Você está na página 1de 6

What can we learn from compressibility experiments?

Lynn S. Bennethum
Department of Mathematics, University of Colorado at Denver, USA

ABSTRACT: Here we examine typical experiments performed on saturated soils (swelling and non-swelling
porous materials) which can be performed on any liquid-solid porous material and relate them to thermodynamic
quantities. Compressibility experiments typically take a porous material sample and control two of the following
four parameters: electro-chemical potential (or pressure for a non-swelling porous material) of the liquid phase,
pressure of the sample, volume of the sample, and mass of liquid leaving the porous material. The other two
variables are measured at equilibrium. In this paper a framework is established which rigorously determines
these compressibility terms with thermodynamic quantities, allowing us to determine the relationships between
what is measurable (compressibility coefficients) and thermodynamic quantities such as the compressibility of
each phase or the change in energy with respect to moisture content. This framework allows us to extend flow
and deformation or wave propagation models to compressible porous materials or for swelling porous materials
such as montmorillonite. Because minimal assumptions are made the results are applicable to modeling wave
propagation or flow and deformation in soils, bio-materials, and bone.

1 INTRODUCTION and held constant (temperature is not considered an


independent variable).
For soils, several compressibility moduli are generally
defined (Wang 2000): drained, undrained, jacketed,
unjacketed, etc. These moduli are defined by the exper-
iment used to measure them and they are then used 2 COMPRESSIBILITY IN POROUS MEDIA
in the effective stress principal to model deformation
and in large-scale tectonic calculations. However pre- In porous materials, there are several definitions for
vious theoretical formulations for the compressibility compressibility, depending on whether one takes the
moduli (e.g. (Zimmerman 1991; Biot 1941; Biot and measurement allowing the liquid phase to drain or not,
Willis 1957)) assume that the volume of the medium is or whether one measures the liquid pressure or loading
only a function of the liquid pressure and differential pressure, or the change in volume of the porous mate-
pressure (the difference between the overburden pres- rial, or the change in volume of the liquid phase only.
sure and liquid pressure). It is also generally assumed If the liquid phase is allowed to drain, the experimen-
that both phases are incompressible. In this paper we tal setup for determining the compressibility is given
propose a framework which extends this work to incor- in Figure 1. The porous material, which has a total
porate compressibility (liquid and/or solid phase) and volume composed of the sum of the volume of the liq-
allows compressibility coefficients to be defined for uid phase, V l , and solid phase V s , is confined by an
a swelling porous material. This is done by carefully impermeable membrane except at one location where
defining which independent variables the energy of the liquid is allowed to leave the system.If the liquid
the system is a funcion of, and then carefully relating phase is continuous so that the liquid can flow freely,
the compressibility definitions with well-defined ther- then the pressure of the liquid in the porous material is
modynamic quantities such as the compressibility of given by pl and this can either be imposed or measured
each phase. indirectly by changing plA . The uniform pressure on
For pedagogical reasons, we make the following the system is sometimes referred to as the confining
assumptions: (1) only uniform pressures and defor- pressure, or as we term here, the total pressure, p. It
mations are considered and hence shear stresses are can be shown by summing the momentum equations
neglected, (2) the soil is isotropic (which is consistent that p = l pl + s ps (Bennethum and Weinstein 2002)
with assumption (1), (3) there is only a liquid and a where l is the volume fraction of the liquid phase,
solid phase (trapped liquid is considered to be a part V l /(V l + V s ), and s = 1 l is the volume fraction
of the solid phase), and (4) temperature is uniform of the solid phase (Bennethum and Weinstein 2004).

617
Copyright 2005 Taylor & Francis Group plc, London, UK
p the fluid energy changes with moisture content. For
confining liquid
a porous material in which there is little interaction
between the liquid and solid phase (e.g. sandstone),
l = 0, and we obtain the equivalence of a balance
soil of the Gibbs potential and pressures (assuming that
sample the density of the liquid phase is the same within the
plA porous medium and out).
pl ps
In this experiment one usually prescribes two of the
semi-permeable quantities {plA , p, V , M l } and measures the response
membrane of the other two. If the liquid phase is not allowed
to drain, then the compressibility is termed undrained
and is determined by measuring the change in pressure,
Figure 1. Experiment for compressibility. p, with respect to (bulk) volume, V .
There are many bulk compressibilities, and here we
There are four variables which vary and are directly mention a few. In measuring the drained compressibil-
observable: the total pressure, p, mass of the liquid in ity the liquid is allowed to drain from the system as it
the porous material, M l , the volume of the bulk mate- is being compressed so that the liquid pressure is held
rial, V = V l + V s , and the pressure, plA , of the liquid constant. Thus measurements are made at constant
phase as measured outside the porous material. For liquid pressure with no solid leaving the system:
non-swelling porous material plA is equal to the pres-
sure of the liquid inside the porous material, plA = pl ,
but for a swelling material, it is the Gibbs potential
or chemical potential, G l , which is constant across
the semi-permeable membrane, i.e. G l = G lA (Callen
1985; Castellan 1983). where M s is the total mass of the solid phase. Thus we
We note that for a non-swelling porous material, the are implicitly assuming that p = p(V , plA , M s ), i.e. that
Helmholtz potential of the liquid phase is a function the external pressure is a function of the total volume,
of liquid volume and liquid mass, i.e. Al = Al (V l , M l ) liquid phase Gibbs potential, and the mass of the solid
or Al = Al (l ), where Al has units of energy of the phase.
liquid phase per unit mass of the liquid phase and l To obtain the unjacketed compressibility the soil
has units of mass of liquid phase per unit volume of sample in Figure 1 is not confined so that the bulk liq-
liquid. For a swelling porous material, in which the uid pressure (and hence confining pressure) and pore
thermodynamic properties of the liquid are a func- liquid are the same, (plA = p). Thus measurements are
tion of the moisture content (or volume fraction of made keeping plA p and M s fixed:
liquid) we have Al = Al (l , l ). Thus, using the fact
2
that pl = ((l ) Al /l )|l (Callen 1985; Bennethum
andWeinstein 2004) (and similarly for plA ) the bal-
ance of the Gibbs potentials at the semi-permeable
membrane implies that
The undrained compressibility is measured allowing
no fluid or solid to leave the system, so that the total
mass of fluid and solid is held fixed. Thus

This implicitly assumes p = p(V , M l , M s ), i.e. the


external pressure is a function of the total volume and
the total mass of the the liquid and solid.
Other compressibilities include, respectively, the
drained pore compressibility, pseudo-bulk compress-
ibility, and pore compressibility (Zimmerman 1991).
We note here that if both phases are compressible
these compressibility coefficients are not welldefined,
even for a non-swelling porous material. To see why,
where l is the swelling potential (Bennethum and note that in terms of intensive variables we have
Weinstein 2002), i.e. it is a measure of how much pl = pl (l ) and ps = ps (s , s ), and p = l pl + s ps .

618
Copyright 2005 Taylor & Francis Group plc, London, UK
This implies that p = p(l , l , s ). In terms of exten- Table 1. Definition of variables.
sive variables, this implies that p = p(V l , V s , M l , M s )
which implies that the confining pressure is a func-
tion of 4 independent variables. Yet all the measurable
compressibility coefficients assume there are only
three independent variables: For example, the drained
compressibility assumes that p = p(V , plA , M s ). The
problem persists if the solid phase is assumed incom-
pressible, for in this case we have that the drained
compressibility is a function of two independent vari-
ables, V and plA , while the thermodynamics tells us
that it should be a function of 3 independent vari-
ables: V l , V s , M l . If both phases are incompressible
then the discrepency is resolved as there are now only to a porous material composed of at least one liquid
two indendent variables. The primary goal of this paper phase it does not apply to composite materials where
is to explain a framework for which we can make use no material leaves the system. However we shall con-
of these compressibilities for porous materials which tinue to include the results for the undrained test for
may be swelling and/or have at least one compressible completeness.
phase. K is the compressibility coefficient for each phase.
In general the compressibility of, e.g. water, is a
function of volume fraction, it is likely difficult to
3 SUMMARY OF RESULTS FOR measure this dependence, so it is generally assumed
NON-SWELLING POROUS MATERIAL that the compressibility is the same as the pure
material.
We define a non-swelling porous material as one in The term (ps /s )|s is a measure of change of
which the properties of the fluid phase do not change pressure with respect to solid phase configuration.
with moisture content. Typically sandstone is consid- If the solid phase is considered incompressible then
ered non-swelling while montmorillonite is considered changing the solid-phase pressure changes the volume
swelling. Thus for a non-swelling porous material fraction i.e. the solid phase will re-arrange itself so
we consider the total pressure to be a function of as to minimize its energy. The thermodynamic defini-
independent variables as follows: tion of pressure is in terms of the Helmholtz potential
(one can think of this as a form of internal energy) is
standard and can be found in a textbook on physical
chemistry (Castellan 1983).
Note that though the liquid pressure is not consid- The terms defined in the last two rows of Table 1 are
ered a function of moisture content (volume fraction well defined in the sense that within this framework
of liquid), the solid phase pressure is, since the con- there are no other independent variables. This means
figuration of the solid phase (how the solid portion is that once a constitutive relationship is determined for
arranged geometrically) affects the pressure. Although these variables this constitutive relationship will hold
the volume fraction cannot capture all the geomet- for a variety of conditions. This is not true of the
ric effects, it is one of the principle factors which unjacketed compressibility and other compressibility
we will consider. Also note that for a non-swelling coefficients. For example consider the undrained com-
porous material, plA = pl as explained in the previous pressibility, (7). This definition assumes p is a function
section. The details of the following calculations (in of V , M l , and M s . However as already explained, p is a
this section) are available in (Bennethum 2005). function of l , l , and s , or, in terms of extensive vari-
To begin with, lets define the quantities which will ables, p is a function of V l , V , M l , and M s . This means
be used in this section; here refers to phase where that the undrained compressibility could depend upon
= l for liquid or = s for solid: an additional variable, and thus when doing an experi-
In regards to Table 1, g(l ) is the change in liquid ment the value measured may change because there is
content as measured by changing the liquid volume an unaccounted for dependence. The same argument
for the drained test (in which the liquid pressure is can be made for the other compressibilities as well as
held fixed). The variables g( l ), g(
l ), g(
l ) are also the change in moisture content variables g(l ), g( l ),
a measure of change in liquid volume with respect to l ) (note that g(
and g( l ) is well defined). The goal of
solid volume, but for different tests. Note that although this work is to relate what is well-defined (the terms
l ), and g(
g(l ), g( l ) are measurable, g(
l ) is not mea- in the last two rows of Table 1), with what is measur-
surable (there is no mass leaving the system to be able (the compressibilities and the change in moisture
measured). This will limit the usefulness of this theory content variables).

619
Copyright 2005 Taylor & Francis Group plc, London, UK
Relating these variables for a non-swelling porous Table 2. Definition of variables.
material has already been accomplished (Bennethum
2005) so here we present some results without
derivation:

assume that Al = Al (l , l ) = Al (M l , V l , V ). The sec-


ond is that at the vicinal water bulk water interface it
is not the liquid pressure which is continuous but the
Gibbs potential. Neverthe less, it is possible.
We again use the quantities defined in Table 1, but
now we have a couple of additional variables. Again
refers to phase where = l for liquid or = s for
solid:
In regards to Table 1 g( l ) is the change in liq-
uid content as measured by changing volume for the
unjacketed test. If the liquid phase is not incompress-
l ) is the change
ible this is not directly measurable. h(
in liquid content as measured by mass. This is directly
measurable if the liquid phase is compressible or for
swelling media. Pressure, p , is defined as before, and
now we have two additional quantities, the swelling
potential, l , and the change in liquid pressure with
respect to volume fraction. The swelling potential can
be directly measured via a reverse-osmotic swelling
test if the solid phase itself cannot support tension
(such as a well-layered montmorillonite sample). It
is zero for a non-swelling porous material (in which
the energy of the liquid is not a function of mois-
ture content) and very high for expansive soils. When
the pressure in the liquid phase changes, it is due to
There are a couple of checks that can be per- a change in volume fraction, density, or both. The
formed to verify these formulas. One is to assume compressibility coefficient K l measures the change in
the solid phase is incompressible, in which case V s pressure due to density changes only, and pl /l |l is
is constant so that g(l ) = g(
l ) = g(
l ) = 1. The result a measure of the change due to changing the volume
should have no dependency upon K s . Another check fraction only.
is to compare the results with what has been derived We begin with the definition of compressibility and
by Zimmerman (Zimmerman 1991; Wang 2000). As try to re-write it in terms of the variables listed in
explained in (Bennethum 2005), these results can be Tables 1 and 2. To begin, we use the fact that plA = p,
obtained by assuming g(l ) = g( l ) = g(
l ) = l . This and equations (1) and (4) we have
too also checks. Its a little more difficult to compare
this formulation with Biot (Biot 1941; Biot and Willis
1957), but this may also be done (see (Bennethum
2005)).

4 EXAMPLE OF FRAMEWORK:
UNJACKETED COMPRESSIBILITY

We go through in some detail the derivation of the Using equations (1) and (4) for dG l we have
equivalent relationship for the unjacketed compress-
ibility, (6), for a swelling porous material. This is quite
a bit more involved for a couple of reasons. One is
that there is an additional independent variable, i.e. we

620
Copyright 2005 Taylor & Francis Group plc, London, UK
Since pl is not a function of l pl directly, we need to
convert the last partial derivative. Beginning with the
total derivative of pl = pl (l , l ) we have

Using l =V 1 /V we have

Substituting (22) into (20), then substituting this result


into (20) and simplifying gives
so that after some simplification equation (13)
becomes

Now we work on the term pl /V |plA p,M s . We begin by


calculating the total differential of pl = pl (V l , M l , V ), Substituting (23) into (15) we finally have
which gives

Finally we need to determine the relationship between


l ). This requires equating plA /V and
l ) and h(
g(
p/V keeping plA p and M s fixed. After pages of
calculations similar to those above we get

We now convert the derivatives in terms of extensive


variables V l , M l , and V into derivatives of intensive
variables: Since V l is held fixed

Similarly

621
Copyright 2005 Taylor & Francis Group plc, London, UK
There are at least two checks we can perform on the volume fraction. Thus if there is a dependence on all 3
combined (24) and (25). The first is to see if, upon intensive variables a constitutive equation which will
simplifying to a non-swelling porous material, we be valid in general is not obtainable for these variables.
recover(10). Thus we set l =pl /l 0, lA = l , To resolve these issues several new, well-defined
and since for the non-swelling porous material the variables are defined: the compressibilities of each
unjacketed test requires pl = p we also have pl = ps . phase, K l and K s , the swelling potential, l , and
Simplifying (24) and (25) using these assumptions the change in pressures with respect to volume frac-
shows we indeed recover (10). The second test is to tion, (pl /l )|l and (ps /s )|s . These quantities are
assume the liquid and solid phases are incomopress- welldefined in the sense that they take into account
ible. In this case g( l ) = l , and lA =l . In
l ) = 1, h( additional dependencies. This implies that once a
this case we should eliminate the dependence of K constitutive relationship is determined for these quan-
upon K l , and it does. tities in terms of the independent variables, it should
Because (24) and (25) are quite complicated addi- remain unchanged (assuming the assumptions of this
tional simplifying assumptions are necessary before framework are valid). However they are not directly
these equations become tractible. However now we measurable. In this paper a framework for the relation-
know what assumptions are being assumed explic- ship between what is measurable (compressibilities
itly. For the case where the liquid phase is considered and change in liquid content quantities, g(l ), etc.)
incompressible(24) becomes and what is well-defined is determined. The relation-
ship has been obtained for the non-swelling porous
materials and an example of the calculations required
is shown for a swelling porous material. It is hoped
that by performing a variety of compressibility exper-
iments the constitutive expressions for the welldefined
which is very tractible. quantities can be determined and thus these quantities
can be reliably used in models involving a deforming
porous material.
5 CONCLUSIONS

We note that quantities defined in terms of only REFERENCES


three extensive variables (the compressibility coeffi-
cients and the measure of change of liquid content: Bennethum, L. S. (2005). Compressibility moduli for porous
l ), etc.), are not well-defined quantities in
g(l ), g( materials incorporating volume fraction. submitted.
the sense that it is very possible that these quanti- Bennethum, L. S. and T. Weinstein (2002). Pressures
in swelling clay soils. In C. Di Maio,T. Hueckel, and
ties could depend on a fourth independent extensive B. Loret (Eds.), Chemo-Mechanical Coupling in Clays,
variable. So for example, the drained compressibility from Nano-Scale to Engineering Applications, pp. 135
assumes the total or confining pressure is a function 148. Tokyo: A. A. Balkema Publishers.
of V , pl , and M s . Since liquid pressure is at least Bennethum, L. S. and T. Weinstein (2004). Three pressures
a function of density, we have that p is a function in porous media. Transport in PorousMedia 54(1), 134.
of p = p(V , l , M s ). Dividing through by V to change Biot, M. A. (1941). General theory of three-dimensional
over into intensive variables, we have p = p(l , s s ). consolidation. Journal of Applied Physics 12, 155164.
However even for a non-swelling porous material we Biot, M. A. and D. G. Willis (1957). The elastic coefficients of
have p = p(l , l , s ) so that the drained compress- the theory of consolidation. J .Appl. Mech. 24, 594601.
Callen, H. B. (1985). Thermodynamics and an Introduction
ibility is not well-defined in the sense that it is a to Thermostatistics. New York: John Wiley and Sons.
function of an additional internal variable which is not Castellan, G. W. (1983). Physical Chemistry.Menlo Park, CA:
accounted for. This dependence on an additional vari- The Benjamin/Cummings Publishing Co.
able is also not directly measurable. This means that Wang, H. F. (2000). Theory of Linear Poroelasticity with
measuring e.g. the drained compressibility in the same Applications to Geomechanics and Hydrogeology. Prince-
system for the same value of 3 independent variables ton: Princeton University Press.
(e.g. V , l , M s for the drained compressibility) could Zimmerman, R. W. (1991). Compressibility ofSandstones.
result in different measurements due to e.g. a changed New York: Elevier.

622
Copyright 2005 Taylor & Francis Group plc, London, UK

Você também pode gostar