You are on page 1of 14

Engineering Geology 106 (2009) 26–39

Contents lists available at ScienceDirect

Engineering Geology
j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / e n g g e o

Unsaturated soil mechanics


Critical review of physical foundations
Rafael Baker ⁎, Sam Frydman
Faculty of Civil & Environmental Engineering, Technion-Israel Institute of Technology, Haifa, 32000, Israel

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

Article history: Most constitutive models for unsaturated soils are based on identification of soil-water suction with the capillary
Received 12 May 2008 component of the matrix potential, ignoring the contribution of adsorption to this potential. Identification of
Received in revised form 15 February 2009 potential (energy per unit volume) with stress (or suction), is questioned, since these quantities have different
Accepted 18 February 2009
physical significance despite their common dimensions. It is suggested that the identification of matrix potential
Available online 6 March 2009
with (ua −uw) results from neglecting the adsorption potential, and adopting an unrealistic pore space model. This
identification was probably motivated by the laboratory axis translation technique, but it is not valid under normal
Keywords:
Unsaturated soil mechanics field conditions where the air pressure is usually atmospheric, and soil water cannot develop high tension without
Soil physics cavitating. Axis translation alters soil behavior by preventing cavitation, thus casting doubt on the relevance of
Matrix potential laboratory results obtained from these tests to actual field conditions. Specifically, in soils having large specific
Suction surface areas, there is a range of conditions, relevant to geotechnical engineering, in which capillary potential
Cavitation appears to account for only a small part of matrix potential, the major contribution resulting from water adsorption
onto the soil particles. Consideration of a double porosity model and cavitation of water under the tension
generated by capillary mechanism appear indispensable for the interpretation of unsaturated soil behavior.
© 2009 Elsevier B.V. All rights reserved.

1. Introduction Bishop (1959) was the first to suggest an expression for effective
normal stress in terms of both pore air pressure, ua, and pore-water
Since the 1950's, there has been growing interest in the mechanical pressure, uw, as follows:
behavior of unsaturated soils. This interest has accelerated during the
last 15 years, and has been accompanied by the development of σ V= ðσ − ua Þ + χ ðua − uw Þ ð3Þ
extensive experimental laboratory testing programs leading to the
where χ is a parameter between 0 and 1, related to the degree of
formulation of various constitutive models for such soils. It has been
saturation of the soil.
generally accepted that the mechanical behavior of unsaturated soils
These attempts to develop effective stress expressions came under
is influenced by the pore-water suction, or negative potential.
criticism in the 1960's (e.g. Jennings and Burland, 1962), following
Research in the 1950's centered on attempts to develop an effective
which their development effectively stopped, but there has been a
stress expression for unsaturated soils. Croney et al. (1958) suggested
significant renewal of activity in this direction over the last 10 years.
that effective normal stress, σ ′, could be expressed as:
For example, Khalili and Khabbaz (1998), and Khalili et al. (2004)
σ V= σ − βuw ð1Þ responded to the previous criticisms and demonstrated supporting
where σ = total normal stress, β = a factor and uw = pore-water evidence for the applicability of Bishop's effective stress expression
pressure. (3), by relating χ to the “suction ratio” defined as the ratio of soil
Aitchison (1961) and Jennings (1961), considering field conditions matrix1 potential, to the air entry value of the soil.
in which the air pressure in the voids is normally atmospheric, Since the 1960's, considerable research has been devoted to the
suggested expressions of the following form: development of constitutive models for the mechanical behavior of
unsaturated soils. The experimental foundation of the majority of
σ V= σ − ψpW ð2Þ suggested constitutive models has been based on the axis translation
where ψ is a parameter, and p″ was defined by Aitchison as the pore- procedure (ATP). The conceptual basis of this technique is related to a
water pressure deficiency, and by Jennings as the absolute value of the simple “bunch of cylindrical capillaries” model of the soil pore structure
negative pore-water pressure.
1
In most publications matrix potential, as referred to here, is termed matric
potential. We find the term matrix potential natural, describing the water potential
⁎ Corresponding author. Tel.: +1 972 4 8292323. generated by the soil matrix, while the term matric potential does not appear to have
E-mail address: Baker@Technion.ac.il (R. Baker). an obvious linguistic source. We thank Prof. John Burland for pointing this out to us.

0013-7952/$ – see front matter © 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.enggeo.2009.02.010
R. Baker, S. Frydman / Engineering Geology 106 (2009) 26–39 27

(see discussion in Tuller et al., 1999). The ATP identifies negative matrix d) The system should be pre-pressurized to high pressure in order to
potential, or suction, with the capillary suction, which is equal to the dissolve all remaining free air.
difference between pore air pressure, ua, and the mechanical stress in the
soil water, uw, i.e. (ua −uw). Under field conditions, where ua is normally They noted that failure to remove all the air from the measuring
atmospheric, the negative matrix potential or suction would then be device results in cavitation and consequent reduction of the
equal to the tensile stress in the pore water, −uw. Attempts to study the maximum measured water tension in the chamber to approximately
influence of suction on the behavior of unsaturated soils, under 100 kPa, (which is the classical cavitation pressure of water).
conditions of ua = 0 (i.e. atmospheric) have been limited by the technical An extensive series of calibrations of the Imperial College tensiometer
difficulty of measuring water tensions of greater than 100 kPa due to was presented by Tarantino and Mongiovi (2001). They concluded that
cavitation of water in the measuring device. In order to solve this cavitation usually occurs in the porous stone, not in the chamber, and
problem, tests are commonly performed utilizing the ATP introduced by found that pre-pressurization of the stone for a year was not sufficient in
Hilf (1956), and involving application of an elevated ua value, which order to completely saturate it. They suggested that the porous stone
“translates” uw to the positive pressure range. As a result of the should be subjected to cycles of pressurization and cavitation in order to
translation of uw to the positive range, the ATP prevents cavitation of remove all the air from it, and discussed the physical mechanisms
the pore water, and allows development of the desired value of (ua −uw) justifying this preconditioning procedure, which is aimed at preventing
which is considered to be a manifestation of the soil matrix negative cavitation in the porous stone. Following this procedure, it has been
potential. In the engineering literature this potential is usually called found possible to measure water tensions in the probe chamber up to the
suction. The pressure plate technique for determination of soil-water air entry value of the stone (commonly 1500 kPa.).
characteristic curves constitutes an early and classical example of this It is noted that a negative soil-water potential of 106 kPa (i.e.
experimental approach. 103 MPa) is generally accepted as the upper limit occurring at zero
The paper questions the conceptual validity of identifying water degree of saturation (e.g. Vanapalli et al., 1997). Experimental evidence
potential with water tension (called “suction” in the geotechnical supporting this value has been provided by several investigators (e.g.
literature) as is done in most developments of a constitutive character- Croney and Coleman, 1961), while thermodynamic support for this
ization for unsaturated soils. In particular, the validity of identifying limiting value was presented by Richards (1965). In view of the
matrix potential with −(ua −uw) under normal field conditions is elaborate procedures required for preventing cavitation in the high air
questioned. The physical basis for questioning this identification includes entry value porous stone discussed above, one is faced with the
two elements: question: how is it possible for the tension in the pore water of
unsaturated soil to reach 103 MPa without cavitating? (after all, real
a) The phenomenon of cavitation, which prevents development of
soils do not satisfy requirements (a) to (d) above, specified by Ridley
high tension in the soil water.
and Burland (1999), or undergo the elaborate conditioning process
b) The existence of water potential associated with adsorption of
described by Tarantino and Mongiovi, 2001). The same question
water onto the mineral skeleton, unaffected by the application of
applies to all values of suction exceeding the water tension at which
elevated air pressure.
cavitation occurs.
The difficulties associated with proper application of advanced
2. Cavitation and the Imperial College tensiometer
tensiometers are due to cavitation of water in the porous stone of the
probe. Similarly, considerable experimental evidence of cavitation in
The problem associated with equating soil suction (mechanical
saturated soils has appeared in the literature. For example, McManus
water tension) with negative water potential becomes immediately
and Davis (1997) and Mokni and Desrues (1998) showed results of
apparent when note is taken of technical difficulties associated with
undrained shear tests on saturated sandy soils. They found that during
recent developments of advanced, high capacity tensiometers which
dilation, pore pressure tended towards a final value of approximately
can measure “suctions” up to 1500–2000 kPa (e g. Ridley and Burland,
−80 kPa, which is the order of the classical cavitation pressure of
1993; Guan and Fredlund, 1997; Tarantino and Mongiovì, 2003; Take
−100 kPa, and remained at that level as the test continued. Following
and Bolton, 2003). The technical difficulties associated with operation
cavitation, there were gas bubbles in the soil water, and the sample was,
of these probes are related to the occurrence of cavitation in the probe
in fact, unsaturated. Mokni and Desrues (1998) noted that under such
if it is not satisfactorily prepared and de aired prior to use.
conditions, the isochoric constraint, εv = 0, characterizing undrained
The basic elements of these tensiometers include a small water
conditions, is no longer binding, and the test becomes, effectively,
reservoir connected on one side to a high air entry value porous stone
(internally), drained. Results reported by McManus and Davis (1997),
which is in contact with the soil, and on the second side to a membrane
similarly showed the transition from undrained to drained behavior
instrumented by strain gages.
following cavitation and desaturation. Brandon et al. (2006), demon-
The common interpretation of data obtained with these tensiometers
strated the difficulties of performing undrained tests on saturated silt at
is to identify the measured tension of water in the chamber with water
low confining pressures, and they attributed this difficulty to the
tension in the soil. However, at equilibrium, it is the water potential,
cavitation phenomenon. Bishop et al. (1975) showed that cavitation also
rather than the water pressure, which is equal at all points of the system.
occurs in saturated clays during the development of negative pore-water
In other words, these, (and in fact all other devices for measuring soil
pressure, resulting in a marked change in stress–strain behavior and
“suction”), actually measure soil-water energy (potential), rather than
shear strength.
mechanical stress in the soil water.
In order to reconcile the inconsistency between the lower limit on
In their discussion of the paper by Guan and Fredlund (1997),
measured potential values (potential of −103 MPa), and the experience
Ridley and Burland (1999) listed the following requirements for
in porous stones and sand, (maximum tension of 100 kPa), it is instructive
proper functioning of their tensiometer:
to review some elements of soil physics.
a) The water and all surfaces within the measurement system must
be pure and clean. 3. Elements of soil physics
b) The surfaces in contact with the water system must be as smooth
as possible to reduce the number and size of crevices which may 3.1. Tensile strength of water
contain cavitation nuclei.
c) The system should be evacuated by vacuum application in order to Based on molecular considerations (hydrogen bonding), it has been
remove the maximum amount of air entrapped in the crevices. concluded (Speedy, 1982) that tension in pure, deaired water cannot
28 R. Baker, S. Frydman / Engineering Geology 106 (2009) 26–39

exceed 160 MPa; this value was confirmed experimentally by Zheng et al. Lu and Likos (2004), (adopted from McQueen and Miller, 1974). Note
(1991). Two points are particularly relevant with respect to this that in this figure, the capillary regime (in which water can be in
observation: tension) extends from a suction of zero to 100 kPa. At lower potentials,
the water is characterized as adsorbed. The potential in adsorbed
a) The tensile strength of 160 MPa is a result of homogeneous cavitation
water does not have stress characteristics; in fact it is doubtful if it is
occurring in cases when the water does not contain cavitation nuclei.
even possible to define the macroscopic variable, “pressure”, in the
Homogeneous cavitation is usually identified with the tensile
thin adsorbed water films.
strength of pure, deaired water. This type of cavitation is obviously
The above review indicates that several authors in soil mechanics
not relevant to unsaturated soils which are mixtures of solids, water,
and physics (e.g. Koorevaar et al., 1983; Gray and Hassanizadeh, 1991;
air and vapor, and therefore always contain cavitation nuclei.
Iwata et al., 1995; Lu and Likos, 2004) suggested that water tension
b) As was pointed out above, the maximum theoretical and experi-
larger then 100 kPa does not exist in soils. Various authors justify their
mental value of negative potential in soil is of the order of 1000 MPa,
doubts on different bases. It seems reasonable to assume that
which is almost an order of magnitude greater than the tensile
heterogeneous and cavitation nuclei rich materials like unsaturated
strength of pure water. It is clear, therefore, that the negative potential
soils can cavitate under the water tension associated with the capillary
in soils cannot be identified with mechanical tensile stress in the soil
potential. At the same time, it appears that the cavitation tension in
water.
fine grained soils may be higher than the classical value of 100 kPa,
Skepticism with respect to the reality of high tensile stress in soil and it is not necessarily equal in all soils. These higher cavitation
water is hardly new. Blight (1967) noted that the extremely large tensions in clays are probably the result of geometrical constraints
negative water potentials indicated by psychometric theory are imposed by the clay network, and various physical–chemical forces
incompatible with the tensile strength of water. He pointed out that which become important only in materials having large specific
“suction, as measured, must be considered merely as a convenient index surface areas. However, while the cavitation tension may be different
of the affinity of the soil for free water” (rather than as a real mechanical in clays than in other materials, there is no doubt that it exists even in
tension). Koorevaar et al. (1983), stated that negative absolute water clays (e.g. Bishop et al., 1975). The exact value of the cavitation tension
pressures (less than −100 kPa.), do not exist in soils. Iwata et al. (1995) is not essential for the main physical arguments discussed here,
were of the opinion that “soil water pressure below −1 bar does not although it will have an effect on the range of validity of the concepts
have physical significance”. For the present purpose, the following presented.
conclusion of Gray and Hassanizadeh (1991) is particularly relevant:
“Although water, under certain conditions exhibits appreciable tensile 3.2. Soil-water potentials
strength, it does not do so in contact with air. When the (absolute) water
pressure approaches the vapor pressure, the water will begin to Buckingham (1907) introduced the notions of total, gravitational
evaporate.” i.e. a cavitation process will be initiated, implying that the and capillary potentials as well as the representation of the SWCC. A
mechanical tension in the soil water is reduced to 80 to 100 kPa. formal, modern definition and discussion of the soil-water total
All experimental methods for the determination of “suction” in soil potential density (i.e. energy per unit quantity of water), φT, at a point
are based on the principle that, at equilibrium, the water potential in the X in a porous medium is given by many authors (e.g. Edlefsen and
soil and the measuring device are equal. Consequently suction, in the Anderson, 1943; Bolt, 1976). This definition may be stated as: “φT at
sense of tension in the soil water, is not a measurable quantity, and point X is the amount of work that must be done per unit quantity of
the actual measured quantity is water potential. Realizing this we use the pure water in order to transport, reversibly and isothermally, an
terminology water potential when referring to experimental results in infinitesimal quantity of water from a pool of pure water at a specific
which the original term used was “suction”, which implies water tension. elevation and atmospheric pressure to the soil water at the point X
The soil-water characteristic curve (SWCC), is a relationship under consideration”.
between water potential and the amount of water in the soil. This Arbitrarily assigning the value φT = 0 to the reference pool, φT(X)
amount can be characterized by the gravimetric or volumetric water represents a potential energy density of soil water at point X. If the “unit
content (ω or θ) respectively, or by the degree of saturation, S. The quantity of water” in the above definition is taken as unit mass, φT is the
SWCC is one of the main constitutive characteristic of unsaturated chemical potential μT which has units of [J/kg]. Considering energy per
soils. Fig. 1 is a schematic soil-water characteristic curve presented by unit weight of water, results in φT being expressed as the total head, hT,
which has units of length [m]. Reference to energy per unit volume of
water yields a measure ψT, which has units of pressure [Pa]. It is noted,
however, that the fact that ψT has pressure dimensions does not
automatically make it a mechanical stress as implied by the geotechnical
term, suction, (just as the head, hT, is not a physical length). The relation
between the three different forms of φT is ψT =ρwμT, μT =gρwhT where:
ρw is the mass density of water, and g is the acceleration of gravity. {ρw, g}
are constants, and the three different forms {μT, ψT, hT} of φT are
physically equivalent.
It is convenient to separate the total potential into a number of
components:

ψT = ψm + ψos + ψgr + N ð4Þ

where ψm is the matrix potential, representing the interaction of soil


water with both the mineral soil skeleton and the gas phase; ψos is the
osmotic potential which depends on the salt concentration in the bulk
water; and ψgr = ρgz is the gravitational potential, with z representing
the elevation relative to an arbitrary reference datum. It is convenient
Fig. 1. Schematic physical characteristic of SWCC (after Lu and Likos, 2004, reproduced to introduce the term internal potential ψint = ψm + ψos which
by permission of John Wiley & Sons, Inc.). represents all potential components associated with sources which
R. Baker, S. Frydman / Engineering Geology 106 (2009) 26–39 29

isms of liquid drainage in a double porosity medium with large and


small pores (r1 and r2 respectively), as shown in Fig. 2. They identified
two basic mechanisms of drainage in such media, namely, air entering
the system under the action of an external elevated air pressure, and
cavitation of water in the pore space associated with high mechanical
tension in the soil water. The main result of their work is presented in
Fig. 3 as a conceptual SWCC, expressed in terms of matrix chemical
potential, μw, versus degree of saturation, S. The dashed line in this
figure represents the situation existing in a pressure plate apparatus
where the external air pressure is increased until the pore section with
the larger radius (r1) is drained, and the equilibrium potential is
controlled by r2. The solid line corresponds to the result of a tension
plate test (or tensiometer) in which pores are drained by the
application of mechanical tension to the soil water. In this case,
water in the large pore cavitates at a certain water tension. Fig. 3
Fig. 2. Mechanical model illustrating mechanism of desaturation in soils (after Or and Tuller, shows that pressure and tension plate tests yield different SWCC's, due
2002, reproduced by the permission of the American Geophysical Union). (a) Desaturation to suppression of cavitation in the pressure plate device, and this
by application of external air pressure. (b) Desaturation due to cavitation.
effect has been observed experimentally (e.g. Chahal and Yong, 1965;
Iwata et al., 1995).
are internal to the soil element. Some devices for measuring water Furthermore, as noted by Marinho et al. (2005), fixing the value of
potential, (e.g. tensiometers), measure matrix potential, while others (ua − uw) in the soil by ATP is justified only if both the water and the
(e.g. psychrometers) measure internal potential. The only component capillary are rigid (Fig. 4). In this case, application of an increased air
not included in ψint is the gravitational potential ψgr which is not pressure does not change the curvature of the air–water meniscus,
relevant to constitutive characterization of the soil. In this paper it is resulting, therefore, with the required value of (ua − uw). However, if
assumed that |ψos|≪|ψm|, which is a reasonable assumption for many, the capillary contains a mixture of both water and air, as described in
non-saline soils. Section 4 of the paper, the water–air mixture can be considered as a
deformable liquid. In this case the application of external air pressure
3.3. Measurements of water potential with particular reference to the increases the curvature of the water meniscus, thus also changing the
axis translation procedure and its limitations magnitude of (ua − uw). This illustrates how the external air pressure
used in the ATP can change the behavior of unsaturated soil even prior
As stated previously, all methods for measuring water “suction” in to cavitation, implying that the ATP is valid only in saturated soils (the
soils actually measure the matrix or internal potentials. Consequently, back pressure technique).
all these methods establish the energy state of water in the soil, and The pressure plate apparatus is a particular case of the ATP, which
suctions (in the sense of tensile stress in the soil water), are never prevents cavitation by increasing the external air pressure acting
actually measured. In some cases (e.g. tension plate, tensiometers), on the soil. The SWCC is one of the basic constitutive elements
the potential of the water in the measuring device is lowered by characterizing the behavior of unsaturated soils. Fig. 3 illustrates,
applying tension to it until this energy is equal to the energy of the soil clearly, that preventing cavitation by the use of the ATP modifies soil
water. However, there is no reason to assume that similar tension behavior, compared to field conditions in which cavitation is possible.
exists in the soil water, as it is the potentials, not the tensions, which The results of McManus and Davis (1997), Mokni and Desrues (1998),
are equal at equilibrium in the sample and the measuring device. and Bishop et al. (1975) show clearly that allowing the cavitation
Consequently even tensiometers do not measure the tension in the process to develop also changes the stress–strain behavior and
soil water, but rather it's potential. strength of the soil (from undrained to drained conditions). Following
In other devices (e.g. osmotic tensiometers and osmotic cells), the cavitation, water tension remains approximately constant, equal to
energy of the fluid in the measuring device is lowered by using salt the cavitation tension which in the case of sand is close to the classical
solutions (e.g. Peck and Rabbidge, 1969; Kassiff and Ben Shalom, value of 100 kPa.
1971). In the filter paper method (Fawcett and Collis-George, 1967;
Mahler, 1998) the energy of water in the filter paper is changed as a
function of its water content, until it equals the soil-water energy.
Psychrometers measure the energy of water vapor in thermodynamic
equilibrium with the soil water (Richards, 1965). In the last three
cases, there is even less obvious relationship between the quantity
being measured and mechanical tension in the soil water; i.e. all these
devices measure water potential rather than water tension.
The axis translation procedure (ATP), and its particular case, the
pressure plate device, are fundamentally different from the above
mentioned procedures. In this experimental approach the energy
state of the soil water is changed by application of external air
pressure to the soil, until this energy is equal to the fixed energy of the
water in the measuring device. Application of external air pressure to
an unsaturated soil may, and probably does, change its state, thus
changing its behavior compared with field conditions in which the air
pressure is usually atmospheric.
The basic limitation of the ATP is that it prevents cavitation of the
soil water (by transferring uw to the positive range), thus modifying
the behavior of the soil. A vivid mechanistic model illustrating this Fig. 3. Effect of cavitation on soil-water characteristic curve (after Or and Tuller, 2002,
effect was presented by Or and Tuller (2002), who studied mechan- reproduced by the permission of the American Geophysical Union).
30 R. Baker, S. Frydman / Engineering Geology 106 (2009) 26–39

Fig. 4. The conceptual justification for the axis translation procedure (after Mahrinho et al., 2005).

Reservations with respect to the validity of the ATP technique have literature, being common in most branches of science dealing with fine
been expressed previously, e.g. Burland and Ridley (1966), who stated grained, porous media. For example, according to Tuller et al. (1999):
that “… caution should be exercised in making use of this technique “recent reviews (e.g. Nitao and Bear, 1996; …) identified the lack of
because changing the absolute water pressure may affect air coming consideration of adsorptive surface forces and liquid films as primary
into and out of solution. Therefore this technique may not be valid…” shortcomings of present theories for flow and transport in unsaturated
In the present work it is suggested that the main limitation of this porous media, particularly at low saturation. Nitao and Bear (1996)
technique is that it prevents cavitation and therefore changes soil suggested that part of the problem lies in the vague definition of the soil
behavior compared to field conditions where the air pressure is not matric potential where capillary and adsorptive forces are lumped
artificially elevated. together without distinguishing individual contributions. In practice, the
matric potential is attributed to capillary forces only, as evidenced from
3.4. Components of the matrix potential the derivation of media pore size distribution from SWC measurements,
assuming liquid is held by capillary forces only in “surfaceless” cylindrical
The basic work concerning liquid adsorption on solid surfaces was pores of various sizes.”
summarized in the fundamental study of Derjaguin et al. (1987). It can The authors agree with Tuller et al.'s interpretation of Nitao and Bear's
be shown (e.g. Philip, 1977; Tuller et al., 1999), that Derjaguin et al.'s analysis, particularly if the water potential is to be used in mechanical
work implies that matrix potential results from the combined effects constitutive equations. Identifying the matrix (or internal) potential, ψm,
of capillary and adsorptive potentials within the soil matrix, and it can with the capillary potential assigns stress attributes to ψm so that it can be
be expressed as the following, augmented Young–Laplace equation: combined directly with stresses in mechanical constitutive equations. In
reality, the adsorption component of ψm does not have such attributes as
ψm = ψad ðhÞ + ψcp ðκ Þ ð5Þ discussed previously in relation to Fig. 1. The present work attempts to
resolve Nitao and Bear's criticism by including two elements not con-
where ψad is the contribution of adsorption to the matrix potential, h is sidered by them:
the thickness of the adsorbed layer, and ψcp is the capillary contribution
to ψm with κ representing the curvature of the water gas interface. It is 1) Taking account of a physical mechanism (cavitation) which limits
noted in passing that the capillary potential, is always equal to: the capillary potential to the range ψcp N −cvtn under field conditions
where ua is atmospheric.
ψcp = −ðua − uw Þ ð6Þ 2) Incorporating into the analysis a more realistic pore space geometry,
thus avoiding the notion of “surfaceless cylindrical pores” mentioned
i.e. Eq. (6) is valid for all capillary shapes, and the function, ψcp(κ), relates by Tuller et al. (1999).
this potential to the shape and size of the capillaries. Note that when
The practice of ignoring the adsorptive components of the matrix
ua = 0, ψcp =uw (both being negative quantities). At a certain tension
potential in the interpretation of test results may be related to the
the water cavitates, resulting in ψcp-cv =uw-cv. We will use the termi-
belief that the short range of adsorptive forces makes the adsorption
nology cavitation tension (cvtn) to designate the capillary potential
term significant only at very low water contents, often not of interest
associated with cavitation of water in the macropores, i.e. ψcp-cv = −cvtn
in engineering applications. In reality, the physical situation is much
(cvtn is by definition a positive quantity).
more complicated, and the relative importance of the adsorption and
Philip (1977) used Eq. (5) in order to establish the shape of the air–
capillary terms depends on both the effective specific surface area
water interface in the presence of an adsorptive field. A simplified
(SSA) of the minerals forming the solid skeleton, and the assumed
form of Philip's analysis was presented by Tuller et al. (1999). Their
geometry of the pore space, as discussed in the following section.
simplification consists, essentially, in superimposing the conventional,
circular air–water interface on a constant thickness h of the adsorbed
4. Pore models
water layer (see Fig. 5b).
Eqs. (5) and (6), show, clearly, that when interpreting measurements
Three simplified models for the geometry of the pore space are
of matrix potential, the geotechnical literature generally ignores the
considered here:
adsorptive component of ψm, assuming that matrix potential is
generated only by capillarity. For example, in the classic book by a) It is common to consider soil pores as a bunch of cylindrical capillaries
Fredlund and Rahardjo (1993), the term, −(ua −uw), (which, according of different radii. A schematic representation of such models is shown
to Eq. (6), represents the capillary potential) is frequently referred to as in Fig. 5a (modified from Tuller and Or, 2004). The characteristic
the matric suction. This practice is not restricted to the geotechnical feature of this model is that any cross section of a capillary must be
R. Baker, S. Frydman / Engineering Geology 106 (2009) 26–39 31

Fig. 5. Models of pore space. a) The bunch of capillaries model (modified from Or and Tuller, 2002, by the permission of Elsevier Ltd.). b) Structure of unsaturated angular macropore.
c) The double porosity model (modified from Hueckel et. al., 2001).

either full of water or full of air, i.e. this model excludes double (1999) developed expressions for ψcp(κ), for angular pores of
occupancy of water and air in the same pore cross section, with the various simple shapes. However information with respect to actual
result that part of the solid surface area is not covered with adsorbed pore shape is usually not available, so these expressions have only
water. limited utility.
b) Tuller et al. (1999) and Or and Tuller (1999) presented evidence c) Double-porosity models — This class of models is a generalization of
refuting the above model, and developed an alternative framework the angular pore space model, in which the large aggregates (Fig. 5c,
of capillarity in an angular pore space. A simplified cross section of modified from Hueckel et al., 2001) are considered to be pods
such a capillary is shown in Fig. 5b. The characteristic feature of containing clay platelets and water. The pore space within the pods is
such a model is that the total surface of the pore is covered with considered as micro-porosity, while the pore space between the pods
adsorbed water, with capillary water concentrated at the corners, represents the macro porosity (e.g. Tovey et al., 1973; Frydman and
and air occupying the central portion of the pore. Tuller et al. Weisberg, 1991). It is usually assumed that at common engineering
32 R. Baker, S. Frydman / Engineering Geology 106 (2009) 26–39

water contents, the pods are saturated and all the water in them is macropores will be reduced. This process will continue until ψcp =
adsorbed. The macropores may be unsaturated, but all the surfaces of −cvtn. At this stage, cavitation of the capillary water in the macropores
the clay plates forming the boundary of the non circular capillary in occurs, and the potential of the capillary water remains at the cavitation
Fig. 5c are covered with adsorbed water, as shown in Fig. 5b. value −cvtn (see results of McManus and Davis, 1997 and Mokni and
Desrues, 1998, presented previously). The adsorption potential of the
It is seen that in the simple, bunch of cylindrical capillaries model,
water in the pods will continue to decrease until it reaches equilibrium
the effective surface area on which water is adsorbed is smaller than
with the external potential, i.e. ψad =ψext. If the external water potential
that suggested by the other models, resulting in a downgrading of the
is lower than −cvtn then there will be a constant potential difference
adsorption term in Eq. (5).
(ψad =ψext) − (ψcp = −cvtn) between the water in the pods and the
macropores. Assuming that the osmotic potential is insignificant, this
5. Adsorbed and capillary water
constant potential difference will cause all the water to migrate from the
capillaries to the pods, and the capillaries will become effectively empty.
A basic element of the present approach is that the cavitation tension,
As a result, ψm =ψad, and ω =ωad (where ω is the water content of the
cvtn, which limits water tension to be smaller then some limiting value,
adsorbed water in the pods) and the capillary potential in the
provides a lower bound on the capillary potential. As a result, matrix
macropores simply does not exist. Now consider the effect of an external
potential significantly lower than −cvtn must have a significant
mechanical stress (compaction), applied to the system. Compaction
adsorption component. The schematic SWCC in Fig. 1 provides an
decreases only the volume of the empty macropores, but does not affect
illustration of the relative importance of the adsorption term in Eq. (5),
the volume, and therefore also not the potential, of the water in the
based on the assumption that the cvtn =100 kPa. According to this figure,
micropores, which are saturated and therefore rigid. As a result, the
water is adsorbed at all potential values lower than −cvtn, and the
water potential is not affected by density changes, and some of the equi-
capillary term is significant only in the range ψcp N −cvtn. This figure
potential lines in Fig. 6 are vertical. The results in this figure show that
supports the argument that the process of cavitation does not allow the
this is the situation for a significant range of water contents of
capillary potential to drop below its cavitation value, and it is consistent
engineering interest. Vertical equi-potential lines indicate that effectively
with the conclusion of Or and Tuller (1999) that “the contribution of
all the water is in the pods, and adsorption potential is the dominant
liquid films often surpasses the capillary contribution (curved menisci)
mechanism controlling ψm. This situation is a consequence of the
by several orders of magnitude”.
limitation imposed on the capillary potential by the cavitation tension,
In the following, the notations (⁎)pd and (⁎)mp are used for quantities
cvtn. Fig 6 indicates that contrary to common belief, adsorption potential
(⁎) associated with the pods and macropores respectively. Since, within
may be significant at relatively high water content. It is noted that the
the framework of the presently used double porosity model, the pods
vertical equi-potential line associated with the highest potential, defines
are always saturated, the capillary potential of water in the pods is equal
the cavitation tension, cvtn. At potentials higher than −cvtn, the equi-
to zero, and (ψm)pd =ψad. The external surfaces of the pods constitute
potential lines are inclined, and the SWCC depends on two variables
the internal surface of the macropores, so it is convenient to consider the
which may be taken as {ω, γd}, {ω, S} or {S, e} where e is the void ratio.
adsorbed water in the macropores as part of the pods water. Adopting
Such a situation indicates that the macropores are not empty, and
this convention, (ψad)mp = 0 and (ψm)mp = ψcp. Thus, as a result of the
capillarity contributes to the matrix potential.
assumed nature of the double porosity model the augmented Young–
The above mechanism was broadly suggested by Romero (1999),
Laplace Eq. (5) is decoupled, and it can be written as:
Romero and Vautat (2000) and Delage (2002); the present contribu-
ðψm Þpod = ψad ð7:1Þ tion to their arguments is only the addition of the criterion that this
phenomenon occurs when ψm = − cvtn. The above discussion implies
that the SWCC has the following representation:
ðψm Þmp = ψcp : ð7:2Þ
 
ψm ðω; γd Þ If ψm = ψad = ψcp N − cvtn
It is noted that these relations do not contradict the augmented ψm = : ð8Þ
ψm ðωÞ If ψm = ψad b − cvtn and ψcp = 0
Young–Laplace equation (Eq. (5)), since they are relevant to different
water “phases”, located in the pods and the macropores respectively.
This representation illustrates the important role played by the
At equilibrium the water in the pods and macropores must be
cavitation tension on the structure of the SWCC. At high values of ψm, the
at the same energy state, i.e. (ψm)pod = (ψm)mp or equivalently ψm =
SWCC is a 2 dimensional function, degenerating into a 1 dimensional
ψad = ψcp. Failure to satisfy this requirement must result in the migra-
function at low values of ψm. −cvtn is the boundary between these two
tion of water from the macropores to the pods or vice versa, until
representations of the SWCC.
potential equality is reestablished. These relations are valid only if the
When all the water is adsorbed in the pods ψm =ψad and ω =ωad.
macropores are not empty of water.
The water content, corresponding to complete coverage of all the surface
Several researchers have reported measured potential in soil
area in and on the boundaries of the pods, can be evaluated from the
during laboratory compaction procedures (e.g. Gens et al., 1995;
expression:
Dineen et al., 1999; Romero, 1999; Tarantino and De Col, 2008). In
some cases the measurements were of internal potential, and in ω = 100 Sa hðψm = ψad Þρw ð9Þ
others of matrix potential. Results of these investigations were plotted
on the background of standard dry density versus moisture content where Sa [m2/kg] is the specific surface area (SSA) of the minerals, h
diagram, and are shown in Fig. 6. A characteristic feature of these (ψm) [m] is the thickness of the adsorbed layers in the pods, and
results is the existence of a range of water contents in which contours ρw = 103[kg/m3] is the mass density of water. Eq. (9) shows that the
of potential are practically vertical, depending on the gravimetric adsorbed water content is directly proportional to the SSA, so that
water content, but not on the dry density or degree of saturation. adsorption is expected to be significant mainly in clays which have
In order to explain this observation, consider an initial equilibrium large values of SSA.
condition in which ψm =ψad =ψcp N −cvtn, so there is water in both the Fig. 6a shows the results of Tarantino and De Col (2008), who tested
pods and the macropores. This system is now subjected to drying by speswhite kaolin. In this figure the potential contours are all practically
exposing it to an external water potential, ψext, lower than −cvtn. At the vertical; the potentials corresponding to full line contours were measured
first stage of the drying process both the adsorption potential of the by Imperial College tensiometers (matrix potential) while those shown
water in the pods, and the capillary potential of the water in the as dashed contours were measured by psychrometers (internal
R. Baker, S. Frydman / Engineering Geology 106 (2009) 26–39 33

potential). Based on these results, it appears that all the water is adsorbed pressure plate equipment, which was criticized earlier in the paper, so
in the pods at least up to a water content of 32% (which was the wettest these contours are not considered here. However, the contours at lower
soil tested). Lack of data at larger water contents (where the equi- potentials were measured using vapor pressure equilibration (internal
potential lines are expected to be inclined) makes it impossible to estimate potential). These contours are, again, effectively vertical, indicating that
the cavitation tension cvtn. However based on this figure it appears that all the water is adsorbed in the pods at least up to a water content of
400 kPa is an upper bound on the cavitation tension and ω=32% is a about 15%. The figure implies that an upper bound on cavitation tension
lower bound on the adsorbed water content. It is noted that the matrix and a lower bound on adsorbed water content are 3 MPa and 15%
and internal equi-potential lines in this figure are consistent, suggesting respectively. The high value of the upper bound on cavitation tension is a
that in this particular soil, the osmotic potential is not significant. consequence of the fact that measurements based on vapor pressure
In Romero's tests on Boom clay (Fig. 6b), potentials higher than equilibrium were performed only at potentials lower than −3 MPa.
−0.45 MPa (water content of up to 20 to 21%) were measured using However the actual value of the cavitation tension may well be

Fig. 6. Compaction data for clays showing equi-suction lines (a) Kaolin (after Tarantino and De Col, 2008); (b) Boom clay (after Romero,1999); (c) Barcelona silt (after Gens et al.,1995); (d)
bentonite enriched sand (after Dineen et al., 1999).
34 R. Baker, S. Frydman / Engineering Geology 106 (2009) 26–39

Fig. 6 (continued).

significantly lower than 3 MPa and the actual value of adsorbed water The above results, showing a unique dependence of suction on
content may, similarly, be larger than 15%. gravimetric water content, independent of dry density, support the
Gens et al. (1995) (Fig. 6c) tested Barcelona silt using psychrometer suggested predominance of adsorption potential, not only at very low
measurements (internal potential). Vertical equi- potential contours water contents. A dependence of potential on both water content and
were obtained at potentials lower than about − 2 MPa, at water dry density (or void ratio) would indicate a significant contribution of
contents below about 11.5%. The contour corresponding to −1 MPa is the capillary component. The fact that the contours of Gens et al.
already inclined, and it scans water contents of 14 to 15%. Based on this (1995) become vertical only at a relatively large suction (between 1
figure it appears that the cavitation tension of Barcelona silt is between and 2 MPa) implies that there is water in the macropores even at these
1 and 2 MPa and the adsorbed water content is between 11 to 15%. low potentials; this may be a result of either an exceptionally high
Dineen et al (1999) (Fig 6d) tested bentonite enriched sand using the value of cavitation tension or of the existence of a very high osmotic
Imperial College tensiometer, and filter paper (matrix potentials). component in the internal potential, or a combination of the two.
Samples were compacted using both static and dynamic compaction. Ridley and Perez-Romero (1998) suggested that the microfabric
Approximately vertical contours of potential were obtained at degrees of compacted and reconstituted clays are similar, and therefore it
of saturation less than about 80%, and water contents up to 14% (the would be expected that, in the range where adsorption is dominant,
highest water content considered), independent of compaction matrix potential should be dependent only on gravimetric moisture
method. In this case an upper bound on cavitation tension appears to content also in reconstituted (and possibly undisturbed) clays, i.e. the
be 110 kPa. This is consistent with the cavitation tension for sand dominant role of adsorption potential may not be restricted only to
discussed previously. compacted clays.
R. Baker, S. Frydman / Engineering Geology 106 (2009) 26–39 35

Keissar et al. (1990) tested compacted Israeli clays using an osmotic which have been discussed previously), the actually measured quantity
cell. They chose to present their results as a system of SWCC in the is the matrix or internal potential rather than the term (ua −uw), and in
coordinates ψm =ψm(S|e), where the vertical line “|” means that their fact, in general the pressures {ua, uw} are not, and cannot be measured or
test were done at constant void ratio. They found that their system of controlled. Plotting, (and using), the SWCC in this way suggests that
SWCC's tended to merge together (becoming independent of e) at a matrix potential is equal to capillary suction, the latter being expressed
degree of saturation below 40% and matrix potential of about −200 kPa, by Eq. (6). Moreover, most of the data surveyed in the present work
implying a cavitation tension of about 200 kPa. suggest that over a significant range of water contents, the matrix
Additional data showing unique relationships between ψm and potential, ψm, is dominated by its adsorptive component, ψad, which
gravimetric moisture content of soils have been reported in the does not have a mechanical interpretation, and that the capillary
literature, from cases in which the potential was applied or measured potential ψcp = −(ua −uw) is limited to the range ψcp N −cvtn, due to
by methods other than the ATP (e.g. (Krahn and Fredlund, 1972; Ridley the phenomenon of cavitation.
and Perez-Romero, 1998; Monroy, 2005; Delage, 2007). A characteristic It may be legitimate to use {ψm, (σ − ua)} as constitutive variables,
feature of all these data is that they show a unique relationship between but ψm should not be confused and combined directly with stress, as is
potential and gravimetric water content (independent of density), up to done in most mechanical constitutive equations for unsaturated soils.
the highest water contents tested. This data may be used to estimate The following section illustrates some conceptual inconsistencies
bounds on cavitation tension and adsorbed water content. resulting from violating this requirement.
Krahn and Fredlund (1972) measured potential in samples of highly
plastic Regina clay, and low plasticity Saskatchewan till, compacted to 6.2. Constitutive models for unsaturated soils
different dry densities and moisture contents. Internal potential was
measured using a thermocouple psychrometer. Their results imply A large number of constitutive models for partially saturated soils
that the cavitation tension of the Regina clay is less than about 400 kPa have been formulated in the last 15 years. Many of these models
and that of the Seskatchawan till is less than about 300 kPa, while the constitute elaboration on, and modifications of, the classical Barcelona
adsorbed water contents are at least 30% and 17% respectively. Basic Model (BBM) introduced by Alonso et al. (1987, 1990), and
Ridley and Perez-Romero (1998), tested compacted, speswhite Navarro and Alonso (2000). The BBM, in all its variants, describes the
kaolin using the Imperial College tensiometer. Their results suggest an stress–strain behavior of partially saturated soils by extending the
upper bound on cavitation tension of 300 kPa, and an adsorbed water modified Cam–Clay model using Fredlund and Morgenstern's (1977)
content of at least 30%. independent stress variables discussed above. This class of theories
Monroy (2005) measured potentials in samples of London Clay, interprets measured or applied potentials as “suction” i.e. tensile
compacted to different dry densities, using the Imperial College ten- water stresses. The BBMs provide a logical framework, explaining
siometer. His results imply an upper bound of 400 kPa on the cavitation various characteristic features of unsaturated soil behavior. However,
tension which occurred at a water content of about 30%. in view of the dependence of these models on concepts related to the
Delage (2007) showed water retention properties of compacted and ATP, which suppresses the cavitation process, they may not be
powder FoCa7 bentonite, in which the potential was applied by vapor relevant to field conditions in which the air pressure is atmospheric
equilibrium. A unique relation was found between potential and water and cavitation is possible. Although this class of models may well
content, up to water content of about 40%, and down to a potential of describe the behavior of laboratory samples where the ATP is applied,
between −100 and −300 kPa, indicating an upper bound on cavitation its applicability to actual field conditions is questionable.
tension of this order. Both soil conditions followed the same relation, A conceptual difficulty introduced by the BBM framework is
indicating negligible influence of sample density. illustrated by the notion of a “neutral line” (Alonso, 1998) which
All the above investigations found unique relationships between represents the boundary between conditions resulting in collapse or
potentials and gravimetric water content, supporting the concept that swelling of the soil (Fig. 7). This line is expressed as s+p =Const, where
even at moderate matrix potentials, the macropores are effectively s = suction and p = mean principle stress. Since “suction” (in the sense
empty, and the soil water is restricted to the pods, under adsorption of negative water pressure) is not a measurable or controlled quantity,
rather than capillary potential. This result indicates that adsorption is it is clear that s actually stands for the internal or matrix potential, so the
significant at water contents relevant to geotechnical engineering. equation of the neutral line should actually be written as: ψm +p =Const.
The values of these upper bounds on the cavitation tension are, in The validity of adding two physically different quantities like ψm — energy
general, consistent with the results shown in Fig. 6, i.e. an upper per unit volume, and p — mean stress is questionable. The source of the
bound on cvtn in the range of 100 to 400 kPa. The experimental results difficulty is that the ψm includes a significant adsorption component
also imply a lower bound on adsorbed water contents in the range of which does not have stress attributes, so ψm itself does not have stress
13 to 40%. The air pressures applied in most pressure plate devices is attributes, and it should not be added to the stress term p.
up to 1.5 to 2.0 MPa. These pressures are significantly larger than the The ATP is not the only experimental technique allowing a study of
implied cavitation tensions reported above, implying that there is a unsaturated soils. Graham and his coworkers (Tang and Graham 2002;
significant pressure range in which the ATP prevents cavitation and
modifies soil behavior.

6. Geotechnical implications

6.1. Independent stress variables

Constitutive modeling of unsaturated soils is presently widely done


in terms of the variables (ua −uw) and (σ −ua), defined by Fredlund and
Morgenstern (1977) as the independent stress state variables for
unsaturated soils. Fredlund and Rahardjo (1993) also identified the
variable (ua −uw) as the matrix suction, equivalent to the negative
matrix potential. For example, many SWCC's are plotted as relations
between (ua −uw) and some measure of the amount of water in the soil.
However, except when using the ATP (the fundamental limitations of Fig. 7. The neutral line concept (after Alonso 1998).
36 R. Baker, S. Frydman / Engineering Geology 106 (2009) 26–39

Tang et al. 2002; Blatz and Graham 2003), developed a variant of the
BBM which is free of some of the above mentioned difficulties. They used
the internal potential, ψin =ψm +ψos, in the modified Cam–Clay model,
and in their experiments this variable was either controlled by a vapor
equilibrium system (using ionic solution) where vapor with a given
potential was provided around the specimen boundary, or was
measured using psychrometers. As a result they did not use the notion
of independent stress variables or the ATP, and their results are free of
the limitations discussed above. They did not use the terminology of
internal potential, calling this variable suction, but their experimental
techniques imply that their “suction” is actually ψin. However, one of the
basic assumptions employed in their model isqthat ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
the equivalent
preconsolidation pressure pe, is given by pe = ψ2in + σ 2. As ψin is
energy per unit volume, while σ represents the average normal stress, it
is doubtful that these physically different variables can be added directly
as if they are two components of the same physical vector.
Khalili and Khabbaz (1998) and Khalili et al. (2004) provided
corroborative evidence for the applicability of Bishop's effective stress
Eq. (3) on the basis of triaxial data obtained using ATP. This effective stress
expression was then incorporated into an elasto-plastic constitutive
model for unsaturated soils (Loret and Khalili, 2000). If it can be
demonstrated that Eq. (3) is valid using Khalili's definition of χ, when the
ATP term (ua −uw) is replaced by ψm, then this approach may represent
an extremely important development in unsaturated soil mechanics.

6.3. Strength of unsaturated soils

Ng at al. (2007) carried out comparative sets of triaxial tests on


unsaturated soils, using osmotic solutions and ATP to develop suctions
in the soil. Although similar effective friction angles were obtained
from the different sets, higher values of apparent cohesion resulted
from the ATP. The potential ranges applied in this investigation were
small — down to − 100 kPa in the ATP tests, and to − 150 kPa using the
osmotic technique. Further testing, over a larger range of potential,
would contribute, significantly, to this comparison.
An alternative approach to studying the strength of unsaturated
soils may be to formulate the model without explicit dependence on
water potential. An example of such a formulation is the work of Tuffour
(1984), who performed slow, direct shear tests on 3 unsaturated Israeli
clays, at different degrees of saturation. Elements of this work were
reviewed by Juca and Frydman (1995). The clays (Soil1: wL = 38%,
IP = 20%; Soil2: wL = 70%, IP = 38%; and Soil3: wL = 49%, IP = 31%) are Fig. 8. Strength of unsaturated soils (reanalyzed data of Tuffour, 1984). (a), Friction
all soils in which the predominant clay mineral is calcium montmor- angle independent of degree of saturation. (b), Cohesion as a function of gravimetric
moisture content for two compacted clays. (c), Cohesion of three soils as a function of
illonite. The samples of Soil 1 were undisturbed samples taken from the
degree of saturation.
remains of a failed water reservoir, while the samples of Soil 2 and Soil 3
were each compacted in the laboratory to a dry density, γd, close to the
maximum Proctor value, at degrees of saturation ranging from 50% to
100%. Frydman (2000) found that most saturated Israeli clays have an The results in Fig. 8a can be summarized by an equation of the
effective frictional angle of approximately 26° and cohesion of approxi- form:
mately 10 kPa, despite a wide range of consistency limits. In Fig. 8a,
Tuffour's strength envelopes for soil 2 have been drawn for different τff = cðωÞ + σ ff tanð/ VÞ ð10Þ
water contents, using an assumed ϕ = 26°. It is noted that the water
contents specified in Fig. 8a are initial values, but measures were taken where the double subscript, ff, indicates failure stresses on the failure
to minimize water evaporation loss during the tests. It is seen that the plane. In principle, the cohesion is probably a function of the matrix
envelopes reasonably fit the test results, indicating that the friction potential, but for conditions in which this potential is dominated by its
angle is practically independent of water content, and therefore also of adsorption component, it is a function only of the gravimetric water
ψm. It appears that Tuffour (1984) was the first to make this content, and it possible to write c(ψm) = c(ψm(ω)) = c(ω), resulting in
observation, but similar results have since been presented by a number Eq. (10).
of authors (e.g. Rahardjo et al., 2004; Lee et al., 2005; Lu and Likos, Fig. 8b shows the functions c(ω) obtained on the basis of Tuffour's
2006). The fact that the ϕ value obtained from slow shear tests on (1984) results for the two compacted clays tested by him. The important
unsaturated soils is independent of ω, and compatible with the effective characteristics of these functions are that they are practically linear and
friction angle, ϕ′, of the saturated soil, is consistent with the proposition parallel to each other. Lu and Likos (2004) showed c as a function of
that unsaturated soils are internally drained during shearing, as degree of saturation, S, and called their function, c(S), a “capillary
suggested earlier when discussing the results of Mokni and Desrues cohesion”. They suggested that it represents the same physical
(1998). information as the conventional SWCC, expressed in a more useful
R. Baker, S. Frydman / Engineering Geology 106 (2009) 26–39 37

form. The same terminology can be applied to Tuffour's c(ω). A linear The above discussion indicates that for certain problems it may be
c(S) function was also found by Miao et al. (2007). more convenient not to include water potentials explicitly in the
In addition, Tuffour (1984) found that when the cohesion values constitutive model. Instead, the use of certain volume–weight relations
obtained from all three clays tested by him were plotted against S, a as the basic constitutive variables may lead to simpler results (e.g. Juca
unique, almost linear function, c(S) was obtained, as shown in Fig. 8c. and Frydman 1995). Such an approach may represent a possible
It is not suggested that this function, c(S), is a universal relation valid alternative to conventional constitutive modeling of certain problems in
for all clays; the source of its uniqueness is not clear, but it may be unsaturated soil mechanics.
related to the fact that, as mentioned previously, all Israeli clays have
practically the same effective stress parameters. 7. Summary and conclusions
Lu (2008) introduced the notion of state variables which he defined as
follows: “In general, a state variable may be considered as any variable The purpose of the present paper was to examine the basic elements
that has a profound effect on a state of interest.” Such a definition is of geotechnical constitutive theory for unsaturated soils against the
legitimate, not suffering from the inconsistencies discussed in Section 6.2. background of modern developments in soil physics. The main physical
In fact the water content ω in the term c(ω) of Eq. (10) is a state variable elements considered in this work may be summarized as follows:
in the sense of Lu (2008). It is important to realize however that Lu's
(2008) state variables and Fredlund and Morgenstern's (1977) indepen- a) All devices for measuring soil “suction” actually measure soil-water
dent stress state variables are entirely different concepts. potential (internal or matrix) rather than the state of pressure
It is interesting to compare Eq. (10) with the representation developed (tension) in the soil water. In fact the actual water tension in
by Fredlund et al. (1978, 1995), Vanapalli and Fredlund (1997), and used unsaturated soil cannot be measured or controlled. The matrix
by many investigators, who expressed the shear strength of unsaturated potential consists of two major components; adsorption and
clays as follows: capillary. Only the capillary component admits interpretation in
terms of mechanical pressure (tension).
 
τff = c V+ ðσ ff −ua Þf tanð/VÞ + ðua −uw Þf tan /
b
ð11Þ b) A basic element of the present approach is that the tensile stresses
generated by the capillary potential can cause cavitation of the
capillary water in the macropore of the proposed double porosity
where the single subscript, f, indicates failure conditions, and ϕb is a
model. This cavitation process limits water tension (capillary
friction angle defining the strength increase with increasing (ua −uw)f.
potential under field conditions, when ua = 0) to a certain limiting
As (ua −uw) is not a measurable quantity unless one uses the ATP, and all
value, cvtn. As a result of the complex nature of clay soils, this value
other “suction” measuring techniques actually measure the matrix or
may be higher than the classical value of 100 kPa. The current
internal potential, (see Section 3.3), (ua −uw)f, which appears in Eq. (11),
experimental information does not allow exact determination of the
should be replaced by the matrix potential −(ψm(ω))f, and ϕb is a
cavitation tensions in soils, allowing only the estimation of upper
friction angle defining strength increase with respect to matrix potential.
bounds on this quantity. Most data sets considered in this review
Considering field conditions, in which ua = 0, it is then possible to write
imply that the upper bound on cavitation tension in soils is in the
Eq. (11) as:
range of 100 to 400 kPa. These data imply that cavitation severely
   limits the magnitude of the capillary potential to ψcp N − cvtn.
b
τff = c V+ σ ff tanð/ VÞ − ψm ðωÞ tan / ðψm ðωÞÞ : ð12Þ Consequently under a significant range of conditions relevant to
f
geotechnical engineering, adsorption may be the main component of
The minus sign in front of the 3rd term is a consequence of the fact the matrix potential. This is particularly significant in soil having high
that ψm is a negative quantity. specific surface area.
Escario and Saez (1986) found that ϕb is not constant, depending c) The geotechnical literature generally ignores the adsorption poten-
on both σff and (ψm(ω))f making the use of Eq. (12) problematic. tial when interpreting “suction” measurements, identifying the
Since ψm is a potential (which, according to the discussion presented matrix potential with its capillary component. This practice implies
previously, may be dominated by its adsorption component), not a identification of matrix potential with soil-water tension and results
stress, it cannot contribute to the strength by a frictional mechanism, in prediction of unrealistically large water tensions which cannot be
and consequently the physical significance of the friction angle, ϕb, is not realized (due to cavitation) under field conditions, when the air
clear. Referring to ϕb, Lu and Likos (2006) wrote: “… conceptual pressure is normally atmospheric.
difficulties associated with determining necessary material variables d) The axis translation technique “transfers” water pressures in soil to
such as ϕb and uncertainties in their uniqueness over a wide range of the positive range, thus eliminating the possibility of cavitation and
saturation have limited the practical applicability of the independent modifying soil behavior.
stress variable approach.” e) The experimental foundation of most constitutive models for
Comparing Eqs. (10) and (12), it is noted that Eq. (10) does not unsaturated soils is based on the axis translation technique. As a
depend explicitly on ψm. However the dependence of c on ω reflects an result, these constitutive models may well describe the behavior of
indirect dependence on the matrix potential. One of the important laboratory samples in which this technique is utilized, but their
differences between Eqs. (10) and (12) is in the choice of independent relevance to field condition, with atmospheric air pressure, requires
variables (ω versus ψm) used to define the strength of unsaturated soils. further validation.
A major physical difference between Eqs. (10) and (12) is that whereas f) Most versions of constitutive laws for unsaturated soils do not
Eq. (12) requires interpretation of the matrix potential as a stress, Eq. distinguish between matrix potential (energy per unit volume)
(10) relates the matrix potential contribution to the cohesive component which is the actually measured or controlled variable and water
of the strength, and therefore it does not necessitate such an identi- tension (suction) which cannot be measured. When the identifica-
fication. Furthermore, Eq. (10) is simpler than Eq. (12) requiring deter- tion of matrix potential with water tension is incorporated into
mination of a single constitutive function c(ω), while Eq. (12) requires mechanical constitutive equations, and they are added directly,
the determination of two such functions, ψm(ω) and the two dimen- conceptual inconsistencies result. Such combinations are, in
sional function ϕb( ψm, σff ) (see results of Escario and Saez, 1986). This principle, illegitimate, having no physical significance.
simplification is a direct consequence of the physical approach adopted g) Experimental results for compacted clays indicate that the matrix
here, which makes a clear distinction between matrix potential and potential depends only on the gravimetric water content over a range
tensile stresses in the pore water. of water contents significant in geotechnical engineering. Since the
38 R. Baker, S. Frydman / Engineering Geology 106 (2009) 26–39

double porosity model adopted in this paper implies that adsorption Delage, P., 2007. Microstructure features in the behavior of engineered barriers for
nuclear waste disposal. Experimental unsaturated soil mechanics. Proceedings in
potential depends only on gravimetric water content, this observa- Physics, Springer 112, 11–32.
tion suggests that in this water content range, adsorption dominates Derjaguin, B.V., Churaev, N.V., Muller, V.M., 1987. Surface Forces. Plenum, New York.
the matrix potential. The capillary potential, on the other hand, Dineen, K., Colmenares, J.E., Ridley, A.M., Burland, J.B., 1999. Suction and volume
changes of a bentonite-enriched sand. Geotechnical Engineering 137, 197–201.⁎⁎
depends on two independent variables; volumetric or gravimetric Edlefsen, N.E., Anderson, A.B.C., 1943. Thermodynamics of soil moisture. Hilgardia 15,
water content (or degree of saturation) and void ratio (or dry unit 31–298.
weight), but this potential is significant only in the matrix potential Escario, V., Saez, J., 1986. The shear strength of partly saturated soils. Geotechnique 36,
453–456.
range down to the cavitation tension. Consequently, combinations of Fawcett, R.G., Collis-George, N., 1967. A filter paper method for determining the
cavitation, adsorption, and a double porosity model consisting of moisture characteristic of soil. Australian Journal of Experimental Agriculture and
saturated soil pods appear to indicate that the adsorption potential Animal Husbandry 7, 162–167.
Fredlund, D.G., Morgenstern, N.R., 1977. Stress state variables for unsaturated soils.
frequently dominates matrix potential in fine grained soils.
Journal of Geotechnical Engineering, ASCE 103, 444–466.
h) A major inconsistency in conventional representation of unsaturated Fredlund, D.G., Rahardjo, H., 1993. Soil Mechanics for Unsaturated Soils. John Wiley and
soil mechanics is that it assigns two different and incompatible Sons, New-York.
meanings to the term (ua −uw). On one hand, this term is considered Fredlund, D.G., Morgenstern, N.R., Widger, R.A., 1978. The shear strength of unsaturated
soils. Canadian Geotechnical Journal 15, 313–321.
to represent the matrix potential, but the same term is considered Fredlund, D.G., Xing, A., Fredlund, M.D., Barbour, S.L., 1995. The relationship of
also as a stress variable. Neither of these interpretations appears unsaturated soil shear strength to the soil–water characteristic curve. Canadian
acceptable. The term −(ua −uw) can be identified with the matrix Geotechnical Journal 32, 440–448.
Frydman, S., 2000. The shear strength of Israeli soils. Israel Journal of Earth Sciences 49, 55–64.
potential only if one ignores adsorption. Furthermore, by definition, Frydman, S., Weisberg, E., 1991. A study of centrifuge modeling of swelling clay. In: Ko,
matrix potential is energy per unit volume, and it cannot be H.-Y., McLean, F. (Eds.), Proc., International Conference — Centrifuge ‘91: Boulder,
considered as a stress (suction) because its adsorption component Colorado. Balkema publishers, pp. 113–120.
Gens, A., Alonso, E.E., Suriol, J., Lloret, A., 1995. Effect of structure on the volumetric
does not have stress attributes. This confusion appears to be a behavior of a compacted soil. Proc., 1st Int. Conf. on Unsaturated Soils, pp. 83–88.
consequence of using the axis translation procedure as the main Gray, W.G., Hassanizadeh, S.M., 1991. Paradoxes and realities in unsaturated flow theory.
experimental technique when studying the behavior of unsaturated Water Resources Research 27, 1847–1854.
Guan, Y., Fredlund, D.G., 1997. Use of tensile strength of water for the direct
soils. measurement of high soil suction. Canadian Geotechnical Journal 34, 604–614.
i) For certain purposes (e.g. soil strength), it may be useful to consider Hilf, J.W., 1956. An investigation of pore water pressure in compacted cohesive soils.
volume–weight relations instead of potentials as basic constitutive Technical memorandum 654. United State Department of the Interior. Bureau of
Reclamation, Denver Colorado.
variables.
Hueckel, T., Loret, B., Gajo, A., 2001. Expansive clays as two-phase, deformable reacting
continua: concepts and modeling options. In: Maio, C.Di., Hueckel, T., Loret, B.
Based on the above arguments, it is concluded that existing consti-
(Eds.), Proc. of the Workshop on Chemo-mechanical Coupling in Clays: From Nano-
tutive models for unsaturated soils, in their present format, are not scale to Engineering Applications. Balkema Pub., pp. 105–120.
consistent with the basic physical elements of the system. Although no Iwata, S., Tabuchi, T., Warkentin, B.P., 1995. Soil–water Interaction: Mechanisms and
viable alternative to the conventional approach is suggested in the applications, 2nd Edition. Lavoisier Pub, France.
Jennings, J.E., 1961. A revised effective stress law for use in the prediction of the behavior
present paper, (except, maybe in the discussion of soil strength, and the of unsaturated soils. Proc., Conf. on Pore Pressures and Suctions in Soils. ICE,
general structure of the SWCC), it is felt that eliminating physical Butterworths, Lond., pp. 26–30.
inconsistencies is a necessary first step towards formulation of realistic Jennings, J.E., Burland, J.B., 1962. Limitation to the use of effective stresses in partly
saturated soils. Geotechnique 12 (2), 125–144.
constitutive models for unsaturated soils. Juca, J.F.T., Frydman, S., 1995. State of the art report — experimental techniques. Proc.,1st
Int. Conf. on Unsaturated Soils. Paris, pp. 1–21.
References Kassiff, G., Ben Shalom, A., 1971. Experimental relationship between swell pressure and
suction. Geotechnique 21, 245–255.
Keissar, I., Uzan, J., Baker, R., 1990. On the characterization of suction in swelling clay.
Aitchison, G.D., 1961. Relationships of moisture and effective stress functions in Transportation Research Record. No. 1277 — Modern Geotechnical Methods:
unsaturated soils. Proc., Conf. on Pore Pressures and Suctions in Soils. ICE, Instrumentation and Vibratory Hammers, pp. 1–7.
Butterworths, Lond., pp. 47–52. Khalili, N., Khabbaz, M.H., 1998. A unique relationship for χ for the determination of the
Alonso, E.E., Gens, A., Hight, D.W., 1987. Special problem soils — general report. Proc. 9th shear strength of unsaturated soils. Geotechnique 48, 681–687.
European Conference on Soil Mechanics and Foundation Engineering. Balkema, Khalili, N., Geiser, F., Blight, G.E., 2004. Effective stress in unsaturated soils: review with
Dublin, Ireland, pp. 1087–1146. new evidence. International Journal of Geomechanics 4, 115–126.
Alonso, E.E., Gens, A., Josa, A., 1990. A constitutive model for partially saturated soils. Koorevaar, P., Menelik, G., Dirksen, C., 1983. Elements of Soil Physics. New York, Elsevier.
Geotechnique 40, 405–430. Krahn, J., Fredlund, D.G., 1972. On total matric and osmotic suction. Soil Science 114 (5),
Alonso, E.E., 1998. Modeling expansive soil behavior. Proc. 2nd Inter. Conf. on 339–348.
Unsaturated soils, Beijing, China, 2. Inter. Academic Pub, pp. 37–71. Lee, I.-M., Sung, S.-G., Cho, G.-C., 2005. Effect of stress state on the unsaturated shear
Bishop, A.W., 1959. The principle of effective stress. Teknisk Ukeblad 106, 859–863. strength of a weathered granite. Canadian Geotechnical Journal 42, 624–631.
Bishop, A.W., Kumapley, N.K., El-Ruwayih, A., 1975. The influence of pore water tension Loret, B., Khalili, N., 2000. A three phase model for unsaturated soils. Int. J. Numer. &
on the strength of clay. Trans., Royal Society of London, Series A 278, 511–555. Analytical Methods in Geomech 24 (11), 893–927.
Blatz, J.A., Graham, J., 2003. Elastic–plastic modeling of unsaturated soil using results Lu, N., Likos, W.J., 2004. Unsaturated Soil Mechanics. John Wiley and Sons Inc.
from a new triaxial test with controlled suction. Geotechnique 53, 113–122. Lu, N., Likos, W.J., 2006. Suction stress characteristic curve for unsaturated soils. Journal
Blight, G.E., 1967. Effective stress evaluation for unsaturated soils. J. Soil Mechanics and of Geotechnical and Geoenvironmental Engineering 132 (2), 131–142.
Foundations Division, ASCE 93, 125–148. Lu, N., 2008. Is matric suction a stress variable? J. Geotech. Geoenviron. Eng. 134 (7),
Bolt, G.H., 1976. Soil physics terminology. International Society of Soil Science, Bulletin 899–905.
49, 16–22. Mahler, C.F., 1998. Measurement of matrix and total in situ suction of porous soils of
Brandon, T.L., Rose, A.T., Duncan, J.M., 2006. Drained and undrained strength interpreta- San-Paulo using the filter paper method. Proceedings of the 2nd International
tion for low plasticity silt. J. Geotech. and Geoenvironmental. Eng., ASCE 132 (2), Conference on Unsaturated Soils, Beijing, China, pp. 402–409.
250–257. Marinho, F.A.M., Take, W.A., Tarantino, A., 2005. State of the art presentation on
Buckingham, E., 1907. Studies on the movement of soil moisture. United States Division tensiometers and axis-translation technique. International Symposium on Advanced
of Agriculture. . Bulletin, vol. 38. Bureau of Soils, Washington, DC. Experimental Unsaturated Soil Mechanics. Trento. Italy. http://www.experus2005.
Burland, J.B., Ridley, A.M., 1966. The importance of suction in soil mechanics. State of the art ing.unitn.it/1_Marinho_Take_Tarantino.pdf.
lecture. Twelfth Southeast Asian Geotechnical Conference. Kuala Lumpur, pp. 27–49. McManus, K.J., Davis, R.O., 1997. Dilation induced pore fluid cavitation in sands.
Chahal, R.S., Yong, R.N., 1965. Validity of the soil water characteristics determined with Geotechnique 47, 173–177.
the pressure apparatus. Soil Science 99, 98–103. McQueen, I.S., Miller, R.F., 1974. Approximating soil moisture characteristics from
Crony, D., Coleman, J.D., Black, W.P.M., 1958. Studies of the movement and distribution limited data: empirical evidence and tentative model. Water Resources Research
of water in soil in relation to highway design and performance. HRB Spec. Rep. 10, 521–526.
No. 40, Washington, D.C. Miao, L., Houston, S.L., Cui, Y., Yuan, J., 2007. Relationship between soil structure and
Croney, D., Coleman, J.D., 1961. Pore pressure and suction in soils. Proc. Conference on mechanical behavior for an expansive unsaturated clay. Can. Geotech. J. 44, 126–137.
Pore Pressure and Suction in Soils. ICE, Butterworths, Lond., pp. 31–37. Mokni, M., Desrues, J., 1998. Strain localization measurements in undrained plane-strain
Delage, P., 2002. State of-the-art-report: experimental unsaturated soil mechanics. biaxial tests on Hostun RF sand. Mechanics of Cohesive-Frictional Materials 4,
Proc., 3rd Int. Conf. on Unsat. Soils, Recife, Brazil. 419–441.
R. Baker, S. Frydman / Engineering Geology 106 (2009) 26–39 39

Monroy, R. (2005). The influence of load and suction changes on the volumetric behavior D'Enginyeria del Terreny I Cartografica, Escola Tecnica Superior D'Enginyers de
of compacted London clay. PhD. Thesis, Imperial College, London University. Camins Canals I Ports, Universsitat Politecnica de Catalonia. Barcelona, Spania.
Ng, G.W.W., Cui, Y., Chen, R., Delage, P., 2007. The axis-translation and osmotic techniques Romero, E.M., Vaunat, J., 2000. Retention curves of deformable clays. In: Tarantino,
in shear testing of unsaturated soils: a comparison. Soils and Foundations 47 (4), Mancuso (Eds.), Conf. on Experimental Evidence and Theoretical Approach in
675–684. Unsaturated Soils. Balkema, Rotterdam, pp. 91–106.
Navarro, V., Alonso, E.E., 2000. Modeling swelling soils for disposal barriers. Computers Speedy, R.J., 1982. Stability-limit conjecture: an interpretation of the properties of
and Geotechnics 27, 19–43. water. Journal of Physical Chemistry 86, 982–991.
Nitao, J.J., Bear, J., 1996. Potentials and their role in transport in porous media. Water Take, W.A., Bolton, M.D., 2003. Tensiometer saturation and the reliable measurement of
Resources Research 32, 225–250. matric suction. Geotechnique 53 (2), 159–172.
Or, D., Tuller, M., 1999. Liquid retention and interfacial area in variably saturated porous Tang, G.X., Graham, J., 2002. A possible elastic–plastic framework for unsaturated soils
media: upscaling from single-pore to sample-scale model. Water Resources Re- with high-plasticity. Canadian Geotechnical Journal 39, 894–907.
search 35, 3591–3605. Tang, G.X., Graham, J., Blatz, J., Gray, M., Rajapakse, R.K.M.D., 2002. Suctions, stresses and
Or, D., Tuller, M., 2002. Cavitation during desaturation of porous media under tension. strengths in unsaturated sand–bentonite. Engineering Geology 64, 147–156.
Water Resources Research 38, (19–1)–(19-4). Tarantino, A., Mongiovi, L., 2001. Experimental procedures and cavitation mechanisms in
Peck, A.J., Rabbidge, R.M., 1969. Design and performance of an osmotic tensiometer for tensiometer measurements. Geotechnical and Geological. Engineering 19, 189–210.
measuring capillary potential. Soil Science Society of America Journal 33, 196–202. Tarantino, A., Mongiovì, L., 2003. Calibration of tensiometer for direct measurements of
Philip, J.R., 1977. Unitary approach to capillary condensation and adsorption. Journal of matric suction. Geotechnique 53, 137–141.
Chemical Physics 66, 5069–5075. Tarantino, A., De Col, E., 2008. Compaction behavior of clay. Geotechnique 58 (3),
Rahardjo, H., Heng, O.B., Choon, L.E., 2004. Shear strength of a compacted residual soil 199–213.
from consolidated drained and constant water content triaxial tests. Canadian Tovey, N.K., Frydman, S., Wong, K.Y.,1973. A study of a swelling clay in the scanning electron
Geotechnical Journal 41, 421–436. microscope. Proc., 3rd International Conference on Swelling Soils, 2, pp. 45–54.
Richards, B.G., 1965. Measurement of the free energy of soil moisture by the psy- Tuffour, Y.A., 1984. Shear strength of partly saturated clays. MSc. Thesis. Faculty of Civil
chrometric technique using thermistors. In: Aitchison, G.D. (Ed.), Moisture Engineering, Technion, Israeli Institute of Technology. Haifa, Israel.
Equilibria and Moisture Changes in Soils Beneath Covered Areas. A Symposium in Tuller, M., Or, D., Dudley, L.M., 1999. Adsorption and capillary condensation in porous
Print. Butterworths &Co. Ltd., Sydney, Australia, pp. 35–46. media: liquid retention and interfacial configurations in angular pores. Water
Ridley, A.M., Burland, J.B., 1993. A new instrument for the measurement of soil moisture Resources Research 35, 1949–1964.
suction. Geotechnique 43, 321–324. Tuller, M., Or, D., 2004. In: Hillel, D. (Ed.), Water retention and characteristic curve.
Ridley, A.M., Perez-Romero, J., 1998. Suction–water content relationships for a range of Encyclopedia of soils in the environment, 4. Elsevier Ltd., Oxford, United Kingdom,
compacted clays. Proc., 2nd International Conference on Unsaturated Soils. Beijing, pp. 278–289.
China, 1. International Academic Publishers, pp. 114–118. Vanapalli, S.K., Fredlund, D.G., 1997. Interpretation of undrained shear strength of
Ridley, A.M., Burland, J.B., 1999. In: Guan, Y., Fredlund, D.J. (Eds.), Discussion of: Use of unsaturated soils in terms of stress state variables. Proc., 3rd Brazilian Symposium
Tensile Strength of Water for the Direct Measurement of High Soil Suction. on Unsaturated Soils, pp. 35–45.
Canadian Geotechnical Journal, vol. 36, pp. 178–180. Zheng, Q., Durben, D.J., Wolf, G.H., Angell, C.A., 1991. Liquids at large negative pressures:
Romero, E.M. (1999). Characterization and thermo-hydro mechanical behavior of water at the homogeneous nucleation limit. Science 254, 829–832.
unsaturated Boom clay: An experimental study. PhD thesis. Departament

You might also like