A Dislocation Model for Seismic Electric Signals 

L. M. SLIFKIN Department of Physics and Astronomy,
University of North Carolina, Chapel Hill, NC 275993255, USA 

It
is suggested that a possible mechanism for production of seismic electric
signals involves the motion of charged dislocation loops upon rapid change
in a shear component of the stress acting on blocks of rock in the
earthquake preparation zone. A calculation, which uses likely values of
the relevant parameters indicates that such a process could produce
significant voltages at the surface of the Earth. 

Introduction Detection of telluric voltages at the Earth's surface has been reported by several groups, such as the VAN team in Greece. In order to relate these signals to impending earthquakes, it would be helpful to understand the mechanisms of both their production and also their transmission in the highly anisotropic and heterogeneous media found in the Earth's crust. A feasible mechanism for transmission of the signals over long distances has been proposed by Lazarus, and is further discussed in a companion paper in these proceedings; the present article is concerned only with the process of the generation of the voltage. It
has been experimentally demonstrated some years ago, by Hoenig, that
severe loading of laboratory specimens of rocks can result in the
appearance of very large voltages at the surface. Piezoelectric effects
and stress induced orientation of impurityvacancy electric dipoles do not
appear capable of giving rise to large enough signals in the usual
polycrystalline materials found in nature. Hoenig proposed that his
observations were the result of a sequence of microfractures, and somewhat
later it was shown by Gershenzon et al that the production and motion of
electrically charged dislocations at the tip of the fracture crack would
produce voltage differences. They were able to estimate the magnitudes of
electromagnetic emission to be expected from this process during crustal
fracturing, and concluded that this phenomenon could produce significantly
large signals. A
number of possible mechanisms, for production of seismic electric signals
have recently been examined by Bernard. He finds that, given the right
conditions, one could perhaps, understand these effects in terms of
electrokinetic phenomena: he suggests that a change of stress may trigger
fluid instabilities that would give rise to large streaming potentials.
One would still require some "channeling" of the electric signal
in order to explain the detection at distances as great as involved in the
VAN experiments. The model proposed by Lazarus would seem to be a possible
explanation of such preferential channeling. It
has been suggested by the present author that, even in the absence of
large amounts of plastic flow or of fracture, an abrupt change in the
shear stress —either increase or decrease— should cause a voltage
difference to appear between opposite surfaces of a specimen of most
minerals. In such materials as silicates, oxides, and the like, lattice
vacancies and aliovalent impurity ions carry effective charges. Moreover,
singlestepped jogs on edge dislocations (essentially, the edges of
partial lattice planes which do not extend completely through the crystal)
are also electrically charged. Because these jogs must establish
simultaneous equilibria with all of the different types of point defects
within the crystal, it follows that, in general, the edge dislocations in
ionic crystals carry a net electric charge. Now,
although the dislocation may not be able to move through large distances
in response to the applied shear stress, segments of it can indeed bow out
between the points at which the dislocation is pinned, as has been
discussed by Granato and Lücke. The pins may be impurity ions, clusters,
or points of intersection with other dislocations. The dislocation loops
between the pinning points respond to applied shear stresses much as if
they were nonHookean elastic bands. Upon a change in stress, the
relaxation rate of the loops is primarily a function of the Peierls stress— the stress required to cause dislocation displacement in an otherwise
perfect lattice. At high temperatures, the dislocation loops can be quite
mobile. The
purpose of this paper is to make an estimate of the magnitude of the
electric field produced by the motion of these dislocation loops when the
stress is rapidly increased or decreased, causing the loops to expand or
contract, respectively. The numerical results are quite sensitive to the
values chosen for the relevant parameters, some of which are uncertain. At
best, then, this discussion is a feasibility test of the question: given a
reasonable set of values for these parameters, is it possible for the
proposed mechanism to account for signals comparable to those that have
been reported? 

Charged
Dislocations The
jogs must therefore establish multiple equilibria —i.e., with each of the
species of point defect. This results in the presence of more jogs of one
sign than the other, so that the dislocations carry a net charge. The
linear charge density on the dislocations is compensated by an
approximately cylindrical space charge around the dislocation, consisting
of an excess of point defects of the opposite sign. The electric potential
difference that is thereby established between the dislocation and the
distant interior just compensates for the differing formation free
energies of the different species of intrinsic point defects. In
materials with significant concentrations of impurities, as in the case of
geophysically interesting minerals, the space charge consists largely of
aliovalent dopant ions. These ions will, in general, be very much less
mobile than dislocation segments bowed out between pinning points. Hence,
after any abrupt change in stress, the bowed loops will quickly respond,
leaving the space charge distribution unrelaxed; the center of the space
charge no longer coincides with the line of the dislocation, and an
electric dipole has been produced. This is the basis of the mechanism
proposed here for the generation of seismic electric signals. 

The
Model The
stressinduced dipole moment lies in the slip plane and is oriented
perpendicular to the dislocation line. Its magnitude per unit length of
dislocation is given by the product of the dislocation charge density and
the change in mean displacement of the bowed segments, caused by the
change in applied stress. For a number of alkali and silver halides,
experimental values have been deduced for the charge density, as a
function of purity, temperature, and dislocation dynamics; these studies
have been reviewed by Whitworth. A typical value is of the order of 0.1
e/b, where e is the electronic charge and b the lattice spacing. If we
assume comparable numbers for the rocks of interest to the present
problem, and take b to be about 5 x 10^{10} m, then the net
charge density is of the order of 3 x 10^{11} C/m. The
second factor determining the dipole moment is the mean displacement of
the dislocation loops bowed out between pairs of pins. If the applied
stress is a significant fraction of the flow or fracture stress, then the
loops will bow out through a distance large compared to b, and we can
approximate the mean displacement by onefourth the spacing between pins
(for example, if the loop were nearly semicircular, then its maximum
displacement would be half the interpin distance). We now need some
reasonable estimate of this interpin spacing, which must depend on the
crystal structure, the presence and state of aggregation of impurities,
the dislocation distribution, and —overall— the thermal and mechanical
histories of the specimen. Analyses of various mechanical properties of a
range of solids have suggestedthat the interpin spacing is often of the
order 10 100 b. If we choose a value in the middle of this range, then
our mean dislocation loop displacement under the maximum supportable
stress is about (50/4) b »
6x10^{9} m. If we now consider a sudden change in stress of magnitude comparable to the stress at failure, the dipole moment produced by the motion of the charged loops, per unit length of dislocation, is equal to the product of these two factors: (3
x 10^{11} C/m) x (6 x 10^{9} m) »
2 x 10^{19}
Cm/m. The resulting electric polarization, the dipole moment per unit
volume, is obtained by multiplying this quantity by the appropriate
density of dislocations (i.e., in the simplest case, the total dislocation
length per unit volume, or the number intersecting a surface of unit
area). The
dislocation density, of course, depends on the mechanical history of the
specimen. While it is possible to grow crystals of such semiconductors as
very pure silicon with virtually no dislocations, most annealed simple
crystals have densities in the range 10^{8} 10^{10}/m^{2},
and heavily deformed material contains 10^{3}  10^{4}
times this. These values, however, are not appropriate to our problem,
because some dislocations have their extra halfplanes coming from above,
while others are from below (one says that they have opposite mechanical
signs). Under an applied stress, the interpin segments of these two types
bow out in opposite directions; hence, they make opposite contributions to
the polarization. Only the net excess of one type over the other produces
a net signal. If the distribution were quite random, this excess would be
approximately equal to the square root of the total number. The
distribution of dislocations in most rock strata is not, however, at all
random, because of the folding they have undergone in earlier history.
Plastic bending of crystalline material is accomplished through
introduction of a set of edge dislocations, all of the same mechanical
sign, to accommodate the differing lengths of the upper and lower
surfaces. Consider, for example, the specimen sketched in Fig. 1, a cube
of edge length 1 m. Suppose that it has been bent through an angle of 1°
(» 0.02 rad). Then the upper edge must be longer than the lower one
by 0.02 m. If we again take b = 5 x 10^{10} m, then the bending
must have resulted from the introduction of 0.02 / (5 x 10^{10})
= 4 x 10^{7} new edge dislocations, all of the same mechanical
sign, as shown in the diagram. In this case, the excess density of
dislocations is then 4 x 10^{7}/m^{2}. This is a quite
modest value, and probably greatly underestimates the dislocation
densities in naturally occurring rock. If,
nevertheless, we use this result for our illustrative example, and recall
our earlier estimate of the stressinduced dipole moment per unit length
of dislocation (i.e., 2 x 10^{19} Cm/m), we then obtain a volume
polarization of (4 x 10^{7}/m^{2}) x (2 x 10^{19}
Cm/m) = 8 x 10^{12} C/m^{2} this is just the charge
density that appears on the surfaces at the ends of the slab (ignoring
such factors as the cosine of the angle between the normal of the end
surfaces and the direction of displacement of the dislocation loops). We
are now ready to calculate the electric field appearing at the Earth's
surface. The numerical result, depends sensitively on the choice of such
parameters as the size and depth of the block. It will also be greatly
perturbed by heterogeneities and anisotropy in the dielectric properties
of the surrounding medium. In particular, the decay of electric field with
the cube of the distance from the dipole is only applicable for a
homogeneous, isotropic medium; this is considered at greater length in the
companion paper in these proceedings by D. Lazarus. With these limitations
in mind, the following estimate can only be taken as an illustrative
example, at best. For
concreteness, consider a horizontal block 1000 m on each side, and 100 m
thick. Suppose that it has been folded about a horizontal axis
perpendicular to one of the end faces, as shown in Fig. 2. If we take the
angle through which the block has been bent to be 1°, as in the
calculation relating the Fig. 1, so that the density of excess
dislocations is 4 x 10^{7}/m^{2}, and if we use the
previously estimated stressinduced dipole moment per unit length of
dislocation (i.e., 2 x 10^{19} Cm/m), then the resulting dipole
moment is (2 x 10^{19} Cm/m) x (4 x 10^{7}/m^{2})
x (10^{5} m^{2}) x (10^{3} m); the last two
factors are the crosssectional area and the length of the dipole,
respectively. The stress change has thus produced an electric dipole
moment of magnitude 8 x 10^{4} Cm. To
estimate the effect of this dipole moment on observations made at the
Earth's surface, we must know the depth of the block and the dielectric
structure of the intervening material. Let us first assume that the medium
is homogeneous, isotropic, and has a dielectric coefficient of unity.
Then, for a dipole of length L, the electric field at a distance d
perpendicular to its axis is given by: E
= (1/4p e_{0}) p (d^{2} + L2/4)^{3/2} where
p is the dipole moment, estimated above to be 8 x 10^{4} Cm, and
the depth d is taken, for sake of illustration, to be 10 km = 10^{4}
m. Ignoring the term L2/4 as compared to d^{2}, and inserting
e_{0} = 9 x 10^{12} C^{2}/Nm^{2}, we
obtain an electric field at the earth's surface of 7 x 10^{6}
V/m, or 7 mV/km. This is in the range of signals obtained for large
earthquakes by the VAN group, but, of course, their electrodes are not
located directly above the active regions. For large distances from the
dipole (d > L), E falls off as l/d^{3}, and one can understand
the experimental observations only in terms of some sort of channeling by
dielectric inhomogeneities. This problem is not unique to the present
mechanism; virtually any reasonable model would have the same requirement
An attractive solution to this puzzle is proposed by D. Lazarus elsewhere
in these proceedings. Another
question relates to the rather long durations of the signals recorded by
the VAN group, as compared to the quite short RC time constant for
relaxation of electric fields in typical wet minerals. It is possible that
the recorded signals consist of unresolved superpositions of many rapid
pulses, generated by the propagation of mechanical relaxations in a
sequence of neighboring blocks. It thus appears that the charged dislocation loop mechanism is a likely candidate but certainly not the only possibility, for explaining stressinduced voltages in rocks. For example, if significant microplasticity occurs, then the longrange motion of the charged dislocations would greatly enhance the electric signal. In any case, ifs waveguiding mechanism for longrange transmission, such as that suggested by Lazarus, proves to be an acceptable means of precluding the l/d^{3 }decay, then the voltages estimated here are indeed consistent with the observations that have been made in Greece. It is clear, however, that the predicted signal is quite sensitive to the chemistry, morphology, and mechanical history of the relevant strata. It thus cannot be expected to be a general phenomenon, independent of the local geology.


Fig. 1. Introduction of edge dislocations by plastic bending. If the angle q = 0.02 radians, the upper edge of the block is longer than the lower by 0.02 m. The dislocations arc indicated by upsidedown "T"s, with the base representing the slip plane and the stem the end of the extra halfplane. (Not to scale)  
Fig. 2. Production of an electric dipole in a previously plastically bent specimen, upon application of a shear stress parallel to the slip plane of the dislocations. 


