechanics notation):
dW  dϑ 
= s ⋅  w  + p' '⋅dε V + q ⋅ dε S (1.115)
1+ e 1+ e 
Equations 1.113 to 1.115 demonstrate that the suction has the conjugate strain rate
dϑw/v, i.e., the moisture ratio ϑw = Sre is the strain state that is work conjugate to matric
suction. The net stress (σ - ua) has conjugate strain rate dε. As Sr → 1, ϑw → e, so the
moisture ratio should be used as a second deformation state variable along with void
ratio, e. Notice also that dϑw = d(Sre) is equal to d(1 + Sre) (cf. eqs. 1.110 and 1.111), so
that we can also derive the conjugate strain rate for matric suction to be dvw/v, where vw =
1 + Sre is the specific water volume defined by Wheeler (1991). The specific water
volume, vw, tends to the state variable v as the degree of saturation Sr approaches unity.
Therefore it is the preferred deformation state variable to be used along with specific
volume, v.
Critical-State Behavior of Agricultural Soils

Critical-state concepts have been used to describe the behavior of agricultural soils
since the 1960’s (e.g., Greacen, 1960; Bailey and VandenBerg, 1968; Kurtay and Reece,
1970). Reece (1977) presented a very simple physically-based account of the basic ideas
of critical-state mechanics directed to a non-engineering scientific audience. Reece
(1977), Hettiaratchi and O’Callaghan (1980, 1985), and Leeson and Campbell (1983)
suggested that the critical-state theory can be applied to the study of compaction and
tillage problems in agriculture. Hettiaratchi (1990) described a combined critical-state
soil mechanics and cellular biomechanics model to describe plant root growth in
compacted soils. Compaction problems can be studied by considering soil behavior on
the Roscoe surface. This involves testing normally-consolidated or lightly over-
consolidated soils. Tillage studies can be considered within the Hvorslev domain by
testing over-consolidated soils. The early pioneering literature was largely concerned
with the theoretical foundations of the theory for agricultural soils. There has been a
concerted effort to experimentally examine the critical-state framework with unsaturated
soils. The experimental evidence indicates that the critical-state theories can be extended
to include unsaturated soils. However, there are some significant differences in the details
of the critical-state concept for such soils. State boundaries for unsaturated agricultural
soils have to take into account the diversity in soil type, as well as transient physical
properties such as water content, organic matter content, and micro-structural state
(Hettiaratchi, 1987). Most of the initial experimental efforts at applying critical-state
theory to the study of agricultural soil behavior have been towards the determination of
the state boundaries. In more recent studies, the theory has been used to predict
agricultural soil behavior. A review of these studies using both total stress and stress state
variable approaches, and a discussion of the major factors affecting critical-state
behavior, are presented next.
State Paths Associated with Agricultural Field Operations
Hettiaratchi and O’Callaghan (1980) used a simple critical-state model and stress
paths in state space to provide a qualitative interpretation of soil behavior in agricultural
soil-machine interactions. A selection of their examples will be presented here. Further
practical examples are provided by Hettiaratchi (1988), Spoor (1975), Stafford (1981),
Kirby (1991a), and Spoor and Godwin (1999).
Slipping Wheel Operating in Loose Soil. As illustrated in figure 1.46, the soil is at
an initially loose state at “a” which could be either sub-critical or super-critical (here a is
shown in the sub-critical domain). A small amount of shearing stress superimposed on a
relatively high vertical stress is sufficient to compact loose soil (Reece, 1977). This is
indicated by the loading path a-b where q increases slowly compared to p. The state path
follows the current unloading-reloading line (url 1) in the v-p plane and reaches the
Roscoe surface Y1 at b. As the soil is further loaded it yields along the Roscoe surface
(sub-critical) and goes critical at c. The yield locus has increased in size (Y2). After
passage of the wheel the soil element falls outside the zone of influence of the moving
stress field and may swell (rebound) along the new url line to d. The soil has undergone
severe compaction as indicated by the significant decrease in specific volume (-∆v) from
a to d.
Slipping Wheel Operating in Firm Soil. The soil is initially in a dense state and lies
in the super-critical domain at a (see fig. 1.47). If the soil is subjected to high shearing
p
0
q0
CSL
T
c Y2
b
Y1
a d p
v v
CSL NCL
a a
b b url 1
∆v T
url 2
d c
c d
q p
Figure 1.46—Approximate state paths associated with a slipping wheel operating in loose soil (after
Hettiaratchi and O’Callaghan, 1980).
rates (q increases rapidly with p) the state path will reach the Hvorslev surface H at b in
the super-critical domain. At this stage the soil “fails,” i.e., it has reached its maximum
value of q. However, as loading continues the soil may traverse the Hvorslev surface to
reach critical at c. The rapid decrease in q along path b-c indicates unstable behavior; the
soil will probably break up into discrete blocks. Notice that specific volume has increased
(+∆v) indicating loosening of the soil.
Wide Cutting Blades. Cutting blades are generally used to loosen firm soil and so the
initial condition a is shown heavily over-consolidated (large initial yield surface Y) with
low specific volume v lying in the super-critical domain (fig. 1.48). Three possibilities
are illustrated, each depending on the relative magnitude of applied isotropic
consolidation pressure p. For low values of consolidation pressure the action is very
p
0
q H
0 CSL
b
T
c
a p
v v
NCL
CSL
T
c c
a a
∆v
b
b url
q p
Figure 1.47—Approximate state paths associated with a slipping wheel operating in firm soil (after
similar to that for a slipping wheel in firm soil as the state path traverses the Hvorslev
surface from b to c. The increase in v as the soil is brought to critical at c indicates that
the cutting blade has loosened the soil. For moderate values of p the state path traverses
the Hvorslev surface at d and goes critical after a very short unstable trajectory from d to
e. There is only a small change in v. For high values of p the state path reaches the yield
surface Y at f in the sub-critical domain. The soil then compacts on its way to critical-
state at g and the cutting blade has actually compacted the soil. High values of p might be
expected for soil elements subjected to large geostatic stresses (e.g., deep layers of a
heavy clay soil).
CSL
g
p0 H d
b Y
e f
q0 c
a p
v v
NCL
CSL
c c
+ a
a
b
- ∆v b
e
f url
d
f
g g
q p
Figure 1.48—Approximate state paths associated with wide cutting blades working compact soil (after
Narrow Tines. Narrow tools load soil in a relatively small region located close to the
blade. The soil element examined is therefore close to a soil free surface and shear
stresses will predominate over hydrostatic stresses during loading. Depending on the
configuration of the tool, three possibilities are considered, as shown in figure 1.49. The
state path from a to b reaches the Hvorslev surface and is similar to the case discussed for
a wide blade and moderate values of p. The state path from a to c reaches the Hvorslev
state boundary at the T-surface. Another state path from a to d lies below this and reaches
the T-surface. Tensile stresses will be set up in the soil, opening up cracks. This agrees
with experience with rigid tines. The good soil loosening effect of sharply raked tines
could also be attributed to the decreasing p regime induced in the soil, as indicated by the
state paths followed. The broken lines trace alternative state path trajectories for the first
two scenarios, in which it is proposed that p might have increased for part of the loading
sequence.
H
b
c CSL
T
p0 d
q0
v a p
v
NCL
CSL
e e
b a
∆v url
d c b a
q p
Figure 1.49—Approximate state paths associated with narrow tines (after Hettiaratchi and O’Callaghan,
1980).
The only processes that will loosen soil are failure by tension or shearing at relatively
low compressive stresses. In practice, tensile stresses in the soil can be generated by
bending (e.g., flow over sweeps and wings) or naturally leaving discs and moldboard
plows as the soil mass is lifted up and unsupported (very low p) and gravity induces
tensile stresses (Spoor and Godwin, 1998). Indeed, the above analysis indicates that for a
tillage tool to effectively loosen the soil it should generate high q in conjunction with low
p (as would be found ahead of a sharply raked blade) to reach and follow the Hvorslev
surface up to critical. Alternatively, the state path must reach the no-tension surface T, in
which case the soil will fracture without overall change in v as the deformations are
largely elastic (Hettiaratchi and O’Callaghan, 1980; Hettiaratchi, 1987, 1988). In both
these cases the state paths lie in the super-critical domain.
Root Growth in Compacted Soils.30 When growing in compacted soils, plant roots
are subjected to mechanical impedance. Roots cannot penetrate pores with a diameter less
than that of the extending zone of the root (Russell, 1977). When the soil pore space
cannot accommodate the extending root system, roots create their own pores of large
enough size to penetrate. Abdalla et al. (1969) hypothesized that the root accomplishes
this by swelling, inducing tensile cracking in the surrounding soil, after which the root
can resume elongation through the newly created pore (fig. 1.50a). As the soil displaces
radially, the induced tensile stresses reduce the ambient value of p. The state path in the
critical-state model will be directed toward the T surface (fig. 1.50b) and the soil will
ultimately fail by tensile cracking. Experimental studies have also shown that roots
growing in compacted soils are thicker and shorter than in uncompacted soils. Theoretical
and experimental studies using mechanical probes to simulate axial (punch) indentation
and radial expansion of roots have shown that the contact stresses mobilized in the course
of radial expansion of a cylinder is always smaller than the stress associated with
penetration of a cylinder (Hettiaratchi, 1990). Kirby et al. (1998a) used a simple critical-
state finite element model to simulate penetration of roots of different thicknesses
through soil (without radial expansion). For a given soil strength, peak stresses (located
near the root tip) were substantially less in the case of 1 mm thick roots compared to 1.8
mm thick roots. The critical-state model thus indicated that thickening of roots also
serves to reduce the arresting stresses at the root tip.
v
(i) (ii) (iii) (i) (ii) NCL
CSL
T
p
(a) (b)
Figure 1.50—Hettiaratchi “variable apex geometry” root growth model. (a) Steps in growth cycle of root
in compacted soil: (i) axial extension arrested by compacted soil in root cap zone, p > pu where pu is the
limiting confining pressure for root growth; (ii) radial expansion of region behind root tip weakening soil
in root cap zone, bringing p < pu ; (iii) axial extension resumed; (i, ii) The cycle then repeats itself. A
indicates radial growth mode, B indicates axial growth mode, C indicates root extension increment (from
Hettiaratchi, 1990). (b) Approximate state path associated with radial expansion of root extension zone
growing in compacted soil (projected on v-p plane).
30
Advances in soil-root dynamics were presented in Volume 1 of this monograph (Bowen et al., 1994).
Verification of Critical-State Concepts for Unsaturated Soils

Much of the research on unsaturated soils, including agricultural soils, has been
directed toward investigating whether the fundamental assumptions underlying critical-
state theories apply to unsaturated soils. For instance, do unsaturated soils undergoing
deviatoric loading reach a true critical state? Can a unique critical-state line be
established for a given soil? Can state boundaries clearly delineating elastic versus plastic
behavior and deformations consistent with “wet of critical” and “dry of critical” zones be
established? Studies investigating these aspects of unsaturated soil behavior will be
reviewed in this section.
Attainment of an Unsaturated Critical-State. McKyes (1985) cautioned that
agricultural soil samples may fail at a peak stress before they reach the critical-state. It is
important to recognize this when applying critical-state models to agricultural problems.
The critical-state line represents an “ultimate” condition in state space toward which
samples will move under deviatoric loading, yet it may not necessarily represent
“failure.” This will be of particular significance for loading states within the super-critical
zone (brittle failure). The formation of rupture planes is associated with stress states
reaching the Hvorslev surface. Under these conditions it is difficult to determine the
values of stresses on the failure plane. Hettler and Vardoulakis (1984) showed that when
testing very dense specimens, shear bands and other non-homogeneous deformations
develop prior to the soil reaching critical-state, introducing uncertainty in the critical-state
parameters obtained under these conditions.
Toll (1990) reported that a true critical-state was not achieved in constant water
content triaxial tests with a Kiunyu gravel. All samples continued to dilate even at large
strains so that a unique specific volume (or void ratio) was not found. The explanation
provided was that the initial soil fabric in unsaturated specimens is not destroyed by
shearing to large strains. Suction was also still changing at the end of each test. Samples
with initially low suctions showed an increase in suction under shear, while samples with
high initial suctions showed a decrease in suction. Similar findings were reported by Shi
(1998) for tests on undisturbed samples of coarse-textured forest soils.
Testing a wide range of agricultural soils, researchers have found that a critical-state
could be identified with respect to hydrostatic stress, deviatoric stress, and volume
changes (e.g., Kirby, 1989; Hettiaratchi et al., 1992; Kohgo et al., 1993b; Petersen, 1993;
Adams, 1996).
In constant suction tests reported by Wheeler and Sivakumar (1995), four state
variables (p´´ = p - ua, q, v, and s = ua - uw) stabilized at the end of the tests indicating that
a critical-state had been attained. Water content still varied by a small amount. For one
series of constant water content (varying suction) triaxial shear tests, p´´, q, and v
stabilized at the end of tests, while pore-water pressure uw was still falling (s rising) at
the end of tests. Adams (1996), using constant water content tests with an agricultural
sandy clay loam, found that under high confining stress regimes suction continued to
decrease (but at a constant rate) even when a critical-state seems to have been attained
(i.e., no further changes in p´´, q, and v). The rate of change of suction with shearing near
or at the “critical-state” increased with increasing confining pressure. In an earlier study,
Larson and Gupta (1980) also found that with increasing normal stress pore-water
pressure became more negative down to a minimum value, after which it began to
increase. It thus appears that there is a transition stress at which negative pore-water
pressure (or matric suction) in an unsaturated soil is neutralized by the applied external
load.
Interestingly, as early as 1954, Croney and Coleman found that complete shearing in
soil produced a suction condition that was dependent solely on the water content of the
soil and was independent of the initial suction before disturbance. In other words, for
each soil a unique relationship (without hysteresis) was found between suction and water
content at the fully sheared condition. They referred to this as the relationship for a
“continuously disturbed soil” and found that it was closely related to the Atterberg Index
values (fig. 1.51) and to the soil shear strength. [Croney and Coleman (1954, 1961) use
the term “suction” to refer to a pore-water pressure deficiency below atmospheric
pressure (i.e., ua is assumed to equal 0 kPa, gauge), in a small sample of soil free from
external stress, and the term “negative pore-water pressure” for any pressure deficiency
(below atmospheric pressure) measured in situ or in the laboratory with the soil subjected
to some non-zero external loading condition.] Croney and Coleman (1961) introduced a
parameter for unsaturated soils, α, to relate the change in negative pore-water pressure to
the change in total stress under constant water content conditions. The concept was
extended by Bishop and Henkel (1962) to more general loading conditions as:
d(ua - uw) = -α d(σ - ua) (1.116)
in which d(ua - uw) = change in matric suction, d(σ - ua) = change in net stress, and α is a
dimensionless soil parameter. The effect of combined normal and shear deformations on
the suction versus water content relationship is little understood, however.
Figure 1.51—Suction versus water content relationships for a heavy clay soil. “Suction” here refers to the
soil suction under zero external loading conditions (from Croney and Coleman, 1961).
Studies Based on a Total Stress Analysis. Bailey and VandenBerg (1968) and Bailey
(1971, 1973) reported the first experimental studies to present agricultural31 soil behavior
using a critical-state framework. Results were presented in terms of mean normal stress
(σm = p), maximum shear stress (τmax), and bulk weight volume (BWV). Remolded
samples compacted at different water contents were used in these studies. Samples
compacted differently had distinctly different strengths, so that a single yield surface in
three-dimensional σm - τmax - BWV space could not be established. These early studies
showed the importance of considering stress history (as indicated by the preconsolidation
stress) in determination of state boundaries for unsaturated soils. In such cases,
uniqueness of the yield surfaces for a particular soil can be examined by plotting data in
terms of stress ratios (see fig. 1.53).
Hettiaratchi and O’Callaghan (1980, 1985) and Hettiaratchi (1987) reported on
experimental investigations conducted at the University of Newcastle upon Tyne and
presented a simple version of the critical-state theory for agricultural soils. The authors
showed that state boundary surfaces could be determined. They demonstrated that the
effects of water content and soil microstrucural state need to be considered in applying
the critical-state concept to unsaturated agricultural soils. They also pointed out the need
to establish appropriate relationships between p and q, which are very difficult to do
except in very simple cases.
Stafford (1981) applied the critical-state framework to a tillage problem involving
rigid tines. Two general modes of failure (i.e., brittle and plastic flow) were identified
within the two regions of the critical-state model. Stafford (1981) identified a critical
range of water content influencing the performance of a rigid tine, and proposed a “wet
side” and “dry side” approach to soil failure about the critical wall. Stafford introduced a
“failure coefficient,” F, which defined a critical value above which soil fails in a flow
mode, and below which it fails in a brittle mode. The coefficient F was shown to be a
function of water content, speed, rake angle, and the reciprocal of density (i.e.,
proportional to specific volume), making a dimensionless form of F impossible. Stafford
(1981) suggested that by identifying other factors, a dimensionless F might be
established.
Leeson and Campbell (1983) presented critical-state data for a loam and a sandy loam.
They investigated the variation of critical-state parameters with water content and
deduced that if it was recognized that the state boundaries varied with water content, then
the critical-state theory could be extended to unsaturated soils. Postulated state boundary
surfaces for saturated and unsaturated soils are shown in figure 1.52. The variation of
reloading κ was shown to be significantly correlated with soil specific volume (and
therefore preconsolidation stress and bulk density) and approached zero at a specific
volume of about 1.5. For both soils, the unloading and reloading gradients were very
different. No relationship was found between unloading gradient and the deformation
state variables v, w or Sr.
Kirby (1989) measured the yield surfaces and established the critical-state condition
for four agricultural soils. An oedometer was used to investigate volume change behavior
and a constant volume direct shear box was used to determine shear behavior. Two of the
soils were tested using undisturbed specimens sampled from the field, while the other two
soils were tested using remolded specimens. Each soil was tested at a single water
31
An artificial soil made up of equal parts of sand and clay and 17% oil as a binder was used in Bailey (1971).
CSL q
(unsaturated
soil)
p Spherical Deviatoric
pressure CSL stress
(saturated
soil)
Tension
NCL Cut-off
(unsaturated)
NCL
(saturated)
Specific
v volume
Figure 1.52—State boundaries of saturated and unsaturated soils in p-q-v space (from Leeson and
Campbell, 1983). The term spherical pressure is synonymous with hydrostatic stress.
content, but with degrees of saturation ranging from 20% to 98% (i.e., soils were tested
over a range of bulk densities and structures). Samples were loaded over a relatively short
period in an attempt to more closely simulate short loading periods in the field. Total
stresses were employed along with the modified the stress path method of Aitchinson
(1973), in which an attempt is made to closely match the soil sample to its in-situ stress
and suction. The soils did approach a critical-state in the tests conducted, and
deformations both “loose of critical” and “dense of critical” were observed. Soils in states
looser than critical compacted in shear on approaching critical-state, while the specimens
in states denser than critical dilated. When data for three series of tests (not identified)
were plotted in terms of the stress ratios (τmax/σ versus σ/p0) the data points for samples
sheared at low normal stress ratio (dilating samples) coincided and those for samples
sheared at a higher stress ratio (compressing samples) coincided as shown in figure 1.53
(Kirby, 1991a). From the points, a single yield surface similar in shape to those depicted
in descriptions of the critical-state concept and a CSL could be constructed. The data in
figure 1.53 show the two forms of deformation behavior (dilating versus compressing)
separated by a normal stress ratio of about 0.25 to 0.3. Note that the Cam clay yield
Figure 1.53—Shear stress ratio (τmax/σ) as a function of normal stress ratio (σ/p0) in a constant volume (v)
shear test. The critical-state line and yield curve are also shown (from Kirby, 1991a).
surface intersects the critical-state line at a normal stress ratio of 0.5, yet it would still
provide a reasonable approximation for the Roscoe surface for this soil. Kirby (1991b)
determined the critical-state parameters for undisturbed samples of 18 different Vertisols
in eastern Australia. The critical-state parameters were dependent on water content,
Atterberg limits, and soil bulk density. The soils generally expanded in failure (dilated)
for normal stress ratios less than about a half, and failed in compression when the normal
stress during shear was more than about half the pre-compression stress.
Harris (1993) proposed a simple two-parameter model for general anisotropic
consolidation of initially very loose soil samples and interpreted compaction in terms of
soil behavior in critical-state space. The model was evaluated using the data of Grisso et
al. (1987) for four soil types, and compared to predictions using the Grisso et al. (1987)
seven parameter hydrostatic compression model. The proposed two-parameter model was
found to predict density changes with comparable accuracy to the seven parameter model
over a wide range of stress ratios. At the highest stress levels, the Grisso et al. (1987)
model was more accurate.
Gupta and Larson (1982) combined stress predictions based on the Boussinesq
equations with soil consolidation relationships obtained from uniaxial compression tests
to predict bulk density changes due to passage of a tire. Smith (1985) developed a
numerical compaction model based on the critical-state concept to include the effect of
stress history in the initial soil conditions.
Petersen (1993) presented data for a loam and a sandy loam. The author expressed
behavior in terms of total stresses and concluded that the role of water content was
expressed in the values of the measured critical-state parameters in a systematic manner.
Volume change behavior was also consistent with the critical-state concept. Petersen
(1994) slightly modified the Modified Cam clay model for unsaturated soils. One
modification was to incorporate an intercept for the CSL line on the q-axis. The proposed
model was used to predict the maximum shear strains for triaxial specimens of two soils
(i.e., a sandy loam and a loam) at three different water contents, for two different stress
paths. The closeness between predicted and observed maximum shear strains depended
on the stress path and soil water content. The model failed to predict maximum shear
strain at stress states close to the critical-state; however, for deviatoric stresses less than
90% of the critical-state values, predictions of maximum shear strain agreed well with
measured values.
Kirby (1994) examined the ability of a critical-state based finite element model to
predict volume change and shear behavior of agricultural soils in laboratory specimens.
The model in this study was an extension of the Modified Cam Clay model which
produces a smoother transition from elastic to plastic behavior using imagined subloading
surfaces (Naylor, 1985), i.e., some plastic behavior is permitted upon reloading within the
yield surface (see the section on Kinematic and Mixed Hardening Models in Part I of this
chapter). Finite element model predictions generally compared well with experimental
data, although some numerical problems were encountered under strain-softening (dry of
critical) conditions.
Considerable scatter was observed for the critical-state parameters determined by
Kirby (1991b) (fig. 1.54). In addition to water content, initial bulk density, Atterberg
limits, and the use of a total stress approach, other sources of variation identified were the
stress history of the soil and the use of field samples, which in themselves are highly
variable. Kirby (1995) investigated the influence of the high variability of soil properties
on model predictions of compaction. The critical-state finite element model described by
Kirby (1994) was incorporated into a Monte-Carlo scheme. In the procedure, the finite
element model was run 200 times, each run using a set of parameters randomly sampled
from the known statistical distributions of the critical-state properties. The output values
from the runs were accumulated to calculate means and variances of predictions. Model
simulations were compared with data from compaction tests performed with small dual
tires in a loamy soil in a small soil bin (0.45 m deep). The tests were modeled as an
axisymmetric problem. Due to the non-linearity of the soil model, the mean rut depth
predicted by the Monte-Carlo model was not the same as the solution obtained by a
single finite element simulation using the mean parameters as inputs. The predicted and
0.5 0.04
0.4
0.03
0.3
κ 0.02
λ
0.2
0.01
0.1
0.0 0.5 1.0 1.5 2.0 0.00 0.5 1.0 1.5 2.0
epc epc
Figure 1.54—Variation of compression (λ), and elastic rebound (κ) indices with precompression void
ratio for some eastern Australian soils (from Kirby, 1991b)
measured shape and sinkage depth for single and multiple passes of the tires are shown in
figure 1.55a. The data generally fell within the predicted ±95% confidence limits. Under
these laboratory conditions, coefficients of variability (i.e., CV = standard deviation/
mean) for predicted void ratio changes were on the order of 1/15 or 7% (see fig. 1.55b).
The influence of variability in soil strength under field conditions was found to be
substantially greater, with CV’s of void ratio changes of 10% to 50% for one field soil.
Chi and Tessier (1994) used a commercial finite element program to simulate soil
compaction under a heavy liquid manure spreader operating on a Ste.-Rosalie clay. An
axisymmetric analysis with uniform pressure at the soil-tire interface was simulated.
Geometric non-linearities due to large deformations were considered in the analysis.
Three constitutive models were compared: a critical-state model (modified Cam clay), a
Drucker-Prager cap model (Drucker and Prager, 1952; Sandler et al., 1976), and a non-
linear elastic hyperbolic model with variable Poisson’s ratio (Duncan and Chang, 1970;
Chi et al., 1992). The three models predicted somewhat different stress paths as well as
distributions and intensities of stresses and strains beneath the applied load. The Cam
clay model predicted higher octahedral normal stresses yet lower volumetric strains (i.e.,
compaction) than the other two models. The predicted octahedral normal stresses tended
to concentrate more directly under the tire and penetrate deeper into the soil for the Cam
clay model (fig. 1.56). Predicted tire sinkage was greatest using the hyperbolic model,
and very similar for the critical-state and the cap models, especially under lower applied
(a) (b)
Figure 1.55—Effect of soil variability on predicted compaction using a critical-state finite element model
in a Monte-Carlo simulation scheme. (a) Predicted sinkage of dual tires in a loam soil. Points show
measured rut depths (initial position z, after 1 pass , after 3 passes S, after 8 passes T). Lines show
predicted mean depths and ± 95% confidence limits. (b) Predicted void ratio changes beneath dual tires
in a loam soil. Solid shaded contours show means of void ratio changes (contour interval -0.05, with -0.1
and -0.2 contours labeled). Dashed contours show standard deviations (contour interval 0.0025, with
0.005 and 0.01 contours labeled). The coefficients of variability are on the order of 1/15 (from Kirby,
1995).
Figure 1.56—Comparison of finite element predictions of octahedral normal stress σoct and volumetric
strains εv beneath a loaded tire on a clay soil using the Cam clay critical-state model and the Drucker-
Prager cap model (from Chi and Tessier, 1994).
surface loads (<90 kPa). Comparisons with field data were not presented. The effect of
parameter variability on such analyses, similar to that reported by Kirby (1995), would
also be useful in determining how significant, for practical management purposes, the
difference is between predictions using various elastoplastic models, such as the Cam
clay and the Drucker-Prager cap models.
Stress State Variable Approach. All the above-cited studies used total stresses to
describe the state of unsaturated agricultural soils. Efforts to develop generalized
constitutive models for unsaturated soil using suction as a state variable are relatively
recent. Alonso et al. (1990, 1993), Toll (1990), Wheeler and Sivakumar (1993, 1995) and
Maâtouk et al. (1995) used independent stress state variables in applying critical-state
theory to the study of unsaturated soils encountered in geotechnical engineering practice.
Adams (1996) used independent stress state variables in the application of the critical-
state theory to the study of an agricultural soil. These studies also supported the validity
of general critical-state concepts to unsaturated soils, such as the existence of a unique
NCL and CSL for a constant suction. Wulfsohn et al. (1996) argued that for agricultural
conditions, knowledge of both the soil water content and the suction is needed to
interpret soil behavior.
Alonso et al. (1990) presented a general constitutive model based on concepts of
yielding and critical-state for unsaturated soils. Proposed state boundaries are shown in
figures 1.57 and 1.58. The model assumes elastic behavior within the state boundary
surfaces and plastic behavior when the surfaces are traversed. Yield can be produced by
an increase in p´´, an increase of q, a reduction of s, or by changing p, q, and s
simultaneously in any suitable fashion (fig. 1.58). If the soil state is within the LC
Figure 1.57—General form of p´´-q-s state boundaries proposed for an unsaturated soil. For a soil
initially at stress state A, yield can be produced by an increase in p´´ (isotropic loading path ABC), an
increase of q (shearing path ADE), a reduction of s (wetting path AFG), or by changing p´´, q, and s
simultaneously in any suitable fashion (from Wheeler and Sivakumar, 1995).
(a) (b)
Figure 1.58—Idealized state boundaries for an unsaturated soil in (p´´, q, s) stress space proposed by
Alonso et al. (1990): (a) Projected yield surface and CSL in q-p´´ plane; (b) Yield surfaces in s-p´´ plane.
(loading collapse) state boundary, wetting produces elastic swelling, whereas if the soil
state is on the boundary wetting leads to a much larger plastic compression (collapse). At
zero suction (i.e., saturation) the state boundary surfaces reduce to the Modified Cam clay
model for saturated soils. Satisfactory agreement was found between computed behavior
and results from suction-controlled laboratory tests on a low plasticity compacted kaolin
(which does not swell extensively) and a sandy clay. In subsequent papers, the authors
showed that the model was able to reproduce many observed features of behavior of
moderate plastic to non-plastic compacted soils, based on their own data as well as other
data published in the literature (i.e., Gens and Alonso, 1992; Josa et al., 1992; Alonso et
al., 1993).
Maâtouk et al. (1993) presented data for isotropic and anisotropic compression tests
and controlled suction triaxial shear tests on a fairly dry collapsible silt, under essentially
constant water content conditions. The data of Maâtouk et al. (1993) are of particular
interest, since unlike the majority of geotechnical engineering studies, they conducted
tests with specimens formed from a natural soil at void ratios leading to a soil with
properties similar to that of a natural loess. Their data generally supported the Alonso et
al. (1990) model. Under isotropic conditions, soil behaved elastically on the left side of
the LC yield curve and plastically on the right. The size of the yield curve increased with
suction. They found that the shape of the yield curve on the q vs. p´´ plane was very
different from that proposed (i.e., based on modified Cam clay), and furthermore, the
shape was different at different suctions.
Wheeler and Sivakumar (1995) proposed a critical-state model for unsaturated soils
(fig. 1.59). The model is qualitatively similar to that of Alonso et al. (1990) but differs in
mathematical details and in some of the underlying assumptions. In the early
development they proposed a very general framework involving five state variables (p´´,
q, s, v and w) (Wheeler and Sivakumar, 1993). In more recent work, Wheeler and
Sivakumar (1995) suggested that water content should not be treated as a state variable,
but that changes in water content could be calculated using a flow rule (analogous to the
Figure 1.59— Projected yield surfaces, CSL and NCL on q-p´´ and v-ln p´´ planes, and LC on s- p´´ plane
as proposed by Wheeler and Sivakumar (1995).
treatment of deviatoric strain εs in critical-state models). They also proposed using

specific water volume vw = 1 + wGs, which is the work conjugate to suction, instead of w
to describe the amount of water within the soil. Experimental data to support the
proposed state boundary relationships were presented for samples of unsaturated
compacted kaolin taken through five different stress paths, namely isotropic consolidation
followed by shearing under conditions of: (1) constant v and constant s; (2) constant p´
and constant s; (3) fully drained and constant s; (4) constant water content and rising s;
and, (5) constant p´´ and falling s. Wheeler and Sivakumar (1995) established the
existence of a unique CSL for any given constant value of suction (fig. 1.60). The
critical-state values of v and q from three constant water content (varying suction) tests
provided additional evidence that the critical-state relationships were independent of
stress path. Shown in figure 1.60c are the critical-state lines deduced from constant
suction shear tests and the locations relative to these lines of the three critical-state data
points from the constant water content tests. Discrepancies between experimental and
model predicted stress paths for constant-volume/constant-suction shearing indicated that
the assumption of elliptical (similar to modified Cam clay) yield surfaces was probably
not correct (fig. 1.60d). In contrast, predicted stress paths for other constant suction stress
paths agreed very well with experimental data. The development of shear strain was also
predicted with reasonable success. The pattern of swelling and collapse observed during
wetting (reduction of s) and the compression observed during isotropic loading (increase
of p´´) supported the existence of an LC yield curve in p´´- s space (fig. 1.59).
Furthermore, the shape of the LC curve derived from the NCL relationship was entirely
consistent with the pattern of behavior observed during wetting and loading. Although
the model now required only four state variables, it involved three elastic parameters plus
six critical-state parameters, which are functions of suction. This compares with two
elastic parameters and five critical-state parameters (obtained once) for saturated soils.
The number of soil parameters would increase if water content (or vw) relationships were
to be incorporated. Wheeler and Sivakumar (1995) recognized the necessity for reducing
the number of parameters that need to be evaluated experimentally for practical
application of the model.
It appears then that the models proposed by Alonso et al. (1990) and by Wheeler and
Sivakumar (1995) can satisfactorily describe a wide range of unsaturated soil behavior,
although there is still insufficient evidence to verify the LC surface for highly swelling
soils. Their application to agricultural machinery problems is limited since constant water
content-varying suction paths cannot be predicted using the current formulations. The
studies that will be reviewed now, all present models that incorporate soil moisture status
in the formulation.
Toll (1990) presented data to support a critical-state model formulated using net stress
and matric suction as independent stress state variables. Degree of saturation was
included as a deformation state variable in the formulation, and indirectly as an indicator
of soil fabric or structure. Thus, the proposed framework was put forward in terms of
three stress variables (p´´ = p - ua, q, s = ua - uw) and two volumetric variables (v and Sr).
Each state boundary was resolved into components due to matric suction and mean net
stress; for example, the critical-state line on the q-p´´ plane was expressed as:
q = Ma(ua - uw) + Mw(p - ua) (1.117)

(a) (b)
(c) (d)
Figure 1.60—Deviatoric stress plotted against mean net stress at critical-states for constant-suction shear
tests: (a) Different stress paths for a test conducted at s = 200 kPa define a unique CSL.; (b) Critical-state
lines obtained for different values of suction; (c) Critical-state values of deviatoric stress for constant
water content tests; (d) Comparison of predicted and experimental test paths for constant-v/constant-s
shear tests (from Wheeler and Sivakumar, 1993, 1995).
Toll (1990) verified the model using data from conducted constant water content
triaxial tests. Model parameters were shown to vary in a well-behaved, consistent manner
with degree of saturation. Wheeler (1991) reanalyzed Toll’s (1990) data and showed that
if the critical-state data were expressed in the form:
q = ƒ(ua - uw) + M(p - ua) (1.118)
the parameter M could be taken as essentially equal to the saturated value, and that the
intercept ƒ varied in a systematic, although complex, way with soil suction. This
approach also eliminated the need for one of the state variables. There is some
uncertainty in interpreting Toll’s data since experiments were performed with specimens
compacted over a wide range of water contents and bulk densities so that different
structures were formed.
A few researchers have accepted that matric suction must be included in constitutive
formulations, but have proposed alternative stress equations in attempts either to simplify
the formulations and/or to account for the effect of soil structure. Kohgo et al. (1993a, b)
presented an elastoplastic model for unsaturated soils within the framework of
unconventional plasticity theory in which a smooth transition between elastic and plastic
behavior is produced using subloading surfaces (Drucker, 1988). The mechanical
behavior was related to three empirical stress equations, two of them suction related.
Their formulation assumed that the critical-state line for unsaturated soils had a zero
intercept with the q-axis, similar to that for saturated soils, and as a result their predicted
failure stresses were consistently lower than experimental values. Model predictions
agreed well with experimental data for soil collapse upon rewetting and an increasing
suction consolidation test for a Lower Cromer Till.
Bolzon et al. (1996) presented a model for unsaturated soil based upon generalized
plasticity and critical-state concepts. The authors used both Bishop’s (1959) stress
equation (rather than net stress) and matric suction as stress variables. They assumed that
the parameter χ was equal to the degree of saturation so that the effect of soil type was
reflected in the Bishop stress equation. The model implicitly uses the relation between the
degree of saturation and suction. They validated aspects of the model by comparing
predictions with some of the published data for unsaturated kaolin. The model was able
to produce qualitative behavior of soil under constant suction isotropic compression (data
of Alonso et al., 1990) and of swelling and collapse under suction decreases (data of
Escario and Sáez, 1973).
There are fundamental drawbacks to the formulations of Kohgo (1993a) and Bolzon et
al. (1996). Material properties are introduced into the stress variable equations in the
former, and a “deformation state” variable (i.e., Sr) is incorporated into a “stress” variable
in the latter. Wheeler and Sivakumar (1992), Fredlund (1995) and Wulfsohn et al. (1996)
argued that constitutive relationships between soil water status and soil suction should be
incorporated in general models for unsaturated soils that include varying suction stress
paths. In the absence of a satisfactory way to quantify soil structure the soil-water
characteristic (i.e., water content versus suction when p ´´ = 0) serves also as an indicator
of soil structure. The nature of volume change constitutive surfaces, e = ƒ[(σ - ua), (ua -
uw)], is well established (see Fredlund and Rahardjo, 1993a). It remains to incorporate
these into the critical-state soil mechanics framework.
Variation of Critical-state Boundaries
Water content, soil suction, and microstructural state are all important transient
variables in unsaturated soils. These must be considered in any critical-state model that is
formulated for agricultural soils. It is important to distinguish between water content,
which is a deformation state variable, and matric suction, which is a stress state variable.
During constant water content deformation for example, the change in matric suction is
more significant than the change in pore-water content, especially under high stresses.
Soil-water relations are in turn closely tied to the soil microstructure.
Variation with Water Content. Larson et al. (1980) determined uniaxial compression
characteristics of soils with initial pore-water pressures of -5 to -100 kPa, while
monitoring changes in pore-water pressure during compression. As water content or
degree of saturation increased, the slopes of the compression lines remained fairly
constant (corresponding to constant λ32) for a given soil texture, while the curves
translated to the left along the stress axis (fig. 1.61), corresponding to a decrease in the
intercept Ν.
Leeson and Campbell (1983) examined the effect of water content and degree of
saturation on critical-state parameters based on laboratory compression tests on a loam
and a sandy loam soil with water contents between 5% to 30% (corresponding to degrees
of saturation of 10% to 40%). The slope of the normal consolidation line (λ) increased
linearly, and its intercept (Ν) decreased, with increasing degree of saturation and water
content. The position of the critical-state line also varied as water content changed. They
concluded that as the soil dries the critical-state line rises and the state boundary surface
shifts away from the origin (fig. 1.52). The state boundary is shown in figure 1.52 to
increase in size as the soil dries and strength increases.
Hettiaratchi (1987) derived critical-state parameters for a silty sand, a loam, and a heavy
clay soil of varying water contents and soil microstructures (fig. 1.62). The rapid decrease
in parameters Γ, λcs, and Μ for soils with high clay contents was shown to coincide with
a steep increase in soil total cohesion, c, indicating that water tension and bond formation
play a crucial role in the position of the CSL. Hettiaratchi and O’Callaghan (1985) and
Hettiaratchi (1987) proposed that the variations in critical-state parameters with water
content could be represented by the location of a “pivot point” in the v-p plane about
which the critical-state and normal compression lines “rotate” and change in slope as the
water content changes (fig. 1.63).
Figure 1.61—Effect of water content on bulk density (1/BWV) versus applied stress relations obtained
from uniaxial compression tests on a Waukegan silt loam (from Gupta and Allmaras, 1987).
32
The principal difference between the “virgin compression line” determined in oedometer and the NCL is the
conventional use of logarithms to base ten in the former. The slopes of the two sets of lines are then simply
related by: C = λ ln 10 ≈ 2.30λ, where C is the slope of the virgin compression line (i.e., the compression
index).
Figure 1.62—Variation of CSL parameters for remolded (a,b,c) and cemented (d,e,f) soils for different
water contents for an agricultural loam (), a heavy clay soil ( ) and a silty sand ( ⋅  ⋅)
(from Hettiaratchi, 1987).
O’Sullivan et al. (1994) derived critical-state parameters for remolded specimens of a

sandy loam, a clay loam, and a clay over a range of water contents. The normal
consolidation lines tended to be linear on the v-ln p plane up to a degree of saturation of
about 85%. At higher saturation, the soils generally became incompressible but highly
deformable, and the CSL was almost identical with the NCL. The slopes of the projected
critical-state lines λcs were slightly higher than those of the normal compression lines λ.
The slope of the CSL on the q-p plane, Μ, remained fairly constant for normalized water
contents up to about 70% to 80%, and then decreased rapidly (for the clay soils in
particular), indicating a loss in shear strength. Data from the three soils indicated that the
projection of the CSL on the q-p plane was a straight line through the origin and that both
the normal compression and critical-state lines pivoted about a point as water content
varied.
Petersen (1993) presented the variation of critical-state parameters with water content
for a loam and a sandy loam (fig. 1.64). For both soils, soil compressibility λ increased in
value with increasing water content up to about 8% to 10%, and decreased with
increasing water content for specimens wetter than approximately 18% for these soils.
Over the intermediate range of water contents, λ did not vary much in the two soils. The
variation of other parameters with water content also seemed to change over these limits.
In general, the slopes of the CSL and NCL lines were different, except in wet soils.
Similar to the findings reported by Hettiaratchi and O’Callaghan (1985) the parameters Γ
and M decreased rapidly with water content. The fairly high value of q0 over a wide range
of intermediate water contents for the sandy loam suggests that the drop in “shear
strength” of the sandy loam soil with increasing water content would be less apparent
Figure 1.63—Proposed pivotal behavior of the CSL and NCL in state space. The + arrows indicate
rotations which increase the slopes of the lines (from Hettiaratchi, 1987).
than at first indicated. The results did not consistently support the existence of a pivot
point.
O’Sullivan et al. (1994) plotted the variation of critical-state parameters versus a
“normalized water content” given by the ratio of the gravimetric water content to the
penetrometer plastic limit (see Campbell, 1976), so that parameters for different soils
could be compared. Similarly, Kirby (1991) found that the water content expressed as a
fractional distance between the liquid and plastic limits explained the variability of
critical-state parameters better than the water content itself.
Effect of Suction. As would be expected, the nature of the effect of suction increases
on variations in critical-state parameters to a large degree parallel that of decreasing water
content. The size of the yield surface increases with suction (Maâtouk et al., 1995;
Adams, 1996). The preconsolidation pressure also increases with increasing suction while
soil compressibility λ(s) decreases as the soil is stiffened (Matyas and Radhakrishna,
1968; Alonso et al., 1993; Wheeler and Sivakumar, 1993; Cui and Delage, 1993;
Maâtouk et al., 1995). The value of N(s) increases with suction. In their study, Wheeler
and Sivakumar (1993) found that the value of λ(s) fell sharply as suction was reduced to
zero.
Most studies have found that when the CSL is expressed in the form q = q0 + Mp´´, the
slope M does not vary significantly with suction and can be taken as essentially equal to
the saturated value, while the strengthening effect of suction is reflected in increasing
values of the intercept (Alonso et al., 1990; Wheeler and Sivakumar, 1995; Adams,
1996). In contrast, data presented by Delage et al. (1987) and Maâtouk et al. (1995) for
some highly frictional soils showed Μ as a strong function of suction over a range of net
stresses.33 As net stress increased the effect of suction on strength declined.
33
These findings are discussed further in the section Choice of Stress State Variables.
Figure 1.64—Measured critical-state parameters versus gravimetric water content for a Tåstrup sandy
loam (O), and a Mårum loam ( •). Note that λ* = λcs (from Petersen, 1993).
Influence of Soil Structure. Cemented and remolded laboratory specimens were used
by Hettiaratchi (1987) and Hatibu and Hettiaratchi (1993) to represent limiting
microstructural states in which a field soil could exist at a given water content. Remolded
microstructures are formed when an unsaturated soil is subjected to a mechanical
disturbance that rearranges the sizes and distribution of the pore contents (Hatibu and
Hettiaratchi, 1993). A cemented microstructure is obtained when an initially loose soil
dries out while the particles maintain contact and a continuity of the air phase is formed.
In-situ soils would exist in conditions intermediate between these states, presumably
having mechanical properties within limits established with remolded and cemented
specimens. It should be noted, however, that a range of different soil fabrics will exist
even within these categorizations of soil “structure,” i.e., different remolded specimens
compacted at different water contents and bulk densities will not have identical
structures, whereas “cemented” specimens initially formed by compaction at the same
water content and bulk density, but then allowed to wet or dry to different water contents
(suctions), may produce very similar structures. This will not be true if there is large
hysteresis of the soil-water characteristic curve, and significant soil swelling or
shrinkage.
Critical-state parameters are strongly influenced by the method of specimen
preparation. Laboratory specimens of the same structure are produced by compacting
them at the same bulk density and water content. Mitchell (1993) and Fredlund and
Rahardjo (1993a) caution that different structures may be produced when the same soil
(texture) is compacted at the same bulk density but different water contents. Peterson
(1988) found that specimens compacted at water contents drier than the optimum water
content34 were stronger, more brittle, and tended to swell, compared to specimens
compacted wet of optimum which were more ductile and tended to consolidate. Based on
experimental data for remolded soil specimens, Hettiaratchi and O’Callaghan (1985)
concluded that there exists a “sensitive” water content range beyond which a totally
different volume change behavior occurs. Soils remolded at higher water contents would
lie in the portion of state space wet of critical and deform plastically along the Roscoe
surface, whereas the same soil texture remolded at a lower water content might lie in dry
of critical-state space and fail as a brittle material. Different methods of compaction (e.g.,
impact, static, kneading, or vibratory) also result in significantly different fabrics in both
sands and clays (Mitchell, 1993). Peterson (1988) found that specimens molded by static
or dynamic compaction were stronger, more brittle, and tended to swell compared to
those prepared by kneading compaction.
Petersen (1993) formed specimens by static compaction from air-dried samples. The
specimens were wetted to zero suction (i.e., saturation) and then brought to the desired
water content by drying on sand tables or ceramic plates. This led to the formation of a
cemented microstructure, similar for all the water contents. Similarly, Wheeler and
Sivakumar (1992, 1993) used specimens of different water contents but similar
structures. Petersen (1993) expressed behavior in terms of total stresses and showed the
variation of critical-state parameters with water content. Wheeler and Sivakumar
considered matric suction as an independent stress and presented the variation of critical-
state parameters with matric suction. Non-zero intercepts of the CSL with the q-axis were
established in both these studies. These trends agree with the observations by Wheeler
and Sivakumar (1993) for a cemented clay soil.
Bailey and VandenBerg (1968), Hettiaratchi (1987), Kirby (1989, for two soils) and
O’Sullivan et al. (1994) presented results for remolded specimens formed at different
water contents. Specimens were thus unidentical in structure and could be considered to
represent “different soils.” The critical-state lines determined using a total stress analysis
had intercepts of zero on the q-p plane (i.e., q = Mp) and M increased as the soil water
content decreased (Hettiaratchi, 1987; O’Sullivan et al., 1994).
Adams (1996) and Adams and Wulfsohn (1997) attempted to resolve the seeming
discrepancies reported in these earlier studies. Three sets of triaxial specimens formed
from a sandy clay loam were tested. Two sets were prepared to produce a cemented
structure. A third set of remolded unsaturated specimens was prepared to produce a
34
The water content at which the bulk density of remolded samples is maximum, with a fixed intensity of
compaction effort (see fig.1.67).
different structure from the first two sets. Comparison of data from the test series with
similar structures but different water contents showed that the slope of the CSL on the q-
p plane, M, remained constant (fig. 1.65a). In contrast, when comparing data from the test
series having both different structures and water contents, the critical-state line could be
said to pivot as water content changed, although it still had a non-zero intercept with the
q-axis (fig. 1.65b). Adams (1996) concluded that the method of specimen preparation
(i.e., initial soil structure) affected the manner in which critical-state parameters varied
with water content. This is perhaps most vividly illustrated by the data of Hettiaratchi
(1987) for remolded and cemented specimens (fig. 1.62).
(a)
(a)
(b)
(b)
Figure 1.65—Effect of soil structure on critical-state lines of a sandy clay loam in the q-p´´ plane: (a)
Data for "identical" specimens compacted wet of optimum (type W) at two matric suctions [ua - uw = 0
(saturated) and 50 kPa (unsaturated)]; V = constant water content tests, U = consolidated drained tests,
= consolidated undrained tests; (b) Data for specimens with dry of optimum structure (open points)
and wet of optimum structure (closed points) for a suction of 50 kPa (from Adams and Wulfsohn, 1997).
Choice of Stress State Variables. Comparison of the published data suggest that the
variation of critical-state parameters and the orientation of the critical-state line with
water content may depend on the manner in which stress is quantified as well as the
influence of soil microstructure.
Hettiaratchi (1987), Wheeler and Sivakumar (1995), Petersen (1993) and Adams and
Wulfsohn (1996), testing cemented specimens, all obtained critical-state lines with
intercepts greater than zero on the q-axis. Hettiaratchi (1987) and Petersen (1993)
expressed stresses in terms of total stresses, while Wheeler and Sivakumar (1995) and
Adams and Wulfsohn (1996) used independent stress state variables. The former studies
found M to be generally a strong function of water content,35 while the latter two studies
found little variation in M from fully saturated (s = 0) to unsaturated (s > 0) conditions.
Several researchers have suggested that φ (to which M is related) and M may be functions
of suction for some soils (Delage et al., 1987; Escario and Juca, 1989; Maâtouk et al.,
1995). However, in its true sense as an intergranular friction, φ should not vary with
suction. Drumright (1989) showed that when the variations in net applied stress were
accounted for, differences in φ with respect to suction were considerably reduced. If
suction, which acts as an omnidirectional normal stress, is not accounted for in
determining normal stresses, then the apparent value of φ will change. It is therefore
possible that the significant variations in M with water content found by Hettiaratchi
(1987) and Petersen (1993) even for “identical” (i.e., the cemented) soil specimens arise
due to the use of a total stress analysis. This deserves further investigation, as there are
advantages to a critical-state formulation in which the CSL can be shown to vary with
suction in a very simple manner (e.g., in which M can be assumed to be constant and q0
increases with suction).
Only a very limited range of soil textures has been investigated using an independent
stress state variable approach. Moreover, with few exceptions (Adams, 1996; Kohgo et
al., 1993b; Maâtouk et al., 1995), researchers have not focused on “natural” soils of
relevance to agricultural researchers. Verification on a wider range of agricultural soil
textures, structures, and suctions is needed.
Transition from Ductile to Brittle Behavior
Various studies have shown that whether an unsaturated soil undergoing loading will
fail in a brittle or ductile manner depends on the current state of the soil in state space and
the manner of loading. Thus, the mode of failure of the soil depends on a combination of
various factors including its initial void ratio or bulk density, suction and water content
(which in turn are closely related to the soil structure), and the confining pressure.
Spoor and Godwin (1979) examined the critical-state behavior of three agricultural
subsoils at different water contents and confining pressures in a series of undrained
triaxial compression tests. At low confining pressures and high water contents (i.e., low
matric suctions), the soils failed in compression (wet side of critical-state line). Brittle
failure was observed on the other hand at low water contents (higher suctions). The
transition from brittle to ductile failure occurred between suctions of 35 kPa and 100 kPa.
Hatibu and Hettiaratchi (1993) carried out a series of triaxial compression tests to
investigate the failure modes of unsaturated specimens of a sand, a loam, and a heavy
clay soil at different water contents, confining pressures, and soil structures. Soil texture,
microstructural state, confining pressure, and water content were all found to influence
35
The silty sand results presented by Hettiaratchi were an exception. For this soil critical-state parameters
remained fairly constant over the entire range of water contents and structures tested (see fig. 1.62).
the soil deformation and failure behavior in a consistent manner (Hatibu and Hettiaratchi,
1993). The confining pressure at which the brittle-ductile transition takes place was found
to be linearly related to water content for a given soil condition. However, the coefficients
of the relationship varied significantly depending on the method used to distinguish
between brittle and ductile failure, i.e., one based on a visual classification (after Price
and Farmer, 1979) versus on the basis of the shapes of the stress-strain curves. They
proposed a “transition surface” separating zones of brittle and ductile failure, as a
function of water content, confining pressure, soil texture and soil structure (fig. 1.66).
For states below the transition surface external loading will cause brittle fracture and
deformation; whereas, states above the surface will behave in a ductile manner. Practical
application of the model, however, requires a way to quantify soil microstructure.
State-Boundaries for Unsaturated Soil
The combined research to date on critical-state soil mechanics for unsaturated soils
can be summarized by the state boundary relations proposed in table 1.3. The equations
given for unsaturated soils reduce to those for saturated soil when s = 0 (e.g., ua = uw so
that p´ = p´´, q0(0) = 0, λ(0) = λcs(0) = λ). Various aspects of the model have been
verified by the experimental research. Some aspects may change significantly depending
on future research findings. For rigor, the equations are expressed in terms of suction;
however, the model can be used to explain soil behavior in terms of total stresses. Note
that for most agricultural situations, the pore-air can be assumed to be interconnected up
to the surface (i.e., ua = 0 kPa, gauge) so that p´´ can be equated to the total stress p.
Figure 1.66—Development of a "transition surface" demarcating conditions conducive to brittle or

ductile behavior. Schematic representation of microstructural state axis: (a) Fully remolded state (R); (b)
Partly remolded state; (c) Fully cemented state (C); (d) Influence of clay content on the configuration of
the transition surface (adapted from Hatibu and Hettiaratchi, 1993).
The yield surface equations are of the form shown in figures 1.57 to 1.59. For a
complete description of unsaturated soil behavior, strength and volume change
constitutive surfaces for water content are also required. It is still not possible to predict
state paths for constant water content, varying suction conditions using the model
presented in table 1.3. Under environmental stresses (e.g., precipitation, evaporation, and
evapotranspiration) the stress-path will traverse specific volume–water content and
suction–water content constitutive surfaces as well.
The constant suction yield surface (1) assumes constant M for all suctions (fig. 1.57a).
Wheeler and Sivakumar (1995) presented a similar expression based upon the modified
Cam clay ellipse which is more general in that it allows for M that varies with suction
(fig. 1.59):
2
q 2 − M (p"+ p s )(p 0 − p" ) (1.119)
*
where
M (s)p x + q 0 (s)
M =
∗ p0 − px
and px = (p0 + ps)/2 is the value of p´´ at which the CSL intercepts the yield surface
(which can be expressed completely in terms of critical-state soil parameters Γ, λcs, and
κ).
Studies have indicated that the shape of yield surfaces for some unsaturated soils in
the q-p´´ plane (specifically, the Roscoe surface on the wet side of critical) may be very
different than the modified Cam clay ellipse (1) (e.g., Bailey and Johnson, 1994; Wheeler
and Sivakumar, 1995), and are also not the same for different suctions or water contents
(e.g., Maâtouk et al., 1995). Nevertheless, expression (1) in table 1.3 (or the more general
equation 1.119) is retained because modified Cam clay type formulations are so widely
used, largely because of their simplicity. Clearly, there is need for further research to
identify the nature of yielding of unsaturated soils.
Table 1.3. State boundaries for saturated and unsaturated soils.
Name Unsaturated soil Saturated soil

2 2
Cam clay yield surface (q-p´´) (1) q − M (p"+ ps )(p 0 − p" ) q − M 2 p ( p 0 − p)
2
Loading-Collapse yield surface (2) (λ (s) − κ) ln(p 0 ) = —

(p´´-s)
(λ(0) − κ) ln(p 0 (0))
+ N(s) − N(0)
+ κ s ln(s + p atm )
Normal Compression Line (3) q=0 q=0

(4) v = Ν(s) - λ(s) ln p´´ v = Ν - λ ln p´
Critical-state Line (5) q = q0(s) + M(s) p´´ q = M p´
(6) v = Γ(s) - λcs(s) ln p´´ v = Γ - λ ln p´
Unloading-Reloading Line (7) v = vκ(s) - κ(s) ln p´´ v = vκ - κ ln p´
p´ = effective stress = p - uw s = matric suction = ua - uw.
p´´ = mean net stress = p - ua Note that ps ≥ 0.
When evaluating intercept parameters, Ν(s), Γ(s) and vκ(s), the selection of a
reference mean stress of 1 kPa may introduce errors, especially if p´´ = 1 falls outside the
range of experimental data. Alonso et al. (1990) and Wheeler and Sivakumar (1993)
expressed all relationships for the state boundaries in terms of stress ratios; for example,
equation (4) for the CSL becomes:
 p" 
v = Γ(s) - λ cs (s) 
 pr  (1.120)
where pr is a reference pressure at which v = Γ. Wheeler and Sivakumar (1995) used

atmospheric pressure as the reference pressure, while Adams (1996) used the
precompression stress of the soil.
Under isotropic consolidation with pressures within the ranges of interest to
agricultural applications, gravimetric water content will change, but by an insignificant
amount. Thus the expressions (3 and 4) given in table 1.3 suffice to describe the NCL
hyperline.
The critical-state line has been shown to be unique for a given suction (Wheeler and
Sivakumar, 1995; Adams and Wulfsohn, 1997b) so that a relationship of the form of
given in expression (5) presented in table 1.3 applies. Under constant water content
conditions (e.g., when an agricultural soil undergoes short-term loading) and “low” stress
regimes, suction also remains essentially constant. This explains why a total stress
approach has proven satisfactory for many agricultural machinery applications. In
contrast, under “high” stress regimes the suction changes as the soil is loaded and pore-
water pressures build up, even under constant water content conditions (Wulfsohn et al.,
1998). The strength of the soil is strongly influenced by matric suction and, consequently,
uniqueness of the CSL cannot be established for a fixed water content (Adams and
Wulfsohn, 1998). When working with total stresses, the NCL and CSL appear to “rotate”
in state space and the intercept of the CSL strength envelope may pass through the origin;
however, this is probably a reflection of changes in the critical-state boundaries as a
“new” soil structure is formed. It becomes important to express the stress state in terms of
the independent stress state variables. The critical stress delineating constant suction and
varying suction conditions seems to be directly related to the in-situ suction of the soil
(Wulfsohn et al., 1996, 1998). For a given soil this matric suction will correspond to
some unique gravimetric water content. Observations by Kirby (1991), O’Sullivan et al.
(1994), Thangavadivelu (1994) and Adams (1996) indicate that the plastic limit (using
Atterberg consistency limits) or the optimum water content may provide an estimate of
this water content.
Determination of Critical-State Parameters

Critical-state parameters are typically determined in laboratory tests with triaxial
testing being the most popular method. Saturated state critical-state parameters can also
be estimated from Atterberg limits. Various empirical relationships for rapidly estimating
saturated critical-state parameters from simple consistency tests have been proposed
(Kenny, 1959; Schofield and Wroth, 1968; Wood, 1990; Atkinson, 1993). Hettiaratchi
(1987) highlighted the need to similarly develop simple experimental techniques to test
unsaturated soil samples. Hettiaratchi and Lang (1987) presented nomograms for
estimating Mohr-Coulomb strength parameters from indentation tests. Presently there is

no rapid method of estimating unsaturated critical-state parameters.
Testing Equipment. Many researchers have analyzed the variation of critical-state
parameters with water content. This means that numerous tests must be conducted over
ranges of practical water contents. If the effect of structural differences is to be evaluated,
it will be necessary to compare the behavior of samples compacted at the same initial
void ratio and different water contents, and tested under identical initial conditions and
stress paths (Alonso et al., 1993; Adams and Wulfsohn, 1997). Most researchers have
used various modifications of the triaxial apparatus (Leeson and Campbell, 1983;
Hettiaratchi et al., 1992; Petersen, 1993; Wulfsohn et al., 1998). On the basis of
principles of critical-state soil mechanics and triaxial data reported in the literature,
Koolen and Kuipers (1983) suggest that soil bulk density varies mainly with σ1, while σ3
has only a small effect. Thus, uniaxial compression tests of loose soil should be sufficient
for determining volume change parameters, while the triaxial apparatus is not required.
Larson et al. (1980) modified the conventional oedometer apparatus to monitor changes
in pore-water pressure during testing. Kirby (1989) has shown that it is possible to use a
combination of one-dimensional compression tests in an oedometer, and direct shear box
tests (in a modified apparatus) to determine critical-state parameters. Greacen (1960)
used a ring shear apparatus. Unlike the triaxial test, use of the shear box makes it possible
to simulate rotation of the principal stresses (Kirby, 1989; Shaw and Brown, 1986).
One advantage of using the triaxial apparatus to determine critical-state parameters is
that volume change and shear strength behavior are integrated. Hettiaratchi (1987)
employed fully drained slow triaxial compression tests. O’Sullivan et al. (1994) and
Kirby and O’Sullivan (1997) used a “constant cell volume” triaxial apparatus (described
by Hettiaratchi et al., 1992) to determine critical-state parameters of agricultural soils
under constant water content conditions. Adams and Wulfsohn (1996) used a suction
controlled triaxial apparatus (Wulfsohn et al., 1998) and constant-water content tests to
determine critical-state parameters. Fredlund and Rahardjo (1993a) gave a detailed
description of triaxial and shear box equipment and testing procedures for unsaturated
soils. Wulfsohn et al. (1998) reviewed triaxial equipment and procedures relevant to
agricultural applications.
Specimen Preparation. Laboratory tests may be performed on undisturbed field
samples or on reconstituted specimens prepared in the laboratory. Undisturbed specimens
represent field conditions better than remolded specimens, but they can be difficult to
obtain. Undisturbed field samples shrink when soil moisture is lost, and there may be
difficulty with stones, vegetative cover, and non-homogeneity in the soil. During testing
of undisturbed specimens, chances of uneven stress distribution and pore pressures within
the soil are more likely. Most critical-state formulations assume the soil is a
homogeneous material. Thus, for fundamental investigations into the nature of soil
behavior (in which the specimen is meant to represent an “element” of soil), reconstituted
specimens are generally used. Wheeler and Sivakumar (1993) found that the
heterogeneous soil structure produced by the compaction process is retained to some
degree after shearing to critical-state, and had a major effect on the specific volume at
critical-states but less so on the deviatoric stress. Nevertheless, since it is practically
impossible to reproduce “natural” structures by reconstituted specimens, there are many
situations in which the use of undisturbed specimens is preferred; for example, when
comparing model predictions with field data or for the evaluation of the variability of soil
properties.
When using reconstituted specimens, only specimens with identical structures or

fabrics should be used to determine unique shear strength parameters for the soil, i.e.,
specimens with the same texture and stress history. If they do not have the same structure,
they may produce different strength parameters. The fabric immediately following
compacting depends on the strength of preexisting structural units, the compaction
method, and the compaction effort (Mitchell, 1993) (fig. 1.67). Specimens prepared by
compaction must be compacted at the same initial water content and bulk density to be
considered as an “identical soil” (Fredlund and Rahardjo, 1993a). Specimens can then be
brought to the desired matric suction or water content. Petersen (1993) brought
specimens to the desired suction using a sand table and ceramic plates. Cui and Delage
(1993, 1996) imposed an initial matric suction by enclosing specimens in semi-permeable
membranes and bringing the water content to equilibrium with a polyethylene glycol
solution of known osmotic pressure (see Zur, 1966; Williams and Shaykewich, 1969).
Adams (1996) estimated matric suction of triaxial specimens by measuring the matric
suction in a thin layer of a specimen using a null-pressure plate apparatus.
Parameter Estimation. When laboratory testing is used to determine model
parameters, a minimum series of test paths must generally be followed. The use of
model-based inverse methods can reduce the amount of tests required (described below).
The frictional constant Μ can be obtained from shear tests on samples first
consolidated isotropically to different consolidation pressures, and then sheared to failure
under a constant confinement pressure. By plotting q versus p values at failure, the slope
of the best-fit straight line is taken as Μ. Alternatively, if the principal stresses at failure
are obtained, then the internal friction angle φ can be obtained from the geometry of the
Mohr’s circle plot and Μ determined using equation 1.73. The intercept value and slopes
in v-p´´ space (Γ and λcs) are obtained from plots of specific volume or void ratio versus
the natural logarithm of p. This requires knowledge of final volume specimens at the end
of tests.
Figure 1.67—Structure of clay specimens compacted at different dry densities and water contents (from
Lambe, 1958). Specimens compacted at A, B, and C represent “different” soils.
The slopes of the NCL and url lines (λ and κ, respectively) and the parameters N and
p0 are obtained from isotropic consolidation tests or from oedometer tests. If the
compression line in v-ln(p´´) space is curved, one has to choose the slope appropriate to
the stress range relevant to the problem to be analyzed. The value of Poisson’s ratio
ν may be obtained from strain measurements in triaxial tests. It is common to assume a
constant value of 0.3 for Poisson’s ratio, although doing so can be a source of much error.
Examples of the determination of Cam clay parameters from conventional test data are
presented by Desai and Siriwardane (1984) and Atkinson (1993). Britto and Gunn (1987)
discuss the choice of critical-state parameters for use in finite element analyses. Both
these references discuss applications for saturated soils. Nevertheless much of their
presentations are relevant for unsaturated soils as well.
Kirby et al. (1998b) proposed an inverse method36 to derive the slope of the critical-
state line, M, the slope of the normal compression line, λN, and the elastic modulus, E,
from constant cell volume triaxial tests conducted entirely in the range of soil plastic
deformation. Kirby (1998) presented quasi-analytical solutions of the Cam clay model for
uniaxial compression, conventional triaxial compression, constant volume triaxial, and
constant load and constant volume shear box tests. These formulations can be used for
parameter estimation using inverse methods, for model evaluation and for testing finite
element solution packages.
Researchers at the University of California, Davis, have developed an inverse
methodology which uses a model based response surface to determine in-situ parameters
from results of simple sinkage plate tests. The method has to date been successfully
implemented for modified Cam clay and Drucker-Prager cap models (Rubinstein et al.,
1994; Upadhyaya and Sime, 1994, 1995).
Critical-state Parameters for Unsaturated Soils. Table 1.4 gives ranges of critical-
state parameters determined for unsaturated soils in some studies. The significant effect
of increases in compressibility and strength of soils with desaturation (increasing s) is
reflected in the λ, λcs, and q0 parameters. Most agricultural soil problems involve large
plastic deformations. In such cases the Cam clay parameters λ, Μ, and Ν play a major
role in controlling deformation, whereas, elastic parameters (ν, G, κ) have very small
effect (Hettiaratchi and O’Callaghan, 1985; Sime and Upadhyaya, 1995). Leeson and
Campbell (1983) show a much greater range of url recompression gradient values than
have been reported by others. A similar range of values was obtained for a sandy loam
soil as well. The high values of reloading κ were obtained for soils that were initially
3
very loose (v > 2.5). At a specific volume of 1.5 (dry bulk density of 1.75 Mg/m ), the
recompression lines of both soils seemed to approach zero. Maâtouk et al. (1995) found
that compressibility λ(s) decreased as the specimen initial void ratio decreased.
Larson et al. (1980) grouped agricultural soils into four compression categories based
on the average slopes of the virgin compression line determined in a uniaxial
compression apparatus (fig. 1.68). Differences in the compressibilities of the soils
resulted from variations in texture, organic matter content, and types of clay minerals (in
the case of types B and C).
36
Inverse methods determine parameters based on an assumed constitutive model describing material behavior.
Parameter estimation usually involves minimizing the errors between observed and model-predicted responses.
Table 1.4. Typical ranges of experimentally determined critical-state parameters.
Average Initial NCL CSL Elastic

Water Matric Initial Dry
Soil Texture Content Suction Density λ Ν λ cs Γ Μ q0 κ
3 -2 -2 -2
(%) (kPa) (Mg/m ) (×10 ) (×10 ) (kPa) (×10 )
Data of Toll (1991) as analyzed by Wheeler (1991)
[a]
Gravel 0 1.6 - 1.8 — — 0.094 2.43 1.6 0 —
17 - 26 <5 - 540 1.44-1.75 — — (—) (—) 1.5 0 - 230 —
Petersen (1993, 1994)
Sandy loam 2 - 25 — — 8 - 18 2.3 - 2.6 9 - 18 2.0 - 2.6 0.5 - 1.7 1 - 50 1.1 - 1.6
Loam 10 - 27 — — 6 - 22 2.2 - 2.7 14 - 24 2.1 - 2.7 0.5 - 2.2 2 - 33 1.2 - 1.9
Leeson and Campbell (1983)
[b]
Loam 12 - 24 — 0.82-1.43 13 - 30 2.0 - 2.2 — — — — 1 - 16
Chi et al. (1993)
Gravely [c]
16 - 24 — 1.1 - 1.3 4.7 - 5.0 — — — 1.5 - 1.6 — 0.5 - 0.7
sandy loam
[c]
Clay 25 - 26 — 1.3 - 1.5 1.8 - 5.9 — — — 0.9 - 1.2 — 0.2 - 0.5
O’Sullivan et al. (1994)
Sandy loam 7 - 19 — — 5 - 10 2.0 - 2.4 1 - 13 1.7 - 2.4 0.9 - 1.4 — —
Clay loam 14 - 26 — — 5 - 25 1.9 - 2.8 5 - 30 1.9 - 2.8 0.2 - 1.7 — —
Clay 19 - 39 — — 17 - 40 3.2 - 4.4 22 - 45 3.4 - 4.4 0.7 - 1.7 — —
Adams (1996)
[c]
Sandy clay 56 [a] 0 1.2 9.2 1.85 9 1.83 1.4 0 0.1
[c]
Loam 16 - 20 50 1.2 25 - 28 2.6 16 - 26 2.2 - 2.4 1.4 - 1.8 40.3 1.2
Maâtouk et al. (1995)
[a]
Silt 25 0 1.6 3 — 3 1.68 1.6 0 0.7
8 80 - 600 1.2 - 1.5 4 - 13 — (—) (—) 0.6 - 1.6 50 - 450 0.2 - 1.0
Wheeler and Sivakumar (1995)
[a]
Kaolin clay 0 1.2 12.8 2.05 11 2.01 0.8 0 —
— 100 - 300 1.2 18 - 20 2.1 - 2.2 11 - 22 2.0 - 2.1 ~0.93 54 - 122 2.0
— Not reported or not applicable. [a] Saturated condition.
(—) Model equations, and thus parameters, were expressed in a non-conventional form. [b] Estimated from reported results.
[c] Wet bulk density.
Figure 1.68—Typical uniaxial compression curves for four soil compression categories at pore-water
potentials of approximately -30 kPa (from Larson et al., 1980). A = amorphous noncrystalline clay
materials such as allophanes (or andisols); B = medium and fine-textured tropical soils; C = medium and
fine-textured temperate soils; D = sandy soils.
(a) (b)
Figure 1.69—Relationships between compression index37 and percentage clay for soils from eight soil
orders: (a) For temperate region, expanding clay (2:1) types (Mollisols, Spodosols, Entisols, Inceptisols,
and Vertisols); (b) For tropical region, nonexpanding (1:1) clay types (Alfisols, Ultisols, and Oxisols)
(from Gupta et al., 1985; reproduced in Gupta and Allmaras, 1987).
Gupta and Allmaras (1987) report on a study by Gupta et al. (1985) of compaction
characteristics of world soils. Empirical relationships established between the
compression index37 and percentage clay sized particles based on 87 remolded samples of
soils from eight soil orders are shown in figure 1.69. It appears that, notwithstanding the
large variability in the data, the average values of compression index are similar between
temperate and tropical soils of the same clay content. Gupta and Allmaras (1987)
suggested that this may be partially due to the destruction of soil structure by grinding of
37
C ≈ 2.30λ. See footnote 32 on page 100.
samples for laboratory specimen preparation and the finite thickness of specimens used in
compression tests. They also note that any effect of soil swelling for the expanding clay
types would be overwhelmed by the large consolidation stresses imposed on the samples
along the virgin compression line.
Concluding Remarks
Elastoplastic soil mechanics provides a systematic, rational framework within which
to describe many aspects of the complex behavior of soils. The elastoplastic framework is
particularly useful because the models can be used to describe soil behavior in a largely
qualitative way with relatively few soil parameters and little mathematics, yet it has the
mathematical rigor needed for implementation in numerical (e.g., finite element)
solutions of boundary value problems. The critical-state theories of soil mechanics in
particular have generated a lot of interest for agricultural soil engineering applications.
The significant contribution of critical-state soil mechanics is that both volume changes
as well as stress changes are considered along with the loading path when analyzing soil
behavior.
The Cam clay critical-state model is one example of an elastoplastic model. It is a
rather crude quantitative model, but has a special qualitative role. The theory can describe
a wide range of soil behavior, including brittle and ductile failure, with a relatively small
number of soil constants. As Naylor (1978b) noted, critical-state soil mechanics linked
several well established but apparently unconnected concepts in soil mechanics: (1) the
existence of a unique yield surface in p-q-e space; (2) the Hvorslev surface representation
of peak strengths for sands and heavily over-consolidated clays; (3) Mohr-Coulomb or
other failure criteria; (4) the irrecoverability of strains in soils consolidated to new stress
levels; and, (5) the critical-state concept. The combination of the flow rule with particular
shapes of the plastic potential surfaces causes dε sp / dε pv to steadily approach zero as the
soil approaches critical-state, automatically satisfying the requirement that continuous
shearing be associated with no further volume changes. No further changes in p and q at
critical-state are also embodied in the model. The models perform best for wet of critical
(i.e., compacting) conditions and for stress ratios producing relatively large deformations.
Unfortunately, deformation predictions are much less reliable for conditions dry of
critical, largely because of instabilities introduced by the formation of slip planes. Since
elastoplastic models assume continuity of the soil material, some of the conditions may
be violated once discontinuities develop. This situation is relevant for many agricultural
conditions. The shear stress corresponding to critical-state for dense soils is also not as
well-defined, considered to be somewhat higher than the residual strength, but lower than
the peak strength (Mitchell, 1993).
There is strong evidence to support the adaptation of the critical-state theory to
agricultural soils. The development of an unsaturated soil mechanics framework in which
the role of matric suction on soil behavior is addressed also offers a powerful tool
towards the understanding and modeling of agricultural soil behavior.
The majority of published studies consider ductile behavior of agricultural soils on the
wet (subcritical) side of the critical-state wall (on or below the Roscoe surface), although
there have been some investigations of brittle behavior on the dry side (on or below the
Hvorslev surface). It appears that the critical-state theories can be extended to include
unsaturated soils, and thus should be capable of predicting mechanical behavior of soil
during tillage and compaction. However, the reported literature also indicates that there
are some significant departures in the details of the critical-state concept for such soils. In
particular, there is need for further research into the nature of yielding of unsaturated soils
and the effect of suction or water content on the shape of the yield surfaces. Several
factors such as specimen preparation methods, testing procedures and stress classification
must be properly addressed to obtain meaningful interpretations of the critical-state
behavior of agricultural soils. The use of a total stress approach is very popular in the
agricultural critical-state soil mechanics literature. Resulting model formulations are
simple, and appear to give satisfactory predictions for practical purposes. Difficulties
arise when such formulations are used to describe soil behavior over a wide range of
applied loads and soil water contents. There is evidence that a total stress approach will
give poor predictions for wet of optimum conditions and that a unique critical-state line
cannot be defined for a given water content. Some of the inconsistencies that arise appear
to have as much or more to do with comparison of samples with different fabrics than
with expressing the stress state in terms of total stresses. It is crucial to recognize that
specimens prepared at different bulk densities and water contents represent different soils.
The use of an independent stress-state variable approach (i.e., separating net stress and
matric suction) is increasingly gaining acceptance. This approach provides important
insight into unsaturated soil behavior, but at the expense of considerable additional
complexity to the model. Simplified formulations are needed for various problems
involving unsaturated soils. Furthermore, the application of such a framework for
conditions relevant to agricultural machinery problems (i.e., constant water content, large
deformations) is only in its infancy.
It is strongly recommended that the soil-water characteristic curve always be presented
along with other basic soil properties. This will facilitate interpretation of results of
various studies as well as comparison between analyses using a total stress or an
independent stress state variable approach. Furthermore, the soil-water characteristic is a
fundamental constitutive relationship for unsaturated soils and may be used to estimate a
range of unsaturated soil behavior functions. Efforts are now being directed towards
formulating generalized models that can describe the paths traversed by an unsaturated
soil undergoing physical as well as environmental stresses. In such a framework, the soil-
water characteristic will become an important relationship in describing the complete
state boundary surface for an unsaturated soil.
Although critical-state soil parameters are readily obtained from standard laboratory
tests, these tests are extremely time-consuming. There is also always the problem of soil
disturbance when taking field samples. For the practical implementation of critical-state
theory to agricultural and forestry soils there is a need for simple experimental techniques
to measure critical-state parameters in situ. Methods based on inverse-solution techniques
such as that described by Rubinstein et al. (1994) show promise. Other approaches of
wide interest include the correlation of critical-state parameters with simple index tests
(Hettiaratchi, 1987).
Another obstacle to practical use of the critical-state model is the lack of knowledge of
the relationship between p and q in all but the simplest soil-implement relations. The
development and implementation of finite element formulations of critical-state models
should make possible the application of critical-state soil mechanics for complex loading
situations. The numerical solution of finite element formulations still pose some
challenges, especially for dry-of-critical or strain-softening soil behavior. As these
challenges are addressed, the application of critical-state theories to solve complex
agricultural and forestry soil problems in tillage, traction, compaction, and plant root
growth can become increasingly popular.
References for Chapter 1
Abdallah, A. M., D. R. P. Hettiaratchi, and A. R. Reece. 1969. The mechanics of root growth in
granular media. J. Agric. Eng. Res. 14: 236-248.
Adachi, T. 1976. A constitutive equation for normally consolidated clays. Numerical Methods in
Geomechanics, ASCE 1: 282-293.
Adachi, T., and F. Oka. 1982. Constitutive equations for normally consolidated clays based on
elasto-viscoplasticity. Soils and Foundations 22(4): 57-70.
Adams, B. A. 1996. Critical-state behavior of an agricultural soil. Ph.D. thesis, University of
Saskatchewan, Saskatoon, Canada.
Adams, B. A., and D. Wulfsohn. 1996. Application of critical-state and unsaturated soil mechanics
principles to an agricultural soil. ASAE Paper No. 96-1065. St. Joseph, Mich.: ASAE.
Adams, B. A., and D. Wulfsohn. 1997. Variation of the critical-state boundaries of an agricultural
soil. Europ. J. Soil Sci. 48(4): 739-748.
Adams, B. A., and D. Wulfsohn. 1998. Critical-state behaviour of an agricultural soil. J. Agric.
Eng. Res. 70(4): 345-354.
Akai, K., T. Adachi, and K. Nishi. 1977. Mechanical properties of soft rocks. In Proc. 9th Intl.
Conference on Soil Mechanics and Foundation Engineering, Vol. 1. Tokyo, Japan.
Alonso, E. 1993. Constitutive modelling of unsaturated soils. In Unsaturated Soils: Recent
Developments and Applications. Civil Engineering European Courses Programme of Continuing
Education, Barcelona, Spain.
Alonso, E. E., A. Gens, and A. Josa. 1990. A constitutive model for partially saturated soils.
Géotechnique 40(3): 405-430.
Alonso, E. E., A. Josa, and A. Gens. 1993. Modeling the behaviour of compacted soils. In
Unsaturated Soils, ed. S. L. Houston and W.K. Wray, ASCE Geotechnical Special Publication
No. 39, 103-114. New York: ASCE.
Al-Tabbaa, A., and D. M. Wood. 1989. An experimentally based ‘bubble’ model for clay. In
Numerical Models in Geomechanics NUMOG III, ed. S. Pietruszczak and G. N. Pande, 91-99.
London: Elsevier.
Araya, K., and R. Gao. 1995. A non-linear three dimensional finite element analysis of subsoiler
cutting with pressurized air injection. J. Agric. Eng. Res. 61: 115-128.
Atakol, K., and H. G. Larew. 1970. Dynamic shearing resistance of dry Ottawa sand. J. Soil Mech.
Found. Div. ASCE 96(SM2): 705-720.
Atkinson, J. 1993. An Introduction to the Mechanics of Soils and Foundations. London: McGraw-
Hill.
Atkinson, J. H., and P. L. Bransby. 1978. The Mechanics of Soils-An Introduction to Critical-state
Soil Mechanics. London: McGraw-Hill.
Awadhal, N. K., and C. P. Singh. 1992. Puddling effects on mechanical characteristics of wet loam
soil. J. Terramechanics 29(4/5): 515-521.
Bailey, A. C. 1971. Compaction and shear in compacted soils. Transactions of the ASAE 14(2):
201-205.
Bailey, A. C. 1973. Shear and plastic flow in unsaturated clay. Transactions of the ASAE 16(2):
218-221, 226.
Bailey, A. C., and C. E. Johnson. 1994. NSDL-AU Model, soil stress-strain behavior. ASAE Paper
No. 94-1074. St. Joseph, Mich.: ASAE.
Bailey, A. C., C. E. Johnson, and R. L. Schafer. 1984. Hydrostatic compaction of agricultural soils.
Transactions of the ASAE 27(4): 925-955.
Bailey, A. C., and G. E. VandenBerg. 1968. Yielding by compaction and shear in unsaturated soil.
Transactions of the ASAE 11(3): 307-311, 317.
Advances in Soil Dynamics, Vol. 2
Baladi, G. Y., and B. Rohani. 1984. Development of an elasto-viscoplastic constitutive relationship

for earth materials. In Mechanics of Engineering Materials, ed. C. S. Desai and R. H. Gallagher,
23-46. New York: Wiley.
Bathe, H. J. 1982. Finite Element Procedures in Engineering Analysis. Englewood Cliffs, N.J.:
Prentice-Hall.
Bishop, A. W. 1959. The principle of effective stress. Tecknisk Ukeblad 106(39): 859-863.
Bishop, A. 1972. Shear strength parameters for undisturbed and remolded soil specimens. In Proc.
Roscoe Memorial Symposium: Stress-Strain Behaviour of Soils, ed. R. H. G. Parry, 3-58. Foulis,
Henley-on-the-Thames.
Bishop, A. W., and D. J. Henkel. 1962. The Measurement of Soil Properties in the Triaxial Test,
2nd ed. London, England: Edward Arnold.
Bjerrum, L., and T. C. Kenney. 1967. Effect of structure on the shear behaviour of normally
consolidated quick clays. In Proc. Geotechnical Conference on Shear Strength Properties of
Natural Soils and Rocks, 2: 19-27, Oslo, Norway.
Bolzon, G., B. A. Schrefleer, and O. C. Zienkiewicz. 1996. Elastoplastic soil constitutive laws
generalized to partially saturated states. Géotechnique 46(2): 279-289.
Bowen, H. D., T. H. Garner, and D. H. Vaughn. Advances in soil-plant dynamics. In Advances in
Soil Dynamics Vol. 1, ed. S. K. Upadhyaya, W. J. Chancellor, J. V. Perumpral, R. L. Schafer,
W. R. Gill, and G. E. VandenBerg, 255-280. ASAE Monograph. St. Joseph, Mich.: ASAE.
Braunack, M. V., and K. W. J. Malafant. 1988. The effect of dynamic impact and water potential
on some surface soil properties. J. Terramechanics 25(2): 149-162.
Britto, A. M., and M. J. Gunn. 1987. Critical-state Soil Mechanics via Finite Elements. New York:
Halsted Press.
Burland, J. B. 1965. Some aspects of the mechanical behaviour of partly saturated soils. In
Moisture Equilibria and Moisture Changes Beneath Covered Areas, ed. G. D. Aitchinson, 270-
278. Sydney, Aus.: Butterworths.
Campbell, D. J. 1976. Plastic limit determination using a drop-cone penetrometer. J. Soil Sci. 27:
295-300.
Casagrande, A., and W. L. Shannon. 1948. Research on stress-deformation and strength
characteristics of soils and soft rocks under transient loading. Soil Mechanics Series No. 31,
Harvard University, Cambridge, Mass., June 1948.
Chancellor, W. J. 1994. Soil physical properties. In Advances in Soil Dynamics Vol. 1, ed. S. K.
Upadhyaya, W. J. Chancellor, J. V. Perumpral, R. L. Schafer, W. R. Gill, and G. E.
VandenBerg, 21-254. ASAE Monograph No. 12. St. Joseph, Mich.: ASAE.
Chancellor, W. J., and J. A. Vomocil. 1970. Relation of moisture content to failure strengths of
seven agricultural soils. Transactions of the ASAE 13(1): 9-13, 17.
Chen, W. 1984. Constitutive modelling in soil mechanics. In Mechanics of Engineering Materials,
ed. C. S. Desai and R. H. Gallagher, 91-120. New York: Wiley.
Chen, W. F., and G. Y. Baladi. 1985. Soil Plasticity. Theory and Implementation. Amsterdam:
Elsevier.
Chi, L., and R. L. Kushwaha. 1990. A non-linear 3-D finite element analysis of soil failure with
tillage tools. J. Terramechanics 27(4): 343-366.
Chi, L., and S. Tessier. 1994. Comparison of nonlinear elastic and elasto-plastic models. ASAE
Paper No. 94-1076. St. Joseph, Mich.: ASAE.
Chi, L., and S. Tessier. 1995. Finite element analysis of soil compaction reduction with high
flotation tires. In Proc. of the 5th North American Conference/Workshop of the Intl. Society for
Terrain-Vehicle Systems, 167-176. University of Saskatchewan, Saskatoon, Canada.
Chi, L., S. Tessier, and C. Laguë. 1993a. Finite element predictions of soil compaction induced by
various running gears. Transactions of the ASAE 36(3): 629-636.
Chi, L., S. Tessier, and C. Laguë. 1993b. Finite element modeling of soil compaction by liquid
manure spreaders. Transactions of the ASAE 36(3): 637-644.
Chi, L., S. Tessier, E. McKyes, and C. Laguë. 1993c. Modeling mechanical behavior of agricultural
soils. Transactions of the ASAE 36(6): 1563-1570.
Chung, T. J., and J. K. Lee. 1975. Dynamics of viscoelastoplastic soil under a moving wheel. J.
Terramechanics 12(1): 15-31.
Coleman, B. D., and W. Noll. 1961. Foundations of linear viscoelasticity. In Continuum Mechanics
III. Foundations of Elasticity Theory, ed. C. A. Truesdell, 3: 297-307. New York: Gordon and
Breach Science Publishers.
Croney, D., and J. D. Coleman. 1954. Soil structure in relation to soil suction (pF). J. Soil Sci. 5(1):
75-84.
Croney, D., and J. D. Coleman. 1961. Soil structure in relation to soil suction (pF). In Proc.
Conference on Pore Pressure and Suction in Soils, 31-37. London: Butterworths.
Croney, D., J. D. Coleman, and W. P. M. Black. 1958. Movement and distribution of water in soils.
Highway Research Board Report 40. London: HMSO.
Crooks, J. H. A., and J. Graham. 1976. Geotechnical properties of the Belfast estuarine deposits.
Géotechnique 26: 293-315.
Cui, Y. J., and P. Delage. 1993. On the elasto-plastic behaviour of an unsaturated silt. In
Unsaturated Soils, ed. S. L. Houston and W. K. Wray, ASCE Geotechnical Special Publication
No. 39, 115-126. New York: ASCE.
Cui, Y. J., and P. Delage. 1996. Yielding and plastic behaviour of an unsaturated compacted silt.
Dafalias, Y. F. 1987. An anisotropic critical-state clay plasticity model. In Constitutive Laws for
Engineering Materials: Theory and Applications, ed. C. S. Desai, E. Krempl, P. D. Kiousis and
T. Kundu, 1: 513-521. Amsterdam: Elsevier.
Dafalias, Y. F., and L. R. Hermann. 1980. A bounding surface plasticity model. In Proc. Intl.
Symposium on Soils Under Cyclic and Transient Loading, ed. G. N. Pande and O. C
Zienkiewicz, 335-345. Rotterdam: Balkema.
Dafalias, Y. F., and L. R. Hermann. 1982. Bounding surface formulation of soil plasticity. In Soil
Mechanics – Transient and Cyclic Loads, ed. G. N. Pande and O. C. Zienkiewicz, 253-282.
New York: Wiley.
Dafalias, Y. F., and L. R. Hermann. 1986. Bounding surface plasticity II: Application to isotropic
cohesive soils. J. Eng. Mech. ASCE 112: 1263-1291.
Dafalias, Y. F., and E. P. Popov. 1975. A model for nonlinearly hardening materials for complex
loading. Acta Mech. 21(3): 173-192.
Davis, R. O., and G. Mullenger. 1978. A rate-type constitutive model for soil with a critical-state.
Intl. J. Num. Anal. Meth. Geomechan. 2: 255-282.
Delage, P., G., P. R. Suraj de Silva, and E. De Laure. 1987. Un nouvel appareil triaxial pour les sols
non-saturés. In Proc. 9th European Conference on Soil Mechanics, 1: 25-28. Dublin, Ireland.
Desai, C. S., and H. V. Phan. 1980. Three-dimensional finite element analysis including material
and geometric nonlinearities. In Computational Methods in Geomechanics, ed. J. T. Oden, 205-
224. Amsterdam, The Netherlands: North-Holland Publishing Company.
Desai, C. S., and H. J. Siriwardane. 1984. Constitutive Laws for Engineering Materials with
Emphasis on Geologic Materials. Englewood Cliffs, N.J.: Prentice-Hall.
Desai, C. S., and D. Zhang. 1987. Viscoplastic model for geologic materials with generalized flow
rule. Intl. J. Num. Anal. Meth. Geomechan.11: 603-620.
DiMaggio, F. L., and I. S. Sandler. 1971. Material model for granular soils. J. Eng. Mechan. Div.
ASCE 97(EM3): 935-950.
Drucker, D. C. 1950. Some implications of work hardening and ideal palsticity. Quart. Appl. Math.
7: 411-418.
Drucker, D. C. 1959. A definition of stable inelastic material. ASME Transactions of the 81: 101-
106.
Drucker, D. C. 1988. Conventional and unconventional plastic response and representation. Appl.
Mach. Rev. 41(4): 151-167.
Drucker, D. C., and W. Prager. 1952. Soil mechanics and plastic analysis or limit design. Quart.
Appl. Math. 10(2): 157-165.
Drumright, E. E. 1989. The contribution of matric suction to the shear strength of unsaturated soils.
Ph.D. thesis, University of Colorado, Fort Collins, Colo.
Duncan, J. M., and C. Y. Chang. 1970a. Non-linear analysis of stress and strain in soils. J. Soil
Mech. Found. Div. ASCE 96(SM5): 1629-1653.
Duncan, J. M., and C. Y. Chang. 1970b. Analysis of soil movement around a deep excavation. J.
Soil Mech. Found. Div. ASCE 96(SM5): 1655-1681.
Elijah, D. L., and J. A. Weber. 1971. Soil failure and pressure patterns for flat cutting blades.
Transactions of the ASAE 14(4): 781-785.
Escario, V., and J. Sáez. 1986. The shear strength of partly saturated soils. Géotechnique 36(3):
453-456.
Feda, J. 1982. Mechanics of Particulate Materials. The Principles. Developments in Geotechnical
Engineering 30. Amsterdam: Elsevier.
Fredlund, D. G. 1995. The scope of unsaturated soil mechanics: an overview. In Proc. 1st Intl.
Conference on Unsaturated Soils, ed. E. E. Alonso and P. Delage, Vol. 1. Paris, France.
Fredlund, D. G. 1997. From theory to practice of unsaturated soil mechanics. In Proc. 3rd Brazilian
Symposium on Unsaturated Soils, Vol. 2. Rio de Janeiro, Brazil.
Fredlund, D. G., and N. R. Morgenstern. 1977. Stress state variables for unsaturated soils. J.
Geotech. Eng. Div. ASCE 104(GT5): 447-466.
Fredlund, D. G., N. R. Morgenstern, and R. A. Widger. 1978. The shear strength of unsaturated
soils. Can. Geotech. J. 15(3): 313-321.
Fredlund, D. G., and H. Rahardjo. 1993a. Soil Mechanics for Unsaturated Soils. New York: Wiley.
Fredlund, D. G., and H. Rahardjo. 1993b. An overview of unsaturated soil behaviour. In
Unsaturated Soils, ed. S. L. Houston and W. K. Wray, 1-31. New York: ASCE.
Fredlund, D. G., H. Rahardjo, and J. K. M. Gan. 1987. Non-linearity of strength envelope for
unsaturated soils. In Proc. 6th Intl. Conference on Expansive Soils, 1: 49-54. New Delhi, India:
Central Board of Irrigation and Power.
Fredlund, D. G., A. Xing, M. D. Fredlund, and S. L. Barbour. 1995. The relationship of the
unsaturated soil shear strength to the soil-water characteristic curve. Can. Geotech. J. 32: 440-
448.
Fredlund, M. D. 1996. Design of a knowledge-based system for unsaturated soil properties. M.Sc.
thesis, University of Saskatchewan, Saskatoon, Canada.
Freitag, D. R. 1985. Soil dynamics as related to traction and transport systems. In Proc. Intl.
Conference on Soil Dynamics: Vol. 4, Soil Dynamics as Related to Traction and Transport
Systems, 605-629. Auburn, Ala.: Auburn University.
Gens, A., and E. E. Alonso. 1992. A framework for the behaviour of unsaturated expansive clays.
Can. Geotech. J. 29: 1013-1032.
Graham, J. 1993. The key material properties in geotechnical engineering need to be measured
instead of being specified. In Predictive Soil Mechanics, ed. G. T. Houlsby and A. N. Schofield,
1-18. London: Thomas Telford.
Graham, J., J. H. A. Crooks, and A. L. Bell. 1983a. Time effects on the stress-strain behavior of
natural soft clays. Géotechnique 33(3): 327-340.
Graham, J., M. L. Noonan, and K. V. Lew. 1983b. Yield states and stress-strain relationships in a
natural plastic clay. Can. Geotech. J. 20: 502-516.
Greacen, E. L. 1960. Water content and soil strength. J. Soil Sci. 11(2): 313-333.
Grisso, R. D., C. E. Johnson, and A. C. Bailey. 1987. Soil compaction by continuous deviatoric
stress. Transactions of the ASAE 30(5): 1293-1301.
Gupta, S. C., and R. R. Allmaras. 1987. Models to assess the susceptibility of soils to excessive
compaction. In Advances in Soil Science, Volume 6, ed. B. A. Stewart, 65-100. New York:
Springer-Verlag.
Gupta, S. C., A. Hadas, W. B. Voorhees, D. Wolf, W. E. Larson, and E. C. Schneider. 1985.
Development of guides for estimating the ease of compaction of world soils. Research Report
submitted to BARD Fund on completion of Project Grant No. US-337-80. (Cited in Gupta and
Allmaras, 1987).
Gupta, S. C., and W. E. Larson. 1982. Modeling soil mechanical behavior during tillage. In
Symposium on Predicting Tillage Effects on Soil Physical Properties and Processes, ed. P. W.
Unger, D. M. Van Doren Jr., F. D. Whisler, and E. L. Skidmore, 151-178. ASA Special
Publication 44. Madison, Wis.: American Society of Agronomy.
Harris, H. D. 1993. A critical-state interpretation of soil compaction by anisotropic consolidation. J.
Agric. Eng. Res. 55: 265-276.
Hassan, A. El-Domiatry, and W. J. Chancellor. 1970. Stress-strain characteristics of a saturated
clay at various rates of strain. Transactions of the ASAE 13(5): 685-689.
Hatibu, N., and D. R. P. Hettiaratchi. 1986. Failure and volume-change behaviour of agricultural
soils. In Proc. 3rd European Conference of the ISTVS, 14-18. Warsaw.
Hatibu, N., and D. R. P. Hettiaratchi. 1993. The transition from ductile flow to brittle failure in
unsaturated soils. J. Agric. Eng. Res. 54(4): 319-328.
Henkel, D. J. 1959. The relationships between the strength, porewater pressure and volume change
characteristic of saturated clays. Géotechnique 9: 119-135.
Hettiaratchi, D. R. P. 1987. A critical-state soil mechanics model for agricultural soils. Soil Use &
Management 3(3): 94-105.
Hetteriatchi, D. R. P. 1988. Theoretical soil mechanics and implement design. Soil & Tillage
Research 11(3&4): 325-347.
Hetteriatchi, D. R. P. 1990. Soil compaction and plant root growth. Phil. Trans. R. Soc. Lond. B
329: 343-355.
Hettiaratchi, D. R. P., and Y. Lang. 1987. Nomograms for the estimation of soil strength from
indentation tests. J. Terramechanics 24(3): 187-198.
Hettiaratchi, D. R. P., and J. R. O’Callaghan. 1980. Mechanical behaviour of agricultural soils. J.
Agric. Eng. Res. 25: 239-259.
Hettiaratchi, D. R. P., and J. R. O’Callaghan. 1985. Mechanical behaviour of agricultural soils. In
Proc. Intl. Conference on Soil Dynamics, 2: 266-281. Auburn, Ala., June.
Hettiaratchi, D. R. P., M. F. O’Sullivan, and D. J. Campbell. 1992. A constant cell volume triaxial
testing technique for evaluating critical-state parameters of unsaturated soils. J. Soil Sci. 43:
791-806.
Hettiaratchi, D. R. P., B. D. Witney, and A. R. Reece. 1966. The calculation of passive pressure in
two-dimensional soil failure. J. Agric. Eng. Res. 11(2): 89-107.
Hettiaratchi, D. R. P., and A. R. Reece. 1967. Symmetrical three-dimensional soil failure. J.
Hettler, A., and I. Vardoulakis. 1984. Behavior of dry sand tested in a large triaxial apparatus.
Hill, R. 1950. The Mathematical Theory of Plasticity. London: Oxford University Press.
Houlsby, G. T. 1997. The work input to an unsaturated granular material. Technical Note,
Hvorslev, M. J. 1937. Über die Festigkeitseigenschaften gestörter bindiger Böden, Danmarks
Naturvidenskabelige Samfund, København, Ingeniørvidenskbelige Skrifter A45. (English
translation. 1969. Physical properties of remolded cohesive soils, U.S. Waterways Experimental
Station, Vicksburg, Miss., No. 69-5).
Hvorslev, M. J. 1960. Physical components of the shear strength of saturated clays. In ASCE
Research Conference on the Shear Strength of Cohesive Soils, 169-273. University of Colorado.
Janbu, N. 1963. Soil compressibility as determined by odeometer and triaxial tests. In European
Conference on Soil Mechanics and Foundation Engineering, 1: 19-25. Wiesbaden, Germany.
Jennings, J. E. B., and J. B. Burland. 1962. Limitations to the use of effective stresses in partly
saturated soils. Géotechnique 12(2): 125-144.
Johansen, O. 1975. Thermal conductivity of soils. Ph.D. thesis, Trondheim, Norway (CRREL Draft
Translation 637. 1977).
Josa, A., A. Balmaceda, E. E. Alonso, and A. Gens. 1992. An elastoplastic model for partially
saturated soils exhibiting a maximum of collapse. In Computational Plasticity — Fundamentals
and Applications, ed. D. R. Owen, E. Oñate and E. Hinton, 815-826. Swansea, England:
Pineridge Press.
Holtz, R. D., and W. D. Kovacs. 1981. An Introduction to Geotechnical Engineering. Englewood
Cliffs, N.J.: Prentice-Hall.
Kaliakin, V. N., and Y. F. Dafalias. 1989. Simplifications to the bounding surface model for
cohesive soils. Short Communication, Intl. J. Num. Anal. Meth. Geomech. 13: 91-100.
Karube, D. 1988. New concept of effective stress in unsaturated soil and its proving test. In
Advanced Triaxial Testing of Soil and Rock, ASTM STP 977, ed. R. T. Donaghe, R. C. Chaney
and M. L. Silver, 539-552. Philadelphia, Pa.: American Society for Testing and Materials.
Kirby, J. M. 1989. Measurements of the yield surfaces and critical-state of some unsaturated
agricultural soils. J. Soil Sci. 40: 167-182.
Kirby, J. M. 1991a. Strength and deformation of agricultural soil: measurement and practical
significance. Soil Use & Management 7(4): 223-229.
Kirby, J. M. 1991b. Critical-state soil mechanics parameters and their variation for Vertisols in
eastern Australia. J. Soil Sci. 42: 487-499.
Kirby, J. M. 1994. Simulating soil deformation using a critical-state model: I. Laboratory tests.
Europ. J. Soil Sci. 45: 239-248.
Kirby, J. M. 1995. The influence of soil variability on predictions of compaction by agricultural
tyres. In Proc. 1st Intl. Conference on Unsaturated Soils, ed. E. E. Alonso and P. Delage, 1:
1079-1084. Paris, France.
Kirby, J. M. 1998. Critical-state analysis of laboratory soil mechanics tests. In Proc. 3rd Intl.
Conference on Soil Dynamics, ed. I. Shmulevich, 119-127. Held in Tiberias, Israel, 3-8 August
1997.
Kirby, J. M., A. G. Bengough, and M. F. O’Sullivan. 1998a. Influence of soil strength on root
growth – Experiments and analysis using a critical-state model. In Proc. 3rd Intl. Conference on
Soil Dynamics, ed. I. Shmulevich, 109-118. Held in Tiberias, Israel, 3-8 August 1997.
Kirby, J. M., M. F. O’Sullivan, and J. T. Wood. 1998b. Estimating critical-state soil mechanics
parameters from constant cell volume triaxial tests. Europ J. Soil Sci. 49: 85-93.
Klausner, Y. 1991. Fundamentals of Continuum Mechanics of Soils. London: Springer-Verlag.
Kohgo, Y., N. Masashi, and T. Miyazaki. 1993a. Theoretical aspects of constitutive modelling for
unsaturated soils. Soils and Foundations 33(4): 49-63.
Kohgo, Y., N. Masashi, and T. Miyazaki. 1993b. Verification of the generalized elastoplastic
model for unsaturated soils. Soils and Foundations 33(4): 64-73.
Kondner, R. L. 1963. Hyperbolic stress-strain response-cohesive soils. J. Soil Mech. Found. Eng.
ASCE 89(SM1): 115-143.
Kondner, R. L., and J. M. Horner. 1965. Triaxial compression of a cohesive soil with effective
octahedral stress control. Can. Geotech. J. 2(1): 40-52.
Kondner, R. L., and J. S. Zelasko. 1963. A hyperbolic stress-strain response of a sand. In Proc. 2nd
Pan-American Conference on Soil Mechanics and Foundations Engineering, 1: 289-324. Brazil.
Krieg, R. D. 1975. A practical two-surface plasticity theory. J. Appl. Mech. 42: 641-646.
Kurtay, T., and A. R. Reece. 1970. Plasticity theory and critical-state soil mechanics. J.
Kutter, B. L., and N. Sathialingam. 1992. Elastic-viscoplastic modeling of rate-dependent behavior
of clays. Géotechnique 42(3): 427-441.
Lade, P. V. 1972. The stress-strain and strength characteristics of cohesionless soils. Ph.D. thesis,
University of California, Berkeley.
Lade, P. V. 1977. Elasto-plastic stress-strain theory for cohesionless soils with curved yield
surfaces. Intl. J. Solids Struct. 13: 1019-1035.
Lade, P. V., and J. M. Duncan. 1973. Cubical triaxial tests on cohesionless soil. J. Soil Mech.
Found. Div., ASCE 99(SM10): 793-812.
Lade, P. V., and J. M. Duncan. 1975. Elasto-plastic stress strain theory for cohesionless soil. J.
Geotech. Eng. Div., ASCE 101(GT10): 1037-1053.
Lade, P. V., and J. M. Duncan. 1976. Stress-path dependent behavior of cohesionless soil. J.
Geotech. Eng. Div., ASCE 102(GT1): 51-68.
Lade, P. V., and R. B. Nelson. 1984. Incrementalization procedure for elasto-plastic constitutive
model with multiple, intersecting yield surfaces Intl. J. Num. Anal. Meth. Geomechan. 8: 311-
323.
Lambe, T. W. 1958. The engineering behavior of compacted clay. J. Soil Mech. Found. Div. ASCE
84(SM2): 1-35.
Larson, W. E., and S. C. Gupta. 1980. Estimating critical stress in unsaturated soils from changes in
pore water pressure during confined compression. Soil Sci. Soc. Am. J. 44: 1127-1132.
Larson, W. E., S. C. Gupta, and R. A. Useche. 1980. Compression of agricultural soils from eight
soil orders. Soil Sci. Soc. Am. J. 44: 450-457.
Lee, E. H. 1962. Viscoelasticity. In Handbook of Engineering Mechanics, ed. W. Flügge, chapter
53, 1-22. New York: McGraw-Hill.
Leeson, J. J., and D. J. Campbell. 1983. The variation of soil critical-state soil parameters with
water content and its relevance to the compaction of two agricultural soils. J. Soil Sci. 34(1): 33-
44.
Lu, Z., and J.-Z. Pan. 1986. A study on the rheological properties of paddy soils in China with
analytical and finite-element method. In Proc. 1st Asian-Pacific Conference of the ISTVS.
Maâtouk, A., S. Lerouel, and P. La Rochelle. 1995. Yielding and critical-state of a collapsible
unsaturated silty soil. Géotechnique 45(3): 465-477.
Matyas, E. N., and H. S. Radhakrishna. 1968. Volume change characteristics of partially saturated
soils. Géotechnique 18(4): 432-448.
McKyes, E. 1985. Soil Cutting and Tillage. Amsterdam: Elsevier.
McKyes, E., and O. S. Ali. 1977. The cutting of soil by a narrow blade. J. Terramechanics 14(2):
43-58.
Mitchell, J. K. 1993. Fundamentals of Soil Behavior, 2nd ed. New York: Wiley.
Mitchell, R. J. 1970. On the yielding and mechanical strength of Leda clays. Can. Geotech. J. 7(3):
297-312.
Mou, C. H., and T. Y. Chu. 1981. Soil suction approach for swelling potential evaluation.
Transportation Res. Rec. 790, 54-60. Washington, D.C.
Moulin, G. 1988. Caractérisation de l’état limit de l’argile de Pornic. Can. Geotech. J. 7: 297-312.
Mroz, Z., V. A. Norris, and O. C. Zienkiewicz. 1979. Application of an anisotropic hardening
model in the analysis of elasto-plastic deformation of soils. Géotechnique 29(1): 1-34.
Mullins, C. E., and K. P. Panayiotopoulos. 1984. The strength of unsaturated mixtures of sand and
Kaolin and the concept of effective stress. J. Soil Sci. 35: 459-468.
Mullins, C. E., D. A. Macleod, K. H. Northcote, J. M. Tisdall, and I. M. Young. 1990. Hardsetting
soils: behavior, occurence and management. In Advances in Soil Science, ed. B. A. Stewart, Vol.
11, 37-108. New York: Springer-Verlag.
Naylor, D. J. 1978a. Finite element methods in soil mechanics. In Developments in Soil Mechanics-
1, ed. C. R. Scott, 1-38. London: Applied Science Publishers.
Naylor, D. J. 1978b. Stress-strain laws for soil. In Developments in Soil Mechanics-1, ed. C. R.
Scott, 39-68. London: Applied Science Publishers.
Naylor, D. J. 1985. A continuous plasticity version of the critical-state model. Intl. J. Num. Meth.
Eng. 21: 1187-1204.
Niyamapa, T. K., and V. M. Salokhe. 1992. Soil failure under undrained quasi static and high speed
triaxial compression tests. J. Terramechanics 29(2): 195-205.
Ohmaki, S. 1979. Strength and deformation characteristics of overconsolidated cohesive soil. In
Proc. 3rd Intl. Conference on Numerical Methods in Geotechn., Vol 1. Aachen, Rotterdam.
Oida, A. 1984. Analysis of rheological deformation of soil by means of finite element method. J.
Oida, A., and T. Tanaka. 1981. Analysis of viscoelastic deformation of soil by means of Finite
Element Method. In Proc. 7th Intl. Conference of the ISTVS, 3: 1473-1492.
Olson, D. J., and J. A. Weber. 1965. Effect of speed on soil failure patterns in front of model tillage
tools. SAE Paper No. 65-0691. Warrendale, Pa.: SAE.
Olszak, W., and P. Perzyna. 1966. On elasto/visco-plastic soils. In Proc. IUTAM Symposium on
Rheology and Soil Mechanics, ed. J. Kravtchenko and P. M. Sirieys, 45-57. Berlin: Springer-
Verlag.
O’Sullivan, M. F., D. J. Campbell, and D. R. P. Hettiaratchi. 1994. Critical-state parameters derived
from constant cell volume triaxial tests. Europ. J. Soil Sci. 45: 249-256.
Pan, J.-Z. 1986. The general rheological model of paddy soils in south China. J. Terramechanics
23(2): 59-68.
Parry, R. H. G. 1958. Correspondence on Roscoe et al. (1958) “On the yielding of soils.”
Parry, R. H. G. 1960. Triaxial compression and extension tests on remoulded saturated clay.
Payne, P. C. J. 1956. The relationship between mechanical properties of soil and the performance
of simple cultivation implements. J. Agric. Eng. Res. 1: 23-50.
Perumpral, J. V., R. D. Grisso, and C. S. Desai. 1983. A soil-tool model based on limit equilibrium
analysis. Transactions of the ASAE 26(4): 991-995.
Perzyna, P. 1963. The constitutive equations for rate sensitive plastic materials. Quart. Appl. Math.
20(4): 321-332.
Perzyna, P. 1966. Fundamental problems in viscoplasticity. Adv. Appl. Mechan. 9: 243-377.
Petersen, C. T. 1993. The variation of critical-state parameters with water content for two
agricultural soils. J. Soil Sci. 44: 397-410.
Petersen, C. T. 1994. Modelling soil distortion during compaction for cylindrical stress load paths.
Europ. J. Soil Sci. 45: 117-126.
Peterson, R. W. 1988. Interpretation of triaxial compression tests on partially saturated soils. In
Advanced Triaxial Testing of Soil and Rock, ASTM STP 977, ed. R. T. Donaghe, R. C. Chaney
and M. L. Silver, 512-538. Philadelphia: American Society for Testing and Materials.
Philip, J. R., and D. E. Smiles. 1969. Kinetics of sorption and volume change in three-component
systems. Austral. J. Soil. Res. 7: 1-19.
Poorooshasb, H. B., J. Holubec, and A. N. Sherborne. 1966. Yielding and flow of sand in triaxial
compression. Part I. Can. Geotech. J. 3(4): 179-190.
Poorooshasb, H. B., J. Holubec, and A. N. Sherborne. 1967. Yielding and flow of sand in triaxial
compression. Parts II and III. Can. Geotech. J. 4(4): 376-397.
Prager, W. 1949. Recent developments in mathematical theory of plasticity. J. Appl. Phys. 20(3):
235-241.
Price, A. M., and I. W. Farmer. 1979. Application of yield models to rock. Intl. J. Rock Mech.
Mining Sci. 16: 157-159.
Qun, Y., and J. Shen. 1988. Research on the rheological properties of wet soils. Part I. Flow
characteristics. In Proc. 2nd Asia-Pacific Conference of the ISTVS, 23-34. Bangkok, Thailand.
Reece, A. R. 1977. Soil mechanics of agricultural soils. Soil Sci. 123: 332-337.
Reicosky, D. C., W. B. Voorhees, and J. K. Radke. 1981. Unsaturated water flow through a
simulated wheel track. Soil Sci. Soc. Am. J. 45: 3-8.
Ridley, A. M., and J. B. Burland. 1994. A new device for the direct measurement of soil suction
over a wide range. In Engineering Characteristics of Arid Soils, ed. P. G. Fookes and R. H. G.
Parry, 289-295. Rotterdam: Balkema.
Ridley, A. M., and W. K. Wray. 1995. Suction measurement: A review of current theory and
practices. In Proc. 1st Intl. Conference Unsaturated Soils, ed. E. E. Alonso and P. Delage, 3:
1293-1322. Paris, September.
Rosa, U. 1997. Performance of narrow tillage tools with inertial and strain rate effects. Ph.D. thesis,
University of Saskatchewan, Saskatoon, Canada.
Rosa, U. A., and D. Wulfsohn. 1999. Constitutive model for high speed tillage using narrow tools.
J. Terramechanics 36(4): 221-234.
Roscoe, K. H., and J. B. Burland. 1968. On the generalized stress-strain behavior of ‘wet’ clay. In
Engineering Plasticity, ed. J. Heyman and F. A. Leckie, 535-609. Cambridge: Cambridge
University Press.
Roscoe, K. H., A. Schofield, and A. Thurairajah. 1963. Yielding of clays in states wetter than
critical. Géotechnique 13(3): 211-240.
Roscoe, K. H., A. Schofield, and C. P. Wroth. 1958. On the yielding of soils. Géotechnique 8: 22-
53.
Rubinstein, D., S. K. Upadhyaya, and M. Sime. 1994. Determination of in-situ properties of soil
using response surface methodology. J. Terramechanics 31(2): 67-92.
Russell, R. S. 1977. Plant Root Systems: Their Function and Interaction with Soil. London:
McGraw-Hill.
Sandler, I. S., F. L. DiMaggio, and G. Y. Baladi. 1976. Generalized cap model for geological
materials. J. Geotech. Eng. Div., ASCE 102(GT7): 683-699.
Schofield, A. N., and C. P. Wroth. 1968. Critical-state Soil Mechanics. London: McGraw-Hill
Book Co.
Scott, R. F. 1985. Plasticity and constitutive relations in soil mechanics. J. Geotech. Eng. Div.
ASCE 111(5): 563-605.
Shames, I. H. 1989. Introduction to Solid Mechanics, 2nd ed. Englewood Cliffs, N.J.: Prentice-
Hall,.
Shaw, P., and S. F. Brown. 1986. Cyclic simple shear testing of granular materials. ASTM Geotech.
Testing J. 9(GTJ): 213-220.
Shen, J., and R. L. Kushwaha. 1995. Investigation of an algorithm for non-linear and dynamic
problems in soil-machine systems. Computers and Electronics in Agriculture 13: 51-66.
Shi, Y.-L. 1998. Prediction of forest soil compaction from harvesting. M.Sc. thesis, University of
Saskatchewan, Saskatoon, Canada.
Sime, M., and S. K. Upadhyaya. 1994. Experimental verification of an inverse solution technique
developed for parameter estimation. ASAE Paper No. 94-1565. St. Joseph, Mich.: ASAE.
Sime, M., and S. K. Upadhyaya. 1995. Development and evaluation of parameter estimation
technique using response surface methodology. In Proc. of the 5th North American
Conference/Workshop of the Intl. Society for Terrain-Vehicle Systems, 249-258. University of
Saskatchewan, Saskatoon.
Smith, D. L. O. 1985. Compaction by wheels: a numerical model for agricultural soils. J. Soil Sci.
35: 621-632.
Smith, G. N. 1990. Elements of Soil Mechanics, 6th ed. Oxford/London: BSP Professional Books.
Spoor, G. 1975. Fundamental aspects of cultivations. In Soil Physical Conditions and Crop
Production. Minn. Agric. Fish. Food Tech. Bull. No. 29, 128-144. London: HMSO.
Spoor, G., and R. J. Godwin. 1999. Role of soil dynamics in improving the efficiency of cultural
practices. In Proc. 3rd Intl. Conference on Soil Dynamics, ed. I. Shmulevich, 305-316. Held in
Tiberias, Israel, 3-8 August 1997.
Stafford, J. V. 1979. The performance of a rigid tine in relation to soil properties and speed. J.
Agric. Eng. Res. 24: 41-56.
Stafford, J. V. 1981. An application of critical-state soil mechanics: The performance of rigid tines.
J. Agric. Eng. Res. 26(3): 387-401.
Stafford, J. V., and D. W. Tanner. 1983. Effect of rate on soil shear strength and soil-metal friction.
I. Shear strength. Soil & Tillage Research 3: 245-260.
Taylor, D. W. 1948. Fundamentals of Soil Behaviour. New York: Wiley.
Terzaghi, K. 1936. The shear strength resistance of saturated soils and the angle between the planes
of shear. In Proc. 1st Intl. Conference on Soil Mechanics and Foundation Engineering, 1: 54-
56. Cambridge, Ma., 22-26 July.
Terzaghi, K., and R. B. Peck. 1948. Soil Mechanics in Engineering Practice. New York, N. Y.:
Wiley.
Thangavadivelu, S., P. Barnes, J. Slocombe, L. Stone, and J. Higgins. 1994. Soil response to track
and wheel tractor traffic. J. Terramechanics 31(1): 41-50.
Toll, D. G. 1990. A framework for unsaturated soil behaviour. Géotechnique 40(1): 31-44.
Turnage, G. W. 1975. Behavior of fine-grained soils under high speed tire loads. Technical Report
TR 3-652 No. 7, U.S. Army Corps of Engineers, Waterways Experiment Station, Vicksburg,
Miss., June 1975.
Upadhyaya, S. K. 1994. Linear and nonlinear elasticity. ASAE Paper No. 94-1071. St. Joseph,
Mich.: ASAE.
Utomo, W. H., and A. R. Dexter. 1981. Soil friability. J. Soil Sci. 32: 203-213.
Vyalov, S. S. 1986. Rheological Fundamentals of Soil Mechanics. Developments in Geotechnical
Engineering 36. Amsterdam: Elsevier.
Warkentin, B. P. 1971. Effects of compaction on content and transmission of water in soils. In

Compaction of Agricultural Soils, ed. K. K. Barnes, W. M. Carleton, H. M. Taylor, R. I.
Throckmorton, and G. E. Vanden Berg, 126-153. St. Joseph, Mich.: ASAE.
Whalley, W. R., P. B. Leeds-Harrison, P. Joy, and P. Hoefsloot. 1994. Time domain reflectometry
and tensiometry combined in an integrated soil water monitoring system. Technical Note. J.
Agric. Eng. Res. 59: 141-144.
Wheeler, S. J. 1991. An alternative framework for unsaturated soil behavior. Géotechnique 41(2):
257-261.
Wheeler, S. J., and V. Sivakumar. 1992. Critical-state concepts for unsaturated soil. In Proc. 7th
Intl. Conference on Expansive Soils, 1: 167-172. Dallas, Tex.
Wheeler, S. J., and V. Sivakumar. 1993. Development and application of critical-state model for
unsaturated soil. In Predictive Soil Mechanics, ed. G. T. Houlsby and A. N. Schofield, 709-728.
London: Thomas Telford.
Wheeler, S. J., and V. Sivakumar. 1995. An elasto-plastic critical-state framework for unsaturated
soil. Géotechnique 45(1): 35-53.
Wiendieck, K. W. 1968. Stress-displacement relations and terrain-vehicle mechanics: A critical
discussion. J. Terramechanics 5(3): 67-85.
Williams, J., and C. F. Shaykewich. 1969. An evaluation of polyethylene glycol (P.E.G.) 6000 and
P.E.G. 20,000 in the osmotic control of soil water matric potential. Can. J. Soil Sci. 49: 397-
401.
Williams, J., and C. F. Shaykewich. 1970. The influence of soil water matric potential on the
strength properties of unsaturated soil. Soil Sci. Soc. Amer. Proc. 34(6): 835-840.
Wong, P. K. K., and R. J. Mitchell. 1975. Yielding and plastic flow of sensitive cemented clay.
Wood, D. M. 1982. Laboratory investigations of the behaviour of soils under cyclic loading: a
review. In Soil Mechanics – Transient and Cyclic Loads, ed. G. N. Pande and O.C. Zienkiewicz,
513-582. Chichester, England: Wiley.
Wood, D. M. 1990. Soil Behaviour and Critical-state Soil Mechanics. New York: Cambridge
University Press.
Woodburn, J. A., J. C. Holden, and P. Peter. 1993. The transistor psychrometer. A new instrument
for measuring soil suction. In Unsaturated Soils, ed. S. L. Houston and W. K. Wray, 91-102.
New York: ASCE.
Wulfsohn, D., B. A. Adams, and D. G. Fredlund. 1996. Application of unsaturated soil mechanics
for agricultural conditions. Can. Agric. Eng. 38(3): 173-181.
Wulfsohn, D., B. A. Adams, and D. G. Fredlund. 1998. Triaxial testing of unsaturated agricultural
soils. J. Agric. Eng. Res. 69(4): 317-330.
Xie, X., and D. Zhang. 1985. An approach to 3-D nonlinear FE simulative method for investigation
of soil-tool dynamic system. In Proc. Intl. Conference on Soil Dynamics, Volume 2, 412-427.
Auburn, Ala., June 1985.
Yong, R. N., and R. D. Japp. 1968. Stress-strain behavior of clays in dynamic compressions. In
Vibration Effects of Earthquakes on Soils and Foundations, ASTM STP 450, 233-246.
Philadelphia, Pa.: American Society for Testing and Materials.
Yong, R. N., and E. Windisch. 1970. Determination of wheel contact stresses from measured
instantaneous soil deformations. J. Terramechanics 7(3&4): 57-67.
Zeng, D., and Y. Yao. 1992. A dynamic model for soil cutting by blade and tine. J. Terramechanics
29(3): 317-327.
Zienkiewicz, O. C. 1976. Plasticity and some of its corollaries in soil mechanics: collapse and
continuing deformation under load repetition. Num. Meth. Geomechan. ASCE 3: 1275-1303.
Zienkiewicz, O. C., C. Humpheson, and R. W. Lewis. 1975. Associated and non-associated visco-
plasticity and plasticity in soil mechanics. Géotechnique 25(4): 671-689.
Zienkiewicz, O. C., and R. L. Taylor. 1988. The Finite Element Method, 4th ed. London: McGraw-
Hill.
Zur, B. 1966. Osmotic control of the matric soil-water potential: I. Soil-water system. Soil Sci.
102(6): 394-398.

Wulfsohn2002 PDF

Uploaded by

Copyright:

Available Formats

Wulfsohn2002 PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Wulfsohn2002 PDF

Uploaded by

Copyright:

Available Formats

Wulfsohn, D., and B. A. Adams. 2002. Elastoplastic soil mechanics.

In Advances in Soil Dynamics Volume 2, 1-

Part I. Constitutive Modeling of Soils

durations. Agricultural engineers are generally interested in the mechanical behavior of

CHARACTER OF CLASSICAL SOIL

Description of State Variables

Verification of Constitutive Relationship

Formulation of Solutions to Practical Problems

or equivalently, in the familiar engineering notation as:

where p is the mean or hydrostatic normal stress:

p = (σx + σy + σz)/3 = (σ1 + σ2 + σ3)/3 (1.6)

or, in terms of the principal stresses, by:

 ε11 ε12 ε13   ε x ε xy ε xz   ε 1 γ xy 1 γ xz 

System Stress Invariants Strain Invariants

General 3-dim J1 = tr σ = I1σ I1 = tr ε = I1ε

+ τ2xy + τ2yz + τ2zx

Octahedral plane σoct = J1/3 εoct = I1/3 = εv/3

Critical-state models p = σoct = (σ1 + σ2 + σ3)/3 εv = ε1 + ε2 + ε3

Cap models J1 I1p / 3

[a] Subscript D denotes invariant obtained from deviatoric tensor.

The strain tensor can be decomposed into two parts:

where w = gravimetric water content, and Sr = degree of saturation. The relationship

J1 = σoct = p = (σ1 + 2σ3)/3 (1.19)

dεoct = dεv = dε1 + 2dε3 (1.21)

ƒ (σ, σ& , ε, ε& ) = 0 (1.27)

[σ] = [C] [ε] (1.28)

[ε] = [D] [σ] (1.29)

where [D] is called the compliance matrix (or tensor).

Volumetric strain, ε ev Deviatoric strain, εse

The inverse relationship (cf. eq. 1.29) is given by:

For triaxial, cylindrical loading conditions, equation 1.38 reduces to:

Integrating equations 1.42 and 1.43, respectively, gives:

Figure 1.14—A hyperbolic representation of stress-strain curve for soils.

unloading-reloading response of soil. For normally consolidated soil, isotropic

(σ1 − σ3)f = Rφ (σ1 − σ3)ult (1.49)

Equation 1.48 can be rewritten as:

describing a surface in three-dimensional stress space. If the material is assumed to be

where h is the hardening parameter.

ƒ = ƒ(σij, Wp). (1.56)

ƒ = ƒ(σij, εp). (1.57)

In the critical-state model, hardening is assumed to be governed by the accumulated

plastic potential plastic potential

dWp = σ d εp > 0 (1.59)

Evidence for Yielding in Soils

Specific Yield Surfaces for Soils

|τf| = c + σn tan φ (1.61)

σ1 - σ3 = sin φ (σ1 + σ3 + 2c cot φ) (1.62)

it can be represented in principal stress space as an irregular hexagonal pyramid (figs

Figure 1.25—Comparison of some experimental results of measurement of strength of disturbed samples

Figure 1.27—Cap models to represent soil strain-hardening behavior.

Critical-State Soil Mechanics

The critical-state theory of soil mechanics was originally developed by researchers at

Volumetric strain εv = ε1 + 2ε3 (1.68)

Deviatoric (shear) strain εs = 2/3 (ε1 - ε3) (1.69)

State boundary critical- state line

sub-critical zone super-critical zone

The Critical-State Line

and, when projected on the v-p plane, by:

for triaxial compression (Schofield and Wroth, 1968).