Sunteți pe pagina 1din 7

Wave propagation

Copyright 2005 Taylor & Francis Group plc, London, UK


Stress and pore pressure dependent anisotropy of elastic waves
in porous structures

Serge A. Shapiro & Axel Kaselow


Freie Universitaet Berlin, Germany

ABSTRACT: Elastic properties of rocks are functions of tectonic stresses. Using the theory of poroelasticity
as a starting point we analyze mechanisms of these dependencies related to changes of the rock-pore-space
geometry. We develop a formalism describing elastic moduli and anisotropy of rocks as non-linear functions of
confining stresses and pore pressure. This formalism appears to be in agreement with laboratory observations.
In a first approximation elastic moduli, seismic velocities as well as the porosity depend on the differential
stress, i.e., the difference between the confining tectonic stresses and the pore pressure only. However, generally,
both, the differential stress tensor and the pore pressure must be taken into account as independent variables.
It is the stress dependent geometry of the pore space that fully controls the stress induced changes of elastic
moduli and seismic velocities. Specifically, the compliant porosity plays the most important role, in spite of
the fact that in many rocks the compliant porosity is just a very small part of the total porosity. Changes of the
compliant porosity with differential stress explain the often observed exponential behaviour of elastic moduli.
We introduce a tensor quantity defining the sensitivity of elastic moduli of rocks to the differential stress and call
it the stress-sensitivity tensor. The stress sensitivity is an important physical property directly related to elastic
non-linearity of rocks.

1 INTRODUCTION dependencies of elastic properties. Moreover, some


of them works in very restricted ranges of pore pres-
Theoretical approaches to understanding stress depen- sure changes or under very restricted geometrical or
dencies of the rock elasticity are necessarily related to geomechanical conditions.
acknowledgement of the fact that rocks behave like From the other hand it is known, that under hydro-
non-linear bodies. Several, quite successful attempts static load the pore pressure-, confining stress- and
to use the formalism of non-linear elasticity theory for differential stress dependencies of seismic velocities
this goal are known from recent literature (Sarkar et al., or elastic moduli are phenomenologically described
2003, Johnson and Rasolofasaon, 1996) However, by the following simple relationship (see e.g.,
these models are restricted to small ranges of stress Zimmerman et al., 1986; Eberhart-Phillips et al., 1989)
changes only, so that the resulting stress dependencies
of elastic parameters principally are liner functions of
stress only.
Such investigations are usually based on a very
general theory of elastic non-linearity. However, it is where P = Pc Pp is the differential pressure, Pc =
intuitively clear that the elastic non-linearity of rocks ii /3 is a confining pressure, ij is a component of
is related to a complex heterogeneity of the rock struc- the total stress tensor (in this paper, the compression
ture. The first-order heterogeneities of rocks are their stress is negative and the summation over repeating
voids, i.e., pores and fractures. A general theory of indices is assumed) and Pp is a pore pressure. The
elastic non-linearity does not take this specifics of coefficients A, K, B and D of equation (1) are fitting
rocks into account. Several following approaches can parameters for a given set of measurements.
be counted to attempts to specify models of porous It is often observed that equation (1) or similar
space geometry in order to arrive to a more spe- equations describing an exponential saturation to a
cific elastic non-linear rock characterization.These are linear trend provide very good approximations for
spherical contacts models (Duffy and Mindlin, 1957) velocities and elastic moduli of dry as well as sat-
or crack contacts models (Gangi and Carlson, 1996). urated rocks in a range of stress changes of several
These approaches are used in geophysical applications. hundreds megapascals. Moreover, it is also observed
These models lead to different quite complex stress that this equation provides a very good approximation

167
Copyright 2005 Taylor & Francis Group plc, London, UK
for elastic properties of anisotropic rocks (see our 2 STRAIN OF PORE SPACE
Figure 1).
In our previous publication (Shapiro, 2003) we In the following we assume that both, the solid as
attempted to explain this equation for isotropic rocks well as the pore space, are continuous. We consider
and hydrostatic loads. We have introduced a rock a thought experiment where a porous pressure as well
property called the elastic piezosensitivity (i.e., pres- as a uniform confining stress ij acting on the external
sure sensitivity; note that this property has nothing surface ! of a rock specimen can be changed. Sur-
to do with characteristics of piezoelectric effects) face ! is the external surface of the solid part of the
which controls exponential terms in equations of the rock. Simultaneously it seals also pores. In a given
type (1). point x of the surface ! the applied traction is then
In this work we consider the case of anisotropic
rocks and an arbitrary symmetry of the tectonic load.
We show that equations of the form (1) can be derived where nj (x) are components of the outward normal
from a rather general consideration. We assume that of !.
the elastic non-linearity of rocks is very weak, so that, Let us assume that the confining stress and/or
for example, seismic wave propagation phenomena porous pressure has changed from the load state
and other small deformation (incremental) effects can (ij0 , Pp0 ) to the load state (ij , Pp ). As a result, points
be considered like purely linear. It is just changes of
of the external surface has been displaced on ui (x).
linear elasticity of propagation media due to changing
The displacement is assumed to be very small in
tectonic loads or pore pressure what must be taken into
comparison with the size of the rock volume under
account. Moreover, we assume that the grain mate-
consideration.
rial is completely linear. We also neglect all effects
For convenience, we assume also that to the surface
of geometrical non-linearity (related to quadratic and
of the pore space not only just a scalar pressure but
higher combinations of displacement gradients). This f
last assumption is justified by the following orderof- rather a uniform stress ij can be applied. Also here,
magnitude estimate. Usually, in well consolidated the stress compressional in respect to the solid phase
rocks strains due to moderate stress changes (few is negative. If the load on the pore space surface is
f
hundred of megapascals) are of the order of 102 or hydrostatic, then ij = Pp ij . The differential stress
less. This is also the order of rotations and rock sam- is defined as:
ple changes. Geometrical non-linearity effects are of
the order 104 . However, relative variations of elastic
velocities are of the order of 101 . Therefore, effects For example, for a completely hydrostatic load
of physical non-linearity must be a dominant reason
of these changes.
Our hypothesis is following. In a range of moderate
changes of tectonic stresses (few hundred of mega- Note also that we use here a different definition for
pascals) the non-linearity of drained rocks is com- the term differential stress than those applied in
pletely controlled by deformations of the pore space tectonophysics, where by such a term usually a dif-
and, therefore, it is related to changes of the geom- ference between minimum and maximum principal
etry of fracture boarder and grain contact vicinities. components of the tectonic stress tensor is denoted.
Changes of contact geometries is directly related to By analogy with the paper of Brown and Korringa
changes of their elastic properties. In this sense our (1975), we introduce compliances of an anisotropic
dr
approach is similar to the works cited above. porous body: Sijkl the compliance of a drained rock,
A significant difference to the previous approaches mt p
Sijkl the compliance of the grain material, Sijkl the
is the following. To account for changes of the pore compliance of the pore space.
space geometry we use the laws of porosity evolu- We analyze the porosity weighted strain tensor of
tion obtained from the theory of poroelasticity. We the pore space ij . This quantity describes geometrical
derive general laws how the geometry of porous changes of the pore space normalized by the rock vol-
space behaves by loading porous systems. In this ume. The trace of this tensor is directly related to the
derivation we keep our considerations in terms of the porosity . If the external surface ! is loaded by an
poromechanics following concepts of Biot (1962) and anisotropic stress and the internal surface of the pore
following and generalizing the approach of Brown space is loaded by a pore pressure Pp , we will obtain the
and Korringa, (1975). Then, we take these laws into following equation for strain changes of pore space:
account for analyzing stress dependencies of elastic
moduli. On this way we arrive at a generalization of the
notion of the piezosensitivity to the stress-sensitivity
tensor for anisotropic rocks and non/hydrostatic loads.

168
Copyright 2005 Taylor & Francis Group plc, London, UK
For the porosity, under a completely hydrostatic load much smaller than the s0 and even than the absolute
these equations are reduced to the following one: value of s . Thus, the following inequality is usually
valid:

For example, in porous sandstones typical orders


of magnitude of these quantities are s0 = 0.1,
Results (5)(6) show that generally the porosity is a |s | = 0.01 and c = 0.001.
function of the both, differential stress as well as pore Analogously with (8) we also represent ij :
p
mt
pressure. However, if Siikk = Siikk (this condition is con-
sistent with the Gassmans equation; see Brown and
Korringa (1975)) then porosity changes depend on
the differential stress only. If porosity is small, then In the following we call quantity ij generalized
both porosity changes and strain of the pore space porosity. Quantity ijs0 denotes the stiff part of ij in
depend on differential stress only. In the following the unloaded state. Quantity ijs denotes changes of
we will work in this approximation. Under hydrostatic the stiff part of the generalized porosity due to a load.
conditions this rule reduces to the porosity depen- Thus, if the load is absent, ijs = 0. Further, ijc denotes
dence on the differential pressure, like it was shown the compliant part of the generalized porosity. It can be
by Zimmerman, et al., 1986: completely closed under a compressional differential
stress of the order of few hundred megapascal. By ijc0
we will denote compliant part of the generalized poros-
ity in the unloaded state. We define also c = 0 for a
configuration with a closed compliant porosity. Then,
where C mt and C dr are bulk compressibilities of the
as a first-order approximation we can use c = iic ,
grain material and of the drained rock, respectively.
s0 = ijs0 , and s = iis .
In the definitions above we introduced asymmet-
ric notations for the stiff and compliant porosities. It
3 STIFF AND COMPLIANT POROSITIES
is justified by different physical behaviour of them.
We assume that under moderate loads considered here
Analogously to Shapiro (2003) we separate the total
the stiff porosity suffers small changes only. In con-
porosity into two parts
trast to this, the compliant porosity can be significantly
changed or even completely closed. Thus, the notation
system introduced above is convenient for describing
such an asymmetric behaviour.
where the first part, c , is a compliant porosity sup- Taking into account that quantities ijs and ijc intro-
ported by thin cracks and grain contact vicinities. duced above are of the order of the strain, it is logic to
According to laboratory observations we expect that assume the first, linear approximations of the skele-
the compliant porosity will close up by the compres- ton compliances as functions of these quantities. The
sional differential stress of few hundred megapascals. Taylor expansion gives:
This corresponds to the porosity with an aspect ratio
(a relationship between the minimum and maximum
dimensions of a pore) less than 0, 01 (see Zimmerman
et al., 1986). The second part, [s0 + s ], is a stiff
porosity supported by more or less isomeric pores (i.e.,
drs
equidimensional or equant pores). The aspect ratio of where Sijkl is the drained compliancy of a hypothetical
such pores is typically larger than 0.1. rock with a closed compliant porosity (i.e., c = 0) and
In turn, we separate the stiff porosity into a part the stiff porosity equal to s0 . Further,
s0 , which is equal to the stiff porosity in the case of
d = 0, and to a part s which is a change of the stiff
porosity due to a deviation of the differential stress
from zero. We assume that the relative changes of the
stiff porosity, s /s0 , are small. In contrast, the relative where the derivatives are taken in points s = 0 and
changes of the compliant porosity (c c0 )/c0 can c = 0, respectively and C drs is the bulk compress-
be very large, i.e., of the order of 1 (c0 denotes the ibility of the hypothetical rock mentioned above.
compliant porosity in the unloaded case d = 0). Note, Quantities are individual for a given piece of rock
however, that c and c0 are usually very small quan- considered in a loading experiment. We assume that in
tities. As a rule, (e.g., in porous sandstones) they are such an experiment as soon as a changing load leads

169
Copyright 2005 Taylor & Francis Group plc, London, UK
to the same ijs and ijc the elastic moduli will also closed or not. If the compliant porosity is closed then
assume the same values. We also assume that the same ijc = 0 and we obtain from (15)
configuration of the load (independent of its history)
will give the same configuration of ijs and ijc . Thus,
non-hysteretic deformations only are considered.
Approximation (11) implies that the quantities of
the form are smaller than 1. Numerous laboratory However, if the assumption above is valid then
experiments and practical experience show that the this relationship will be valid also for an arbitrary
drained compressibilities depend strongly on changes (however, because of other assumptions, small) c .
in the compliant porosity, and depend much weaker Therefore,
on changes in the stiff porosity. We will express this
empirical observation by the restriction

A complete analysis of equation system (17) is a chal-


lenging problem. We consider orthorhombic media
If so, the approximation (11) further can be used in the only. We assume also that the main stress compo-
following simplified form: nents I iid (I is equal to 1, 2 or 3 denoting one
of index combinations 11, 22 or 33; note also that
there is no summation over i here) act perpendicular
to the symmetry planes of the orthorhombic system.
Moreover, we will naturally assume that the medium
Beside this approximation we use also following sim- stay orthorhombic with the same symmetry plains in
plifying assumptions. In direct additive terms we a loaded state (this also includes possibilities for the
neglect in comparison with 1, i.e., we will con- medium to become transversely isotropic, cubic or
sider rocks with moderate or small porosity of the isotropic).
order of 0.1 or less. We assume that S mt and S p can The most simple, isotropic stress-sensitivity tensor
be neglected in comparison with S dr . This is quite is characterized by only 3 independent components. A
a realistic assumption for rocks at least in the upper detailed analysis of equations (17) shows that in real-
Earth crust. Finally, we will assume that magnitude istic for rock situations only one of these components
of differential stress changes is not smaller than this is not vanishing. In this case, changes of quantities
of pore pressure changes. This corresponds to realistic c c c
11 , 22 and 33 are completely independent from
geological and technical processes in the underground. each other. This is equivalent to an assumption that
Under the assumptions summarized above we mod- there are three non-intersecting components of the
ify equation (5): compliant porosity independently changing by apply-
ing corresponding uniaxial differential stresses. Such
a situation is possible for rocks. Thus, we analyse it
further.
Equations (11) and (15) provide the searched for Moreover, if we assume that changes of quanti-
c c c
description of stress dependencies of rock elastic mod- ties 11 , 22 and 33 are completely independent from
c
uli. The last equation shows that tensor quantity klijmn each other even for lower symmetries of the stresssen-
sitivity tensor, our consideration will become more
controls these dependencies.
c general and still very simple. This assumption (note
Tensor quantity klijmn is analogous to the scalar
that in the case of an isotropic stress sensitivity, it is
dimensionless quantity called piezosensitivity in not an assumption but rather the only possible realistic
(Shapiro, 2003). Further we will call this quantity the situation) gives:
tensor of stress sensitivity. In the first approximation
the stress-sensitivity tensor possess symmetries coin-
ciding with the symmetries of the third-order elastic
coefficient tensor of non-linear elastic media.

4 STRESS DEPENDENCIES OF ELASTIC Here we introduced new notations 1c , 2c and 3c for


c c c
PROPERTIES 111111 , 222222 and 333333 , respectively.
We consider now an arbitrary elastic characteristics
We assume that stiff porosity changes with stress are  (e.g., a seismic velocity, a stiffness or a compliance)
independent of the changes of the compliant porosity. of a porous drained body. We assume that the character-
This means also, that changes of the stiff porosity are istic  is such that in the point where porosity is equal
independent on the fact if the compliant porosity is to s0 function  can be expanded in the Taylor series

170
Copyright 2005 Taylor & Francis Group plc, London, UK
(in a way similar to equation (11)) relative to the poros- Here, BNNM =C drs NNM c
Mc0
/SNN drs
(no summation over
ity (this should be valid for all such characteristics like N , M ), as well as Ei = e ic i C drs
(no summation over i).
seismic velocities and elastic moduli): Note that in the contracted notation components NKM c
drs dr n
as well as SNK and SNK are 2 times the correspond-
ing tensor coefficient, where n is the number of
where we have kept only linear part of the Taylor appearances of 4, 5 or 6 as a subscript.
expansion. Further, To discuss stress dependence of anisotropy we
consider Tsvankins parameters (so-called anisotropy
measures , and ) for an orthorhombic medium
under drained conditions (we omit in the following
the index dr). Usually these parameters are given in
and the derivatives are taken at s = 0 and c = 0, terms of stiffnesses (see e.g., Sarkar et al., 2003).
respectively. Substituting equations (16) and (18) into To consider a stress induced elastic anisotropy
equation (19) we obtain: drs
we assume that the compliances tensor Sijkl is
isotropic only. In addition, we assume that the
stress sensitivity tensor and ijc0 are also isotropic.
This last assumption leads to the following equiva-
lences for non vanishing NNM c c
: 111 = 222
c
= 333
c
=
442 = 443 = 551 = 553 = 661 = 662 = , and also
c c c c c c c
where K and I can assume one of values 1, 2 or 3 1c0 = 2c0 = 3c0 = c0 /3. In this approximation
denoting 11, 22 and 33, respectively. In the exponent Tsvankins parameters simplify to:
there is no summation over repeating indices.
A comparison of this result with equation (1) shows
that they have the same form. In the case of a uniaxial
stress N of a given direction N all coefficients D are
identic to a single exponent DN characterizing all elas-
tic quantities changing by such a load. A result of an
arbitrary three axial load is equivalent to a simple sum
of changes due to corresponding uniaxial stresses.

drs drs
5 STRESS INDUCED ANISOTROPY Here, A1 = 23 S44 (S
S11 C
44 4S11 )
and A2 = 16 CS44 . If stress
changes are small the exponential functions Ei in these
Significant changes of elastic characteristics of rocks equations can be expanded in Taylor series. Then, in
are produced by changes of compliant porosities only. the linear approximation with respect to stress the
Thus, for the further analysis it is reasonable to neglect resulting formulas will be completely analogous to
the contributions of the stiff porosity. We also recall equations (19)(23) of Sarkar et al., (2003). However,
that we consider orthorhombic and higher symmetry. in contrast to their results the equations above are also
Then, in the case of a uniaxial stress N of a given direc- valid for moderate stress changes.
tion N compliances changing due to such a load are In the case of a hydrostatic load
dr dr dr dr
SNN (no summation over N here), S44 , S55 and S66 . For
a three-axial stress we obtain then the following stress
dependencies of the compliances (from this point on
all capital indices can accept one of the values 1, 2, 3, Obviously, in this case anisotropy does not change
4, 5 or 6 and we use the standard contracted notation): with stress. For initially isotropic rocks the situation
reduces to the one considered in (Shapiro, 2003).
Quantity c is then the piezosensitivity. Moreover, even
in the case of initially anisotropic rocks only one sin-
gle quantity D = c C drs controls the exponential parts
of the pressure dependency of any compliance, of
any stiffness and of any elastic wave velocity. In fact,
there are anisotropic rocks, that show this interesting
behaviour. Figure 1 gives an example for such a rock.
This metamorphic sample was taken from the German
KTB deep drill hole. Three P- and six correspond-
ing S-wave velocities were simultaneously measured
in orthogonal directions over a differential pressure

171
Copyright 2005 Taylor & Francis Group plc, London, UK
stress and the pore pressure only. Equations of the
type V (P) = A + KP B exp (PD) are quite gen-
eral. The stress dependence of the porosity controls
the elastic moduli and velocity changes with stress.
Here, the most important role is played by the com-
pliant porosity which can be just a very small part
of the total porosity. Closing compliant porosity with
increasing differential stress explains the experimen-
tally observed exponentially saturating increase of
seismic velocities. Coefficients of this relationship are
defined by the compliant porosity dependence of the
drained bulk modulus.
The dimensionless tensor quantity c defines the
sensitivity of the elastic characteristics to the differ-
ential stress. In parallel with elastic moduli it is an
additional characteristic of rocks. We propose to call
it the tensor of stress sensitivity. It is defined by the
compliant porosity of rocks. A single stress sensitivity
is enough in the case of isotropic non-linearity and a
hydrostatic load. In this case the anisotropy does not
change under the load.

REFERENCES

Biot, M. A. (1962). Mechanics of deformation and acoustic


propagation in porous media. Journal of Applied Physics
33, 14821498.
Brown, R. J. S. and J. Korringa (1975). On the depen-
dence of the elastic properties of a porous rock on the
compressibility of the pore fluid. Geophysics 40, 608616.
Figure 1. Seismic velocities versus confining stress in a dry Detournay, E. and A.-D. Cheng (1993). Fundamentals
sample of a German KTB crystalline rock. The data are taken of poroelasticity. In J. Hudson (Ed.), Comprehensive
from Kern, (1997). The fitting curves are given by equations Rock Engineering: Principles, Practice and Projects,
of the type (1) with a single D. Chapter 5, pp. 113171. Oxford: Pergamon Press.
Duffy, J. and R. D. Mindlin (1957). Stress-strain relations
and vibrations of a granular medium. Journal of Applied
Mechanics 24, 585593.
range up to 600 MPa. We were able to fit all velocities Eberhart-Phillips, D., D.-H. Han, and M. D. Zoback (1989).
with an equation of type (1), where the fit parameter Empirical relationships among seismic velocity, effec-
tive pressure, porosity and clay content in sandstone.
D = 0.0093 MPa1 was identical for all velocities.
Geophysics 54, 8289.
Exactly as found by Sarkar et al., (2003) the results Gangi, A. F. and R. L. Carlson (1996). An asperity-
above also describe elliptical changes of anisotropy deformation model for effective pressure. Tectonophysics
due to the stress only. However, anisotropy changes 256, 241251.
are not elliptical in more general situations of a Johnson, P. A. and P. N. Rasolofosaon (1996). Nonlinearl
non isotropic stress-sensitivity tensor. An anelliptic- elasticity and stress-induced anisotropy in rock. Journal
ity remains non vanishing also in the case of uniaxial of Geophysical Research 101, 31133124.
stress and even in the case of a hydrostatic load. Sarkar, D., A. Bakulin, and R. L. Krantz (2003). Anisotropic
inversion of seismic data for stressed media: Theory and a
physicalmodeling study on Berea sandstone. Geophysics
43, 805830.
6 CONCLUSION Shapiro, S. A. (2003). Elastic piezosensitivity of porous and
fractured rocks. Geophysics 68, 482486.
In the first approximation elastic moduli, seismic Zimmerman, R. W., W. H. Somerton, and M. S. King (1986).
velocities as well as the porosity depend on the differ- Compressibility of porous rocks. Journal of Geophysical
ential stress, i.e., the difference between the confining Research 91, 1276512777.

172
Copyright 2005 Taylor & Francis Group plc, London, UK

S-ar putea să vă placă și