Kyoto University
Kyoto University
Kyoto University
Title Interactions
URL http://hdl.handle.net/2433/124882
Right
Textversion publisher
Kyoto University
Bull. Disas. Prey. Res. Inst., Kyoto Univ., Vol. 29, Part 1, No. 259, July, 1979 1
By Akio SAKAI
(ManuscriptreceivedMarch31, 1979)
Abstract
1. Introduction
In the so-called crack model of seismic sources (e.g., Kostrov (1966)), the start and
stop of fracture motions are controlled by proper boundary conditions on a fracture
region, which is usually composed of one fracture plane. In order to obtain more
realistic features, the model should have a more complicated structure. For instance,
we may consider an increased number of fracture planes. These lines of complications,
however, would need the enormous memory of computers and hence there should be
alternative ways of complication, retaining high precision in numerical computations.
For this reason, models have been simplified under the assumptions that some compli-
cated structures are concentrated on one fracture plane. (e.g., Otsuka (1971), Mikumo
and Miyatake (1978))
We discuss here an extension of conventional source models (which we call "one
shed model") To overcome some difficulties and restrictions imposed by computer
capacity, we introduce a new freedom of motion into the problem. This parameter
could be an adequate representation of the nature of fault gouge material for which much
attention has been paid in laboratory experiments. (e.g., Byerlee et al. (1978))
A model of fault gouge is formulated under assumptions that the internal energy
of the system of an aggregate of deformable particles is composed of two variables, that
is, strains on a particle and rigid rotations independent of averaged movements. An
2 A. SAKAI
intuitive exposition of a model and the necessity of dissipative forces are discussed.
Experimental results are cited in support of these formulations. From an observational
point of view, there remains some problems to be solved, that is, pre-seismic or post-
seismic time dependent deformations and coseismic quasi-elastic deformations. The
solution of the above formulated equations gives necessary information for a possible
interpretation of these phenomena.
As will be subsequently described in a separate paper, some extension of a model
are attempted in a thermodynamical approach (Part II) and in a differential geometrical
approach (Part III). The idea of independent-rigid rotations first took rise in Cosserat
(1909) whose research had its recaptulation in Mindlin and Tiersten (1962), and
independently in Oshima (1953). Our present formulation is more elementary and as
explanatory as possible to leave out the ambiguity existing in former researches.
2. Mechanism of friction
Recently much work has been done on the problem of friction in relation to rock
mechanics such as Brace and Byerlee (1966), Brace (1972), Scholtz (1968), Brune
(1973), Johnson et al. (1975) and others. In frictional sliding experiments, there are
several controlling parameters and conditions, e.g., surface roughness, thickness of gouge
layers, confining pressure, temperature etc. Brace (1972) has shown that a transition
occurs from stick-slip to stable-sliding motions under appropriate P-T conditions.
Dieterich (1972) has also shown the time-dependent recovery of fracture strength or
static frictions after stick-slip events under some controlled normal pressures. How-
ever no complete dynamical experiments wherein the various parameters are controlled
have been performed to date, so the above results may reflect only one aspect of the facts.
There has been much discussion about the stress drops estimated from the analysis
of seismic waves ranging some ten to one hundred bars, which are one to two orders of
magnitude lower than the results from laboratory experiments. When we compare the
experimental results with the field data, we must pay attention not only to scaling effects
(e.g., Maeda (1977)) but to the effects of compound material including several weak
zones. Here we introduce gouge as a possible way of representing these effects. One
sheet model with properly controlled surface conditions may be one primitive model
along this line.
Recent development in experimental techniques has revealed the nature of gouge
in sliding on a microscopic scale. (e.g., Byerlee et al. (1978)) Although the formation
process of gouge is not yet clear at present, we have the following concept in constructing
our model. After a virgin fracturing in the earth and several repetitions of sliding on
or near the same surface, the slipped segments will have been crushed to pieces and lose
their elasticity as a whole. An influx of water from the surrounding region may pro-
mote further degradation. The region of slippage is called gouge. Although the
whole gouge has lost elastic properties, the constituents of the gouge, which we call
particles, still have elasticity as in the surrounding region. Even if gouge is substitut-
A Model of Fault Gougewith Dissipative Rotational Interactions 3
ed by different minerals due to the influx of water, it is usual that each of the particles
are still elastic. For this reason, it is now necessary not only to elucidate the nature of
friction mechanism of each contacting particle on the gouge material but to model the
aggregation of particles.
particles and the continuum medium, the dis- Fig. 1. Two particles which constitute
sipation due to rotations without slippage maythe part of gouge contacting
each other without mutual
be negligible, while with internal contacting local rotations.
4 A. SAKAI
CI ) 1, CIO
points, the dissipation due to slippage is not. (We call rotations without slippage as
the motion exerted under static type of friction and rotations with slippage as the
motion exerted under dynamic type of friction.) Deformation without dissipation
cannot be realized in such a multi-particle system as in Fig. 4. This is contrary to the
case of a model of perfect elasticity or a perfect fluid without internal structures.
Since we cannot usually expect a favorable configuration of particles such as shown in
Fig. 3, we assume here the model of gouge as in Fig. 4.
When the slippage initiates after the stress exceeds the critical strength, it is to be
noted that even when the nature of contacting points that are crushed primarily deter-
mines the type of dissipation mechanism, the averaged nature of dissipation would be
somewhat different from an elementary process. For instance, in crystallography, vis-
cous medium is inserted within grain boundaries as a model of the mechanism of gliding
or attenuation of acoustic vibrations, or in other words, as a model of grain boundaries.
(Smoluchowski (1952)) Since grain boundaries should be expressed as an aggregation
of dislocation lines (e.g., Cottrell (1953)), this way of modelling has a phenomenological
30
0
°C)
0
D8.OC
c
DO•11100
Fig. 3. A multi-size particle system of which Fig. 4. A schematic exposition of a model of
motion is generated by mutual fault gouge: particles rotate in the same
rotation without slippage. direction after being triggered by fault
movement.
A Modelof Fault GougewithDissipativeRotationalInteractions 5
significance in that the mechanical behavior of the model properly approximates those
of the matter.
In this paper we adopt Smoluchowski's model of grain boundaries as a model of
dissipation for the system of particles, based on the reasoning stated above, although
our particles do not correspond to his crystallographic grains.
4. Formulation of models
We next derive the equations of motion for the system with dissipative forces. As
stated in the preceding section, we supposed that gouge is composed of particles with a
finite size. This finiteness can be put in another way; inhomogeneities of distribution
of relative rotations as modelled below. Each of the particles can deform elastically.
Motions of a particle can be decomposed into a relative rotating motion around the
center of gravity and a deformation of the system with respect to the center of gravity.
This deformation can also be decomposed into strains and rotations within linear transfor-
mations.
In the following calculations we apply Einstein's summation convention. And
the process of mixing over p indices by constructing all p isomers obtained by permuta-
tions of these indices, addition of these isomers and division of /5I is denoted by a round
bracket ( ):
/ „ , A,
e.g..Aar=
A similar process with the only difference that all isomers obtained by odd permutation
get a negative sign is denoted by a square bracket [ ]. If indices have to be singled out,
the sign I I is used :
/
e.g., Briiiik]= (DiJk—Bhp)
Here we assume that the type of interaction between particles is that described
below: Strains on an average of particles are not primary and interacting forces be-
tween the particles only act through mutual rotations. This assumption will be debated
later with a supplemented extension of a model.
First we assume that dissipative forces act on or near the surface of a particle.
Hence the equations of motion for the system of the center of gravity are expressed below,
1
2-f plo;
kidV=f ac[kGiiii]dS;±
f MaidSt (4-3)
(4-4)
where MI ki is the surface distribution of torque which has an arm in the k-direction and
the force exerted in the i-direction on a surface element with a normal in the 1-direction.
As stated previously, Mipi is determined by mutual rotations of neighboring particles.
I is the distribution of a moment of inertia per unit mass, which is properly transformed
into that in the principal axis, and here assumed to be isotropic for convenience's sake.
After some calculations,
X[kGliji3d.51=
f (Glki]d-pX[A61])d
V
If we presume that
a; J=a0 (4-5)
and
G[a]=F[ii] (4-6)
then
1r•
—
2'maw —pX[kVii (4-7)
The above expressions (4-7) show that the force is decomposed into the purely conserv-
ative part and the dissipative part. Usually the assumption (4-5) is identical to the
law of the conservation of angular momentum. Since we introduced the interacting
term represented by Mai, our new form of conservation of angular momentum is ful-
filled together with the assumptions (4-5) and (4-6). Especially the assumption (4-6)
comes from the fact that dissipative forces should be tangent to the surfaces of particles.
These aspects are easily confirmed by formalistic calculations. With the voluminal
distribution of torque hu and that of body force in newly introduced, the equations of
motion for inertia moment are expressed as,
If x[kpvi]e/
=f x[kG[if
/id&
+f MikidS:
+f xikg
nelV+f hodV (4-8)
and the translational part of motion is expressed as,
f xckp;
;idV=f xikal
iGil
Ad V+f x[kg
AdV
=f f Gitod
V+f xckg
ijdV (4-9)
From expressions (4-8) and (4-9), we get
A Model of Fault Gouge with Dissipative Rotational Interactions 7
AftkidSt+
f hfrid17+
f Guind
V=0 (4-10)
Hence, if Mu/=0 and hki=0, then GEkii=0 and vice versa.
The expressions (4-7) and (4-10) are identities derived from kinematical relations.
Usually since we may presume
17•
2preokt=px[hvii,
expression (4-10) is identical to (4-7). If the discrete nature of mass distributions is
essential, such as of atoms in a crystal, plthki may have proper meaning. However,
since we assume the smoothness of mass distributions, we get the next expression.
Fcii3+,3,Atii=0 (4-7')
We must assume dynamical relations between kinematics and deformation quan-
tities — constitutive equations. The law of termodynamics — energy conservation and
entropy production — must be fulfilled. We express by iii, the eigen rotations inde-
pendent of the rotational strains referred at the particle frame. Internal energy te for
each of the particles or per unit volume is a function of strains s;j, relative rotations
Cif*approximated below and entropy S, that is,
5) (4-11)
where
hilt 843(065]-0O (4-12)
The khi;are defined on the absolute frame, so that aLian—oil have only approximate
meanings. Formalistic extension of expressions (4-11) and (4-12) are, of course,
conceivable. For instance, as a substitute for Cho,
Zif=aiitif]-0i;
may be variables of the internal energy. But since we now concentrate our discussions
on the model with inhomogeneous distributions of rotations, we do not adopt such models
These points will be discussed in Part III.
The law of energy conservation
dE=8W-1-8Q (4-13)
where
dE: the increment of the sum of internal energy U and kinetic energy K.
W: the work done by the exterior system.
8Q: the inflow of heat (non-mechanical external action)
The expression (4-13) is rewritten as evolutional forms.
dE dU
at dt
8 A. SellCAI
dE SW, SQ
dt dt
K=Kor--
J 1pitihaiV,
SW dK
dt f(audvd-FifhpdV (4-13')
die=a0,;;),
and
SW
dt
=f tiferifdSi+
f itFudS,
Here we furthermore add the effect of rotational type of interaction, that is,
SW=
f [(ak,d-FH)auf-Ftlfki,SCif]dSk
=f (akiSaou,)+FA,856ki+Mai,a(SkCi,)-FpU,SundV
(4-14)
using expressions (4-1), (4-2) and (4-7'). We notice the discrepancies between (4-13')
and (4-14), which originates from newly introduced terms.
The increment of internal energy is
auau
dU=clei•+au dCki•H-as dS (4-15)
where
au
as =T (4-16)
dS812 (4-17)
ds—a, (4-18)
where Qi is heat flux from a unit surface element with a normal in the 1-direction. If
we substitute (4-13), (4-14), (4-15), and (4-16) into (4-18), noting
A Model of Fault Gougewith Dissipative Rotational Interactions 9
(4-19)
then
2uau
(an—wf-)deng±(Mki)—didtku
±Fsgelkbii+Qeae 7,1 0 (4-20)
This inequality must be fulfilled for all variation of EH,Cub Oi; and (2i. Hence
ail= au (4-21)
au;
Mkii= au
u6hinr(4-22)
j
Fj;=2rl(j;)E.+k (4-23)
-Fir4ai(anik)-Endui=-2naiii2"±ld-pu- (4-27)
27/56i
.,2"±I=C4(54ii—a[jUi]) (4-28)
The duality between the image of an aggregation of particles with a finite size and
that of properly smoothed but inhomogeneous distributions of relative rotations will
not be rectified, but be used accordingly.
The difference in the time derivative with respect to the moving frame (Lagrangean
description) and to the absolute frame (Eulerian description) will be negligible when
the spatial variations of the physical quantities are small, or when the initial field small.
10 A. SAKAI
LI
VPp (55)
Here we define A=1=12. c If AC tends to zero as L approaches zero. In
this case the system of equations cannot be reduced to that for perfect elasticity as long
as friction terms have not been left out. >1=0 should be substituted for expressions
(4-27)and (4-28), if we want to eliminate the effects of friction alone. Then the equations
of motion for perfect and isotropic elasticity are derived. In addition, it is necessary
that d[pco-¢;1 are harmonics. Since this part if the equations is time-independent,
c6ig=eciuz)should be kept. if held at an initial time.
There are two possibilities; (i) C tends to zero or (ii) L tends to infinity, if A tends
to infinity. In case (ii), as the size of particles increases, we describe a one-particle
approximation, that is, on the earth we may let a maximum value of it be a plate size.
Of course, this is not directly applicable to the plate motions, because of the assumption
(5-3). Under the assumption Ppazni<10;;), by similar reasoning, we get
A_ 2p-Tu_a 2 L2 u
1? L IS C LO
While in the case of Idpufil= kind, we have
A=Cas
p2qTu L I ck-nu
Hence we can determine the law of similarity of the system of equations (4-27) and (4-28)
by parameters;
A
(Model
L )2 ofFault Gouge
withDissipative
Rotational
Interactions (5-6)11
p. \T-11
277/2 (5-7)
C T
and
L2 (=A) (5-8)
(1pizir/V)'=(1
FfidS1)1 (5-10')
The left hand side of (5-10') is the time derivative of the linear momentum and the right
hand side is that of the purely dissipative forces acting on the surface of particles.
Then acted forces are purely dissipative. If A<1, there are two possibilities; (i) C tends
to infinity or (ii) L tends to zero. In the case (ii) for which the particle size gets smaller
and smaller, particle motions reduce to that of microscopic scales, i.e., random motions
by thermal agitations. If L tends to zero under the condition of C, p, p, and n fixed,
similarity of motions will not be maintained due to expressions (5-6) and (5-7) so far as
T is fixed. As previously stated, to determine similarity relations, parameter A is
directly relevant. Here is given some expositions. If L and C simultaneously tend to
zero; for expression (5-7),
27)
C L2T1TI
=0(aqzoo.)
owing to the conditions of diffusivity and for expression (5-6), the same inequality as
(5-5) is maintained owing to the conditions of stability of hyperbolic type of equations
and for expression (5-8),
12 A. SAKAI
A 0 1 00
C (°'j)Fault 0
elasticity
oo
Microscopic
motionGouge
G one-particle
(platetectonics)
L2=' OW,
32v aav
ekz=pv,
2n9C2=c- ax +Os) (6-4)
A Model of Fault Gouge with Di SSi',Wive Rotational Interactions 13
ILars
ax ax 24.-6ax.1ax
32/3 +0Y) (6-5)
The first part of (6-3) shows plane P-wave propagation. This P-wave is charac-
terized by non-dispersive and non-attenuative nature. If we give an initial condition
for the second part
qc,(x,0)=8(x),
then we have
0,(x,0= 1r exp It
x2
t
) (6-6)
lair=
This equation has the same form as that for heat conduction and diffusion equation.
If we assume
9Sx=c6x°-exp(i(kx—11)},
----Oro•exp74-
.x)expft(1/14--x
—ft)) (6-7)
P-h I
rt -
v
i 1.0E7 .•
_, at ) •'''
/0E4
Q01
1
Q0/ 01
.400 1
Fig. 5 (1). Relation between f and imaginary parts of wave number divided
to
o C
14 A. SAKAI
It is easily shown that equations (6-4) and (6-5) have the same type of solutions. If we
assume for (6-4) that
v=vo•exp {i(kx—ft)} and cd,=0.°•exp (i(kx—ft)) (6-8)
then we have the next characteristic equation.
pf2—µk2 —nfk =0 (6-9)
—iCk3 2ifn—C42
Dispersion and attenuation relations derived from this equation are shown in Fig. 5.
The abscissa of the figures gives the angular frequency (f), and the ordinate gives the
magnitude of the real and imaginary parts of the wave number (4) divided by f with
parameters C and a varied. The density p and rigidity p are assumed to be unity for
simplicity. It is seen from this figure that two branches of solution exist, one or both
of which have the properties of a velocity higher than that of S waves. One of the two
bracnhes can be properly approximated by the (6-7) in a wide range of combinations
of parameters C and ,j.
There appears a maximum in each of the Im(k)11 vs. f curves within the frequency
range 0.001 to 10.0 rad/sec, as 12increases. In the case when C is relatively larger and
n is smaller, the value off at which each of the Im(k)If curves of the non-purely dissi-
lc)
to
- 5
1.0E4
KC:1
am-
tOE7
0001 001 01
Fig. 5(2). Relation between J and /m(k)/f-second branch of solutions, under
q=25 and C varied. For higher values of C (104 and 107), Re(k)/f
is nearly equal to Im(4)11
}
A Modelof Fault Gougewith DissipativeRotationalInteractions 15
pative branch of the solutions takes a maximum is roughly proportional to lin. (e.g.,
c=107, 7i=102/4, j= 0.069; C=105, n=10/4, f=0.7 ; C=106, -9=1/4, f=6.9; C=104,
n=1/4 f =7.0 ; C=103, 77=1/4, 1=6.9) If the condition C27? is satisfied, one of the two
branches of solutions shows a close agreement between Re(k) and Im(k). Comparative-
ly speaking under such combinations of C and n as C<7, there appear discrepancies
between Re(k) and Im(k). For both circumstances, the velocity necessarily becomes
lower than that of S waves at lower frequencies.
As C becomeslarger with?) left fixed, the velocityof the dissipative branch generally
becomes higher, and the motion gradually gains non-attenuative properties. The
higher velocityis mainly due to the dissipative forces and results in rigidly coupled motion
of the medium. In field observations, we might not detect this type of motion directly
through the determination of local distribution of higher propagation velocities, but
moderately averaged motions through several instrumental distortions and filtrations.
It may be appropriate for models with a sudden start of fracturing in the case of strong
interaction between particles. Conversely, as C becomes smaller with n left fixed, lower
velocities and highly attenuative branches become predominant, that is, in the case of
weak interaction between particles.
Next we consider some special configuration of equations (9-27) and (4-28), that is,
1.0E-4
1
1D
10
rt
_0.1 - 1.0E .4
0.1 1 10
Fig. 5(3). Relation betweenf and real parts of wave number divided by j
(Re(k)/f)-first branch of solutions,under n=25 and C varied.
16 A. SAKAI
u=u(x, y, t), v=v(x, y, t), zef=0 and 0, y, t), 1Si 53. These equations are
reduced to
24.= cdox (6-10)
2inks.—Cdoku
274:=C(tlikz—a,Av±a„Au) (6-11)
(Ä+143.,(8.u+a,o+taite-94,---pie (6-12)
0+14ar(aru+3,0+144v+-04.--pi; (6-13)
Forthe conditionforinitialvalues,wegetfromequation(6-10)and theequationforw,
10
1.0E-4
1>/-
0.1
1.0
10
001-
a 1 10
Fig.5(4). Relation
between
j andRe(k)/I-second
branchofsolutions,
undern=25andC varied.
A Modelof Fault GougewithDissipativeRotationalInteractions 17
Cdox4y—aoz>=0 (6-14)
If
ax0y=ask. (6-14')
is fulfilled at infinity or at initial time, then azor=a06, is identically realized at any time.
We tentatively call this time-independent identity for 0; the compatibility condition of
itn. Then and 0„ field is expressed as a gradient field like 4's=axx and 021-aa.
From the expression (6-10), x field is purely dissipative. Here if we introduce new
functions W and 0 for the sake of convenience,
as u=aro-kayw
(6-15)
v=avo—axlm
then
274„—C(44W
+400 (6-16)
(A+4)40 p0 (6-17)
W-2/2Sz
=Pql (6-18)
If Ctendsto infinitywithn keptfinite,expressions
(6-16)and (6-18)aretransformed
into
PAW-HA qfr-PP (6-19)
Then the fields0 and Ware decoupledintoexpression(6-17)and expression(6-19).
The fieldOs,is constrainedby the fieldV' in the sameformas conventional
continuum
field,i.e.,
44.= axv—apt,
Theequationsystemof (6-17)and (6-19)are a specialtypeof Voigtsolid(Voigt(1892),
Sezawa(1927),etc) In theschemeof Table1,weshouldadd theabovetypeof medium.
As in the caseof planewaves,P wavesare non-dispersive and non-attenuative.
It is easilyseenthat monochromatic
2-dimensionalwaves
P=W0•exp {i(kx+ly—ft)}
95.=the •exp {i(kx+ly—ft))(6-20)
satisfythe characteristic
equation
442+19—pf2 (6-21)
C(k2+/2)2 c(k2+12)-21-v
Thisrelationis similarto that of (6-9)if k2is substitutedfork2+12.
Secondly in the caseof a=0, v=0, w--w(x,y, t) and#i---44(x, y, t), 1Si53, then
we have,
18 A. SAKAI
Cdch =0 (6-27)
and from the remaining equations we obtain
C(-3,42v-i-Asbo=0 (6-28)
cia,ziw-Hioy>=0
The relations (6-27) and (6-28) can be summarized using vector notation as
Cd(rot U-0)=0 (6-29)
As discussed in §. 5, 0—rot U are harmonics and
0=rot U (6-30)
holds if initially holds. Then in this case the system of equations can be reduced to
those for perfect and isotropic elasticity.
If
0 (6-31)
in theequation(6-22)shouldhavea form
sit=s1r(t) (6-32)
and simultaneously
fromthe equations(6-22)we have
(42=0 (6-33)
The remaining equations are
Civ=p4w+17(-3,4+4,)
2.74x=C(-8,4.,4—drk.) (6-34)
27A,=C(8rdze
—24,)
It is easily seen that monochromatic 2-dimensional waves
•
A Modelof Fault Gougewith DissiOalive
RotationalInteractions 19
F32==2 18103(0i/is
—03)-822141-822(asus+01)) (6-40)
(6-41)
ciss—Maiut-i-asus)+2ta3113
M 331=-2Vn3Wn3141-80,3-02)
Cf (6-42)
Fat=—C—312(aatit-a,u3-02)+82202 —493203/11
—alma—001 (6-43)
The above decomposition has essentially the same nature as the previous one derived
by a rather heuristic manner. Using this 2-dimensional type decomposition, we can
determine not only the wave number vs. frequency relation, but the absolute amplitudes
of displacement and eigen rotation due to the applied external forces. However, the
latter approach is not performed in this paper partly because there have not been much
data for the estimate of boundary conditions both in the field and in the laboratory
except for seismic moment calculated from far field seismograms.
The main reason is that, if a step-wise propagation after the drop of moment is
included in calculations, the totality of the nature of the fracturing process would be
2
20 A. SAKAI
obscured by the combination of several parameters and would not be obtained from a
rather simplified analytic consideration. Our simplified model has two characteristics :
relaxation of initial moment distribution can be realized, and delay or deceleration
effects which are inherent from the existence of strength are not taken into consideration
hereafter.
It is easily inferred from the preceding discussions that type (i) corresponds to an
extention of SH-type deformation, and that type (ii) corresponds to that of P and S V-
type deformation in conventional continua. The motion of type (ii) can be illustrated
as Fig. 4, which is composed of aggregates of cylinders.
Here we notice the coventional definition of the dislocation field. The right
hand side of expression (6-39), as an instance, can be transformed as,
f as(3,u2-1,3)dx3A
dx=izeti
—f stsdxt (6-44)
The first term of the right hand side of expression (6-44) corresponds to a non-integra-
bility effect (screw-type dislocation) and the second term corresponds to the Volterra
type of dislocation. From the relation
and expression (6-44), it is shown that the well-known equivalency between the exertion
of double-couples and the occurrence of an infinitestimal dislocation loop in elasticity
theory (e.g. Maruyama (1963)) is valid in an extended form.
Generally it cannot be expected that the total rotational field disappears in the
medium composed of aggregations of coarse particles. The interaction we have adopted
in this paper is decomposed into such translational and rotational terms as in expression
(6-44). Especially, according to the independence of displacements and eigen rotations,
we have formulated dissipative rotational interactions among neighbouring particles
with the expression (4-7') utilized as an intermediary relation.
What differs from other models with porous media (e.g., Biot (1941)) is the assump-
tion that there is no macroscopic flow in a viscous liquid and that the liquid is incom-
pressible. Following this assumption, the propagation of P waves is characteristic
of nondispersion as well as non-attenuation. The macroscopic flow of liquid becomes
important in lower frequencies namely, beyond which the assumption of local relaxation
breaks down. (O'Connell and Budianski (1977)) Of cource there remains an upper
bound for the frequency, at which the wavelength becomes the size of a pore. As stated
previously, it is assumed to be of no consequence that there might exist micro-fluctu-
ations of density distribution, e.g., mixture of bulk water and particles. Hence the
effect of the configuration or shape of pore and matrix is replaced only with the distri-
bution of one parameter C. This assumption is a simplification for complicated rock
structure. Although the influence of the existence of a thin liquid layer between the
particles can be partly evaluated by the effect of dissipation in our formulation, there are
no effects on the P wave propagation from neglecting the effects of compressibility and
draining of water.
A Model of Fault Gougewith Dissipative Rotational Interactions 21
7. Discussions
The synthesis of seismograms in the regions located near or far from the seismic
source has been successfully performed by one sheet model of dislocation. (e.g., Aki
(1968)) Furthermore, the synthetic seismograms in the near-field have been constructed
not only for direct body waves (e.g., Kanamori (1973)), but for surface waves, surface-
associated waves such as SP phase and reflected waves at the Moho-discontinuity.
(Kawasaki et al. (1972)) If we make use of inversion techniques, the distribution of
dislocation could be determined within a statistical allowance. (e.g., Matsu'ura (1977))
Some questions arise here as to what the distribution of dislocations tectonically
means and how the movement of faults is developed, if these are controlled by some
physical parameters. Some calculations concerning these questions have been
performed under approximations by 2-dimensional or 3-dimensional motions of a discrete
system composed of spring-mass. (Burridge and Knopoff (1967), Yamashita (1977))
Yamashita (1977) successfully showed the effects of inhomogeneous distribution of the
initial stress on the fracture patterns. For the initial stress with spatially linear and
quadratic functions, the fracture is more localized in the latter distribution than the
former, in the case of uniform distribution of the fracture strength. Such localization
of the fracture nucleation for a quadratic distribution will be discussed in Part H,
from a somewhat different standpoint.
Tiltmeters and strainmeters have been installed in regions usually far from faults
except in the case of the San Andreas Fault, the Yamasaki Fault in Japan etc., where a
dense array of these instruments is operated along these faults. (Savage et al. (1976),
Oike (1977)) For instance, residual tilts associated with teleseismic earthquakes have
been reported by Nishimura (1953) and others, although the elastic strain energy change
required to produce the observed tilts is impossibly high. (e.g., Press (1965)) The
instruments must be examined for mechanical instability accompanied by ground shak-
ing due to seismic waves such as large-amplitude Rayleigh waves. Even if these instru-
mental problems remained to be solved, it is not regarded to be unreasonable to assume
one sheet model for an earthquake source, especially for shallow earthquakes, as a whole.
Probably, it is no exaggeration to say that all facts are hidden in the darkness except
these coseismic dynamics which are considered as rather convincing among the observed
earthquake phenomena.
It is also an important problem whether laboratory experiments could properly
simulate or reproduce earthquake phenomena or not. Some laboratory experiments of
fracture—A. E., Kaiser effects etc.—as well as frictional sliding, have been performed
with recent advancements in electronics, i.e., extension of frequency band, which will
elucidate the problem of dilatancy and some related phenomena. Concerning frictional
sliding experiments, it is pointed out that stick-slip motions accompany inhomogeneous
distribution of strains and local rotations whithin the gouge and that except in these
narrow regions the material does not deform too much. (e.g., Byerlee et al. (1978))
These experiments are considered to give support to our model.
It must be pointed out that no reliable interpretations have been offered particularly
22 A. SAKAI
with respect to long-period and transient parts of coseismic and associated movements,
that is,
( i ) aseimic creep motions along faults
( ii) tilt and strain steps
Although concerning (ii), it seems that some consensus, as stated previously, exists under
one sheet model, there remains many problems to be solved. It may be said that residual
tilts and strains could be investigated more thoroughly by using faster recording speed
of signals. Since residual tilts independent of elastic strain change have an important
meaning in our model, dynamic behaviors of the model are directly reflected in obser-
vation of dynamic behaviors of immediate neighboring regions of faults. It is to be
reckoned in our model that the condition under which the fault movement occurs is
identicalto the conditionthat 44,,*0.
It might be assumed here that there exists a threshold of torque similar to the
strength in the case of a discrete system of a mass-spring. The value of matter constant
C determines the roughness of an aggregation of particles. If we neglect the deceler-
ation effect due to the threshold of torque at each point, the fracture pattern will be
mainly determined by the distribution of matter constants C and 71. The above simplified
assumption will make the analysis of an initial value problem easier. The assumption
will be satisfied in case (i). Then the nucleation or initiation of fracture is characterized
by the existence of matter constant C. The derived equations are of a non-causal type
due to rigid rotation and dissipation. By means of numerical procedures, we get two
branches of solutions of the characteristic equation, i.e., almost purely dissipative
motions and coupled motions with a velocity higher than S waves. If the branch of
(6-6)or (6-7) is applied to case (i) as
/c—y(7-1)
It71
Le--f—C
Cif(7-2)
where v: the expanding rate of fracture region
L: the magnitude of fractured region,
it follows that f For estimates of v=1 to 10 km/day for propagating creep
velocity obtained for the San Andreas Fault (Nason (1969)), we get Ch-107 cm2/sec.
It is to be noted that similar results can be obtained, except for extreme cases of C<'j,
by means of the quasi-decoupled branch of solutions as well as the dissipative branch.
From observations of the decay of strain and tilt step amplitudes, one sheet model
is a good approximation of the far field data. (e.g., Takemoto (1972)) It may be
considered that the propagation velocity of the main part of tilt step movements along
faults is of the order of S wave velocities. The nature of propagation for (ii) cannot
be regarded as purely dissipative as in the case of (i). Another branch of solution pro-
vides the properties of propagation, as
R >1 (7-3)
e(k)
A Modelof Fault Gougewith DissipativeRotationalInteractions 23
(7-4)
(when.Knm
where the velocity of S waves is assumed to be unity for the convenience of numerical
calculations. It can be understood that the attenuation effectshows a strict dependence
on frequency as in the type of Voigt model within the above frequency range considered.
In the frequency range f>1/11, fm(k)-- const. A Maxwell type of liquid possesses such
an attenuation property. Our model possesses two extreme states in respect to attenu-
ation: At higher frequencies it behaves like a liquid, on the other hand, at lower
frequencies it behaves like a solid. The weak dispersion of coupled type waves may he
non-detectable and negligible at lower frequencies.
8. Conclusions
Acknowledgements
The autor expresses his sincere thanks to Professor Takeshi Mikumo, Disaster
Prevention Research Institute, Kyoto University, who encouraged him to complete
the work, made valuable suggestions and critically read the manuscript. The author
is also grateful to Dr. Takuo Maruyama for helpful suggestions concerning many
aspects of this work.
References
Aki, K., Seismic displacements near a fault. J. Geophys. Res., 73, 5359-5376, 1968
Biot, M. A., General theory of three-dimensional consolidation, J. Appl. Phys., 12, 155-164, 1941
Brace, W. F., Laboratory studies of stick-slip and their applications to earthquakes, Tectonophysics,
14, 189-200, 1972
Brune, J. N.. Earthquake modeling by stick-slip along precut surfaces in stressed form rubber, Bull.
Seism. Soc. Am., 63, 2105-2119, 1973
Burridge, R. and L. Knopoff, Model and theoretical seismology, Bull. Seism. Soc. Am., 57, 341-371,
1967
Byerlee, J., V. Mjachkin, R. Summers and 0. Voevoda, Structures developed in fault gouge during
stable sliding and stick-slip, Tectonophysics, 44, 161-171, 1978
Cosserat, F. et F., Theorie des Corps Deformables, A. Hermann & Fils, Paris, 1909
Cottrell, A. H., Dislocation and Plastic Flow in Crystals, Oxford, New York, 1953
Courant, R., K. 0. Friedrichs and H. Lewy, Uber die partiellen Differenzengleichungen der mathe-
matischen Physik, Math. Ann. 100, 32, 1928
Dieterich, J. H., Time-dependent friction as a possible mechanism for aftershocks, J. Geophys. Res.,
77, 3771-3781, 1972
Dieterich, J. H., Time-dependent friction in rocks, J. Geophys. Res., 77, 3690-3697. 1972
Johnson, T., F. Wu and C. H. Scholtz, Source parameters for stick-slip and for earthquakes, Science,
179, 278-279, 1973
Kanamori, H., Mode of strain release associated with major earthquakes in Japan, AIM. Rev. Earth
Planet. Sci., 1, 213-239, 1973
Kawasaki, I., Y. Suzuki and R. Sato, Seismic waves due to double couple source in a semi-infinite
space Part I, Zisin (J. Seismol. Soc. Japan), 25, 207-217
Part II, 25, 333-342, 1972 (in Japanese)
Love, E. H., A Treatise on the Mathematical Theory of Elasticity, Cambridge, 1927
Maeda, I., On the preshock, preslip, and the time dependent effect of stick-slip motion, Zisin (J. Seismol.
Soc. Japan), 30, 55-72, 1977 (in Japanese)
Maruyama, T., On the force equivalents of dynamic dislocations with reference to the earthquake
mechanism, Bull. Earthq. Res. Inst., 41, 467-486, 1963
Matsu'ura, M., Inversion of geodetic data Part II, optimal model of conjugate fault system for the 1927
Tango Earthquake, J. Phys. Earth, 25, 233-255, 1977
Mikumo, T. and T. Miyatake, Dynamical rupture process on a three-dimensional fault with non-uni-
form friction and near-field seismic waves, Geophys. J. R. astr. Soc., 54, 417-438, 1978
Mindlin, R. D. and H. F. Tiersten, Effects of couple-stresses in linear elasticity, Arch. Rational Mech.
Anal., 11, 415-448, 1962
Nishimura, E., On some destructive earthquakes observed with a tiltmeter at a great distance, Bull.
Disast. Prey. Res. Inst., Kyoto Univ., 6, 1-16, 1953
Oike, K., Seismic activities and crustal movements at the Yamasaki Fault and surrounding regions in
the southwest Japan, J. Phys. Earth, 25 (supplement), 31-41, 1977
Oshima, N., Asymmetrical stress tensor and its application to a granular medium, Proc. 3rd Japan
national congress for applied mechanics, 77-83, 1953
A Model of Fault Gouge with Dissipative Rotational Interactions 25
Otsuka, M., A simulation of earthquake occurrence, Part 1, A mechanical model, Zisin (J. Seismol.
Soc. Japan), 24, 13-25, 1971 (in Japanese)
O'Connell, R. J. and B. Budiansky, Viscoelastic properties of fluid-saturated cracked solids, J. Geophys.
Res., 82, 5719-5735, 1977
Press, F., Displacements, strains, and tilts at teleseismic distances, J. Geophys. Res., 70, 2395-2412,1965
Sakai, A., A model of fault gouge (Part II, Part III) (in preparation)
Savage, J. C., M. A. Spieth and W. H. Prescott, Preseismic and coseismic deformation associated with
the Hollister, California, Earthquake of November 28, 1974, J. Geophys. Res., 81, 3567-3574, 1976
Scholz, C. H., Microfracturing and the inelastic deformation of rock in compression, J. Geophys. Res.,
73, 1417-1432, 1968
Sezawa, K., On the decay of waves in visco.elastic solid bodies, Bull. Earthq. Res. Inst., 3, 43-54, 1927
Smoluchowski, R., Movement and diffusion phenomena in grain boundaries, Imperfections in Nearly
Perfect Crystals, John Wiley & Sons Inc., 1952
Takemoto, S., Strain steps associated with the earthquakes of March 12, 1966, Off East Coast of Taiwan,
Zisin (J. Seimil. Soc. Japan), 24, 354-362, 1972 (in Japanese)
Voigt, W., titer innere Reibung fester Korper, Ann. Physik, 47, 671-693, 1892
Yamashita, T., On the dynamical process of fault motion in the presence of friction and inhomogeneous
initial stress, Part I. rupture propagation, J. Phys. Earth, 24, 417-444, 1976, Part II. relation
between tectonic field and source parameters, 25, 283-301, 1977