1249

Challenges to modelling heave in expansive soils

Hung Q. Vu and Delwyn G. Fredlund

Abstract: There are challenges associated with the numerical modelling of unsaturated expansive soils. The challenges
are primarily related to the quantification of the void ratio constitutive surface, the characterization of the void ratio
constitutive surface at low stresses and (or) suction, and the solution of coupled equations with several nonlinear unsat-
urated soil property functions. This study suggests that the void ratio constitutive surface of an expansive soil subject
to a monotonic wetting path can be estimated from volume change indices obtained from conventional laboratory tests.
The constitutive surfaces for both the soil structure and the water phase can be described using mathematical equations
that allow net normal stress and suction to be reduced to zero. The solutions for two typical volume change problems
are presented using both a coupled approach and an uncoupled approach. The first example problem simulates water
leakage from a pipe under a flexible cover. The second example problem simulates the infiltration of water at ground
surface. The results of the analyses are in accordance with anticipated behaviour. The results also show that the an-
swers from an uncoupled analysis compared well with those from a coupled analysis. It is suggested that an uncoupled
analysis may be adequate for most prediction of heave problems involving unsaturated expansive soils.
Key words: heave prediction, numerical modelling, expansive soil, constitutive surface, uncoupled analysis, matric suc-
tion.
Résumé : Il y a des défis associés avec la modélisation numérique de sols non saturés gonflants. Les défis sont princi-
palement reliés à la quantification de la surface constitutive du rapport de vides, à la caractérisation de la surface cons-
titutive de rapport des vides à faibles contraintes et (ou) succion, et à la solution des équations couplées avec plusieurs
fonctions non linéaires des propriétés des sols non saturés. Cette étude suggère que la surface constitutive du rapport
des vides d’un sol gonflant soumis à un cheminement de mouillage monotonique peut être estimée à partir des indices
de changement de volume obtenus par des essais conventionnels en laboratoire. Les surfaces constitutives tant pour la
structure du sol que pour la phase d’eau peuvent être décrites au moyen d’équations mathématiques qui permettent de
réduire à zéro la contrainte normale et la succion nettes. On présente les solutions pour deux problèmes typiques de
changement de volume au moyen tant d’approche couplée que non couplée. Le premier exemple de problème simule la
perte d’eau d’un tuyau sous un couvert flexible. Le second exemple simule l’infiltration de l’eau à la surface du ter-
rain. Les résultats des analyses sont en accord avec le comportement anticipé. Les résultats montrent également que les
réponses de l’analyse non couplée se comparent bien avec ceux de l’analyse couplée. L’on suggère que l’analyse non
couplée peut être adéquate pour la plupart des problèmes de prédiction de soulèvement impliquant des sols gonflants
non saturés.
Mots clés : prédiction de soulèvement, modélisation numérique, sol gonflant, surface constitutive, analyse non couplée,
matrice de succion.

[Traduit par la Rédaction] Vu and Fredlund 1272

Introduction
The behaviour of an unsaturated expansive soil can be for-
mulated using the theory of unsaturated soils, formulated us-
ing two independent stress state variables, the constitutive
relationships for the soil structure and water phase, and flow
laws for the water phase (Alonso et al. 1990, Gens and
Alonso 1992, Fredlund and Rahardjo 1993, Vu and Fredlund
Received 29 September 2005. Accepted 15 May 2006.
Published on the NRC Research Press Web site at
http://cgj.nrc.ca on 5 January 2007.
H.Q. Vu.1,2 Clifton Associates Ltd., 340 Maxwell Crescent,
Regina, SK S4N 5Y5, Canada.
D.G. Fredlund. Golder Associates Ltd., 145 – 1st Ave N.,
Saskatoon, SK S7K 1W6, Canada.
1
Corresponding author (e-mail: hung_vu@golder.com).
2
Present address: Golder Associates Ltd., 145 – 1st Ave. N.,
Saskatoon, SK S7K 1W6, Canada
Can. Geotech. J. 43: 1249–1272 (2006) doi:10.1139/T06-073
2004). The two stress state variables are net normal stress,
(σ – ua), and matric suction (ua – uw), where σ is total normal
stress, ua is pore-air pressure, and uw is pore-water pressure.
Changes in the void ratio and degree of saturation of an un-
saturated soil can be expressed as functions of the stress state
variables to form two three-dimensional constitutive surfaces
(Matyas and Radhakrishna 1968).
The elasticity parameter functions required for a volume
change analysis can be computed from the constitutive sur-
faces. Testing techniques are available for directly measuring
the soil properties associated with the constitutive surfaces
(Fredlund and Rahardjo 1993; Pham et al. 2004); however,
the tests are costly, laborious, and time consuming and may
not be justifiable for routine engineering projects. Approxi-
mate volume change (and water content) coefficients should
be adequate for most engineering analyses. Procedures for
the approximation of the volume change coefficients assist
in the implementation of unsaturated soil mechanics into
geotechnical engineering practice (Fredlund 2000a).
© 2006 NRC Canada
1250 Can. Geotech. J. Vol. 43, 2006

The primary interactive processes involved with a volume volume, REV, assuming that water flow follows Darcy’s law
change analysis of an unsaturated expansive soil are stress with a nonlinear hydraulic conductivity (Vu and Fredlund
deformation and water flow. The assumption is made that 2004):
isothermal conditions exist and the air phase is continuous
and at atmospheric pressure. The stress-deformation process ∂ εv ∂ (ua − uw) ∂  x ∂  uw 

[1] βw1 + βw2 = kw + Y
is governed by static equilibrium, while the water flow is ∂t ∂t ∂ x  ∂ x  ρ wg 

governed by water continuity equation. Solutions for volume
change require that both the equilibrium equation and the
∂  y ∂  u 
continuity equation be solved. The solutions can be obtained + kw  w + Y 
∂y  
using either a coupled or an uncoupled approach.  ∂ y  ρ wg 
Coupled solutions are difficult to obtain in part because of
the nonlinear soil property functions associated with both where kwx and kwy are the hydraulic conductivities in the x-
water flow and stress deformation. An uncoupled solution is and y-directions, respectively; t is time; ρ w is the density of
obtained by independently solving the water flow process water, g is the gravitational acceleration, Y is the elevation
and the stress-deformation process. Uncoupled solutions can head, and
be more easily achieved than coupled solutions because the m1w E
soil property functions involved in each process (i.e., water βw1 = , or
flow or stress deformation) are considered to be independent m1s
(1 − 2µ) Ew
of one another. Coupling between the stress deformation and
m1wm 2s 1 3E
the water flow process takes place under transient condi- βw2 = m 2w − , or −
tions. Under steady-state conditions, the coupling disappears m1s Hw (1 − 2µ)HEw
and pore-water pressure conditions can be determined
where E is an elasticity parameter for the soil structure with
through the solutions of the continuity equation.
respect to a change in the net normal stress; H is an elastic-
The objective of this paper is to investigate difficulties
ity parameter for the soil structure with respect to a change
that have been encountered when modelling heave in expan-
in matric suction; Ew is an elasticity parameter for the
sive soils when using unsaturated soil theory. The specific
change in the amount of water in the soil with respect to a
objectives of this paper involve the: (i) estimation of the void
change in the net normal stress; Hw is an elasticity parameter
ratio constitutive surface of an unsaturated expansive soil
for the change in the amount of water in the soil with respect
from swelling indices obtained from conventional oedometer
to a change in matric suction; µ is Poisson’s ratio; coeffi-
tests; (ii) review of existing constitutive surface equations
cients of volume change, m1s and m 2s, are slopes of the soil
for both the soil structure and the water phase; (iii) presenta-
structure constitutive surface; and coefficients of volume
tion and solution of problems associated with low net nor-
changes, m1w and m 2w, are slopes on the constitutive surface
mal stresses and (or) low matric suctions when an expansive
for the water phase. The general form for the unloading con-
soil swells; (iv) examination and evaluation of proposed
stitutive surfaces for both the soil structure and the water
equations for the constitutive surfaces; and (v) presentation
phase are shown in Fig. 1.
and comparison of uncoupled and coupled solutions for two
The governing partial differential equations for soil struc-
typical volume change problems using the constitutive sur-
ture equilibrium, written in terms of displacements in the x-
faces suggested in this study.
and y-directions (i.e., u and v) for the plane strain case of an
isotropic, nonlinear elastic soil are as follows (Vu and
Governing partial differential equations Fredlund 2004):
Vu and Fredlund (2004) presented the formulations of ∂  ∂u ∂v  ∂  ∂u ∂v 
[2] c + c + c +
∂x  ∂x ∂y  ∂y  ∂y ∂x 
governing partial differential equations for both saturated– 11 12 33
unsaturated seepage and stress deformation. The equations
were derived based on the following assumptions: (i) the air ∂ (ua − uw)
phase is continuous and at atmospheric pressure; (ii) the soil − ds + bx = 0
∂x
is isotropic and nonlinear elastic (i.e., incrementally elastic);
(iii) strains are small; (iv) the pore water is incompressible; ∂  ∂u ∂v  ∂  ∂u ∂v 
and (v) the effects of air diffusing through water, air dissolv- [3] c33  +  +  c12 + c22 
ing in the water, and the movement of water vapour are neg- ∂x  ∂y ∂x  ∂y  ∂x ∂y 
ligible. The incremental elasticity equations have a form ∂ (ua − uw)
similar to those presented by Biot (1941) for a soil with oc- − ds + by = 0
∂y
cluded air bubbles.
Let us consider a two-dimensional field with x and y as where bx and by are body forces in the x- and y-directions,
the rectangular Cartesian coordinates (i.e., x for the horizon- respectively; and
tal direction and y for the vertical direction) and with u and v
(1 − µ) E
as components of the displacement vector (i.e., u and v for c11 = c22 =
the x- and y-directions, respectively). The governing partial (1 + µ) (1 − 2µ)
differential equation for water flow through a heterogeneous,
µE
anisotropic, saturated–unsaturated soil can be derived by sat- c12 =
isfying conservation of mass for a representative elemental (1 + µ) (1 − 2 µ)

© 2006 NRC Canada

Vu and Fredlund 1251

Fig. 1. Constitutive surfaces for (a) soil structure and (b) water phase of an unsaturated expansive soil.

E dεv 1 de
c33 = [5] m 2s = =
2 (1 + µ) d (ua − uw) 1 + e0 d (ua − uw)

E where dεv = de/(1 + e0), e0 is the initial void ratio of the soil
ds = (i.e., referential element), e is the void ratio of the soil, and σmean
(1 − 2µ) H
is the mean net total stress (i.e., σmean = (σx + σ y + σ z)/3).
The elasticity parameters for the soil structure, E and H,
The seepage equation (eq. [1]) shows the influence of the can be calculated as follows:
compressibility and the rate of the volume change of the soil
structure on the transient water flow process in an expansive (1 − 2µ)
[6] E =3
soil. Solutions to the seepage equation (eq. [1]) and soil m1s
structure equilibrium equations (eqs. [2] and [3]) can be ob-
tained through the use of either an uncoupled or a coupled 3
solution. Procedures for uncoupled and coupled analyses are [7] H =
m 2s
presented later in this paper.
The coefficients of water volume change, m1w and m 2w, can
be obtained by differentiating the constitutive surface for the
Soil properties required for volume change water phase
predictions in an expansive soil
d θw
Soil properties required for a volume change analysis in- [8] m1w =
d (σmean − ua )
clude: (i) Poisson’s ratio, µ; (ii) an elasticity parameter (i.e.,
elasticity function) for the soil structure with respect to net d θw
normal stress, E; (iii) an elasticity parameter (i.e., elasticity [9] m 2w =
d (ua − uw)
function) for the soil structure with respect to matric suction,
H; (iv) an elasticity “type” parameter for the water phase where θw = Vw/V0 is the volumetric water content.
with respect to net normal stress, Ew; (v) an elasticity “type” The coefficients of water volume change can also be cal-
parameter for the water phase with respect to matric suction, culated from the void ratio constitutive surface and degree of
Hw; and (vi) the hydraulic conductivity function, kw. The soil saturation, S, constitutive surface as follows:
properties are functions of both net normal stress and matric
suction. Assuming a value of Poisson’s ratio, the elasticity S de e dS
[10] m1w = +
parameters, E, H, Ew, and Hw, can be calculated from the co- 1 + e0 d (θmean − ua ) 1 + e0 d (σmean − ua )
efficient of volume change, m1s, m 2s, m1w, and m 2w, respec-
tively. S de e dS
[11] m 2w = +
At a particular stress state, the coefficients of volume 1 + e0 d (ua − uw) 1 + e0 d (ua − uw)
change, m1s and m 2s, can be obtained by differentiating the
constitutive surface for the soil structure and written as a The elasticity parameters for the water phase, Ew and Hw,
function of void ratio: can be calculated as follows:
dεv 1 de 3
[4] m1s = = [12] Ew =
d (σmean − ua ) 1 + e0 d (σmean − ua ) m1w

© 2006 NRC Canada

1252
1 rithm of matric suction converge at a net normal stress equal
[13] Hw =
m 2w
It is suggested that the void ratio surface can be estimated
from the swelling indices obtained from extreme planes (i.e.,
net normal stress plane and matric suction plane). The elas-
ticity parameters can then be calculated from the estimated
void ratio constitutive surface. Soil data required to estimate
the void ratio surface are shown in Fig. 2; namely, (i) the
swelling index with respect to net normal stress, Cs; (ii) the
swelling index with respect to matric suction, Cm; (iii) the
initial void ratio, e0, and (iv) the swelling pressure, Ps. The
swelling indices, Cs and Cm, are the slopes of the void ratio
versus logarithm of net normal stress and logarithm of
matric suction, respectively.
Estimation of void ratio constitutive surface
from volume change indices
The following assumptions are made with regard to the
void ratio constitutive surface: (i) there is a linear relation-
ship of void ratio versus logarithm of net normal stress at the
extreme net normal stress plane (i.e., suction equal to s0);
(ii) there is a linear relationship of void ratio versus loga-
rithm of matric suction at an extreme suction plane (i.e., net
normal stress equal to (σ – ua)0); (iii) the void ratio versus
logarithm of net normal stress and void ratio versus loga-
 P Cs / Cm
s0  s  × 10 ( e − e0 )/ Cm − 1 [Ps × 10 ( e − e0 )/ Cs − (σ − ua )]
 a 
[14] (ua − uw) =  ( e − e0 )/ Cs
Ps × 10
The net normal stress and matric suction in eq. [14] can
be obtained either from a one-dimensional loading test (i.e.,
(σy − ua) and (ua – uw)) or from an isotropic loading test
(i.e., (σmean – ua) and (ua – uw)). While the swelling indices
are essentially the same for both loading conditions (Graham
and Li 1985; Al-Shamrani and Al-Mhaidib 2000; Vu 2003),
appropriate values for the parameters, (σ – ua)0 and s0, must
be used.
A mathematical equation is needed to describe the entire
void ratio surface for modelling and analysis purposes. This
equation can be used to fit experimental data or data gener-
ated from swelling indices (i.e., data generated from
eq. [14]).
Review of equations for the void ratio constitutive
relationship
Mathematical equations have been used to describe the
variation in void ratio or water content with respect to
changes in the stress state of a soil. These mathematical
equations can be divided into two groups. The first group in-
cludes those equations that are capable of representing the
shape of the soil property function. These equations are de-
termined for a particular soil under specific testing condi-
tions using a best fit of measured data. These equations
contain fitting parameters that need to be determined. Exam-
ples are the equations proposed to best fit the properties of
Can. Geotech. J. Vol. 43, 2006
to (σ – ua)0 (i.e., to be defined) and a suction equal to s0 (i.e.,
to be defined), and (iv) a constant void ratio plane intersects
the void ratio constitutive surface as a straight line on an
arithmetic plot of the stress state variables.
The assumption of a linear relationship between net nor-
mal stress and matric suction at a constant void ratio on an
arithmetic scale is supported by the work of Escario (1969)
(Fig. 3) and Matyas (1969) (Fig. 4). The same lines form as-
ymptotic curves on a logarithmic scale. The shape of these
lines suggests the form for the void ratio surface on a three-
dimensional plot.
Figure 5 illustrates the first, second, and third assumptions
mentioned above. These assumptions are reasonable and
widely accepted (Holtz and Gibbs 1956; Richards et al.
1984; Ho et al. 1992; Fredlund and Rahardjo 1993).
Swelling pressures, Ps, obtained at different suctions are not
the same; however, at low matric suctions, the difference is
negligible. Figures 5 and 6 are used to derive the void ratio
relationship with respect to net normal stress and matric suc-
tion. It should be noted that Fig. 5 is on a semilogarithmic
scale while Fig. 6 is on an arithmetic scale.
The derivation of an equation for the calculation of matric
suction corresponding to any void ratio and net normal
stress, or the calculation of net normal stress from the void
ratio and matric suction is presented in detail in Appendix
A. The equation has the following form:


+ s0
− (σ − ua ) 0
soils related to the water phase such as the soil-water char-
acteristic curve (SWCC) and the hydraulic conductivity
functions. These equations can be used to describe a wide
range of soil properties. The Fredlund and Xing (1994)
equation, along with a correction factor, can be used to de-
scribe the SWCC with soil suctions varying up to a limiting
value of 106 kPa.
The second group includes equations that use other soil
properties as parameters in the equation. Examples include
the equations for shear strength (i.e., using cohesion and an-
gle of shearing resistance) and volume changes (i.e., using
volume change indices). The parameters in these equations
usually have a clear physical meaning; however, the equa-
tions might not be suitable to describe the entire range of
stress states. The linear equation on a semilogarithmic plot,
which is commonly used to define the volume change index,
cannot be used at extremely low net total stress conditions
and low soil suctions. These equations need to be modified
to provide greater flexibility to best fit measured data.
Soil properties for an unsaturated soil are known to be
functions of both stress state variables (i.e., net total stress
and matric suction). Sometimes equations are proposed to
relate a soil property to only one dominant stress state vari-
able. These equations can be expanded to be functions of
both stress state variables provided the soil properties can be
evaluated. This section reviews mathematical equations that
© 2006 NRC Canada
Vu and Fredlund
Fig. 2. Soil property functions required for volume change analysis of an unsaturated expansive soil.
can be used to describe the void ratio constitutive surfaces
associated with an unsaturated swelling soil.
Terzaghi (1943) and Casagrande (1936) noted that the vir-
gin compression curve and the rebound curve of the void ra-
tio versus logarithm of effective stress form essentially a
straight line. The void ratio equation has the following form:
[15] e = a + b log(σ)
where σ is the designation of the stress state, and a and b are
constants (i.e., fitting parameters).
Fredlund (1979) suggested that the void ratio constitutive
surface for an unsaturated soil could be linearized over a
wide range of stress changes using the logarithm of the
stress state variables. The void ratio under any set of stress
conditions has the following form:
[16] e = a + b log(σ − ua ) + c log(ua − uw)
where a, b, and c are constants (i.e., fitting parameters).
Lloret and Alonso (1985) studied a number of mathemati-
cal equations for the description of the volume change con-
stitutive surface of unsaturated soils subjected to confined or
isotropic compression. The equations were used to best fit
experimental results on different soil types, and the optimum
equations were selected on the basis of minimum fitting er-
rors. For a limited range in the total stress, Lloret and
Alonso (1985) suggested that a suitable analytical expres-
sion for the void ratio constitutive surface is as follows:
[17] e = a + b (σ − ua ) + c log(ua − uw)
+ d (σ − ua ) log(ua − uw)
1253
If the range in the stress variation is large, a more suitable
equation for the void ratio surface was given by (Lloret and
Alonso 1985)
[18] e = a + b log(σ − ua ) + c log(ua − uw)
+ d log(σ − ua ) log(ua − uw)
Fredlund (2000b) proposed a four-parameter equation that
can be used to represent an overconsolidated soil that con-
tains both a recompression curve and a virgin compression
branch. The equation of void ratio as a function of stress has
the following form for particular soil suction:
  σ 2    σ 2 
[19] e = a + b ln 1 +    + d ln 1 +   
  c     f  
where a, b, c, d, and f are constants (i.e., fitting parameters)
at a constant suction.
Low net normal stress and zero suction
problem
A practical engineering problem is encountered in the low
net normal stress and zero suction ranges when the elastic
moduli are calculated from the swelling index, Cs, or other
volume change indices as illustrated in Fig. 7. On the ex-
treme planes, the relationships between void ratio and the
logarithm of net normal stress or matric suction are essen-
tially linear (Fig. 7a). When the semilogarithm relationship
is converted to an arithmetic plot (i.e., void ratio versus net
normal stress or matric suction), the void ratio tends to in-
crease to infinity as the stress state variable approaches zero
(Fig. 7b). Therefore, the calculated elastic modulus with re-
© 2006 NRC Canada
1254 Can. Geotech. J. Vol. 43, 2006

Fig. 3. Cross-sections of void ratio surfaces plotted on arithmetic scale and logarithmic scale for a remoulded Madrid clay (modified from
Escario 1969).

Fig. 4. Matric suction versus isotropic net normal stress at con-

stant void ratio (modified from Matyas 1969).
spect to changes in net normal stress, E, becomes extremely
small (and approaches zero) at low net normal stresses
(Fig. 7c). The elastic modulus with respect to a change in
matric suction, H, also becomes extremely small at low
matric suctions (or zero suction) (Fig. 7c). These unrealisti-
cally small values of elastic moduli at low net normal stress
and matric suction result in an unstable solution in numeri-
cal modelling. As well, the solutions produce unreasonably
large deformations.
A part of this study was directed at finding a procedure
that could adequately describe void ratio changes at low net
normal stresses and matric suctions. The procedure should
involve the use of an equation in a mathematically continu-
ous form that would adequately describe the void ratio in the
low net normal stress range and low matric suction range.
The equation should also produce a continuous, smooth,
physically reasonable, and differentiable form over the entire
range of net normal stresses and matric suctions.
The equation can then be used to fit a set of measured
data or data calculated from swelling indices. This equation
needs to be able to include a value of void ratio at zero suc-
tion and zero net normal stress for modelling and analysis
purposes as illustrated in Fig. 8. Using this approach, unreal-
© 2006 NRC Canada
Vu and Fredlund 1255

Fig. 5. Linear relationships between void ratio and logarithm of stress state variables at extreme planes.

Fig. 6. Relationship between net normal stress and matric suction at constant void ratio plane.

istically large values of void ratio at low net normal stresses It can be noted that parameter a is the void ratio at zero
and matric suctions can be avoided. net normal stress and matric suction; if this parameter is
known, then the number of fitting parameters is reduced by
Proposed equations for void ratio one. Parameters c, d, f, and g take on non-negative values to
satisfy the logarithmic equation. A higher number of fitting
constitutive surface
parameters will increase the fit of the equation to the data,
Vu (2003) proposed six functions to fit the void ratio con- but the physical interpretation of the parameters is rapidly
stitutive surface of an unsaturated, expansive soil. Two of the lost (Lloret and Alonso 1985).
functions; namely unsat-1 with three fitting parameters
(eq. [20]) and unsat-6 with six fitting parameters (eq. [21]) Evaluation of equations for the void ratio
are as follows: constitutive surface
[20] e = a + b log[1 + (σ − ua ) + c (ua − uw)] Equations for the void ratio constitutive surfaces were eval-
1 + c (σ − ua ) + d (ua − uw)  uated to determine the most suitable equation for a swelling
[21] e = a + b log   soil analysis. Soil data for Regina clay (Shuai 1996) was used
1 + f(σ − ua ) + g (ua − uw)  to evaluate these equations for the void ratio surface. The ex-
© 2006 NRC Canada
1256
Fig. 7. Illustration of the problem associated with low net nor-
mal stress and zero suction.
perimental data are obtained under K0-loading conditions. A
dataset for one-dimensional loading conditions was gener-
ated, using eq. [14], from the swelling indices for Regina
clay. Data used to generate this dataset include: (i) initial
void ratio, e0 = 0.955; (ii) swelling index with respect to
change in net normal stress, Cs = 0.088; (iii) swelling index
with respect to change in matric suction, Cm = 0.080; and
(iv) swelling pressure, Ps = 320 kPa. The two curves on the
extreme planes meet at a vertical net normal stress equal to
1 kPa and a suction equal to 1 kPa. The results are presented
in Fig. 9 using both a semilogarithmic scale and an arithme-
tic scale. This dataset corresponds to wetting unloading test
conditions.
SigmaPlot software (Systat 2000) was used for fitting the
equations to the data points. The R2 criterion and Akaike In-
formation Criterion (AIC) (Akaike 1974) were used to eval-
uate the mathematical equations. The AIC criterion takes
into account the number of data points, the sum of the
weighted squared residuals, and the number of fitting param-
Can. Geotech. J. Vol. 43, 2006
eters. The value of R2 closest to 1, with the lowest value of
AIC, indicates the best fit.
Several equations are considered for the best-fit analysis
of the void ratio data generated from the swelling indices;
namely Fredlund (1979); Lloret and Alonso (1985);
Fredlund (2000b); and the equations suggested in this study
(i.e., unsat-1 and unsat-6 functions). The best-fit results
from different void ratio equations to the generated void ra-
tio data (Fig. 9) are presented in Table 1. The Fredlund
(2000) equation has been modified to allow best fit in three
dimensions. Statistical results on the proposed void ratio
equations are presented in Table 2. Figure 10 presents the
comparison of different equations using both the AIC crite-
rion and the R2 criterion. It can be seen that the unsat-6
function appears to provide the best fit to this set of data.
The fitting results using the unsat-6 function are shown in
Fig. 11.
Vu (2003) also evaluated mathematical equations for the
description of the measured void ratio data and measured
water content data for the constitutive surfaces. The results
of the evaluation suggest that the unsat-6 function is satis-
factory for describing both the soil structure and water phase
constitutive surfaces.
Example problems using the soil properties
of Regina clay
Two example problems are used to illustrate the suggested
approach to volume change analyses. Experimental data ob-
tained from tests on compacted specimens of Regina clay
are used for the analysis. The void ratio surface for three-
dimensional unloading conditions is estimated from the
swelling indices (eq. [14]) and described using the unsat-6
equation (eq. [21]). The initial matric suction in the soil
mass is assumed to be constant and equal to 400 kPa for
both examples.
Soil properties characterization
Lytton (1994) presented typical values for coefficients of
earth pressure at-rest, K0, back calculated based on field ob-
servations of heave and shrinkage. A value of the coefficient
of earth pressure at rest equal to 0.67 was suggested for wet-
ting conditions when cracks in the soil are essentially closed.
A Poisson’s ratio equal to 0.4 is calculated under this con-
sideration from the following relation:
K0
[22] µ=
1 + K0
Void ratio data associated with general three-dimensional
swelling were generated from the following parameters: ini-
tial void ratio, e0 = 0.955; swelling index with respect to
change in net normal stress, Cs = 0.088; swelling index with
respect to change in matric suction, Cm = 0.080; swelling
pressure, Ps = 320 kPa; and the two curves on extreme
planes meet at vertical net normal stress equal to 1 kPa and a
suction equal to 1 kPa.
Figure 12a presents a set of generated void ratios in a
semilogarithmic plot with an assumed Poisson’s ratio equal
to 0.4. The best-fit surface for the void ratio dataset is pre-
sented in an arithmetic plot in Fig. 12b. The degree of satu-
© 2006 NRC Canada
Vu and Fredlund 1257

Fig. 8. Proposed procedure to solve the low net normal stress and suction problem.

Fig. 9. Generated data for the void ratio constitutive surface for Regina clay. e0 = 0.955; Cs = 0.088; Cm = 0.080; Ps = 320 kPa.

© 2006 NRC Canada

1258 Can. Geotech. J. Vol. 43, 2006

Table 1. Best-fit results of different void ratio equations to void ratio data for Regina clay.
ID Function Fitting parameters
Fredlund (1979) e = a + b log ( σ − ua ) + c log ( ua − uw ) a = 1.071; b = –0.027; c = –0.024
Lloret and Alonso (1985) e = a + b log ( σ − ua ) + c log ( ua − uw ) a = 1.186; b = –0.091; c = –0.083; d = 0.034
+ d log ( σ − ua ) log ( ua − uw )
 a1 = 1.205; a2 = –5.2×10–5; b1 = –0.724;
 σ    σ 
Fredlund (2000b) 2 2
e = a + b ln 1 +    + d ln 1 +    b2 = 0.292; c = 0.011; d1 = 0.705;
  c     f   d2 = –0.287; f = 0.010
where a = a1 + a2(ua – uw); b = b1 + b2 ln (ua – uw);
d = d1 + d2 ln (ua – uw)
Unsat-1 e = a + b log [1 + ( σ − ua ) + c( ua − uw )] a = 1.186; b = –0.092; c = 0.610
Unsat-6 1 + c ( σ − ua ) + d ( ua − uw )  a = 1.350; b = –0.095; c = 49.722;
e = a + b log   d = 34.571; f = 4.38×10–4;
 1 + f( σ − ua ) + g ( ua − uw )  g = 8.22×10–4

Table 2. Statistical results of void ratio constitutive equations for analyses. Table 4 presents various parameters associated
void ratio data for Regina clay. with plane strain coupled and uncoupled approaches.
Number of fitting
ID parameters R2 AIC Coupled approach
Fredlund (1979) 3 0.5299 –102 In the coupled approach, the water phase continuity (i.e.,
Lloret and Alonso (1985) 4 0.9469 –174 seepage) equation and the equilibrium (i.e., stress deforma-
Fredlund (2000b) 8 0.9622 –179 tion) equations are solved simultaneously; meaning that the
Unsat-1 3 0.9890 –228 dynamic interdependence between the seepage and deforma-
Unsat-6 6 0.9979 –288 tion problems is fully realized. Except for Poisson’s ratio, all
other soil parameters (i.e., E, H, Ew, Hw, and kw) are consid-
Note: AIC, Akaike Information Criterion.
ered to be functions of both net normal stress and matric
suction. Boundary conditions for both the water continuity
equation (i.e., pore-water pressure and water flux) and equi-
ration surface can be calculated for general three- librium equations (i.e., displacements and load) must be de-
dimensional swelling as shown in Fig. 12c. fined. The results show displacements, induced stresses,
The hydraulic conductivity function is predicted from the pore-water pressures, and water fluxes obtained at any time
saturated hydraulic conductivity and the SWCC (Fredlund during the transient process.
and Rahardjo 1993). Shuai (1996) presented the hydraulic
conductivity function for Regina clay using the Gardner
(1958) equation Uncoupled approach
kw0e b In the uncoupled approach, the water phase continuity
[23] kw = n (i.e., seepage) equation is solved independently from the
(u − uw)  equilibrium (i.e., stress deformation) equations. The interde-
1+a  a  pendent solutions of the equations is undertaken using an it-
 ρ wg  erative manner where the flow portion of the formulation is
where kw0 = 0.4 × 10–8 m/s; e is void ratio; b = 18.5; a = solved for a given time period and the resultant pore-water
0.01; and n = 1.1. pressure changes are used as input in a deformation analysis.
The hydraulic conductivity constitutive surface is pre- In turn, volume changes and induced stresses from the de-
sented graphically in Fig. 12d for the void ratio surface as formation analysis are used in the computation of the soil
shown in Fig. 12a. properties for the next time period in the seepage analysis.
The fitting results for the generated void ratio and degree In the uncoupled approach, the dependent variables and
of saturation constitutive surfaces are shown in Table 3. The nonlinear soil properties are separated into two analyses;
coefficients of volume change associated with the generated namely, a seepage analysis and a stress-deformation analy-
void ratio data (Fig. 12a) and an assumed degree of satura- sis. For the seepage analysis, the dependent variable is pore-
tion surface (Fig. 12c) are presented in Fig 13. The elasticity water pressure. At each given time period, the elasticity pa-
parameter functions are presented graphically in Fig 14. rameters, E and H, for soil structure are calculated at initial
conditions of current period and assumed to remain un-
Coupled and uncoupled analyses changed over the time increment. Net normal stress is as-
sumed to be unchanged in the seepage analysis; therefore,
Equations [1], [2], and [3] can be solved for three vari- the elasticity parameters, Ew and Hw, for the water phase, as
ables; namely, horizontal displacement, u; vertical displace- well as the hydraulic conductivity, kw, are taken to be func-
ment, v; and pore-water pressure, uw. Swelling is a time- tions of matric suction. Boundary conditions for the seepage
dependent process involving nonlinear soil properties. Initial analysis can be either pore-water pressure (or hydraulic
conditions must be specified for both coupled and uncoupled head) or water flux. The results of the seepage analysis pro-

© 2006 NRC Canada

Vu and Fredlund
Fig. 10. Averaged AIC and R2 criterion for generated data for the void ratio constitutive surface.
Fig. 11. Best-fit void ratio constitutive surface using proposed unsat-6 function for generated void ratio data of Regina clay.
vide the pore-water pressure and water flux with time for the
time period under consideration.
The dependent stress-deformation variables are horizontal
displacement, u and vertical displacement, v. In addition to
Poisson’s ratio, two elasticity parameters, E and H, are re-
quired for the soil structure. These properties are a function
of matric suction when the initial net normal stress is un-
changed. The elasticity parameters, Ew and Hw, for the water
phase and hydraulic conductivity, kw, are no longer needed
for stress deformation analysis. Boundary conditions for the
stress-deformation analyses can be of the displacement type
or load type. Results of the stress-deformation analysis pro-
vide displacements and induced stresses due to applied
boundary conditions and changes in pore-water pressure.
Solutions using the uncoupled approach depend on the
magnitude of selected time periods for the seepage analysis.
Short time periods allow the stress state in the soils and the
soil properties to be described more accurately. Conse-
quently, the results provide more accurate values for pore-
water pressures and displacements.
Several approaches have been tested in the uncoupled
analysis. Differences arise from the stress state used for the
seepage and stress-deformation analyses. Uncoupled solu-
tions obtained from four types of analyses are called UCS1,
UCS2, UCS3, and UCS4. The solution obtained using a cou-
1259
pled approach is named as CS. The stress paths followed in
each type of analysis are shown in Figs. 15a and 15b for the
seepage analysis and the stress-deformation analysis, respec-
tively. It should be noted that the initial stress state may vary
from one type of uncoupled analysis to another while bound-
ary conditions remain the same.
In the UCS1 uncoupled analysis, the soil properties ob-
tained for the extreme plane are used for the entire constitu-
tive surface. Therefore, net normal stress is not taken into
consideration for both the seepage and stress-deformation
analysis. Volume changes and changes in stress were not
considered in the seepage analysis. A net mean stress of
0.78 kPa (i.e., the mean stress at which the volume change
index, Cm, was obtained) was used for the UCS1 uncoupled
analysis.
In the UCS2 uncoupled analysis, the variation in the stress
state in the soil profile under initial stress state conditions
was considered. Volume changes and changes in net normal
stress were not considered in the seepage analysis. Change
in net normal stress was not considered in the stress-
deformation analysis.
In the UCS3 uncoupled analysis, the assumptions related
to the seepage analysis are the same as those used in the
UCS2 analysis; however, changes in net normal stress are
considered in the stress-deformation analysis.
© 2006 NRC Canada
1260
Fig. 12. Constitutive surfaces for Regina clay.
In the UCS4 uncoupled analysis, changes in volume and
changes in net normal stress are considered in both the seep-
age and stress-deformation analysis. It should be noted that
variations in the stress state in the soil better reflect the
stress state conditions related to the problem.
The examples presented in this paper are analyzed using
various types of uncoupled analyses and the results are then
compared with the coupled solutions.
Computer programs
A finite element computer program, COUPSO, was used
to obtain the coupled solution and a general-purpose partial
differential equation solver, FlexPDE (PDE Solutions Inc.
2001), was used to obtain uncoupled solutions in this study.
The COUPSO program was developed by Pereira (1996) to
solve a coupled problem involving a small earth dam associ-
ated with unsaturated collapsing soils. Vu (2003) modified
the program to accommodate a coupled analysis for volume
change problems associated with an unsaturated expansive
Can. Geotech. J. Vol. 43, 2006
soil. A detailed description of the COUPSO program and its
verification were presented by Pereira (1996) and Vu
(2003).
FlexPDE is a scripted finite element model builder and
numerical solver for both two- and three-dimensional prob-
lems. FlexPDE performs the operations necessary to turn a
description of the partial differential equation system into a
finite element model and then solves the system of equa-
tions. FlexPDE has automatic mesh generation and refine-
ment, adaptive time step design, and time refinement. A
more detailed description of the FlexPDE program can be
found in PDE Solution Inc. (2001).
Computer results and discussions
Example 1: Infiltration of water from ground surface
This example considers the hypothetical case of a 5 m
thick deposit of swelling clay. The surface is partially cov-
ered with a flexible cover. Figure 16 presents the geometry
and key variable for this problem. The transient wetting pro-
© 2006 NRC Canada
Vu and Fredlund 1261

Table 3. Fitting parameter results for general three-dimensional constitutive surfaces for Regina clay
Fitting parameters
Constitutive surface a b c d f g
Void ratio, e 1.2492 –0.0979 4.8240 3.3330 0.0009 0.0012
Degree of saturation, S 1.0000 –0.0725 0.0125 11.7265 0.0125 0.0071

Fig. 13. Coefficient of volume change functions for Regina clay.

cess is introduced by imposing a water infiltration rate equal
to 2 × 10–8 m/s at the uncovered portion of the ground sur-
face. Such a wetting condition simulates water infiltration
into the soil mass due to the watering of a lawn or a light
rain. The analysis is performed to observe the swelling soil
behaviour and matric suction changes as the transient wet-
ting front advances into the soil mass.
Figures 17 and 18 show the change of heave at ground
surface with time and the change of heave versus depth with
time, respectively, obtained from a coupled analysis. A max-
imum differential heave of about 30 mm can occur between
the cover location and the far point of the uncovered portion.
A maximum heave of about 58 mm could take place at the
surface and heave would essentially be complete after about
375 days of infiltration.
Figures 19 and 20 compare the changes of matric suction
and vertical displacement at the monitoring points in the soil
mass for various types of analysis. The matric suction below
the cover reduced rapidly in the first 50 days. The rate of
change in matric suction in the soil outside the cover re-

© 2006 NRC Canada

1262
Fig. 14. Elasticity parameter functions for Regina clay.
duced rapidly with time. Corresponding to the change in
suction, the horizontal stresses increased rapidly during the
first 50 days. Horizontal stresses increased more near ground
surface where the soil has a higher swelling potential. A to-
tal increase of about 58 kPa in horizontal stress was com-
puted for point A. A total increase of about 20 kPa in the
horizontal stress was computed for point C.
The changes in matric suction at any elapsed time appear to
be overpredicted in the uncoupled analyses, resulting in the
differences in horizontal and vertical displacements. The un-
coupled analyses in this study used elastic parameters at net
normal stresses that are lower than the actual net normal
stresses. The stiffness of an unsaturated expansive soil de-
creases with a decrease in net normal stress, resulting in a
larger amount of heave in uncoupled analyses. The magnitude
of the differences between uncoupled and coupled solutions,
therefore, depends on the stress paths followed in the uncou-
Can. Geotech. J. Vol. 43, 2006
pled solutions. The uncoupled solution, UCS4 appear to be
essentially the same as that of the coupled solution, CS.
Figures 21, 22, and 23 present and compare the distribu-
tions of matric suction, horizontal displacement, and vertical
displacement, respectively, in the soil profile at day 53. Imme-
diately after wetting was introduced below the uncovered sur-
face, water flowed downward and to the left in the soil
domain. Matric suction near the uncovered surface reduced to
less than 100 kPa. Horizontal displacements decreased with
depth with a maximum value of 11 mm at ground surface
near to the cover. About 33 mm of heave took place at the un-
covered location. The results of the analyses also show a sig-
nificant increase in horizontal stress near ground surface,
particularly within the uncovered portion. An increase of
about 60 kPa in the horizontal stress was predicted for the soil
near ground surface where the soil had a high potential to
swell. There were increases in vertical stress because of the
© 2006 NRC Canada
Vu and Fredlund 1263

Table 4. Summary of uncoupled and coupled model for two-dimensional swelling analysis associated with an expansive soil.
Description
Governing PDE(s)
Computer program
Dependent variables
Initial conditions
Soil properties as constants
Soil properties as functions
Boundary conditions (valued)
Boundary conditions (natural)
Output
Analysis for
Uncoupled approach
Seepage model
Water continuity equation
FlexPDE
Pore-water pressure, uw
(σmean – ua)i
(ua – uw)i
µ µ
E, H at [(σmean – ua)i ; (ua – uw)i]
Ew = fn(ua – uw) at (σmean – ua)i
Hw = fn(ua – uw) at (σmean – ua)i
kw = kw(ua – uw) at (σmean – ua)i
Pore-water pressure, uw
Water flux, q
Pore-water pressure, uw
A change in time, ∆t
Stress–strain model
Stress equilibrium equations
FlexPDE
Horizontal displacement, u
Vertical displacement, v
(σmean – ua)i
(ua – uw)i
E = fn(σmean – ua) at (ua – uw)i
H = fn(σmean – ua) at (ua – uw)i
Displacement, u, v
Applied load
Displacements, u, v
Resulting stresses
A change in suction, ∆uw
Coupled approach
Stress equilibrium equations
Water continuity equation
COUPSO
Horizontal displacement, u
Vertical displacement, v
Pore-water pressure, uw
(σmean – ua)i
(ua – uw)i
µ
E = fn[(σmean – ua), (ua – uw)]
H = fn[(σmean – ua), (ua – uw)]
Ew = fn[(σmean – ua), (ua – uw)]
Hw = fn[(σmean – ua), (ua – uw)]
kw = fn[(σmean – ua), (ua – uw)]
Displacement, u, v
Pore-water pressure
Applied load
Water flux
Displacements, u, v
Pore-water pressure, uw
Resulting stresses
Volumetric water content
A change in time, ∆t

Fig. 15. Stress path followed in seepage model and stress-deformation model of various types of uncoupled analyses.

Fig. 16. Illustration of the geometry and key variables for Example 1: Infiltration of water from ground surface.

© 2006 NRC Canada

1264 Can. Geotech. J. Vol. 43, 2006

Fig. 17. Change of heave at ground surface with time, Example 1: Infiltration of water from ground surface, coupled solution.

Fig. 18. Change of heave versus depth with time, Example 1: Infiltration of water from ground surface, coupled solution.

Fig. 19. Comparison of matric suction development with time for points A, B, and C; Example 1: Infiltration of water from ground
surface.

© 2006 NRC Canada

Vu and Fredlund 1265

Fig. 20. Comparison of vertical displacement development with time for points A, B, and C; Example 1: Infiltration of water from
ground surface.

Fig. 21. Comparison of matric suction distribution at day 53; Example 1: Infiltration of water from ground surface.

Fig. 22. Comparison of horizontal displacement distribution at day 53; Example 1: Infiltration of water from ground surface.

increase of the soil self weight due to saturation, but these
changes were insignificant in comparison with the change in
horizontal stress. Most of the increases in void ratio and de-
gree of saturation took place within the uncovered portion
near to the ground surface where infiltration took place.
Example 2: Leakage of water below a flexible cover
Example 2 considers the hypothetical case of a 5 m layer
of swelling clay below a flexible cover (Fig. 24). Initial
matric suction is taken to be constant throughout the depth
and equal to 400 kPa. It is assumed that a leaking water line

© 2006 NRC Canada

1266 Can. Geotech. J. Vol. 43, 2006

Fig. 23. Comparison of vertical displacement distribution at day 53; Example 1: Infiltration of water from ground surface.

Fig. 24. Illustration of the geometry and key variables for Example 2: Leakage of water below a flexible cover.

Fig. 25. Change of heave at ground surface with time, Example 2: Leakage under a flexible cover, coupled solution.

produced zero pore-water pressure under the cover. Defor-
mation and matric suction profiles versus time were com-
puted.
Figures 25 and 26 show the change of heave at ground
surface with time and the change of heave versus depth with
time, respectively, obtained from a coupled analysis. The
heave patterns indicate that most of the heave below the
cover occurred in the first 100 days after wetting com-
menced. Most of the differential heave (i.e., about 65 mm)
took place in the first 56 days. In the period from day 56 to
day 150, heave increased gradually at the same rate through-
out the entire soil mass. After this period, heave developed
faster along the right side of the soil mass. A maximum
heave of 112 mm was predicted at day 350.
© 2006 NRC Canada
Vu and Fredlund 1267

Fig. 26. Change of heave below the cover versus depth with time, Example 2: Leakage under a flexible cover, coupled solution.

Fig. 27. Comparison of matric suction development with time; Example 2: Leakage under a flexible cover.

Fig. 28. Comparison of horizontal displacement development with time; Example 2: Leakage under a flexible cover.

© 2006 NRC Canada

1268 Can. Geotech. J. Vol. 43, 2006

Fig. 29. Comparison of vertical displacement development with time; Example 2: Leakage under a flexible cover.

Fig. 30. Comparison of matric suction distribution at day 150; Example 2: Leakage under a flexible cover.

Fig. 31. Comparison of horizontal displacement distribution at day 150; Example 2: Leakage under a flexible cover.

Figures 27, 28 and 29 compare the changes of matric suc-
tion, horizontal displacement, and vertical displacement, re-
spectively, at a selected point in the soil mass for various
types of analyses. Figure 27 shows that water appeared to
move slightly faster in the uncoupled analyses, when
changes in stress and deformation were not considered in the
seepage equation (i.e., UCS2 and UCS3). The changes in
matric suction at any elapsed time were overpredicted in
these analyses resulting in slight differences in horizontal
and vertical displacements. It should be noted that changes

© 2006 NRC Canada

Vu and Fredlund
Fig. 32. Comparison of vertical displacement distribution at day 150; Example 2: Leakage under a flexible cover.
in stress were not considered in solution UCS2, but were
considered in solutions UCS3 and UCS4. The UCS4 uncou-
pled solution compares well with the coupled solution, CS.
Figures 30, 31, and 32 compare the distribution of matric
suction, horizontal displacement, and vertical displacement,
respectively, in the soil profile at day 150. At day 150, the
matric suction value below the cover reduced to less than
50 kPa. The wetting front reached the extreme ends of the
soil mass. Matric suctions varied from 0 kPa to 300 kPa in
the soil mass at this time. Cumulative horizontal and vertical
displacements had the same patterns as those at day 56;
however, soil in the right portion of the soil mass was
pushed to the left as matric suctions reduced significantly in
this portion. A maximum cumulative heave of 105 mm took
place at the centre of the cover. About 25 mm of heave could
be observed at the far side of the soil mass.
Concluding remarks
The difficulties associated with numerical modelling of an
unsaturated expansive soil have been presented and dis-
cussed. The characterization of the swelling portion of the
void ratio surface for an unsaturated expansive soil from
swelling indices provides a practical approach for engineer-
ing practice. Two mathematical equations were proposed to
describe the void ratio constitutive surface even when the net
normal stress and matric suction approached zero values.
The equations were shown to successfully characterize the
constitutive surfaces for both the soil structure and water
phase of Regina clay.
Solutions to the two example problems associated with
unsaturated expansive soils have been obtained using both
the uncoupled and the coupled analyses. Several types of un-
coupled analyses, arising from various assumptions related
to the stress path followed in the water flow analysis and
stress-deformation analysis in the uncoupled solutions have
been presented. The coupled analysis generally produces
slightly smaller displacements when compared to the uncou-
pled analyses. Changes in the stress and deformation associ-
ated with the water flow analysis in the swelling soils should
be taken into account as the soil approaches saturation. The
effect of the changes in stress and deformation appeared to
be minor when the soil has high matric suctions. The cou-
pled seepage and stress-deformation analysis provides a
1269
more rigorous understanding of the swelling behaviour of
expansive soils and forms a reference for the evaluation of
various uncoupled analyses. When changes in net normal
stress and volumetric deformation are considered in an un-
coupled analysis, the uncoupled and coupled solutions pro-
vide similar results.
References
Akaike, H. 1974. A new look at the statistical model identification.
Institute of Electrical and Electronics Engineers Transactions on
Automatic Control, 19(6): 716–723.
Alonso, E.E, Gens, A., and Josa, A. 1990. A constitutive model for
partially saturated soils. Géotechnique, 40(3): 405–430.
Al-Shamrani, M.A., and Al-Mhaidib, A.I. 2000. Swelling behav-
iour under eodometric and triaxial loading conditions. In Ad-
vances in Unsaturated Geotechnics: Proceedings of Sessions of
Geo, Denver, Colo., 5–8 August 2000. Edited by C.D.
Shackelford, S.L. Houston, and N.-Y. Chang. American Society
of Civil Engineers, Geotechnical Special Publication 99, Reston,
Va. pp. 344–360.
Biot, M.A. 1941. General theory of three-dimensional consolida-
tion. Journal of Applied Physics, 12(2): 155–164.
Casagrande, A. 1936. The determination of the preconsolidation
load and its practical significance. Discussion D-34. In Proceed-
ings of the 1st International Conference on Soil Mechanics and
Foundation Engineering, Cambridge, Mass. Vol. 3, pp. 60–64.
Escario, V. 1969. Determination of the geotechnical characteristics
of expansive soils. In Proceedings of the 2nd International Con-
ference on Expansive Soils, Texas A&M University, 17–19 Au-
gust 1969. Edited by S.J. Buchanan. Texas A&M Press, College
Station, Tex. pp. 114–120.
Fredlund, D.G. 1979. Second Canadian Geotechnical Colloquium:
Appropriate concepts and technology for unsaturated soils. Ca-
nadian Geotechnical Journal, 16(1): 121–139.
Fredlund, D.G. 2000a. The 1999 R.M. Hardy Lecture: The imple-
mentation of unsaturated soil mechanics into geotechnical engi-
neering. Canadian Geotechnical Journal, 37(5): 963–986.
Fredlund, D.G., and Rahardjo, H. 1993. Soil mechanics for unsatu-
rated soil. John Wiley & Sons, New York, 560 p.
Fredlund, D.G., and Xing, A. 1994. Equations for the soil-water
characteristic curve. Canadian Geotechnical Journal, 31(4):
521–532.
Fredlund, M.D. 2000b. The role of unsaturated soil property func-
tions in the practice of unsaturated soil mechanics. Ph.D. disser-
tation, University of Saskatchewan, Saskatoon, Sask.
© 2006 NRC Canada
1270
Gardner, W.R. 1958. Some steady state solutions of the unsaturated
moisture flow equation with application to evaporation from a
water table. Soil Science, 85: 228–232.
Gens, A., and Alonso, E.E. 1992. A framework for the behaviour
of unsaturated expansive clays. Canadian Geotechnical Journal,
29:1013–1032.
Graham, J., and Li, E.C.C. 1985. Comparison of natural and
remolded plastic clay. Journal of Geotechnical Engineering,
American Society of Civil Engineers, 111(7): 865–881.
Ho, D.Y.F., Fredlund, D.G., and Rahardjo, H. 1992. Volume
change indices during loading and unloading of an unsaturated
soil. Canadian Geotechnical Journal, 29(2): 195–207.
Holtz, W.G., and Gibbs, H.J. 1956. Engineering properties of ex-
pansive clay. Transactions of the American Society of Civil En-
gineers, 121: 641–663.
Lloret, A., and Alonso, E.E. 1985. State surfaces for partially satu-
rated soils. In Proceedings of the 11th International Conference
on Soil Mechanics and Foundation Engineering, San Francisco,
Calif., 12–16 August 1985. Edited by Publication Committee of
XI ICSMFE, A.A. Balkema, Rotterdam, The Netherlands.
Vol. 2, pp. 557–562.
Lytton, R.L. 1994. Prediction of movement in expansive clay. In
Vertical and Horizontal Deformations of Foundations and Em-
bankments: Proceedings of Settlement ‘94, College Station,
Texas, 16–18 June 1994. Edited by A.T. Yeung and G.Y. Feaalio.
American Society of Civil Engineers, Geotechnical Special Publi-
cation 40, pp. 1827–1845.
Matyas, E.L. 1969. Some properties of two expansive clays from
western Canada. In Proceedings of the 2nd International Confer-
ence on Expansive Soils. Texas A&M University, College Sta-
tion, Texas, 17–19 August 1969. Edited by S.J. Buchanan. Texas
A&M Press, College Station, Tex. pp. 263–278.
Matyas, E.L., and Radhakrishna, H.S. 1968. Volume change char-
acteristics of partially saturated soils. Géotechnique, 18(4): 432–
448.
PDE Solutions Inc. 2001. FlexPDE 3 Reference manual. PDE So-
lutions Inc., Antioch, Calif.
Pereira, J.H.F. 1996. Numerical analysis of the mechanical behav-
ior of collapsing earth dams during first reservoir filling. Ph.D.
dissertation, University of Saskatchewan, Saskatoon, Sask.
Pham, Q.H., Fredlund, D.G., and Padilla, J.M. 2004. Use of the
GCTS apparatus for the measurement of soil-water characteris-
tic curve. In Proceedings of the 57th Canadian Geotechnical
Conference, Quebec City, Que. 24–27 October 2004. pp. 1–6.
Richards, B.G., Peter, P., and Martin, R. 1984. The determination
of volume change properties in expansive soils. In Proceedings
of the 5th International Conference on Expansive Soils,
Adelaide, Australia, 21–23 May 1984. Edited by P.W. Mitchell,
Organized by the Australian Geomechanics Society and the In-
stitution of Engineers, Australia, National Conference Publica-
tion No. 84/3. pp. 179–186.
Shuai, F. 1996. Simulation of swelling pressure measurements on
expansive soils. Ph.D. dissertation, University of Saskatchewan,
Saskatoon, Sask.
Systat Software Inc. 2000. Sigmaplot user’s guide. Systat Software
Inc., Point Richmond, Calif.
Terzaghi, K. 1943. Theoretical soil mechanics. John Wiley & Sons,
New York.
Vu, H.Q. 2003. Uncoupled and coupled solutions of volume change
problems in expansive soils. Ph.D. thesis, Department of Civil
Engineering, University of Saskatchewan, Saskatoon, Sask.
Vu, H.Q., and Fredlund, D.G. 2004. The prediction of one-, two-,
and three-dimensional heave in expansive soils. Canadian
Geotechnical Journal, 41: 713–737.
Can. Geotech. J. Vol. 43, 2006
List of symbols
a, b, c, d, e, f, g fitting parameters
bx, by body forces in x- and y-directions, respectively,
[ML–2T–2]
Cm swelling index, slope of the void ratio versus
logarithm of matric suction
Cs swelling index, slope of the void ratio versus
logarithm of net normal stress
e void ratio of the soil, [decimal]
ef initial void ratio of the soil at a net normal stress
equal to (σ – ua)0 and a matric suction equal to
s0, [decimal]
e0 initial void ratio of the soil, [decimal]
E elasticity parameter function for the soil struc-
ture with respect to a change in net normal
stress, [ML–1T–2]
Ew elasticity parameter function for the change in
the amount of water in the soil with respect to a
change in net normal stress, [ML–1T–2]
g gravitational acceleration, [LT–2]
Gs specific gravity
H elasticity parameter function for the soil struc-
ture with respect to a change in matric suction,
[ML–1T–2]
Hw elasticity parameter function for the change in
the amount of water in the soil with respect to a
change in matric suction, [ML–1T–2]
x y
kw kw the hydraulic conductivities in the x- and y-di-
,
rections, respectively, [LT–1]
K0 coefficient of earth pressure at rest
m1w coefficient of water volume change with respect
to a change in net normal stress, [M–1LT2]
m2w coefficient of water volume change with respect
to a change in matric suction, [M–1LT2]
s
m1 coefficient of volume change with respect to a
change in net normal stress, [M–1LT2]
m2s coefficient of volume change with respect to a
change in matric suction, [M–1LT2]
Ps swelling pressure, [ML–1T–2]
s0 matric suction at which the swelling index Cs
was obtained, [ML–1T–2]
ss swelling suction (i.e., swelling pressure equiva-
lent value of on the suction plane, [ML–1T–2]
S degree of saturation, [decimal]
Si initial degree of saturation, [decimal]
t time, [T]
u, v displacements in the x- and y-directions, respec-
tively, [L]
ua pore-air pressure, [ML–1T–2]
uw pore-water pressure, [ML–1T–2]
(ua – uw) matric suction, [ML–1T–2]
(ua – uw)i initial matric suction, [ML–1T–2]
V0 initial overall volume, [L3]
Vw volume of water, [L3]
x, y, z rectangular Cartesian coordinates
Y elevation, [L]
γ t total or moist unit weight, [ML–2T–2]
εv volumetric strain
θw volumetric water content, [decimal]
µ Poisson’s ratio
© 2006 NRC Canada
Vu and Fredlund 1271

ρ w density of water, [ML–3] The swelling suction, ss, can be estimated from eq. [A3]
σx , σy , σz net total stresses in x-, y-, and z- direction, re- as follows:
spectively, [ML–1T–2] Cs / Cm
σmean mean net total stress, σmean = ( σx + σy + σz)/3,  Ps 
[ML–1T–2]
[A4] ss = s0  

(σ – ua) net normal stress, [ML–1T–2]
 (σ − ua ) 0 
(σ – ua)i initial net normal stress, [ML–1T–2] Let us assume that e is the void ratio at a net normal stress
(σ – ua)0 net normal stress at which the swelling index equal to X, and matric suction equal to Y. The change in void
Cm was obtained, [ML–1T–2] ratio (Fig. A1) can be written as
[A5] ∆e = e − e0
Appendix A. Derivation of equation for the From the net normal stress plane at a matric suction equal
estimation of void ratio surface from to b, (Fig. 5), the change in void ratio can be expressed as
volume change indices follows:
M
The void ratio, ef, at a net total stress equal to (σ – ua)0, [A6] ∆e = C s log
Ps
and a matric suction equal to s0, can be calculated from ei-
ther the net normal stress plane or the matric suction plane, Rearranging eq. [A6], gives
as shown in Fig. A1 and the following equations.
[A7] M = Ps × 10∆e / Cs where (σ – ua)0 < M < Ps
(σ − ua ) 0
[A1] ef = e0 + C s log From the matric suction plane at a net normal stress equal
Ps to (σ – ua)0 (Fig. A1), a change in void ratio can be written
as follows:
s0
[A2] ef = e0 + C m log [A8] ∆e = C m log
N
ss ss
where Ps is swelling pressure; ss is swelling suction (i.e., Rearranging eq. [A8] gives
equivalent value of swelling pressure on the suction plane).
Suction “s0” and net total stress “(σ – ua)0” are values of [A9] N = ss × 10∆e / Cm where s0 < N < Ss
stress points at which the void ratio versus logarithm of suc- For a constant void ratio plane at a void ratio equal to e,
tion and the void ratio versus logarithm of net normal stress the following relationship can be obtained (Fig. A2):
converge (Fig. A1). These values are measurable in a labora-
tory program. A value of 1 kPa was assumed for the refer- N − s0 Y − s0
[A10] =
ence stress states of suction “s0” and net total stress “(σ – M − (σ − ua ) 0 M − X
ua)0” in this study.
Equating eqs. [A1] and [A2] gives Rearranging eq. [A10] gives
(σ − ua ) 0 s N − s0
[A3] C s log = C m log 0 [A11] Y = ( M − X ) + s0
Ps ss M − (σ − ua ) 0
Substituting eqs. [A4], [A5], [A7], and [A9] into
eq. [A11] gives the relationship amongst net normal stress X,
matric suction Y, and void ratio, e

 P Cs / Cm 
s0  s  × 10 ( e − e0 )/ Cm − 1 (Ps × 10 ( e − e0 )/ Cs − X )
 a  
[A12] Y =  ( e − e0 )/ Cs
+ s0
Ps × 10 − (σ − ua ) 0
Equation [A12] can be rewritten setting the X variable to net normal stress and the Y variable to matric suction
 P Cs / Cm 
s0  s  × 10 ( e − e0 )/ Cm − 1 [Ps × 10 ( e − e0 )/ Cs − (σ − ua )]
 a  
[A13] (ua − uw) =  ( e − e0 )/ Cs
+ s0
Ps × 10 − (σ − ua ) 0

© 2006 NRC Canada

1272 Can. Geotech. J. Vol. 43, 2006

Fig. A1. Linear relationships between void ratio and logarithm of stress state variables at extreme planes.

Fig. A2. Relationship between net normal stress and matric suction at constant void ratio plane.

© 2006 NRC Canada