Fracture of disordered solids in compression as a critical phenomenon: I. Statistical mechanics formalism
Extracted source text ยท arXiv cond-mat/0209124
%% LyX 1.1 created this file. For more info, see http://www.lyx.org/.
%% Do not edit unless you really know what you are doing.
%\documentclass[aps,twocolumn,pre,groupedaddress,showpacs]{revtex4}
\documentclass[aps,twocolumn,pre,groupedaddress,showpacs,showkeys]{revtex4}
\usepackage[T1]{fontenc}
\usepackage[latin1]{inputenc}
\usepackage{graphics}
\makeatletter
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% LyX specific LaTeX commands.
\providecommand{\LyX}{L\kern-.1667em\lower.25em\hbox{Y}\kern-.125emX\@}
%% Special footnote code from the package 'stblftnt.sty'
%% Author: Robin Fairbairns -- Last revised Dec 13 1996
\let\SF@@footnote\footnote
\def\footnote{\ifx\protect\@typeset@protect
\expandafter\SF@@footnote
\else
\expandafter\SF@gobble@opt
\fi
}
\expandafter\def\csname SF@gobble@opt \endcsname{\@ifnextchar[%]
\SF@gobble@twobracket
\@gobble
}
\edef\SF@gobble@opt{\noexpand\protect
\expandafter\noexpand\csname SF@gobble@opt \endcsname}
\def\SF@gobble@twobracket[#1]#2{}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% User specified LaTeX commands.
\usepackage[T1]{fontenc}
\usepackage[latin1]{inputenc}
\usepackage{graphics}
\makeatletter
\usepackage{natbib}
\usepackage{graphicx}
\usepackage{bm}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsmath}
\def\dd{\! : \!}
\def\ep{{\bm \varepsilon}}
\makeatother
\makeatother
\begin{document}
\title{Fracture of disordered solids in compression as a critical phenomenon:\\
I. Statistical mechanics formalism}
\author{Renaud Toussaint}
\email[email: ]{Renaud.Toussaint@fys.uio.no}
\altaffiliation[Present address: ]{ Department of Physics,
University of Oslo, P.O. Box 1048 Blindern, 0316 Oslo 3, Norway}
\affiliation{G\'eosciences Rennes, Universit\'e de Rennes 1, 35042 Rennes Cedex, France}
\author{Steven R. Pride}
\email[email: ]{Steve.Pride@univ-rennes1.fr}
\affiliation{G\'eosciences Rennes, Universit\'e de Rennes 1, 35042 Rennes Cedex, France}
\date{\today}
\begin{abstract}This is the first of a series of three
articles that treats fracture localization as a critical phenomenon.
This first article establishes
a statistical mechanics
based on ensemble averages when fluctuations through time play no role in defining the ensemble.
Ensembles are obtained by dividing a huge rock sample into many mesoscopic volumes.
Because rocks are a disordered
collection of grains in cohesive contact, we expect that once shear strain is
applied and cracks begin to arrive in the system, the mesoscopic volumes will have a wide
distribution of different crack states. These mesoscopic volumes are the members of our ensembles.
We determine the probability of observing a mesoscopic volume to be in a given crack state
by maximizing Shannon's measure of the emergent crack disorder subject to constraints
coming from the energy-balance of brittle fracture. The laws of thermodynamics, the
partition function, and the quantification of temperature are obtained for such
cracking systems.
\end{abstract}
\pacs{46.50.+a, 46.65.+g, 62.20.Mk, 64.60.Fr}
\keywords{localization, fracture, critical phenomena}
\maketitle
\section{INTRODUCTION}
When rocks and other disordered-solid materials are in compression and then
have an additional deviatoric strain applied to them, small stable cracks irreversibly
appear at random throughout the material. Each time the deviatoric strain is
increased, more cracks appear. In the softening regime following peak stress,
a sample will unstably fail along a plane localized at an angle relative to
the principal-stress direction. We have accumulated evidence suggesting that
such localization is a continuous
phase transition.
This is the first of three articles that develops
a statistical
mechanics that allows the possible phase transitions in a cracking solid to be
investigated.
Many studies have assumed that fracture is a thermally-activated process and
have used a statistical mechanics based on thermal fluctuations {[}1-5{]}.
\nocite{SWG+91}\nocite{RK89}\nocite{WLS+91}\nocite{GP95}\nocite{Mea90}
However, our interest here is with ``brittle fracture'' in which cracks appear
irreversibly and in which thermal fluctuations play no role. For this problem,
the statistics of the fracture process is entirely due to the initial quenched
disorder in the system.
A considerable literature has developed for so-called ``breakdown'' phenomena
in systems having quenched disorder and zero temperature {[}6-23{]}.
\nocite{BH98}\nocite{DAH89}\nocite{DLB87} \nocite{AS93a}\nocite{SA96}
\nocite{fracture-o:1989}\nocite{KHH97}\nocite{burst-aval:1994}
\nocite{the-distri:1992}\nocite{failur-of-:1996}\nocite{MGP00}
\nocite{DAHR89}\nocite{HHR91b}\nocite{HR90}\nocite{ZRS+97}\nocite{ZRS+99}
In particular, the burned-fuse {[}6-8{]},\nocite{BH98}\nocite{DLB87} spring-network
{[}9-11{]},\nocite{AS93a}\nocite{fracture-o:1989} and fiber-bundle {[}12-17{]}
\nocite{KHH97}\nocite{MGP00}
analog models for fracture have all been shown to yield various types of scaling
laws prior to the point of breakdown {[}18-23{]} \nocite{DAHR89}
\nocite{ZRS+97}\nocite{ZRS+99}.
Our work is different in that we directly treat the fracture
problem (not an analog model of it) assuming that all of the statistics
is due to quenched disorder. We obtain
the probability of
emergent damage states by maximizing Shannon's entropy subject to appropriate
constraints. This approach has recently been proven exact in the special
case of fiber bundles \cite{PT02}.
The principal conclusion of our present theory is that at a critical-strain
point, there is a continuous phase transition from states where cracks
are uniformly distributed to states where coherently oriented cracks are grouped
into conjugate bands. Several facts justify classifying
such band formation as a critical phenomenon.
First, the localization of the cracks into bands spontaneously breaks both the
rotational and translational symmetry of the material even though
our model Hamiltonian preserves these same symmetries.
The entropy of the material remains continuous and the
ensemble of the most probable states becomes degenerate at the localization
transition; {\em i.e.}, prior to localization, the most probable state
is the intact state, while right at the
transition, certain banded states acquire
the same probability as the intact state.
Further, an autocorrelation length associated with the aspect
ratio of the emergent crack bands diverges in the approach to the critical point.
Unfortunately, quantitative laboratory measurements of how the bands of cracks
coalesce and evolve in size and shape prior to the final localization point do
not presently exist.
We speculate in the third article of
this series on how such measurements might be performed.
Our explanation of localization based on the physics of interacting
cracks is distinct from
the bifurcation analysis of
Rudnicki and Rice \cite{RR75} in which localization is a
consequence of a proposed phenomenological elasto-plasticity law.
Our work provides a method for obtaining such a plasticity law
from the underlying physics.
\section{THE PROBABILISTIC NATURE OF THE FRACTURE PROBLEM}
Rocks are a disordered collection of grains in cohesive
contact. The grains have varying shapes and sizes with typical grain sizes
in the range of 10--100 \( \mu \)m but sometimes considerably larger. The
contacts between the grains are generally weaker than the grains themselves
and have strengths and geometries that vary from one contact to the next. When
deviatoric (\emph{i.e.}, shear) strain is applied to a rock, grain contacts
begin to break. In what follows, a broken grain contact will be called a ``crack''.
Such a break is a stress-activated irreversible process. Once a grain contact
is broken, there is no significant healing that occurs. Cracks are not arriving
and disappearing due to thermal fluctuations. This fact makes our definition
of statistical ensembles quite different from that in the usual application
of statistical mechanics to molecular systems as we now go on to discuss.
\subsection{Creating a statistical ensemble}
\label{statens}
We imagine dividing a huge (formally infinite) system into mesoscopic
volumes that will be called ``mesovolumes''. Because the materials of interest
here have a wide range of grain-scale disorder, many different crack states
will emerge in the various mesovolumes once energy has been put into the system
and cracking begins. These various mesovolumes and the crack states they contain
comprise the ensembles in our theory.
In order to be specific with our ideas, we now introduce a simple model of the
initial disorder and emergent crack states. The purpose of this special
model in the present paper is to motivate how ensembles are formed; however,
the model Hamiltonian developed
in Paper II will be based upon it.
In the model, each mesovolume
is divided into \( N \) identical cells where a cell has dimensions on the order of a
grain size and where \( N \) is a large number such as \( 10^{2D} \)
or more with \( D \) the system's dimension.
In each cell, only a single grain contact is allowed to break.
The local order parameter (explicitly defined in Paper II) characterizes
both the orientation and length of such a broken grain contact. In
the present paper, an order-parameter description is not yet necessary.
Prior to breaking, all cells are assumed to have the same elastic moduli.
The quenched disorder is in
how the grain-contact breaking energy \( {\mathcal{E}}({\bm {x}}) \)
is distributed in the cells \( {\bm {x}} \) of a mesovolume. We
assume that only a fraction of the nominal grain-contact area is actually
cemented together, and that the degree of cementation from one contact to the
next is random.
Thus, the breaking energies
\( {\mathcal{E}}({\bm {x}})\)
are random variables independently sampled from a distribution $\pi({\cal E})$
having support on
$[0,\Gamma d^{D-1}]$ where
\( \Gamma \) is the surface-energy density
of the mineral, $d$ is the nominal linear dimension of a grain contact,
and $d^{D-1}$ is the grain-contact area in $D$ dimensions.
The quenched-disorder distribution $\pi({\cal E})$ can have any assumed form.
We now define an infinite collection of distinct mesovolumes by allowing for
every conceivable way that \( {\cal E} ({\bm {x}}) \) may be distributed
in a mesovolume.
Putting this collection together forms the infinite rock mass whose properties
we are interested in determining. Each mesovolume so defined is a deterministic
system and upon slowly applying the same strain tensor \( {\bm \varepsilon } \)
to all the mesovolumes, each will undergo a deterministic cracking scenario
and end up in a well defined crack state. We denote each of the possible final
crack states with an index \( j \).
A principal goal of the present paper is to obtain the occupation probabilities
\( p_{j} \) of these various crack states which are simply the fraction of
the mesovolumes in the system that are in the state \( j \).
We can understand how the various crack states
emerge by appealing to a form of Griffith's \cite{Gri20} criterion. A cell
will break only if the change in the elastic energy due to the break is greater
than or equal to the bond-breaking energy \( {\mathcal{E}}({\bm {x}}) \).
If \( {\textbf {C}}_{a} \) is the effective elastic-stiffness tensor of the
entire mesovolume that holds after the break occurs and if \( {\textbf {C}}_{b} \)
is the stiffness-tensor that held before the break,
Griffith's criterion
can be stated
\begin{equation}
\label{Grifeq}
\ell ^{D}\, {\bm \varepsilon }\dd ({\textbf {C}}_{b}-{\textbf {C}}_{a})\dd {\bm \varepsilon }/2>{\mathcal{E}}
\end{equation}
where \( {\bm \varepsilon } \) is the strain tensor characterizing the entire
mesovolume at the moment of the break and \( \ell ^{D} \) is the volume of
a mesovolume. This particular statement is an approximation
based on an assumed linear elasticity and absence of residual strain after unloading,
but a general statement will be derived in Section \ref{work}. Since the mesovolume
with an extra crack is more compliant than without it, the weakest cells will
begin to break even after the slightest of applied strain.
Yet an emergent crack state is not just a trivial consequence of
the ${\cal E} ({\bm x})$ distribution in a mesovolume. Cracks
aligned along bands concentrate stress allowing
even large barriers ${\cal E} ({\bm x})$ to be overtaken along the band.
In the present model, this means that placing cracks along
bands produces a larger change in the elastic moduli of the mesovolume than
placing cracks in more random positions.
Thus, at least above
some applied strain level, we expect the banded states to emerge as the ones that
are significantly present in a rock system.
Non-banded states at large strain are much more special.
They can come only from mesovolumes
in which the weak cells making
up the state are all surrounded by strong cells.
A key idea here is that each mesovolume embedded in the system experiences
the same global strain tensor and, as such, has a crack state statistically
independent from the other mesovolumes. This is only valid so long
as the emergent bands
of organized cracks have a dimension \( \xi \) that is small relative to the
size \( \ell \) of the mesovolume. Screening effects due to destructive strain
interactions between incoherently oriented cracks cause the far-field strain
from a local crack structure to fall off with distance \( r \) even more rapidly
than the \( (\xi /r)^{D} \) fall off in an uncracked material.
But even in the thermodynamic limit of infinite system sizes,
the required statistical independence of the mesovolumes
breaks down right
at the critical strain where
divergent bands of cracks become important. The conclusion is that although
our ensemble-based statistics is valid in the approach to localization, it is
incapable of describing the post-localization physics.
\subsection{Macroscopic observables}
\label{macroobs}
In the laboratory experiments to which we apply our theory,
a sample is immersed in a reservoir from which either uniform
stress or strain conditions can be applied to the sample's exterior
surface $\partial \Omega$.
The macroscopic strain tensor ${\bm \varepsilon}$ is defined in
terms of the
displacement \( {\textbf {u}} \)
at points on \( \partial \Omega \) as
\begin{equation}
{\bm \varepsilon }=\frac{1}{L^{D}}\int _{\partial \Omega }{\textbf {n}}{\textbf {u}}\, dS
\end{equation}
where \( {\textbf {n}} \) is the outward normal to the sample's surface and
\( L^{D} \) is the volume of the sample in \( D \) dimensions. This definition
of deformation thus corresponds to the volume average of the local deformation
tensor \( \nabla {\textbf {u}}({\bm {x}}) \) defined at interior points
\( {\bm {x}} \) of the sample. It will soon be shown to be
conjugate to the macroscopic stress tensor ${\bm \tau}$ in the expression
for the work carried out on the sample. If strain (rather
than stress) is the control variable, the
displacements at points ${\bm x}$ of the external surface $\partial \Omega$ are given by
${\bf u} = {\bm x} \cdot {\bm \varepsilon}$.
\begin{figure}
{\par\centering \resizebox*{7.5cm}{5.5cm}{\includegraphics{fig2.eps}} \par}
\caption{Stress-strain data courtesy of David Lockner of the USGS Menlo Park. The slope
measured upon loading a sample is defined by \protect\protect\( {\textbf {D}}\protect \protect \)
while that measured upon unloading and/or reloading the sample is defined by
\protect\protect\( {\textbf {C}}\protect \protect \). }
\label{Lockner}
\end{figure}
As shown in Fig.\ \ref{Lockner}, a typical compression experiment starts with
the sample in a pure hydrostatic pressure state and then systematically increases
the deformation in the axial direction, keeping the radial ``confining'' pressure
\( p_{c} \) constant. Other ways of controlling the radial stress during
the experiment are to keep a constant ratio between axial and radial stress,
or to impose a constant radial deformation. So long as the confining pressure
does not become so large as to induce a brittle-to-ductile transition
\cite{Pat78}, these various experiments all result in the same type of
localized structure at large axial strains.
When axial strain is monotonically increased, cracks arrive at each
strain increment and the deformation and stress changes are related as
\begin{equation}
d{\bm \tau }=\frac{d{\bm \tau }}{d{\bm \varepsilon }}\dd d{\bm \varepsilon }
={\textbf {D}}\dd d{\bm \varepsilon }
\end{equation}
where the fourth-order tensor \( {\textbf {D}} \) is called the tangent-stiffness
tensor. This tensor defines the slopes between the various stress and strain
components as the sample is being loaded and is an experimental observable.
If at some point in the stress history the axial pressure is reduced, we follow
a different deformation path as seen in the figure due to the fact that no new
cracks are created. Such an unloading experiment defines the elastic (or secant)
stiffness tensor \( {\textbf {C}} \)
\begin{equation}
d{{\bm \tau }}={\textbf {C}}\dd d{\bm \varepsilon }.
\end{equation}
We model the unloading/reloading paths as being entirely reversible and in so doing
neglect the small hysteresis due to
friction along the opened cracks.
In order to distinguish loading paths (with crack creation) from unloading paths
(without crack creation), all
properties are explicitly taken to depend on two strain variables; namely,
the maximum strain \( {\bm \varepsilon }_m \) having been applied to a sample,
and the current strain
\( {\bm \varepsilon } \) that is different than the maximum only if the sample
has been subsequently unloaded.
Note that even if \( {\bm \varepsilon }\) and \( {\bm \varepsilon }_m \)
are written as tensors, they each correspond to only one scalar
degree of freedom along the loading/unloading paths, since the radial components
can always be expressed in terms of the axial components via the type of radial
control employed
({\em e.g.}, \( p_{c} = \rm{const} \)
in a standard
triaxial test).
The stress tensor ${\bm \tau}$ in the theory corresponds to the volume average
of the local stress tensor ${\bf T}({\bm x})$ that satisfies
$\nabla \cdot {\bf T}(\bm x) = 0$ at interior points ${\bm x}$; {\em i.e.},
\( {\bm \tau }=L^{-D}\int _{\Omega }{\textbf {T}}({\bm {x}})\, dV \) and
is a function of the current and maximum strains ${\bm \tau} =
{\bm \tau} ({\bm \varepsilon}, {\bm \varepsilon}_m)$ as shown in Fig.\ 1.
By averaging the elastostatic identity \( \nabla \cdot ({\textbf {T}}\, {\bm {x}})
={\textbf {T}} \) over the mesovolume we further have that
\( {\bm \tau }=L^{-D}\int _{\partial \Omega }{\bf n} \cdot {\textbf {T}}{\bm {x}}\, dS \).
The work density \( dU \) performed on the sample when there is an increment
in strain \( d{\bm \varepsilon } \) is in both cases of loading and unloading
\begin{eqnarray}
dU & = & \frac{1}{L^{D}}\int _{\partial \Omega }{\textbf {n}}\cdot
{\textbf {T}}\cdot d{\textbf {u}}\, \, dS\label{work1} \\
& = & {\bm \tau }\dd d{\bm \varepsilon }.\label{work2}
\end{eqnarray}
To obtain Eq.\ (\ref{work2}) from (\ref{work1}), we have written the
controlled displacements on a sample's surface as $d{\bf u} = {\bm x} \cdot
d{\bm \varepsilon}$ where the strain increment $d{\bm \varepsilon}$ is uniform over $\partial
\Omega$.
Thus, \( dU \) corresponds to the volume average of the local
work density \( {\textbf {T}}({\bm {x}})\dd d\nabla {\textbf {u}}({\bm {x}}) \).
The total energy \( U \) per unit sample volume that goes into the sample during
the loading up to a maximum strain-tensor \( {\bm \varepsilon _{m}} \) is then
\begin{equation}
U({\bm \varepsilon _{m}})=\int _{\bm \varepsilon _{0}}^{\bm \varepsilon _{m}}{\bm \tau }
({\bm \varepsilon }',{\bm \varepsilon }')\dd d{\bm \varepsilon }'.
\end{equation}
where \( {\bm \varepsilon _{0}} \) is the strain associated with the initial
isotropic stress.
If after loading to
\( {\bm \varepsilon }_{m} \), the sample is unloaded back to a current strain of
\( {\bm \varepsilon } \), we have the general expression
\begin{equation}
U({\bm \varepsilon },{\bm \varepsilon }_{m})=U({\bm \varepsilon _{m}})
+\int _{\bm \varepsilon _{m}}^{\bm \varepsilon }{\bm \tau }(
{\bm \varepsilon }',{\bm \varepsilon }_{m})\dd d{\bm \varepsilon }'.
\end{equation}
If the sample is unloaded back to the initial stress, corresponding to a possibly
non-zero residual strain \( \varepsilon^{\rm res} \), a last experimental observable
is the energy \( Q({\bm \varepsilon }_{m})=U({\bm \varepsilon }^{\rm res},{\bm \varepsilon }_{m}) \)
(per unit sample volume) that went into crack creation and that is
lost during the loading process .
\subsection{Ergodic hypothesis}
We have shown above that the experimentally-measurable variables of energy density
\( U \), deformation \( {\bm \varepsilon } \), and applied stress \( {\bm \tau } \)
correspond to volume averages of each field throughout a system. Our ergodic
hypothesis amounts to assuming that the systems we work with are sufficiently
large that such volume averages can be replaced by ensemble averages
\begin{equation}
\label{ensaverage}
U=\sum _{j}p_{j}E_{j},\, \, \, \, \, {\bm \varepsilon }
=\sum _{j}p_{j}{\bm \varepsilon }_{j},\, \, \, \, \, {\bm \tau }=\sum _{j}p_{j}{\bm \tau }_{j}.
\end{equation}
Here, \( E_{j} \) is the average work per unit mesovolume required to take
an initially uncracked mesovolume from zero strain to the strain tensor \( {\bm \varepsilon }_{j} \).
A similar definition holds for \( {\bm \tau }_{j} \). In both the definition
of \( E_{j} \) and $\tau_j = dE_j/d{\bm \varepsilon}_j$, the average is over the
initial quenched-disorder distribution.
So long as each mesovolume contains crack states that have no significant influence
on the neighboring mesovolumes
(formally valid only in the thermodynamic limit), the sum
over the collection of mesovolumes (ensemble averaging) is equivalent
to a volume integral over the entire system.
In practice, we will only ever consider ensembles that have by definition
$ {\bm \varepsilon }_{j}={\bm \varepsilon} $; however, we could equivantly
immerse each mesovolume in a uniform stress-tensor reservoir and allow
$ {\bm \varepsilon }_{j}$ to vary from state to state.
\section{THERMODYNAMICS OF CRACK POPULATIONS}
\subsection{Fundamental postulate}
The fracture-mechanics problem of counting how many of the initial mesovolumes
can be led to the same crack state appears to be hopelessly intractable. Fortunately,
it also appears to be unnecessary for systems containing
initial quenched disorder. Upon putting deviatoric strain energy into such
a system, the emergent crack states $j$ will on the one hand attempt to
mirror this quenched disorder with weakest cells breaking first; however,
due to the energetics of the crack interactions,
many different types of initial mesovolumes may be led to the same crack state
which results in non-uniform crack-state probabilities \( p_{j} \) even if the
quenched-disorder distribution is uniform.
We state our fundamental postulate as follows: \emph{The probability \( p_{j} \)
of observing a mesovolume to be in crack state \( j \) can be determined by
maximizing Shannon's \cite{Sha48} measure of disorder}
\begin{equation}
\label{shannon}
S=-\sum _{j}p_{j}\ln p_{j}
\end{equation}
\emph{subject to constraints involving the macroscopic observables that derive
from the energetics of the fracture mechanics.} That entropy is to be maximized
can be expected since the quenched disorder allows all states to be present
in a sufficiently large system.
In recent work \cite{PT02}, we have demonstrated that this postulate yields
exact results for the special case of fiber bundles with global-load sharing.
The constraints are what give the dimensionless function \( S \) defined
by Eq.\ (\ref{shannon}) all the thermodynamic information about our cracking
system and must explicitly involve the independent variables of $S$.
Such independent variables are determined by establishing the first law of
thermodynamics
for a system cracking in compressive shear.
\subsection{The work of creating a crack state}
\label{work} To obtain the first law, it is first necessary
to define the detailed energy balance for each crack state and to understand
how the work \( E_{j} \) required to create state \( j \)
depends on both the
actual strain \( {\bm \varepsilon } \) and on the maximum-achieved strain
\( {\bm \varepsilon}_m \).
\subsubsection{Griffith's criterion and crack-state energy}
Consider a given mesovolume with a deterministic distribution of breaking energies
\( {\mathcal{E}}({\bm {x}}) \) assigned to each cell \( {\bm {x}} \)
of the mesovolume. Starting from a state of isotropic strain \( {\bm \varepsilon }_{0} \),
we slowly apply an additional axial deformation and monitor how one crack after
another enters the mesovolume until the final strain tensor \( {\bm \varepsilon } \)
and final crack state \( j \) are arrived at. Lets say that this state \( j \)
has a total of \( {\mathcal{N}} \) cracks associated with it.
\begin{figure}
{\par\centering \resizebox*{7.5cm}{4cm}{\includegraphics{fig4.eps}} \par}
\caption{The heavy line is the actual path followed during the steady application of
axial strain. Each vertical drop in stress corresponds to the arrival of a
crack.}
\label{crackhistory}
\end{figure}
Figure \ref{crackhistory} details the history of how the stress (and, therefore,
work) might evolve in the mesovolume as strain is applied and cracks arrive.
Initially, the mesovolume will elastically deform according to the stiffness
tensor \( {\textbf {C}}_{0} \) (no cracks yet present) until the first crack
arrives at the strain tensor \( {\bm \varepsilon }_{1} \) with an associated
drop in the mesovolume's stress. Lets say the bond-breaking energy of this first
crack was \( {\mathcal{E}}_{1} \). The mesovolume will now have a different
overall stiffness tensor \( {\textbf {C}}_{1} \) and will elastically deform
with these new moduli until the second crack arrives and so on until all \( {\mathcal{N}} \)
cracks have entered and the mesovolume has attained its final stiffness tensor
of \( {\textbf {C}}_{j}={\textbf {C}}_{\mathcal{N}} \).
The final tensor \( {\textbf {C}}_{j} \)
depends on both the location and orientation of these \( {\mathcal{N}} \)
cracks in addition to their number.
At some intermediate stage having \( n \) cracks, the stress tensor
\( {\bm \tau }_{n}({\bm \varepsilon }) \) is defined by integrating
\( d{\bm \tau }={\textbf {C}}_{n}({\bm \varepsilon }')\dd d{\bm \varepsilon }' \)
from \( {\bm \varepsilon }_{n}^{\rm res} \) to \( {\bm \varepsilon } \),
where \( {\bm \varepsilon }_{n}^{\rm res} \)
is the ``residual'' deformation observed upon unloading the
sample back to zero stress as shown in the figure. We have
\begin{equation}
{\bm \tau }_{n}({\bm \varepsilon })=
\int _{{\bm \varepsilon }_{n}^{\rm res}}^{{\bm \varepsilon }}{\textbf {C}}_{n}\dd d{\bm \varepsilon '}.
\end{equation}
The elastic energy density corresponding to this state at deformation \( {\bm \varepsilon } \)
is similarly
\begin{equation}
E_{n}^{\text {el}}({\bm \varepsilon })=
E^{\rm res}_{n}+\int _{{\bm \varepsilon }_{n}^{\rm res}}^{{\bm \varepsilon }}{\bm \tau }_{n}
({\bm \varepsilon '})\dd d{\bm \varepsilon '}
\label{elastic1}
\end{equation}
where \( E^{\rm res}_{n} \) represents the residual elastic energy that remains in the system
when the state with \( n \) cracks is unloaded to zero applied stress.
These residual (zero-stress) quantities are present whenever
plastic deformation occurs within a grain-contact.
After a sample elastically returns to zero applied
stress, such plastic deformation remains
and, accordingly, there is an elastic stress field surrounding any crack that experienced
plastic deformation. The strain energy associated with such local residual
stress is what constitutes the residual energy \( E_{n}^{\rm res} \).
When the \( n^{\textrm{th}} \)
crack arrives in a strain-controlled experiment, there is no change in the
strain \( {\bm \varepsilon }_{n} \) and thus no external work performed.
However, there is a change in stiffness (and possibly residual strain)
resulting in an associated stress drop \( \Delta {\bm \tau }_{n}={\bm \tau }_{n-1}
({\bm \varepsilon }_{n})-{\bm \tau }_{n}({\bm \varepsilon }_{n}) \),
and a drop in the stored elastic energy density \( \Delta E^{\text {el}}_{n}=E_{n-1}^{\text {el}}
({\bm \varepsilon }_{n})-E_{n}^{\text {el}}({\bm \varepsilon }_{n}) \).
Energy conservation requires the elastic energy reduction to exactly
balance the work performed in opening the crack so that
\begin{equation}
\label{deltaEn}
-\Delta E^{\text {el}}_{n}+\frac{{\mathcal{E}}_{n}+K_{n}}{\ell ^{D}}=0
\end{equation}
where \( {\mathcal{E}}_{n} \) is the bond-breaking work performed at
the grain contact of the \( n^{\textrm{th}} \) crack, \( K_{n} \) is the
energy that went into acoustic emissions when the crack arrived and/or expended
in any mode II frictional sliding or plastic deformation at the grain contact
(\( K_{n} \) is a positive ``loss'' term), and, as earlier, \( \ell ^{D} \)
is the volume of a mesovolume. Because \( K_{n} \) is positive, we can rewrite
Eq.\ (\ref{deltaEn}) as an inequality
\begin{equation}
\frac{K_{n}}{\ell ^{D}}=\Delta E^{\text {el}}_{n}-\frac{{\mathcal{E}}_{n}}{\ell ^{D}}\geq 0
\end{equation}
which is a general statement of Griffith's criterion. Upon appealing to linear
elasticity (elastic stiffnesses independent of strain level) and putting the
residual deformation to zero (no plasticity inside the cracks), we arrive
at the convenient statement \( \ell ^{D}\, {\bm \varepsilon }_{n}\dd
({\textbf {C}}_{n-1}-{\textbf {C}}_{n})\dd {\bm \varepsilon }_{n}/2\geq {\mathcal{E}}_{n} \)
given earlier.
The work performed between the arrival of the \( n^{\rm th} \) and the \( (n+1)^{\rm th} \)
crack is defined
\begin{equation}
\label{eq:intermediatework}
W_{n}=\int ^{{\bm \varepsilon }_{n+1}}_{{\bm \varepsilon }_{n}}{\bm \tau }_{n}
({\bm \varepsilon '})\dd {\bm \varepsilon '}=E_{n}^{\text {el}}({\bm \varepsilon }_{n+1})
-E_{n}^{\text {el}}({\bm \varepsilon }_{n}).
\end{equation}
Thus, the total work required to reach the final strain \( {\bm \varepsilon } \)
is the sum ({\em c.f.}, Figure \ref{crackhistory})
\begin{equation}
E^{p}_{j}=\sum ^{{\mathcal{N}}}_{m=0}W_{n}
\end{equation}
where by convention \( W_{{\mathcal{N}}} \) is the work performed after the
arrival of the last crack to get to the final deformation ${\bm \varepsilon}$.
The superscript
\( p \) on \( E_{j}^{p} \) is simply indicating that this is the work for
one particular realization of the quenched disorder. Rewriting the sum by introducing
Eqs.\ (\ref{eq:intermediatework}) and
(\ref{deltaEn}), then gives
\begin{eqnarray}
E^{p}_{j} & = & E^{\text {el}}_{{\mathcal{N}}}({\bm \varepsilon })
-E^{\text {el}}_{0}({\bm \varepsilon }_{0})
+\sum ^{{\mathcal{N}}}_{n=1}\Delta E^{\text {el}}_{n}\nonumber \\
& = & E^{\text {el}}_{j}({\bm \varepsilon })+\sum ^{{\mathcal{N}}}_{n=1}
\frac{{\mathcal{E}}_{n}+K_{n}}{\ell ^{D}}- E^{\text {el}}_{0}({\bm \varepsilon }_{0})
\label{total2}
\end{eqnarray}
where \( E^{\text {el}}_{0}({\bm \varepsilon }_{0}) \) is the small and
physically-unimportant amount of energy that is stored in the initial isotropic
strain field.
Equation (\ref{total2}) is the natural statement
that the work performed
in creating state \( j \) at strain \( {\bm \varepsilon } \) is the sum of
the elastic energy density stored in the material in the final state plus the
energy irreversibly expended during the opening of each crack.
Both the loss term \( K_{n} \) and the residual energies \( E_{j}^{\rm res} \)
(contained in
\( E_{j}^{\text {el}} \)) are potentially a function of the point in
strain history at which a grain contact actually breaks; \emph{e.g.}, most models
one might propose for plastic deformation at a grain contact
are dependent on the applied stress level. However, modeling such plastic
processes seems uncertain at best.
We thus assume that at least
for those crack states significantly contributing to any phase transition (states
with lots of cracks), the stress-history dependence of \( K_{n} \) is, on
average, negligible. Further, since the residual strain in
brittle-fracture experiments is never more than a few percent of the peak-stress
deformation
and since the essence
of the localization process does not seem to lie in \( E_{j}^{\rm res} \),
we assume that \( E_{j}^{\rm res}\ll \sum _{n}{\mathcal{E}}_{n} \).
With these approximations,
the work density \( E^{p}_{j} \)
depends only on the final state \( j \), the final strain \( {\bm \varepsilon } \)
(through \( E_{j}^{\text {el}} \)), and the breaking energies
\( {\mathcal{E}}_{n} \).
The energy density \( E_{j} \) needed later in our probability
law is obtained by further averaging over the quenched-disorder in
the breaking energies \( {\mathcal{E}}_{n} \) to give
\begin{equation}
\label{Ej;averaged}
E_{j}= E^{\text {el}}_{j}({\bm \varepsilon }) + \gamma _{j}({\bm \varepsilon}_m) \frac{{\mathcal{N}}_{j}}{\ell ^{D}}
- E^{\text {el}}_{0}({\bm \varepsilon }_{0}).
\end{equation}
Here, \( {\mathcal{N}}_{j}={\mathcal{N}} \) is the total number of
cracks in state \( j \) and \( \gamma _{j} \) is the average energy
required to break a single grain contact where the average is over all cells
throughout all mesovolumes led to state \( j \).
This \( \gamma _{j} \) can be different for different final crack states. It
will also be greater at greater values of the maximum strain \( \varepsilon_{m} \)
because, according to Griffith, the cells comprising \( j \) can break at higher
energy levels when the strain is greater.
The first term in Eq. (\ref{Ej;averaged}) corresponds to the purely reversible
elastic energy and therefore depends only on the actual strain state \( {\bm \varepsilon } \).
\subsubsection{Specific expression for $E_j$}
To facilitate the development in Paper II and to be more specific,
we now use Griffith's criterion to develop an
expression for $E_j$ that is based
on linear elasticity.
When the \( n^{\textrm{th}} \)
crack arrives, the linear-elastic variant of the Griffith criterion gives that
\begin{eqnarray}
{\mathcal{E}}_{n} & < & \ell ^{D}\, {\bm \varepsilon }_{n}\dd ({\textbf {C}}_{n-1}-{\textbf {C}}_{n})\dd {\bm \varepsilon }_{n}/2\\
& < & \ell ^{D}\, {\bm \varepsilon}_m \dd ({\textbf {C}}_{n-1}-{\textbf {C}}_{n})
\dd {\bm \varepsilon }_m/2\label{inequality2}
\end{eqnarray}
where as earlier \( {\bm \varepsilon }_{n} \) is the strain point on the load curve
where the \( n^{\textrm{th}} \)
crack arrives while \( {\bm \varepsilon }_m \) is the final maximum strain level of the
experiment.
The second inequality follows from the first since an extra crack always reduces the
stiffness of a mesovolume.
For any particular mesovolume in state \( j \), the average
energy required to break a contact \( \gamma _{j}^{p} \) thus satisfies
\begin{equation}
\label{gamma_j_p}
\gamma _{j}^{p}\equiv \frac{1}{{\mathcal{N}}_{j}}
\sum _{n=1}^{{\mathcal{N}}_{j}}{{\mathcal{E}}_{n}}<\frac{\ell ^{D}}{2{\mathcal{N}}_{j}}
{\bm \varepsilon }_m \dd \left( {\textbf {C}}_{0}-{\textbf {C}}_{j}\right) \dd {\bm \varepsilon}_m
\end{equation}
where the right-hand side comes from summing Eq.\ (\ref{inequality2}). Since
this inequality is independent of the history, every mesovolume that is led
to state \( j \) must satisfy it. We may thus write \( \gamma _{j} \) in the
form
\begin{equation}
\label{gamma_{j}}
\gamma _{j}=f_{j}\, \frac{\ell ^{D}}{2{\mathcal{N}}_{j}}
{\bm \varepsilon }_m\dd \left( {\textbf {C}}_{0}-{\textbf {C}}_{j}\right) \dd {\bm \varepsilon }_m
\end{equation}
where the fraction \( f_{j} \) is bounded as \( 0<f_{j}<1 \).
We next demonstrate that the variation of \( f_{j} \) from one state
to the next is so small as to be neglected altogether.
A tighter lower bound for \( f_{j} \) is obtained by considering crack states \( j \)
having \( {\mathcal{N}}_{j} \) non-interacting cracks. Since the cracks do
not interact to concentrate stress, all of the \( {\mathcal{N}}_{j} \) cells
that broke had their breaking energies somewhere in the range $0 \le {\cal E} \le
\delta E = \ell^D {\bm \varepsilon}_m \dd \delta {\bf C} \dd {\bm \varepsilon}_m/2$ where
$\delta {\bf C}$ is the change in the stiffness tensor due to
the arrival of a single non-interacting crack and $\delta E$ is the associated change in
the elastic energy.
Since the breaking energies are independent random variables taken from
the distribution $\pi({\cal E})$, we obtain
\begin{equation}
\gamma_j =
\frac{\int_0^{\delta E} e \pi(e) \, de} {\int_0^{\delta E} \pi(e) \, de}
\label{gammagen}
\end{equation}
for non-interacting crack states $j$.
We now appeal to a specific form for the probability distribution $\pi({\cal E})$.
Initially, our rocks are intact and it is expected that more grain contacts
are entirely bonded (${\cal E} = \Gamma d^{D-1}$) than entirely unbonded
(${\cal E}=0$). We thus assume a monotonic distribution ${\cal E}^k$
with $k>0$ satisfying the normalization $\int_0^{\Gamma
d^{D-1} } \pi(e) de =1$ so that
\begin{equation}
\pi({\cal E}) = \frac{(k+1)}{\Gamma d^{D-1}} \left(\frac{\cal E}{
\Gamma d^{D-1}}\right)^k = c \, {\cal E}^k.
\label{pi}
\end{equation}
Using this $\pi$, the average energy required to break a contact in a non-interacting
crack state is
\begin{equation}
\gamma_j = \frac{k+1}{k+2} \delta E = \frac{q}{2}
{\ell^D {\bm \varepsilon}_m \dd \delta {\bf C} \dd {\bm \varepsilon}_m}
\label{gammaun}
\end{equation}
where we have defined $q = (k+1)/(k+2)$. All dependence on
the underlying quenched-disorder distribution in our theory is confined to
the constant $q$ which for any $k>0$ is in the range $[0.5,1]$.
Since for non-interacting states
\( {\textbf {C}}_{0}-{\textbf {C}}_{j}
={\mathcal{N}}_{j}
\delta {\textbf {C}} \), a comparison of Eqs.\ (\ref{gammaun}) and
(\ref{gamma_{j}}) shows that
$f_j = q$ for all the non-interacting states.
For the interacting states, the prefactor \( f_{j} \) must be slightly
greater because now stress concentration
can allow stronger
cells to break. It is thus concluded that for all states,
the \( f_{j} \) of Eq.\ (\ref{gamma_{j}}) are bounded
as \( q \le f_{j}<1 \) which when compared to how \( {\mathcal{N}}_{j} \) varies
from state to state can be considered negligible. From here on, we simply
take $f_j=q$ for all states.
The essential physics for the average amount of work that goes into building up
any given crack state \( j \) is thus captured by
\begin{eqnarray}
E_{j}({\bm \varepsilon },{\bm \varepsilon }_m) & = &
E_{j}^{R}({\bm \varepsilon }) +
E_{j}^{I}({\bm \varepsilon }_m)
\label{hamiltonian} \\
E^{R}_{j}({\bm \varepsilon }) & = & \frac{1}{2}{\bm \varepsilon }\dd
{\textbf {C}}_{j} \dd {\bm \varepsilon }
\label{revenergy}\\
E_{j}^{I}({\bm \varepsilon }_m) & = & \frac{q}{2}{\bm \varepsilon }_m\dd
\left( {\textbf {C}}_{0}-{\textbf {C}}_{j}\right) \dd {\bm \varepsilon }_m\label{irrevenergy}
\end{eqnarray}
where the superscripts \( R \) and \( I \) denote respectively the reversible
and irreversible part of the energy.
The intact hydrostatic
energy \( E^{\text {el}}_{0}({\bm \varepsilon }_{0}) \)
has been neglected since it does not involve cracks and,
therefore, cannot influence the probability of the various crack states.
\subsection{The laws of our crack-based thermodynamics}
Using the ergodic hypothesis discussed earlier, the average energy density
in a disorded solid can be written \( U=\sum _{j}p_{j}E_{j} \).
We are interested in how $U$ changes when increments in ${\bm \varepsilon}$
and ${\bm \varepsilon}_m$ are applied to the system.
In general, a small increment in $U$ can be written
\begin{equation}
\label{dU1}
dU= \sum_j E_j dp_j + \sum _{j}p_{j}dE_{j}.
\end{equation}
The first term involving the probability change is entirely due to crack creation.
Some mesovolumes that were in less cracked states prior to the increment, are transformed
to state $j$ during the increment, while mesovolumes that were in state $j$, are
transformed to other more cracked states. If in the increment,
the number of mesovolumes arriving in state $j$
is different than the number leaving, there is a change $dp_j$ in the occupational probability
of that state. Such changes are the only way to change the disorder in the system, so that
\begin{equation}
\label{TdS}
\sum _{j}E_{j}dp_{j}=TdS
\end{equation}
is the work involved in changing the system's disorder via crack production.
The proportionality constant $T$ is formally a temperature and will be treated
in detail.
Using the decomposition $E_j({\bm \varepsilon}, {\bm \varepsilon}_m)=
E_j^R({\bm \varepsilon}) + E_j^I({\bm \varepsilon}_m)$,
we can write
the second term of Eq. (\ref{dU1}) as
\begin{equation}
\label{dU2}
\sum _{j}p_{j}dE_{j}=\sum _{j}p_{j}dE_{j}^{R}+\sum _{j}p_{j}dE^{I}_{j}.
\end{equation}
The first part is due to purely elastic (reversible) changes in each mesovolume and
may be further written
\begin{equation}
\sum _{j}p_{j}dE_{j}^{R}={\bm \tau }\dd d{\bm \varepsilon }
\end{equation}
where ${\bm \tau}$ is the average stress tensor acting on the mesovolumes.
This result can be verified by appealing either to Eq.\ (\ref{revenergy}) or
to the more general statement of Eq.\ (\ref{elastic1}).
The second part $\sum_j p_j dE^I_j$ represents
the average work performed in creating cracks in just the final strain increment
\(d{\bm \varepsilon }_m \).
Some of the initial
mesovolumes
led to state $j$ at maximum strain \({\bm \varepsilon}_m + d{\bm \varepsilon }_m \)
had all their cracks in place before the final strain increment, while others had
cracks arrive in the final increment. We write
\begin{equation}
\sum _{j}p_{j}dE^{I}_{j}={\bm g}\dd d{\bm \varepsilon }_m
\end{equation}
where the tensor ${\bm g}$ has units of stress but is quite distinct from the
stress tensor ${\bm \tau}$.
The ``first law'' for the rock mass is then
\begin{equation}
\label{firstlaw}
dU={\bm \tau }\dd d{\bm \varepsilon }+{\bm g}\dd d{\bm \varepsilon }_m+TdS
\end{equation}
with the formal definitions
\begin{equation}
{\bm \tau} = \left. \frac{\partial U}{\partial {\bm \varepsilon}} \right|_{
{\bm \varepsilon}_m, S} ,
\, \, \, \, \,
{\bm g} = \left. \frac{\partial U}{\partial {\bm \varepsilon}_m} \right|_{
{\bm \varepsilon}, S},
\, \, \mbox{ and } \, \,
{T} = \left. \frac{\partial U}{\partial {S}} \right|_{
{\bm \varepsilon},{\bm \varepsilon}_m}.
\end{equation}
The natural variables of the fundamental function \( U \)
are \( (S,{\bm \varepsilon },{\bm \varepsilon }_m) \). Equivalently if
\( S \) is treated as the fundamental function,
then \( S=S(U,{\bm \varepsilon },{\bm \varepsilon }_m) \) which means
that the constraints placed on the maximization of $S$ must involve
$U$, ${\bm \varepsilon}$, and ${\bm \varepsilon}_m$.
The ``second law'' of this crack-based thermodynamics is that $dS \ge 0$
(equal to zero only if $d{\bm \varepsilon}_m =0$ so that no cracks are created)
while a
``third law'' may be proposed by simply defining \( T=0 \)
when \( S=0 \). The system will have zero emergent disorder before cracks begin
to arrive and so our third law states that the temperature \( T \) starts at
zero and then increases in magnitude as the number of cracks in the system increases from
zero. The justification for this postulate comes \emph{a posteriori} when it
is found that in order to have zero probability for a mesovolume being in anything
but the uncracked state (\( S=0 \)), we must have that \( T=0 \).
\subsection{The probability distribution}
To obtain the probability of observing a mesovolume to be in crack state \( j \),
we maximize \( S =- \sum_j p_j \ln p_j \) subject to the
constraint that \( \sum _{j}p_{j}=1 \),
and to the additional constraints that
${\bm \varepsilon }_{j}={\bm \varepsilon }$, ${\bm \varepsilon }_{mj}={\bm \varepsilon }_m$,
and
$\sum _{j}p_{j}E_{j}=U$.
These constraints define our canonical ensemble. Other ensembles can be defined by considering
other constraints involving $ {\bm \varepsilon }$, ${\bm \varepsilon }_{m}$, and $U$;
however, since all ensembles yield identical average properties in the thermodynamic limit,
we elect to work only with the canonical ensemble due to its analytical
convenience.
This maximization problem is solved
using Lagrange multipliers to obtain the
Boltzmannian
\begin{equation}
\label{probaform}
p_{j}=\frac{e^{-E_{j}/T}}{Z}\mbox {\hskip 2mmwhere\hskip 2mm}Z=\sum _{j}e^{-E_{j}/T},
\end{equation}
and where the parameter \( T \) is exactly the partial derivative $ \partial U/\partial S
|_{
{\bm \varepsilon}, {\bm \varepsilon}_m}$ called ``temperature''.
\subsection{The free energy and its derivatives}
Any equilibrium physical property that depends on the distribution of cracks
throughout the system can be obtained from the partition function \( Z \) given
by Eq.\ (\ref{probaform}).
To do so, a thermodynamic potential \( F \) called
the free-energy density is introduced that is related to \( Z \) by
\begin{equation}
\label{free}
F({\bm \varepsilon}, {\bm \varepsilon}_m,T)=-T\,
{\ln Z({\bm \varepsilon}, {\bm \varepsilon}_m,T)}.
\end{equation}
This potential \( F \) is the Legendre transform with respect to \( S \)
of the total-energy density \( U=U({\bm \varepsilon },{\bm \varepsilon }_m,S) \)
as can be seen from
\begin{equation}
U-TS = \sum _{j}p_{j}E_{j}+T\sum _{j}p_{j}\ln p_{j}
= -T\, {\ln Z}\sum _{j}p_{j}=F\label{Legendre}
\end{equation}
where we used that \( \ln p_{j}=-E_{j}/T-\ln Z \).
When \( ({\bm \varepsilon },{\bm \varepsilon }_m,T) \) are the independent
variables, the first law can be obtained by taking the total
derivative of Eq.\ (\ref{free})
\begin{eqnarray}
dF & = & -T\frac{dZ}{Z}-\ln Z\, dT\nonumber \\
& = & -T\sum _{j}\left[ -\frac{dE_{j}({\bm \varepsilon },{\bm \varepsilon }_m)}{T}+E_{j}\frac{dT}{T^{2}}\right] p_{j}-\ln Z\, dT\nonumber \\
& = & (F-U)\frac{dT}{T}+\sum _{j}p_{j}\left[ dE^{R}_{j}({\bm \varepsilon })
+dE^{I}_{j}({\bm \varepsilon }_m)\right] \nonumber \\
& = & -SdT+{\bm \tau }\dd d{\bm \varepsilon }+{\bm g}\dd d{\bm \varepsilon }_m\label{derF}
\end{eqnarray}
where we have used the definitions that
\( {\bm \tau }_{j} = dE_j^{R}({\bm \varepsilon })/ d{\bm \varepsilon } \)
and
\( {\bm g}_{j}=dE_{j}^{I}({\bm \varepsilon }_m )/d{\bm \varepsilon }_m \).
% as well as
% \( {\bm \tau} =\sum _{j}p_{j}{\bm \tau}_{j} \) and
%${\bm g} = \sum_j p_j {\bm g}_j$.
With $\beta=1/T$, the various thermodynamic functions are related to the
partial derivatives of $\ln Z({\bm \varepsilon }, {\bm \varepsilon }_m, \beta)$ as
\begin{eqnarray}
-\frac{\partial \ln Z}{\partial \beta } & = & \sum _{j}E_{j}p_{j}=U\\
-\frac{1}{\beta }\frac{\partial \ln Z}{\partial {\bm \varepsilon} } & = &
\sum _{j}{\bm \tau} _{j}p_{j}={\bm \tau} \\
-\frac{1}{\beta }\frac{\partial \ln Z}{\partial {\bm \varepsilon}_{m}} & = &
\sum _{j}{\bm g}_{j}p_{j}={\bm g}.
%-\frac{\partial ^{2}\ln Z}{\partial \beta ^{2}} & = &
%\sum _{j}(U^{2}-E_{j}^{2})p_{j} = \frac{1}{\beta} \frac{\partial S}{\partial \beta} =
%\frac{\partial U}{\partial \beta}\\
%-\frac{1}{\beta }\frac{\partial ^{2}\ln Z}{\partial {\bm \varepsilon} \partial
%{\bm \varepsilon}} & = &
%\sum _{j}\frac{d{\bm \tau}_{j}}{d{\bm \varepsilon} }p_{j}
%+\beta \sum _{j} ({\bm \tau}{\bm \tau} - {\bm \tau}_j {\bm \tau_j}) p_{j}
%\\
%-\frac{1}{\beta }\frac{\partial ^{2}\ln Z}{\partial {\bm \varepsilon}_{m} \partial
%{\bm \varepsilon}_m} & = &
%\sum _{j}\frac{dg_{j}}{d{\bm \varepsilon}_{m}}p_{j}+\beta \sum _{j}({\bm g}{\bm g}
%- {\bm g}_{j}{\bm g}_{j}) p_j \\
%-\frac{\partial ^{2}\! \ln Z}{\partial {\bm \varepsilon} \partial \beta } \!\!&\!\! = \!\!&\!\!
%{\bm \tau} +\beta \sum _{j}E_{j}({\bm \tau} -{\bm \tau}_{j})p_{j}
%= {\bm \tau} + \frac{1}{\beta} \frac{\partial S}{\partial {\bm \varepsilon}}
%= \frac{\partial U}{\partial {\bm \varepsilon}}\\
%\!\!\!\!-\frac{\partial ^{2}\!\ln Z}{\partial {\bm \varepsilon}_{m}\partial \beta }
%\!\!\!& \! \!\!= \!\!&\!\!
%{\bm g}+\beta \sum _{j} \!\!E_{j}({\bm g}-{\bm g}_{j})p_{j}
%= {\bm g} \!+ \!\frac{1}{\beta} \frac{\partial S}{\partial {\bm \varepsilon}_m}
%\!= \!\frac{\partial U}{\partial {\bm \varepsilon}_m}\!.
\end{eqnarray}
These results, along with $S=\ln Z + \beta U$, are used in Paper III.
\section{the temperature}
\label{tempstatus}
The temperature is a well-defined essential part of our quenched-disorder
statistics. Through the probability law $p_j= e^{-E_j/T}/Z$,
the temperature quantifies
the energy scale that separates probable from improbable states and how
this energy scale evolves with strain. No other
meaning should be read into $T$. We now demonstrate how to exactly obtain $T$.
\subsection{Evolution of temperature with strain}
The only way energy enters the system is by performing work on
the external surface. Thus, the general relation
$dU = {\bm \tau} \dd d{\bm \varepsilon}$ holds for either loading or
unloading situations. This previously unused fact provides
a differential equation for $T=1/\beta$
that permits everything
about our system to be exactly known once an order-parameter based
model for $E_j({\bm \varepsilon},
{\bm \varepsilon}_m)$ is determined and the functional sums
defining $Z({\bm \varepsilon},
{\bm \varepsilon}_m, \beta) = \sum_j e^{-\beta E_j({\bm \varepsilon},
{\bm \varepsilon}_m)}$ are performed.
The temperature and entropy only evolve along
load paths defined by ${\bm \varepsilon} =
{\bm \varepsilon}_m$ and only such paths need be considered
in what follows.
Using
\( dU={\bm \tau }\dd d{\bm \varepsilon } \), the first law
{[}Eq.\ (\ref{firstlaw}){]} can then be rewritten
\begin{equation}
\label{eq:balance,irreversibles}
TdS+{\bm g}\dd d{\bm \varepsilon }=0.
\end{equation}
Since it always requires energy to break contacts, we have that
\( {\bm g}\dd d{\bm \varepsilon }>0 \)
and consequently \( TdS<0 \). Furthermore, since the entropy (disorder)
necessarily grows during the crack-creation process (at least initially), the
temperature of our
system is negative (at least initially).
The load path of a standard triaxial experiment is when
axial strain $\varepsilon_z$ monotonically increases while the
radial confining stress $\tau_x=\tau_y = -p_c$ remains constant.
Along this path, all properties evolve only as a function of $\varepsilon_z$.
With $Z({\bm \varepsilon},
{\bm \varepsilon}_m, \beta)$ considered as known,
the radial deformation components can be expressed in terms of the axial
deformation by using the two equations
$$
\beta p_c = \left.
\frac{\partial \ln Z}{\partial \varepsilon_x} \right|_{
{\bm \varepsilon}_m = {\bm \varepsilon}}
=\left.
\frac{\partial \ln Z}{\partial \varepsilon_y} \right|_{
{\bm \varepsilon}_m = {\bm \varepsilon}}$$
to obtain the two functions
\begin{equation}
\varepsilon_x = f_x(\beta, \varepsilon_z)
\, \mbox{ and } \, \, \varepsilon_y = f_y(\beta, \varepsilon_z)
\end{equation}
that are valid only along the load path.
We now write $dU$ in two different ways. First,
$dU = {\bm \tau} \dd d{\bm \varepsilon}$ is evaluated along the load path to obtain
\begin{equation}
dU= \tau_z d\varepsilon_z - p_c (df_x + df_y).
\label{dUv1}
\end{equation}
Second, we use the fact that $U=U({\beta, \bm \varepsilon}, {\bm \varepsilon}_m)$
to obtain
\begin{eqnarray}
\!\!\!\!\!\!
&&dU= \frac{\partial U}{\partial \beta} d\beta +
\left(\frac{\partial U}{\partial \varepsilon_z}
+ \frac{\partial U}{\partial \varepsilon_{mz}}\right)
d \varepsilon_z \nonumber \\
&& \mbox{ \hskip3mm }+ \!
\left(\frac{\partial U}{\partial \varepsilon_x}
+ \frac{\partial U}{\partial \varepsilon_{mx}}\right)\!
d f_x + \!
\left(\frac{\partial U}{\partial \varepsilon_y}
+ \frac{\partial U}{\partial \varepsilon_{my}}\right)\!
d f_y.
\label{dUv2}
\end{eqnarray}
Upon equating Eqs.\ (\ref{dUv1}) and (\ref{dUv2}) we obtain
a first-order non-linear differential equation for $\beta$
\begin{equation}
a(\beta, \varepsilon_z) \frac{d\beta}{d \varepsilon_z} + b(\beta, \varepsilon_z) = 0
\label{tempde}
\end{equation}
where $a$ and $b$ are given by
\begin{eqnarray}
a= \frac{\partial U}{\partial \beta} &+& \left(p_c + \frac{\partial U}{\partial \varepsilon_x}
+ \frac{\partial U}{\partial \varepsilon_{mx}} \right) \frac{\partial f_x}{\partial \beta} \nonumber \\
&+& \left(p_c + \frac{\partial U}{\partial \varepsilon_y}
+ \frac{\partial U}{\partial \varepsilon_{my}} \right) \frac{\partial f_y}{\partial \beta}
\label{a} \\
b= -\tau_z &+& \frac{\partial U}{\partial \varepsilon_z} +
\frac{\partial U}{\partial \varepsilon_{mz}} \nonumber \\
&+&\left(p_c + \frac{\partial U}{\partial \varepsilon_x}
+ \frac{\partial U}{\partial \varepsilon_{mx}} \right) \frac{\partial f_x}{\partial \varepsilon_z}
\nonumber \\
&+&\left(p_c + \frac{\partial U}{\partial \varepsilon_y}
+ \frac{\partial U}{\partial \varepsilon_{my}} \right) \frac{\partial f_y}{\partial \varepsilon_z}.
\label{b}
\end{eqnarray}
We are to use $\tau_z = - \beta^{-1} \partial \ln Z/\partial \varepsilon_z$
and $U= - \partial \ln Z/\partial \beta$ in these expressions for $a$ and $b$ once
the function $Z({\bm \varepsilon}, {\bm \varepsilon}_m, \beta)$ has been determined.
Furthermore,
all partial derivatives are to be evaluated along the load curve; {\em i.e.}, at
$\varepsilon_{mx} =
f_x(\beta, \varepsilon_z)$, $\varepsilon_{my} =
f_y(\beta, \varepsilon_z)$, and $\varepsilon_{mz}=\varepsilon_z$.
\subsection{Initial conditions}
In order to integrate Eq.\ (\ref{tempde}),
initial conditions must be provided.
The initial conditions of our so-called ``third law'' ({\em i.e.},
the intact conditions that $\beta=-\infty$
when $\varepsilon_z = 0$) are not well-defined for $\beta$.
Thus, Eq.\ (\ref{tempde}) must be integrated not from the intact state,
but from a state that contains at least a few cracks so that $\beta\neq -\infty$.
Accordingly, we define ``one-crack'' initial conditions by considering
the point in strain history where on average throughout the ensemble of
mesovolumes, there is one crack
in each mesovolume. If there are $N$ cells in a mesovolume,
the probability of any given cell to be broken somewhere in the ensemble is then
$P_1=1/N$.
This same probability can also be obtained from
Griffith's criterion by integrating the
quenched-disorder
distribution of Eq.\ (\ref{pi}) to obtain
$P_1 = [\delta E_1/(\Gamma d^{D-1})]^{k+1}$ where
$\delta E_1=\ell^D {\bm \varepsilon}_1 \dd \delta {\bf C}
\dd {\bm \varepsilon}_1/2$ is the elastic
energy change due to a single isolated crack and where ${\bm \varepsilon}_1$ is the
strain tensor at which on average there is a single crack in each mesovolume.
Thus,
we have ${\bm \varepsilon}_1 \dd \delta {\bf C} \dd {\bm \varepsilon}_1=
2 \Gamma d^{D-1}/(N^{1/(k+1)} \ell^D)$ which can be used to obtain an expression for
the initial axial strain $\varepsilon_{z1}$ at which on average there is
one crack per mesovolume.
To obtain the inverse temperature $\beta_1$ corresponding to this
initial strain, the exact probability of observing
a particular type of crack state is determined and compared
to our temperature-dependent Boltzmannian. The particular states we choose
to analyze are, for simplicity, those having
precisely one broken cell.
The probability $p_j$ of a state consisting of
one broken cell and $N-1 $ unbroken cells can be written
\begin{equation}
p_j = P_1 (1-P_1)^{N-1} \Pi_{\bm x}\left[1- \delta P({\bm x})\right]
\end{equation}
where $P_1$ is again the probability of having a single broken cell and $(1-P_1)^{N-1}$
is the probability of having $N-1$ broken cells in the absence of other
cracks.
Thus, the product
$\Pi_{\bm x}\left[1- \delta P({\bm x})\right]
$ is the probability that no cells broke due to the strain perturbations
caused by the presence of
a first broken cell where ${\bm x}$ represents distance from this first broken cell.
We define $\delta E_2({\bm x})$
as the elastic energy change in a mesovolume when a second cell breaks solely
in the perturbed strain field
emanating from a first broken cell. This energy
varies with the separation distance $|{\bm x}|$ between the two cracks as $|{\bm x}|^{-D}$.
We have
\begin{equation}
\delta P({\bm x}) = \int_{0}^{ \delta E_2({\bm x})} \!\!
\pi(e) \, de = \left(\frac{\delta E_2({\bm x})}{\Gamma d^{D-1}}\right)^{k+1}
= \frac{c_2}{|{\bm x}|^{D(k+1)}}
\end{equation}
where Eq.\ (\ref{pi}) was used for $\pi$ and where
$c_2$ depends on both the overall applied strain and the angle from
the first crack's orientation to the second crack. Since $\delta P$ is small
compared to one
(restricting to models where cracks are smaller than the cell size $\Lambda$, since the separation distance $|{\bm x}|$ always exceeds it), we have
\begin{equation}
\Pi_{\bm x} [1-\delta P({\bm x})] =
1- \frac{1}{\ell^D}\int_{|{\bm x}|>\Lambda} \frac{c_2} {|{\bm x}|^{D(k+1)}} d^{D}{\bm x}
\end{equation}
and since $k>0$, this spatial integral over the mesovolume can be neglected in
the thermodynamic limit.
The conclusion is that
\begin{equation}
p_j = P_1(1-P_1)^{N-1} = p_0 \frac{P_1}{1-P_1} = p_0 e^{-\ln{(N-1)}}
\end{equation}
where $p_0=(1-P_1)^N$ is the probability of the entirely intact state.
This can be compared to our probability law where,
from Eqs.\ (\ref{hamiltonian})--(\ref{irrevenergy}), we have
\begin{equation}
p_j = p_0 \exp{\left[\beta_1 \frac{(1-q)}{2}
{\bm \varepsilon}_1 \dd \delta {\bf C} \dd {\bm \varepsilon}_1\right]}.
\end{equation}
Thus, the
inverse temperature that holds
when ${\bm \varepsilon}= {\bm \varepsilon}_1$ is
\begin{equation}
\beta_1 = - \frac{\ell^D N^{1/(k+1)} \ln (N-1) }{(1-q) \Gamma d^{D-1}}.
\end{equation}
\subsection{Approximate approach to the temperature}
The approach just taken in defining the initial conditions
suggests a convenient way of obtaining an approximate
expression for the temperature.
Consider ``dilute'' states $j$ where cracks do not significantly interact.
In this case, the probability $P_m$ that any one cell has broken when the maximum
strain tensor is at ${\bm \varepsilon}_m$ is again just the cumulative
distribution $P_m = [\ell^D {\bm \varepsilon}_m \dd \delta {\bf C} \dd {\bm \varepsilon}_m/
(2 \Gamma d^{D-1})]^{k+1}$. In this case, the probability of observing a non-interacting
state $j$ consisting of ${\cal N}_j$ cracks is $p_j = P_m^{{\cal N}_j} (1-P_m)^{(N-{\cal N}_j)}$
where we have forgone the analysis of the preceeding section demonstrating
that the unbroken-cell probabilities are negligibly influenced by the strain
perturbations from the ${\cal N}_j$ broken
cells (at least for $k>0$).
We may write
\begin{equation}
p_{j}
= p_{0}\exp \left[ -\ln \left( \frac{1}{P_{m}}-1\right) {\mathcal{N}}_{j}\right]
\label{eq:asympt,distrib,integree}
\end{equation}
where \( p_{0}=(1-P_m)^{N} \) is the probability of the unbroken state.
For such dilute states, the Hamiltonian of Eq.\ (\ref{hamiltonian}) is written
(with \( {\bm \varepsilon_m}={\bm \varepsilon } \))
\begin{equation}
E_{j}=\frac{1}{2} {\bm \varepsilon_m}\dd {\bf C}_0 \dd
{\bm \varepsilon_m} -\frac{(1-q)}{2}{\bm \varepsilon_m}\dd \delta {\textbf {C}}\dd
{\bm \varepsilon_m}\, {\mathcal{N}}_{j}
\end{equation}
so that our probability law predicts
\begin{equation}
\label{eq:asympt,distrib,shannon}
p_{j}=p_{0}\exp \left[ \frac{\beta(1-q)}{2}
{\bm \varepsilon}_m \dd \delta {\textbf {C}}\dd {\bm \varepsilon_m}
{\mathcal{N}}_{j}\right] .
\end{equation}
Upon using
\( 1/P_m=[2\Gamma d^{D-1} /(\ell ^{D}{\bm \varepsilon_m}
\dd \delta {\textbf {C}}\dd {\bm \varepsilon_m}) ]^{k+1}\)
and equating Eqs.\ (\ref{eq:asympt,distrib,shannon})
and (\ref{eq:asympt,distrib,integree}), the temperature is identified
\begin{equation}
\label{temp}
\beta({\bm \varepsilon_m})
=\frac{- 2 \ln \left\{ [2\Gamma d^{D-1} /\left( \ell ^{D}{\bm \varepsilon _m}
\dd \delta {\textbf {C}}\dd {\bm \varepsilon _m}\right)]^{k+1} -1\right\} }
{(1-q) {\bm \varepsilon_m}\dd \delta {\textbf {C}}\dd {\bm \varepsilon _m}}.
\end{equation}
This expression for \( \beta \)
has the expected behavior that \( \beta=-\infty \) when \( {\bm \varepsilon }_m=0 \), and
that \( \beta \)
is a negative and increasing function of \( {\bm \varepsilon}_m \) up to
the strain point $P_m=1/2$ where it smoothly goes to zero. For $P_m > 1/2$,
$\beta$ is a positive and increasing function of \( {\bm \varepsilon}_m \).
Our probability law with $\beta$ negative
predicts the intact state to have the greatest probability,
while when \( P_{m} > 1/2 \) and $\beta$ is positive,
the most probable state jumps to every
cell being broken. Although such a phase transition occurs in fiber bundles \cite{PT02},
we demonstrate in
Paper III using the exact differential equation for temperature,
that the localization transition always occurs prior to this
divergent-temperature
transition.
We emphasize that Eq.\ (\ref{temp}) is an approximation to the extent that
due to the long-range nature
of elastic interactions, one can never truly define a non-interacting state.
We use it to obtain an order-of-magnitude idea of the temperature at a given strain.
But it should always be considered preferable to obtain the temperature by integrating
the exact Eq.\ (\ref{tempde}) from the first-crack (or other exact) initial conditions.
\section{CONCLUSIONS}
The present theory of fracture in disordered solids works from the postulate
that the probability \( p_{j} \) of observing a mesovolume in a given emergent-crack
state \( j \) and at a given applied strain can
be determined by maximizing Shannon's measure of the emergent-crack disorder
subject to constraints that come from the energy balance of brittle fracture.
These constraints are what allow non-uniform probability distributions to occur.
The validity of this postulate can be demonstrated in simpler cases \cite{PT02}
by integrating the probability distribution through history, but its general
validity in the case of rocks with interacting cracks remains an open problem.
Our approach to answering this question is to use the statistical mechanics
that follows from our maximal-disorder postulate to make predictions about the
physical properties of real systems and to compare such predictions to laboratory
data.
\begin{acknowledgments}
The authors thank S.\ Roux and M.\ Holschneider for useful discussions in the early stages
of this work, and D.\ Lockner for sharing both his data and knowledge of the fracture process.
R.T.\ received financial support from the TMR network "Fractal Structures and Selforganization"
through EEC grant FMRXCT980183.
\end{acknowledgments}
\bibliographystyle{apsrev}
\begin{thebibliography}{27}
\expandafter\ifx\csname natexlab\endcsname\relax\def\natexlab#1{#1}\fi \expandafter\ifx\csname bibnamefont\endcsname\relax \def\bibnamefont#1{#1}\fi \expandafter\ifx\csname bibfnamefont\endcsname\relax \def\bibfnamefont#1{#1}\fi \expandafter\ifx\csname citenamefont\endcsname\relax \def\citenamefont#1{#1}\fi \expandafter\ifx\csname url\endcsname\relax \def\url#1{\texttt{#1}}\fi \expandafter\ifx\csname urlprefix\endcsname\relax\def\urlprefix{URL }\fi \providecommand{\bibinfo}[2]{#2} \providecommand{\eprint}[2][]{\url{#2}}
\bibitem[{\citenamefont{Selinger et~al.}(1991)
\citenamefont{Selinger, Wang, Gelbart, and Ben-Shaul}}]{SWG+91}
\bibinfo{author}{\bibfnamefont{R.L.}~\bibnamefont{Blumberg Selinger}},
\bibinfo{author}{\bibfnamefont{Z.G.}~\bibnamefont{Wang}},
\bibinfo{author}{\bibfnamefont{W.M.}~\bibnamefont{Gelbart}}, \bibnamefont{and} \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Ben-Shaul}}, \bibinfo{journal}{Phys. Rev. A} \textbf{\bibinfo{volume}{43}}, \bibinfo{pages}{4396} (\bibinfo{year}{1991}).
\bibitem[{\citenamefont{Rundle and Klein}(1989)}]{RK89} \bibinfo{author}{\bibfnamefont{J.B.}~\bibnamefont{Rundle}} \bibnamefont{and} \bibinfo{author}{\bibfnamefont{W.}~\bibnamefont{Klein}}, \bibinfo{journal}{Phys. Rev. Lett.} \textbf{\bibinfo{volume}{63}}, \bibinfo{pages}{171} (\bibinfo{year}{1989}).
\bibitem[{\citenamefont{Wang et~al.}(1991)
\citenamefont{Wang, Landman, Selinger, and Gelbart}}]{WLS+91}
\bibinfo{author}{\bibfnamefont{Z.G.}~\bibnamefont{Wang}},
\bibinfo{author}{\bibfnamefont{U.}~\bibnamefont{Landman}},
\bibinfo{author}{\bibfnamefont{R.L.}~\bibnamefont{Blumberg Selinger}}, \bibnamefont{and}
\bibinfo{author}{\bibfnamefont{W.M.}~\bibnamefont{Gelbart}},
\bibinfo{journal}{Phys. Rev. B} \textbf{\bibinfo{volume}{44}}, \bibinfo{pages}{378} (\bibinfo{year}{1991}).
\bibitem[{\citenamefont{Golubovic and Peredera}(1995)}]{GP95}
\bibinfo{author}{\bibfnamefont{L.}~\bibnamefont{Golubovi\v{c}}} \bibnamefont{and}
\bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Peredera}}, \bibinfo{journal}{Phys. Rev. E} \textbf{\bibinfo{volume}{51}}, \bibinfo{pages}{2799} (\bibinfo{year}{1995}).
\bibitem[{\citenamefont{Meakin}(1990)}]{Mea90} \bibinfo{author}{\bibfnamefont{P.}~\bibnamefont{Meakin}}, in \emph{\bibinfo{booktitle}{Statistical models for the fracture of disordered media}}, edited by \bibinfo{editor}{\bibfnamefont{H.}~\bibnamefont{Hermann}} \bibnamefont{and} \bibinfo{editor}{\bibfnamefont{S.}~\bibnamefont{Roux}} (\bibinfo{publisher}{Elsevier}, \bibinfo{year}{1990}).
\bibitem[{\citenamefont{Batrouni and Hansen}(1998)}]{BH98} \bibinfo{author}{\bibfnamefont{G.G.}~\bibnamefont{Batrouni}} \bibnamefont{and} \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Hansen}}, \bibinfo{journal}{Phys. Rev. Lett.} \textbf{\bibinfo{volume}{80}}, \bibinfo{pages}{325} (\bibinfo{year}{1998}). \bibitem[{\citenamefont{{de Arcangelis} and Herrmann}(1989)}]{DAH89} \bibinfo{author}{\bibfnamefont{L.}~\bibnamefont{{de Arcangelis}}} \bibnamefont{and} \bibinfo{author}{\bibfnamefont{H.J.}~\bibnamefont{Herrmann}}, \bibinfo{journal}{Phys. Rev. B} \textbf{\bibinfo{volume}{39}}, \bibinfo{pages}{2678} (\bibinfo{year}{1989}).
\bibitem[{\citenamefont{Duxbury et~al.}(1987)\citenamefont{Duxbury, Leath, and Beale}}]{DLB87} \bibinfo{author}{\bibfnamefont{P.M.}~\bibnamefont{Duxbury}}, \bibinfo{author}{\bibfnamefont{P.L.}~\bibnamefont{Leath}}, \bibnamefont{and} \bibinfo{author}{\bibfnamefont{P.D.}~\bibnamefont{Beale}}, \bibinfo{journal}{Phys. Rev. B} \textbf{\bibinfo{volume}{36}}, \bibinfo{pages}{367} (\bibinfo{year}{1987}).
\bibitem[{\citenamefont{Arbabi and Sahimi}(1993)}]{AS93a} \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Arbabi}} \bibnamefont{and} \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{Sahimi}}, \bibinfo{journal}{Phys. Rev. B} \textbf{\bibinfo{volume}{47}}, \bibinfo{pages}{695} (\bibinfo{year}{1993}); \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{Sahimi}} \bibnamefont{and} \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Arbabi}}, \bibinfo{journal}{Phys. Rev. B} \textbf{\bibinfo{volume}{47}}, \bibinfo{pages}{703} (\bibinfo{year}{1993}); \bibinfo{journal}{Phys. Rev. B} \textbf{\bibinfo{volume}{47}}, \bibinfo{pages}{713} (\bibinfo{year}{1993}).
\bibitem[{\citenamefont{Sahimi and Arbabi}(1996)}]{SA96} \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{Sahimi}} \bibnamefont{and} \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Arbabi}}, \bibinfo{journal}{Phys. Rev. Lett.} \textbf{\bibinfo{volume}{77}}, \bibinfo{pages}{3689} (\bibinfo{year}{1996}).
\bibitem[{\citenamefont{Hermann et~al.}(1989)\citenamefont{Hermann, Hansen, and Roux}}]{fracture-o:1989} \bibinfo{author}{\bibfnamefont{H.J.}~\bibnamefont{Herrmann}}, \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Hansen}}, \bibnamefont{and} \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Roux}}, \bibinfo{journal}{Phys. Rev. B} \textbf{\bibinfo{volume}{39}}, \bibinfo{pages}{637} (\bibinfo{year}{1989}).
\bibitem[{\citenamefont{Kloster et~al.}(1997)\citenamefont{Kloster, Hansen, and Hemmer}}]{KHH97} \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{Kloster}}, \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Hansen}}, \bibnamefont{and} \bibinfo{author}{\bibfnamefont{P.C.}~\bibnamefont{Hemmer}}, \bibinfo{journal}{Phys. Rev. E} \textbf{\bibinfo{volume}{56}}, \bibinfo{pages}{2615} (\bibinfo{year}{1997}).
\bibitem[{\citenamefont{Hansen and Hemmer}(1994)}]{burst-aval:1994} \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Hansen}} \bibnamefont{and} \bibinfo{author}{\bibfnamefont{P.C.}~\bibnamefont{Hemmer}}, \bibinfo{journal}{Phys. Lett. A} \textbf{\bibinfo{volume}{184}}, \bibinfo{pages}{394} (\bibinfo{year}{1994}).
\bibitem[{\citenamefont{Hemmer and Hansen}(1992)}]{the-distri:1992} \bibinfo{author}{\bibfnamefont{P.C.}~\bibnamefont{Hemmer}} \bibnamefont{and} \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Hansen}}, \bibinfo{journal}{J. Appl. Mech.} \textbf{\bibinfo{volume}{59}}, \bibinfo{pages}{909} (\bibinfo{year}{1992}).
\bibitem[{\citenamefont{Zhang and Ding}(1996)}]{failur-of-:1996}
\bibinfo{author}{\bibfnamefont{S.D.}~\bibnamefont{Zhang}} \bibnamefont{and}
\bibinfo{author}{\bibfnamefont{E.J.}~\bibnamefont{Ding}},
\bibinfo{journal}{Phys. Rev. B} \textbf{\bibinfo{volume}{53}},
\bibinfo{pages}{646} (\bibinfo{year}{1996}). \bibitem{Sor89a} \bibinfo{author}{\bibfnamefont{D.}~\bibnamefont{Sornette}}, \bibinfo{journal}{J. Phys. A} \textbf{\bibinfo{volume}{22}}, \bibinfo{pages}{L243} (\bibinfo{year}{1989}).
\bibitem[{\citenamefont{Moreno et~al.}(2000)\citenamefont{Moreno, G\'{o}mez, and Pacheco}}]{MGP00}
\bibinfo{author}{\bibfnamefont{Y.}~\bibnamefont{Moreno}},
\bibinfo{author}{\bibfnamefont{J.B}~\bibnamefont{G\'{o}mez}}, \bibnamefont{and}
\bibinfo{author}{\bibfnamefont{A.F.}~\bibnamefont{Pacheco}}, \bibinfo{journal}{Phys. Rev. Lett.} \textbf{\bibinfo{volume}{85}}, \bibinfo{pages}{2865} (\bibinfo{year}{2000}); \bibinfo{journal}{Physica A} \textbf{\bibinfo{volume}{296}}, \bibinfo{pages}{9} (\bibinfo{year}{2001}).
\bibitem[{\citenamefont{{de Arcangelis} et~al.}(1989)\citenamefont{{de Arcangelis}, Hansen, Herrmann, and Roux}}]{DAHR89} \bibinfo{author}{\bibfnamefont{L.}~\bibnamefont{{de Arcangelis}}}, \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Hansen}}, \bibinfo{author}{\bibfnamefont{H.J.}~\bibnamefont{Herrmann}}, \bibnamefont{and} \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Roux}}, \bibinfo{journal}{Phys. Rev. B} \textbf{\bibinfo{volume}{40}}, \bibinfo{pages}{877} (\bibinfo{year}{1989}).
\bibitem[{\citenamefont{Hansen et~al.}(1991)\citenamefont{Hansen, Hinrichsen, and Roux}}]{HHR91b} \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Hansen}}, \bibinfo{author}{\bibfnamefont{E.L.}~\bibnamefont{Hinrichsen}}, \bibnamefont{and} \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Roux}}, \bibinfo{journal}{Phys. Rev. B} \textbf{\bibinfo{volume}{43}}, \bibinfo{pages}{665} (\bibinfo{year}{1991}).
\bibitem[{\citenamefont{Herrmann and Roux}(1990)}]{HR90} \bibinfo{editor}{\bibfnamefont{H.J.}~\bibnamefont{Herrmann}} \bibnamefont{and} \bibinfo{editor}{\bibfnamefont{S.}~\bibnamefont{Roux}}, eds., \emph{\bibinfo{title}{Statistical models for the fracture of disordered media}} (\bibinfo{publisher}{North-Holland}, \bibinfo{address}{Amsterdam}, \bibinfo{year}{1990}). \bibitem[{\citenamefont{Zapperi et~al.}(1997)\citenamefont{Zapperi, Ray, Stanley, and Vespignani}}]{ZRS+97} \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Zapperi}}, \bibinfo{author}{\bibfnamefont{P.}~\bibnamefont{Ray}}, \bibinfo{author}{\bibfnamefont{H.E.}~\bibnamefont{Stanley}}, \bibnamefont{and} \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Vespignani}}, \bibinfo{journal}{Phys. Rev. Lett.} \textbf{\bibinfo{volume}{78}}, \bibinfo{pages}{1408} (\bibinfo{year}{1997}).
\bibitem[{\citenamefont{Zapperi et~al.}(1999)\citenamefont{Zapperi, Ray, Stanley, and Vespignani}}]{ZRS+99} \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Zapperi}}, \bibinfo{author}{\bibfnamefont{P.}~\bibnamefont{Ray}}, \bibinfo{author}{\bibfnamefont{H.E.}~\bibnamefont{Stanley}}, \bibnamefont{and} \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Vespignani}}, \bibinfo{journal}{Phys. Rev. E} \textbf{\bibinfo{volume}{59}}, \bibinfo{pages}{5049} (\bibinfo{year}{1999}).
\bibitem{ASL97}
\bibinfo{author}{\bibfnamefont{J.V.}~\bibnamefont{Andersen}}, \bibinfo{author}{\bibfnamefont{D.}~\bibnamefont{Sornette}} \bibnamefont{and} \bibinfo{author}{\bibfnamefont{K.T.}~\bibnamefont{Leung}}, \bibinfo{journal}{Phys. Rev. Lett.} \textbf{\bibinfo{volume}{78}}, \bibinfo{pages}{2140} (\bibinfo{year}{1997}).
\bibitem[{\citenamefont{Pride and Toussaint}(2002)}]{PT02} \bibinfo{author}{\bibfnamefont{S.R.}~\bibnamefont{Pride}} \bibnamefont{and} \bibinfo{author}{\bibfnamefont{R.}~\bibnamefont{Toussaint}}, \bibinfo{journal}{Physica A} \textbf{\bibinfo{volume}{312}}, \bibinfo{pages}{159} (\bibinfo{year}{2002}) .
\bibitem[{\citenamefont{Rudnicki and Rice}(1975)}]{RR75}
\bibinfo{author}{\bibfnamefont{J.W.}~\bibnamefont{Rudnicki}} \bibnamefont{and}
\bibinfo{author}{\bibfnamefont{J.R.}~\bibnamefont{Rice}}, \bibinfo{journal}{J. Mech. Phys. Sol.}
\textbf{\bibinfo{volume}{23}},
\bibinfo{pages}{371} (\bibinfo{year}{1975}).
\bibitem[{\citenamefont{Griffith}(1920)}]{Gri20} \bibinfo{author}{\bibfnamefont{A.A.}~\bibnamefont{Griffith}}, \bibinfo{journal}{Philos. Trans. Roy. Soc. London} \textbf{\bibinfo{volume}{A221}}, \bibinfo{pages}{163} (\bibinfo{year}{1920}).
\bibitem[{\citenamefont{Paterson}(1978)}]{Pat78} \bibinfo{author}{\bibfnamefont{M.S.}~\bibnamefont{Paterson}} , \emph{\bibinfo{title}{Experimental rock deformation - the brittle field}} (\bibinfo{publisher}{Springer-Verlag}, \bibinfo{address}{Berlin}, \bibinfo{year}{1978}).
\bibitem[{\citenamefont{Shannon}(1948)}]{Sha48} \bibinfo{author}{\bibfnamefont{C.}~\bibnamefont{Shannon}}, \bibinfo{journal}{Bell Systems Technical Journal} \textbf{\bibinfo{volume}{27}}, \bibinfo{pages}{373} (\bibinfo{year}{1948}).
\end{thebibliography}
\end{document}