You are on page 1of 35

PUBLICATIONS

Journal of Geophysical Research: Solid Earth


RESEARCH ARTICLE Induced electromagnetic field by seismic waves
10.1002/2014JB010962
in Earth’s magnetic field
Key Points: Yongxin Gao1,2, Xiaofei Chen1,2, Hengshan Hu3, Jian Wen1,2, Ji Tang4, and Guoqing Fang5
• EM wavefields arising from to the
motional induction effect are simulated 1
Laboratory of Seismology and Physics of Earth’s Interior, School of Earth and Space Sciences, University of Science and
• The motional induction effect is
effective to generate observable Technology of China, Hefei, China, 2National Geophysical Observatory at Mengcheng, Anhui, China, 3Department of
EM signals Astronautics and Mechanics, Harbin Institute of Technology, Harbin, China, 4Institute of Geology, China Earthquake
• The motional induction effect Administration, Beijing, China, 5Earthquake Administration of Shanghai Municipality, Shanghai, China
is compared with the
electrokinetic effect
Abstract Studied in this article are the properties of the electromagnetic (EM) fields generated by an
Supporting Information: earthquake due to the motional induction effect, which arises from the motion of the conducting crust
• Readme across the Earth’s magnetic field. By solving the governing equations that couple the elastodynamic
• Data Set for Figure 3
equations with Maxwell equations, we derive the seismoelectromagnetic wavefields excited by a single-point
• Data Set for Figure 5
• Data Set for Figure 6 force and a double-couple source in a full space. Two types of EM disturbances can be generated, i.e., the
• Data Set for Figure 7 coseismic EM field accompanying the seismic wave and the independently propagating EM wave which
• Data Set for Figure 8
arrives much earlier than the seismic wave. Simulation of an Mw6.1 earthquake shows that at a receiving
• Data Set for Figure 9
• Data Set for Figure 10 location where the seismic acceleration is on the order of 0.1 m/s2, the coseismic electric and magnetic fields
• Data Set for Figure 11 are on the orders of 1 μV/m and 0.1 nT, respectively, agreeing with the EM data observed in 2008 Mw6.1
• Data Set for Figure B1
Qingchuan earthquake, China, and indicating that the motional induction effect is effective enough to
generate observable EM signal. We also simulated the EM signals observed by Haines et al. (2007) which were
Correspondence to:
X. Chen, called the Lorentz fields and cannot be explained by the electrokinetic effect. The result shows that the
xfchen1@ustc.edu.cn EM wave generated by a horizontal force can explain the data well, suggesting that the motional induction
effect is responsible for the Lorentz fields. The motional induction effect is compared with the electrokinetic
Citation: effect, showing the overall conclusion that the former dominates the mechanoelectric conversion under
Gao, Y., X. Chen, H. Hu, J. Wen, J. Tang, low-frequency and high-conductivity conditions while the latter dominates under high-frequency and
and G. Fang (2014), Induced
low-conductivity conditions.
electromagnetic field by seismic waves
in Earth’s magnetic field, J. Geophys.
Res. Solid Earth, 119, doi:10.1002/
2014JB010962.
1. Introduction
Received 17 JAN 2014 When a conductor moves across a magnetic field, an electromotive force or electric potential will be
Accepted 17 JUN 2014
Accepted article online 25 JUN 2014 generated across the conductor due to the well-known Faraday’s law of induction. In fact such a
phenomenon also occurs during an earthquake event because of the existence of the Earth’s magnetic field
and the conductive property of the Earth’s crust. In this situation ground motions in the ambient
geomagnetic field can yield an electromotive force as well as a motional induction electric current, which
then gives rise to electromagnetic (EM) fields. This kind of mechanoelectric coupling is referred to as the
motional induction (MI) effect [Gershenzon et al., 1993; Yamazaki, 2012] or the seismic dynamo effect
[Matsushima et al., 2002] and has been proposed to be a possible mechanism for the anomalous EM
disturbances observed during earthquake events [Gershenzon et al., 1993; Gershenzon and Bambakidis, 2001;
Matsushima et al., 2002; Ujihara et al., 2004; Honkura et al., 2004; Yamazaki, 2012].
Anomalous EM fields that appeared in earthquake events have drawn a lot of attentions because they were
thought to be related to the conceiving and occurring of the earthquake, and were expected to be useful for
earthquake prediction or early warning. Studies have been conducted to connect the EM disturbances with
the earthquake from the viewpoint of the motional induction effect and to try to explain the observed data.
Gershenzon et al. [1993] developed a dipole model for the earthquake-driven EM source and estimated the
amplitude of the earthquake-induced EM field arising from the induction (dynamo) effect as well as the
electrokinetic effect and the piezomagnetic effect. Iyemori et al. [1996] proposed that the conducting slab
(which corresponds to the area of many faults) moving with a velocity v in the Earth’s magnetic field B0 can
generate an electric field v × B0, which drives an electrical current across the slab and causes magnetic
disturbances that can be measured at a distant observation site. They suggested that such a crustal dynamo
mechanism is a possible reason for the magnetic variations observed during the 1995 Hyogoken-Nanbu

GAO ET AL. ©2014. American Geophysical Union. All Rights Reserved. 1


Journal of Geophysical Research: Solid Earth 10.1002/2014JB010962

M 7.2 earthquake. Matsushima et al. [2002] recorded the coseismic EM variations that accompanied the
seismic waves in 1999 Mw7.4 İzmit earthquake and its aftershocks. They proposed that those EM variations
might be caused by the motion of the conductive ground in the geomagnetic field. Similar coseismic EM
fields were also observed in other places [Ujihara et al., 2004; Honkura et al., 2004; Abdul Azeez et al., 2009;
Tang et al., 2010]. Yamazaki [2012] studied the coseismic magnetic field arising from the motional induction
effect from the plane-wave point of view. His numerical result showed that the coseismic magnetic field can
reach the order of 0.1 nT, which is detectable. Recently, reports on the observed circularly polarized coseismic
electric field [Honkura et al., 2009; Kuriki et al., 2011; Matsushima et al., 2013] provided evidences for the role of
the Earth’s magnetic field in generating EM disturbances. Moreover, observations of the tsunami-induced
magnetic variations [Manoj et al., 2011; Toh et al., 2011; Utada et al., 2011; Ichihara et al., 2013] were further
evidences which indicate that the conductive seawater moving across the Earth’s magnetic field can also
generate observable variations in the magnetic field.
Besides the coseismic EM signals, EM waves in connection with the motional induction effect were also
measured. Haines et al. [2007] conducted a series of field experiments for the purpose of understanding
seismoelectric phenomenon of electrokinetic nature. By striking a sledgehammer on the hammer plate
which can be taken as a vertical point force, they detected the expected seismoelectric direct field (actually
EM waves arising from the electrokinetic effect) which can be explained by a vertical electric dipole.
Fortuitously, they detected another kind of EM waves which cannot be explained by the electrokinetic effect.
Haines et al. [2007] called this kind of EM waves the Lorentz fields and proposed that it was generated by the
motion of the metal hammer plate with velocity v in the Earth’s magnetic field B0 because such a kind of
Lorentz field cannot be measured if a plastic hammer plate is used. They explained the Lorentz fields by a
horizontal electric dipole which is orientated along the direction v × B0 and the radiation pattern of this
horizontal dipole agreed with the data well.
These previous studies have supported the motional induction effect to be an effective mechanism to
generate earthquake-induced EM fields. However, understanding of the coupling between the EM field and
the elastic wave arising from such a mechanism is not sufficient. Yamazaki [2012] studied the magnetic
variations accompanying the seismic wave, but the formulations he presented were based on the plane-
wave theory. Note that EM signals were often observed near the earthquake sources. For example, the
measurement sites where Matsushima et al. [2002] recorded the coseismic EM signals were very close to the
epicenter, from several kilometers to several tens of kilometers. The plane-wave assumption turns to be
inappropriate when the source-receiver distance becomes very small and in this situation the source effect
needs to be taken into account. Therefore, in this study we investigate the property of the EM disturbances
resulting from the motional induction effect in an alternative way, that is, to solve the closed-form solutions
of the electric and magnetic fields in a uniform full space. By modeling the earthquake as a double-couple
source, we solve the Green’s functions of the electric and magnetic fields arising from the motional induction
effect. The process is similar to solving the seismoelectromagnetic fields due to the electrokinetic effect
[Gao and Hu, 2010].
Note that there are also some other candidate mechanisms that can convert the mechanical energy into
the EM variations, e.g., the piezoelectric effect [Ogawa and Utada, 2000a, 2000b; Huang, 2002], the
piezomagnetic effect [Stacey and Johnston, 1972; Yamazaki, 2011, 2013], and the electrokinetic effect
[Pride, 1994; Haartsen and Pride, 1997; Gao and Hu, 2010; Ren et al., 2012; Gao et al., 2013a, 2013b]. It is possible
that part or all of these mechanisms contribute to the mechanoelectric conversion simultaneously, but
contribution from each mechanism differs. Therefore, comparison between the EM fields arising from
different mechanisms is indispensable, and it can help distinguishing one mechanism from the other and
also benefit the explanation as well as the understanding of the anomalous EM changes observed in real
earthquake events. In this article we shall compare the motional induction effect with the electrokinetic
effect since the latter is a very popular mechanism associating the EM fields with the seismic waves and has
been studied a lot.
The present article is organized as follows. In the next section we first present the governing equations that
couple the elastodynamic equations with Maxwell equations and solve the Green’s functions of the electric
and magnetic fields generated by a single-point force. Then we derive the expressions of the electric and
magnetic fields generated by a moment tensor source which is frequently used to model an earthquake

GAO ET AL. ©2014. American Geophysical Union. All Rights Reserved. 2


Journal of Geophysical Research: Solid Earth 10.1002/2014JB010962

caused by a fault slip. In section 3 we make numerical simulations and apply them to explaining the coseismic
EM data observed during actual earthquake and the Lorentz fields observed in the field experiments made
by Haines et al. [2007]. In section 4 comparison is made between the motional induction effect and the
electrokinetic effect.

2. Theoretical Formulations
2.1. Green’s Functions due to a Single-Point Force
In this study we describe the motional induction effect by coupling the elastodynamic equations governing
the propagation of the elastic waves with Maxwell equations controlling the propagation of the EM waves.
Assuming a time dependence of e iωt, the coupling equations can be expressed as
ðH  GÞ∇∇  u þ G∇2 u þ ω2 ρu ¼ F; (1)
J ¼ σ0 E þ σ0 vB0 ; (2)
∇E ¼ iωμH; (3)
∇H ¼ iωεE þ J þ Jc ; (4)
where u is the displacement vector; E is the electric field; H is the magnetic field; F is the body force density;
B0 is the ambient geomagnetic field; ω is the angular frequency; G and H = λ + 2G are the elastic moduli of the
medium with λ representing the Lamé constant; and ρ, σ0, ε, and μ are the density, conductivity, dielectric
permittivity, and permeability of the medium, respectively. All the parameters of the medium are set to be
constant since we shall solve the wave fields in a full-space model. v =  iωu is the velocity, and σ0v × B0
represents the induction electric current via which the elastic and EM energies are coupled. Jc is an applied
current-density source, and here we ignore it since a mechanical source is used in this study. We also
ignore the feedback of the induced EM fields on the seismic wave because these induced EM fields are
usually weak (typically on the orders of 1 μV/m and 0.1 nT, respectively) and they would exert negligible
impacts on the seismic wavefields. Therefore, the one-way-coupling governing equations (1)–(4) is
appropriate since we focus on the EM responses induced by a mechanical source.
By substituting equations (2) and (4) into equation (3), one can obtain

∇∇  E  ∇2 E  ω2 eεμE ¼ ω2 μσ 0 uB0 ; (5)

where eε ¼ ε  σ0 =ðiωÞ is the equivalent dielectric permittivity. Equations (1) and (5) are in the frequency-
space domain. It is convenient to solve these two equations by transforming them into the wave number
domain. We define a pair of forward and inverse Fourier transforms as follows:

u ∫
eðkÞ ¼ uðrÞeikr dr; (6)

∫ueðkÞe
1
uðrÞ ¼ ikr
dk; (7)
ð2πÞ3
where r = xiei is the position vector and ei is the coordinate unit vector. Here we use an orthogonal coordinal
system. k ¼ k ^k is the wave vector in which ^k denotes the direction of the wave propagation. Let the source
be a single-point force which is located at position r′ ¼ xi′ ei

F ¼ F0 δðr  r′ Þ: (8)

By applying the forward Fourier transform defined in equation (6) to equations (1) and (5), we have
  
eðkÞ ¼ F0 eikr′ ;
Gk 2 þ ω2 ρ I  ðH  GÞkk  u (9)
  
k 2  ω2eεμ I  kk e e  B0 ;
EðkÞ ¼ ω2 μσ0 u (10)

where I is the identity matrix. Using the following identity

1
ðI  αkkÞ1 ¼ I  kk (11)
k 2  1α

GAO ET AL. ©2014. American Geophysical Union. All Rights Reserved. 3


Journal of Geophysical Research: Solid Earth 10.1002/2014JB010962

(α is an arbitrary constant), we solve equations (9) and (10), and obtain the displacement and the electric field
2 3
 
1 1 1
eðkÞ ¼ 4  2
u  I  2 2 kk þ  kk5 F0 eikr′ ; (12)
G k  ω2 s2s ω ss H k ω s s ω
2 2 2 2 2
p p

 
e ω2 μσ0 kk
EðkÞ ¼ I  eB0 Þ;
ðu (13)
k  ω2 εeμ
2 ω2 εeμ

where ss and sp are the slownesses of the S and P waves, respectively, and they are expressed as
rffiffiffi
ρ
ss ¼ ; (14)
G
rffiffiffi
ρ
sp ¼ : (15)
H

Substituting equations (12) into (13), we rewrite the electric field as

e
EðkÞ ¼
½  2
1
 2
G k  ω sem k  ω ss
2 2 2 2
 B0 I   2
1

Gss ω k  ω s2em k 2  ω2 s2s
2 2 2
 B0 kk

1
þ   2 B0 kk
Hs2p ω2 k 
2
ω2 s2em k  ω2 s2p

 
1

ω2 s2em G k 2  ω2 s2em k 2  ω2 s2s

 
 kkB0  F0 ω2 μσ0 eikr′ ; (16)

where
pffiffiffiffiffiffi
sem ¼ εeμ (17)

is the slowness of the EM wave. eεμ has two square roots that are conjugated since it is complex. Here we
choose the root which has a nonnegative imaginary part. This constrain guarantees that the EM wave will
have nonincreased amplitude during its propagation. The displacement and electric field shown in equations
(12) and (16) are in the frequency-wavenumber domain. To get their responses in frequency-space domain,
we apply the inverse Fourier transform defined in equation (7) to equations (12) and (16), respectively. After a
lengthy algebraic manipulation (detailed in Appendix A), one gets

uðrÞ ¼ GFu ðrj^rÞF0


n 1 eiωss r 1 eiωsp r
¼ ðI  ^r^rÞ þ ^r^r
G"4πr H 4πr ! #
  iωss r o
1 i 1 e 1 i 1 eiωsp r
þ  2 2 2   2 2 2 ðI  3^r^rÞ  F0 ; (18)
G ωss r ω ss r 4πr H ωsp r ω sp r 4πr

EðrÞ ¼ GFE ðrj^rÞF0   iωsem r


n eiωsem r ^  i 1 e  
¼ Aem1 B 0 I  ^r^r ^B 0 þ Aem1  I  ^B 0  3 ^r^r^B 0
4πr ωsem r 2 ω2 s2em r 3 4π
  iωsem r
eiωsem r ^ i 1 e  
 Aem2 B 0  ^r^rþ Aem2  ^B 0 I  3 ^B 0 ^r^r
4πr ωsem r 2 ω sem r
2 2 3 4π
  iωss r
eiωss r i 1 e  
 As1 ^r^r ^B 0 þ As1  I ^B 0  3 ^r^r ^B 0
4πr ωss r 2 ω2 s2s r 3 4π
  iωss r
eiωss r ^  i 1 e ^ 
þ As2 B 0 I  ^B 0 ^r^r þ As2  B 0 I  3 ^B 0 ^r^r
4πr ωss r 2 ω2 s2s r 3 4π
!
eiωsp r ^ i 1 eiωsp r ^ o
þ Ap B 0 ^r^r Ap  2 2 3 B 0 I  3 ^B 0 ^r^r F0 ; (19)
4πr ωsp r 2 ω sp r 4π

GAO ET AL. ©2014. American Geophysical Union. All Rights Reserved. 4


Journal of Geophysical Research: Solid Earth 10.1002/2014JB010962

where ^B 0 ¼ BB00 ¼ ^Bi ei and B0 = |B0| are the direction vector and amplitude of the ambient geomagnetic
field, respectively. r = |r  r′| is the source-receiver distance, and ^r ¼ rr ^
r ¼ ri ei is the unit vector pointing

from the source to the receiver. Gu ðrj ^rÞ and GE ðrj ^rÞ are the Green’s functions of the displacement and
F F

the electric field, respectively. Aem1, Aem2, As1, As2, and Ap are amplitudes, and their expression are detailed
in Appendix A. Since the electric field has been solved, one can easily get the magnetic field by using
equation (3):

1
HðrÞ ¼ ∇EðrÞ ¼ GFH ðrjr′ ÞF0
iωμ
1 n iωsem eiωsem r ^   eiωsem r ^  
¼ Aem1 B 0^r  ^B 0 ^r I  Aem1 2
B 0^r  ^B 0 ^r I
iωμ 4πr 4πr
iωsem eiωsem r ^ 
^
  eiωsem r  ^     
 Aem2 B 0^r B 0 ^r ^r^r þ Aem2 2
2 B 0^r þ ^r^B 0 þ ^B0 ^r I  6 ^B 0 ^r ^r^r
4πr  4πr
eiωsem r i 1      
þ Aem2  3^B 0^r þ 3^r^B 0 þ 3 ^B 0 ^r I  15 ^B 0 ^r ^r^r
4π ωsem r 3 ω sem r
2 2 4

iωss eiωss r ^    eiωss r ^     


þ As2 B 0 ^r ^r^r  ^B 0 ^r I þ As2 2
B 0^rþ ^r^B 0 þ 2 ^B 0 ^r I  6 ^B 0 ^r ^r^r
4πr   4πr
eiωss r i 1      
þ As2  3^B 0^rþ 3^r^B 0 þ 3 ^B 0 ^r I  15 ^B 0 ^r ^r^r
4π ωss r 3 ω ss r
2 2 4

iωsp eiωsp r ^    eiωsp r  ^     


þ Ap B 0^r ^B 0 ^r ^r^r  Ap 2B 0^rþ ^r^B 0 þ ^B 0 ^r I  6 ^B 0 ^r ^r^r
4πr ! 4πr 2

eiωsp r i 1      o
 Ap  2 2 4 3^B 0^rþ 3^r^B 0 þ 3 ^B 0 ^r I  15 ^B 0 ^r ^r^r F0 : (20)
4π ωsp r 3 ω sp r

GFH ðrj^rÞ is the Green’s function of the magnetic field generated by the single-point force. One can easily
prove that

  iωss r
eiωss r i 1 e  
∇ As1 ^r^r^B 0 þ As1  I^B 0  3^r^r^B 0 ¼ 0: (21)
4πr ωss r 2 ω2 s2s r 3 4π

Thus, the terms with the amplitude As1 appearing in the expression of the electric field in equation (19) vanish
in that of the magnetic field in equation (20).
The propagation terms eiωss r and eiωsp r in equation (18) indicate that the seismic disturbance can propagate
with the velocities of the S and P waves. This is a phenomenon as known to all that a single-point force source
can generate elastic S and P waves. Notice that the terms eiωss r and eiωsp r also exist in equations (19) and (20),
implying that there are electric and magnetic perturbations which propagate at the S and P wave velocities.
In other words, the P and S waves can produce electric and magnetic signals during their propagations.
Furthermore, there are also propagation terms eiωsem r in equations (19) and (20), indicating that the electric
and magnetic fields can travel at the EM wave velocity. This implies that a single-point force can generate an
EM wave arising from the motional induction effect as well.
2.2. Wave Field due to Moment Tensor Sources
Since the EM signals were usually measured during actual earthquake events [e.g., Matsushima et al., 2002], it
is necessary to consider an earthquake source. Most earthquakes are generated by fault slips and can be
described mathematically by the moment tensors. In seismology [Aki and Richards, 2002], the displacement
due to a moment tensor in the frequency domain is expressed as

∂GFu;ij
ui ¼ Mjk ; (22)
∂r ′k

where Mjk is the jkth component of the moment tensor and GFu;ij is the Green’s function of the displacement
due to a single-point force.

GAO ET AL. ©2014. American Geophysical Union. All Rights Reserved. 5


Journal of Geophysical Research: Solid Earth 10.1002/2014JB010962

Equation (22) is also valid for the displacement u, the electric field E, and the magnetic H for the elastic-
EM-coupling problem. Using equation (22) and the Green’s function of u shown in the brace in equation (18),
one easily gets that
1 iωss eiωss r     1 iωsp eiωsp r  T 
uðrjr′ Þ ¼ M^r  ^rT M^r ^r þ ^r M^r ^r
G 4πr H 4πr
iωss r    
1e
þ trðMÞ^r þ MT^rþ 2M^r  6 ^rT M^r ^r
G 4πr 2
1 eiωsp r    
 trðMÞ^r þ MT^r þ M^r  6 ^rT M^r ^r
H 4πr 2
 
1 eiωss r i 1    
þ  3trðMÞ^r þ 3MT^r þ 3M^r  15 ^rT M^r ^r
G 4π ωss r 3 ω2 s2s r 4
!
1 eiωsp r i 1    
  3trðMÞ^r þ 3MT^r þ 3M^r  15 ^rT M^r ^r ; (23)
H 4π ωsp r 3 ω2 s2p r 4

where tr(M) = M11 + M22 + M33 is the trace of the moment tensor. The electric field can be obtained in a
similar way
 
iωsem iωsem r 1 eiωsem r 2 eiωsem r i 1 3
Eðrjr′ Þ ¼ Aem1  e Eem1 þ E þ  E
4πr 4πr 2 em1 4π ωsem r 3 ω2 s2em r 4 em1
 
iωsem eiωsem r 1 eiωsem r 2 eiωsem r i 1 3
þ Aem2 Eem2 þ E þ  E
4πr 4πr 2 em2 4π ωsem r 3 ω2 s2em r 4 em2
iωss r iωss r
 
iωss iωss r 1 e 2 e i 1 3
þ As1 e Es1 þ E þ  E (24)
4πr 4πr 2 s1
4π ωss r 3 ω2 s2s r 4 s1
 
iωseiωss r 1 eiωss r 2 eiωss r i 1 3
þ As2  Es2 þ E þ  E
4πr 4πr 2 s2 4π ωss r 3 ω2 s2s r 4 s2
" ! #
iωsp eiωsp r 1 eiωsp r 2 eiωsp r i 1 3
þ Ap  Ep  E   E ;
4πr 4πr 2 p 4π ωsp r 3 ω2 s2p r 4 p

where
h T
Eem1 ¼ ^B 0 ðM^r Þ  ^r^B 0 M^r ^r;
1
(25)

 T    h T i
Eem1 ¼ 2^B 0 ðM^r Þ þ ^r^B 0 M þ ^B 0 M : I ^r  6 ^r^B 0 M^r ^r;
2
(26)

 T    h T i
Eem1 ¼ Es1 ¼ 3^B 0 ðM^r Þ þ 3 ^r^B 0 M þ 3 ^B 0 M : I ^r  15 ^r^B 0 M^r ^r;
3 3
(27)

 
Eem2 ¼ Ep ¼ ^rT M^r ^B 0 ^r;
1 1
(28)

     
Eem2 ¼ Ep ¼ ^B 0 ðM^r Þ þ trðMÞ^B 0 ^r þ ^B 0  MT^r  6 ^rT M^r ^B 0 ^r ;
2 2
(29)
   
Eem2 ¼ Es2 ¼ Ep ¼ 3^B 0 ðM^r Þ þ 3trðMÞ^B 0 ^r þ 3^B 0  MT^r  15 ^rT M^r ^B 0 ^r;
3 3 3
(30)
h T i
^r^B 0 M^r ^r;
1
Es1 ¼ (31)

 T    h T i
Es1 ¼ ^B 0 ðM^r Þ þ ^r^B 0 M þ ^B 0 M : I ^r  6 ^r^B 0 M^r ^r;
2
(32)

 
Es2 ¼ ^B 0 ðM^r Þ  ^rT M^r ^B 0 ^r;
1
(33)

   
Es2 ¼ 2^B 0 ðM^r Þ þ trðMÞ^B 0 ^r þ ^B 0  MT^r  6 ^rT M^r ^B 0 ^r:
2
(34)

GAO ET AL. ©2014. American Geophysical Union. All Rights Reserved. 6


Journal of Geophysical Research: Solid Earth 10.1002/2014JB010962

The symbol “:” in the third term of the right side of equation (26) denotes a double inner product, that is,
 
^ 0 M : I ¼ B^l Mjk εljm em ek : δst es et ¼ B^l Mjk εljm δst ðem  es Þðek et Þ ¼ B^l Mjk εljm δmk ;
B (35)

where εljm is the permutation symbol. Similarly, we obtain the magnetic field generated by the moment tensor
2 2 iωsem r  
  1 ω sem e 1 eiωsem r iωsem 1 2
H rjr′ ¼ Aem1 Hem1 þ  2 þ 3 Hem1
iωμ 4πr 4π r r
2 2 iωsem r  
1 ω sem2 e 1 iωs em2 e iωsem r
2 eiωsem r 3 eiωsem r i 1 4
þ Aem2  Hem2  H  H   H
iωμ 4πr 4πr 2 em2
4πr 3 em2 4π ωsem r 4 ω2 s2em r 5 em
2 2 iωss r  
1 ω ss e 1 iωss e iωss r
2 e iωss r
3 e iωss r
i 1 4
þ As2 Hs2 þ H þ H   H
iωμ 4πr 4πr 2 s2
4πr 3 s2
4π ωss r 4 ω2 s2s r 5 s2
" ! #
1 ω2 s2p eiωsp r 1 iωsp eiωsp r 2 eiωsp r 3 eiωsp r i 1 4
þ Ap Hp þ H þ H þ  Hp : (36)
iωμ 4πr 4πr 2 p
4πr 3 p 4π ωsp r 4 ω2 s2p r 5

In equation (36),
   
Hem1 ¼ ^rT M^r ^B 0  ^rT ^B 0 M^r;
1
(37)
   
Hem1 ¼ trðMÞ^B 0  3 ^rT M^r ^B 0  M^B 0 þ 3 ^rT ^B 0 ðM^r Þ;
2
(38)
    
Hem2 ¼ Hp ¼ ^rT M^r ^B 0  ^rT M^r ^rT ^B 0 ^r;
1 1
(39)
 T    
Hem2 ¼ Hp ¼ trðMÞ^B 0 þ ^B 0 M^r ^r þ 4ð^rT M^r Þ^B 0 þ ^rT M^B 0 ^r þ ^rT ^B 0 trðMÞ^r
2 2

      
þ ^rT ^B 0 MT^r  10ð^rT M^r Þ ^rT ^B 0 ^rþ ^rT ^B 0 ðM^r Þ; (40)
 T  
Hem2 ¼ Hp ¼ 2trðMÞ^B 0  6 ^B 0 M^r ^r  9ð^rT M^r Þ^B 0  6 ^rT M^B 0 ^r
3 3

      
 6 ^rT ^B 0 trðMÞ^r  6 ^rT ^B 0 MT^r þ 45ð^rT M^r Þ ^rT ^B 0 ^r
 
þ MT ^B 0 þM^B 0  6 ^rT ^B 0 ðM^r Þ; (41)
 T
Hem2 ¼ Hs2 ¼ Hp ¼ 3trðMÞ^B 0  15 ^B 0 M^r ^r  15ð^rT M^r Þ^B 0
4 4 4

      
 15 ^rT M^B 0 ^r  15 ^rT ^B 0 trðMÞ^r  15 ^rT ^B 0 MT^r
   
þ 105ð^rT M^r Þ ^rT ^B 0 ^r þ 3MT ^B 0 þ 3M^B 0  15 ^rT ^B 0 M^r; (42)
    
Hs2 ¼ ^rT M^r ^rT ^B 0 ^r  ^rT ^B 0 M^r;
1
(43)
 T  
Hs2 ¼  ^B 0 M^r ^r  ð^rT M^r Þ^B 0  ^rT M^B 0 ^r  ð^rT B0 ÞtrðMÞ^r
2

      
 ^rT ^B 0 MT^r þ 10ð^rT M^r Þ ^rT ^B 0 ^r þ M^B 0  4 ^rT ^B 0 ðM^r Þ; (44)
3  
Hs2 ¼ trðMÞB0 þ 6 B0 T M^r ^r þ 6ð^rT M^r ÞB0 þ 6ð^rT MB0 Þ^r þ 6ð^rT B0 ÞtrðMÞ^r
 
þ 6ð^rT B0 Þ MT^r  45ð^rT M^r Þð^rT B0 Þ^r  MT B0  2MB0 þ 9ð^rT B0 ÞðM^r Þ: (45)

3. Numerical Simulations and Application to Data Analysis


3.1. Application to Coseismic EM Signal Observed in 2008 Qingchuan Mw6.1 Earthquake
We have derived the frequency-space-domain expressions of the electric and magnetic fields that are
generated by a point force source and a moment tensor source in the former section. The time responses of the
EM fields can be easily obtained by applying the discrete Fourier transform with respect to the frequency. The
propagation terms eiωsem r, eiωss r, and eiωsp r in both the expressions of the electric and magnetic fields in equations
(19), (20), (24), and (36) imply that the EM disturbances will travel at the velocities of the EM, S, and P waves,
respectively. In the following sections, we shall make numerical simulations to illustrate these EM disturbances.
Before conducting the simulation, we introduce an example of real EM data that were measured during
actual earthquake. On 25 May 2008 an Mw6.1 earthquake (star symbol in Figure 1a) took place at 16:21:49.3
(local time) in Qingchuan County of Sichuan province, China. It is an aftershock of the 12 May 2008 Wenchuan
earthquake that also took place in Sichuan Province. The location of the epicenter is (32.57°N, 105.42°E).

GAO ET AL. ©2014. American Geophysical Union. All Rights Reserved. 7


Journal of Geophysical Research: Solid Earth 10.1002/2014JB010962

Figure 1. (a) Distributions of the epicenter (star), the seismic station (solid circle), and the MT measurement site (triangle).
Their locations are (32.57°N, 105.42°E), (33.35°N, 104.99°E), and (33.34°N, 105.02°E), respectively. (b) Full-space geological
model for the theoretical simulation in this study. A Cartesian coordinates system is set up on the Earth’s surface with x, y,
and z indicating the north, east, and downward directions, respectively.

During this aftershock, a seismic station (solid circle in Figure 1) which is located at (33.35°N, 104.99°E) and
about 99 km northwest of the epicenter, recorded the seismic waves by an accelerometer with a sampling
frequency of 200 Hz. In the meanwhile, a magnetotelluric (MT) monitoring instrument (triangle in Figure 1a)
that is located at (33.34°N, 105.02°E) and about 3 km southeast to the seismic station, recorded the electric,
and magnetic variations. The sampling frequency of the electric and magnetic signals is 24 Hz. The MT
instrument was equipped and began working 3 days before the occurrence of this earthquake by
the researchers of the Institute of Geology, China Earthquake Administration [Tang et al., 2010]. The topmost
three plots of Figure 2 show the acceleration signals recorded at the seismic station, from which we can see that
the P wave arrived about 15 s after the occurrence of the earthquake. The remaining five plots of Figure 2
display the two components of the electric field and three components of the magnetic field measured by
the MT equipment. We can see clear coseismic EM disturbances that approximately share the same arrival
time with the P wave since the MT equipment is very close to the seismic station (only 3 km). The S wave that
arrived about 25 s after the earthquake occurrence produced more significant coseismic EM responses
because of its larger amplitude. The acceleration corresponding to the S wave is on the order of 0.1 m/s2,
while the coseismic electric and magnetic fields are on the order of 1 μV/m and 0.1 nT, respectively. In this
section we shall make a simple estimation of these coseismic EM signals on the basis of the motional
induction effect. The purpose is to investigate whether the EM fields caused by such an effect can reach the
amplitudes of the real EM data.
Here we consider a full-space geological model because the theoretical formulations we derived in the
former section is a full-space solution. As shown in Figure 1b a Cartesian coordinate system (o, x, y, z) is
constructed with the original point o locating at the hypocenter, and x, y, and z denoting the positive north,
positive east, and downward directions, respectively. The coordinates of the seismic station are (x = 86.8 km,
y =  40 km, z =  27 km) and those of the MT measurement site are (x = 85.7 km, y =  37.5 km, z =  27 km).
The parameters of the medium for the calculation are listed in Table 1 (elastic medium 1). For the source
mechanism of this earthquake, we use the global centroid-moment tensor solution which gives the strike,
dip, rake, and depth of the source as 59°, 84°, 178°, and 27 km, respectively. The scalar moment is
M0 = 1.59 × 1018 N  m corresponding to an Mw6.1 earthquake. The time function we use is a Ricker wavelet
with a dominant frequency of f0 = 0.6 Hz and a time delay t0 = 1/f0:

h i 22 2
s0 ðtÞ ¼ 1  2π2 f 20 ðt  t 0 Þ2 eπ f 0 ðtt0 Þ : (46)

Using the formulas shown in equations (23), (24), and (36) as well as the parameters listed in Table 1, we
calculate the seismic and EM signals and plotted their waveforms in Figure 3. Figures 3a and 3b show the

GAO ET AL. ©2014. American Geophysical Union. All Rights Reserved. 8


Journal of Geophysical Research: Solid Earth 10.1002/2014JB010962

synthetic seismic accelerations at the


seismic station, from which we see that
the P wave arrives at about t = 15.4 s
while the S wave arrives at about
t = 25.6 s. Figures 3d–3h display the
simulated electric and magnetic signals
recorded at the MT measurement site
(black lines). The electric and magnetic
disturbances starting at about t = 15.4 s
are the coseismic EM fields caused by
the P wave, while those beginning at
about t = 25.6 s are the coseismic EM
fields produced by the S wave.
Notice that the waveforms shown in
Figure 3 are computed from the full-
space analytical expressions. The
calculated seismic or EM signals cannot
be identical to the recorded signals,
because the real signals contain a
variety of wave components due to
multiple reflections and refractions
resulting from the complexity of the
strata. However, we find that the
amplitudes of the simulated signals can
reach the orders of the real signals. As
can be seen from Figures 3a–3c, the
amplitude of the seismic acceleration is
on the order of 0.1 m/s2, the same with
the observed signal (Figure 2). The
Figure 2. Coseismic electromagnetic signals recorded during the Qingchuan simulated coseismic electric field is on
Mw 6.1 earthquake on 25 May 2008. The uppermost three records are the the order of 1 μV/m, sharing same
acceleration signals ax, ay, and az measured by the accelerometer in the
order of the amplitude with the
seismic station with x, y, and z denoting the north, east, and downward
directions, respectively. The fourth and fifth ones are the horizontal electric observed one as well. In addition, the
fields Ex and Ey measured by the MT instrument. The lowermost three are the simulated and observed coseismic
magnetic fields Hx, Hy, and Hz. The left vertical dashed line that is marked by magnetic fields are also on the same
the symbol “EQ” denotes the occurrence time of the earthquake, that is, order, that is, 0.1 nT. This indicates that
16:21:49.3 local time. The symbols “P” and “S”indicate the arrival times of the P
the motional induction effect is
and S waves, respectively.

Table 1. Parameters of the Elastic Media for Calculation


Value

Solid Rock 2 Equivalent to the Porous


Parameter Variable (Unit) Solid Rock 1 Medium (Sandstone) in Table 2

Elastic modulus H(GPa) 117.9 38.1


Shear modulus G(GPa) 42.6 14.7
3
Density of the medium ρ(g/cm ) 2.8 2.3
Velocity of the P wave Vp( km/s) 6.5 4.0
Velocity of the S wave Vs( km/s) 3.9 2.5
 11  11  11
Permittivity ε(F/m) 7.35 × 10 5.15 × 10 ~ 7.26 × 10
5
Conductivity σ(S/m) 0.1 11.3 × 10 ~ 1.4
Ambient Earth’s magnetic field B0(nT) B0x = 33445.7 B0x = 33445.7
B0y =  1195.2 B0y = 1195.2
B0z = 39535 B0y = 39535

GAO ET AL. ©2014. American Geophysical Union. All Rights Reserved. 9


Journal of Geophysical Research: Solid Earth 10.1002/2014JB010962

effective enough to generate


observable coseismic EM disturbances.
Figure 3 also shows the EM wave (red
lines) radiated from the source, which
shows long-period variation compared
with the coseismic EM fields. We can
see that the electric and magnetic fields
are of the orders of 10 16 V/m and
10 10 nT, respectively, much smaller
than those of the coseismic EM field
which are of the orders of 1 μV/m and
0.1 nT. This is because in such a
conductive medium (σ = 0.1 S/m in the
simulation) the EM wave is highly
dispersive and attenuated. In addition,
at low frequencies (less than several
Hertz), the EM wave behaves more like
diffusion than propagation [Løseth
et al., 2006] and its phase velocity could
be lower than that of a seismic wave.
Therefore, we find that the source
signal arrives a little earlier than the
P arrival rather than at the onset time of
the earthquake.
3.2. Explanation of the Observed
Lorentz Field in Haines et al. [2007]
Figure 3. Synthesized signals calculated from the motional induction effect:
Equations (19), (20), (24), and (36) imply
(a–c) seismic accelerations at the seismic station, (d–e) electric field, and
(f–h) magnetic field at the MT measurement site. Note that not only the that a mechanical source like a point
force or a moment tensor source can
simulated accelerations are on the same order of the observed seismic data
but also the electric and magnetic fields share the same orders of the generate an EM wave due to the
amplitudes with the observed EM data (Figure 2). The red lines in motional induction effect. Actually, a
Figures 3d–3h represent the EM wave radiated from the source, which show
kind of EM wave that might result from
much longer period and smaller amplitude than the coseismic EM field.
such a mechanism was observed in
the field experiment conducted by
Haines et al. [2007] for electrokinetic investigation. They hit a metal hammer plate (aluminum cylinder shown
in Figure 4b) by a sledgehammer and measured the electric field by 24 electrode pairs spaced around a circle
of 4.3 m radius (Figure 4a). The sledgehammer impact on the hammer plate can be regarded as a vertical
force, and it excited seismic waves.
Due to the electrokinetic effect, the vertical force generated an EM wave (referred to as the direct field by
Haines et al. [2007]). It acted like a vertical electric dipole so that the EM wave radiated by the source is
axisymmetric about the vertical line through the source point. As a result, the radial electric fields Er measured
at electrode pairs along the radial direction are equal while the tangential electric fields Et along the
tangential direction are zeroes. Aside from the EM wave resulting from the electrokinetic effect, they also
measured another kind of EM wave which they referred to as the Lorentz field and showed different
properties from the electrokinetic direct field. The amplitudes of Er (red dotted line) and Et (green dotted line)
of the Lorentz field recorded at different positions are plotted in Figure 5, from which we can see that they
are direction dependent.
As indicated by Haines et al. [2007], the Lorentz field cannot be explained by the electrokinetic effect, because
the EM wave generated by the vertical sledgehammer impact (that can be taken as a vertical force) due to the
electrokinetic effect should be axisymmetric. To confirm this, we simulated the EM wave generated by a
vertical force by using the full-space analytic solution given by Pride and Haartsen [1996] and the parameters
for a porous medium listed in Table 2. The salinity is chosen as Cf = 0.1 mol/L, and the corresponding

GAO ET AL. ©2014. American Geophysical Union. All Rights Reserved. 10


Journal of Geophysical Research: Solid Earth 10.1002/2014JB010962

Figure 4. Layout of the seismoelectric experiment conducted by Haines et al. [2007]. (a) Circular electrode array with 12
radial and 12 tangential channels in a horizontal plane. Small black circles represent electrode, and dashed lines indicate
the pairs across which the voltage was measured. (b) Side and (c) map views of the creation of the Lorentz field. The
horizontal component of the hammer plate velocity v crossed with the vertical component of the Earth’s magnetic field B0
creates an electric field EF in the conductive hammer plate [after Haines et al., 2007].

conductivity is σ0 = 0.069 S/m. The time


function used is also a Ricker wavelet
defined in equation (46) but with a
higher center frequency f0 = 300 Hz. The
radial and vertical distances between
the source and electrode are set to be
4.3 m and 0.5 m, respectively. As can
be seen from Figure 6a, the radial
electric fields Er received along the
circular electrode array are identical and
direction independent. The tangential
electric fields Et are zero at all positions
so that we did not plot them. The
amplitudes of the synthetic Er and Et are
compared with the data collected by
Haines et al. [2007] in Figure 5a, showing
that the Lorentz field was not caused by
the electrokinetic effect. Considering
the possibility pointed out by Haines
et al. [2007] that the hammer impact
might deviate from the vertical
direction, as shown in Figure 4b, there
might be a horizontal component of the
impact on the cylindrical hammer plate.
Figure 5. Comparison between the amplitude patterns of the observed Such a horizontal impact acted as a
Lorentz field (dotted lines) and the modeled EM waves (solid lines). horizontal force which is oriented
(a) The electrokinetic (EK) EM waves excited by a vertical force FV and (b) a
40° east of north and perpendicular the
horizontal force oriented 40° east of north FH. (c) The motional induction
(MI) EM waves excited by FV and (d) FH. Red and olive lines denote the hammer plate orientation. Figures 6b
radial and tangential components of electric fields, respectively. and 6c display the simulated electric

GAO ET AL. ©2014. American Geophysical Union. All Rights Reserved. 11


Journal of Geophysical Research: Solid Earth 10.1002/2014JB010962

Table 2. Parameters of the Porous Medium (Sandstone) Used for Simulating the Seismic and EM Wavefields due to the
Electrokinetic Effect
Parameter Symbol Value or Range (Unit)
a
Porosity ϕ 0.19
a  15 2
Permeability κ0 70 × 10 (m )
a
Formation factor F 17.7
Tortuosity α∞ 3.36
Solid grain bulk modulus Ks 35.7 (GPa)
b
Frame bulk modulus Kb 15 (GPa)
Shear modulus G 14.7 (GPa)
Fluid bulk modulus Kf 2.25 (GPa)
b 3
Solid grain density ρs 2650 (kg/m )
3
Fluid density ρf 1000 (kg/m )
Solid relative permittivity εs 4
5
Salinity of the pore fluid Cf 10 ~ 5 (mol/L)
c
Fluid relative permittivity εr 78.24 ~ 36.1
 11  11
Effective permittivity of the porous medium ε 5.15 × 10 ~ 7.26 × 10 (F/m)
c
Fluid viscosity η 1.9 ~ 3 (Pa s)
c
Zeta potential ζ  89.9 ~  31.1 (mV)
c 5
Rock conductivity σ0 11.3 × 10 ~ 1.4 ( S/m)
c  10
Low-frequency electrokinetic coupling coefficient L0 13.9 ~ 1.88 (×10 sC/kg)
a
The porosity, the permeability, and the formation factor are chosen as the same values with the sandstone used in
the work by Jaafar et al. [2009].
b
The frame bulk modulus Kb is calculated from the porosity according to the relations between frame modulus and
porosity given by Vernik [1998].
c
εr, η, ζ , σ0, and L0 vary with the pore fluid salinity Cf, and they are calculated from their corresponding equations in
Appendix B.

fields Er and Et generated by such a horizontal force due to the electrokinetic effect. By comparing their
amplitude patterns (Figure 5b) with the data collected by Haines et al. [2007], we find even a horizontal force
cannot explain the data, either.
Haines et al. [2007] proposed that the Lorentz field is caused by the motion v of the metal hammer plate
crossing the Earth’s magnetic field B0, i.e., v × B0 (in their experiments, no Lorentz field was measured when a
plastic hammer plate was used). Furthermore, they pointed out that the aluminum cylindrical hammer plate
has more freedom to move horizontally than vertically so that the horizontal velocity of cylinder vH can be
higher than vertical one vV even though the impact on the cylinder has a smaller horizontal component than
the vertical component. As a result, the Lorentz field was proposed to be created by the horizontal motion vH
crossing with the vertical component of the Earth’s magnetic field B0z and seemed to be radiated by a
horizontal dipole with an orientation 50° west of north (i.e., the hammer plate orientation in Figure 4c) which
is perpendicular to both the directions of vH and B0z.
To evaluate their explanation of the Lorentz field, we calculate the electric fields generated due to the
motional induction effect using equation (19). The source we utilize is also a vertical force as well as a
horizontal force that is oriented 40° east of north. The horizontal force is used to generate the horizontal
motion of the hammer plate (Figure 4c). Note that the parameters used for the electrokinetic simulations
in Figures 6a–6c are for a porous medium (in Table 2) but our formulations for the motional induction
effect require an elastic medium. We then treat this porous medium as a nonporous solid, and get the
equivalent P velocity, S velocity, and density, which are 4.0 km/s, 2.5 km/s, and 2.3 g/cm3, respectively. The
conductivity and relative permittivity are chosen to be σ0 = 0.069 S/m and ε = 7.2 × 10 11 F/m, respectively.
The Earth’s magnetic field are chosen with components Bx = 2.3 × 10 5 T, By = 0, and Bz = 4 × 10 5 T, the
same value with Haines [2004]. The radial and tangential electric fields of the EM wave generated by a
vertical force are displayed in Figures 6d and 6e, from which we can see that they are direction dependent
while those resulting from the electrokinetic effect (Figure 6a) are not. The EM wave generated by the
horizontal force is also directionally dependent (Figures 6f and 6g). In Figures 5c and 5d, we plot the
amplitude patterns of the motional induction EM waves as a function of receiving position and compare
them with the Lorentz field collected by Haines et al. [2007]. It can be seen that the amplitude patterns of

GAO ET AL. ©2014. American Geophysical Union. All Rights Reserved. 12


Journal of Geophysical Research: Solid Earth 10.1002/2014JB010962

Figure 6. Waveforms of the modeled EM waves arising from the electrokinetic (EK) effect received at the 12 positions in
Figure 4a: (a) Radial electric fields excited by a vertical force FV and (b) radial and (c) tangential electric fields excited by
a horizontal force FH oriented 40° east of north. Waveforms of the modeled EM waves arising from the motional induction
(MI) effect: (d) radial and (e) tangential electric fields excited by FV and (f) radial,and (g) tangential electric fields excited by
FH. In each subgraph, the waveforms are normalized by the max amplitude among the 12 signals.

the EM wave resulting from the horizontal force (Figure 5d) fit the data better than those by the vertical
force (Figure 5c). This confirms the conclusion made by Haines et al. [2007]; that is, the horizontal
motion of the cylinder in the Earth’s magnetic field are responsible for the observed field. By the way,
Figure 6 also illustrates that a mechanical source can generate observable EM wave due to the motional
induction effect.

3.3. Approximate Expression of the EM Wave Component


To obtain a further understanding of the EM wave excited by a point force, we study its radiation pattern. In
the analytical solution of the electric field in equation (19), the terms with eiωsem r denote the EM wave, in which

GAO ET AL. ©2014. American Geophysical Union. All Rights Reserved. 13


Journal of Geophysical Research: Solid Earth 10.1002/2014JB010962

the far field with 1r dependency and the near



field with ωsemi r2  ω2 s12 r3 dependency can
em

be identified. There are two amplitude


coefficients for both the far and near fields,
i.e., Aem1 and Aem2. Figure 7 shows the ratio
of |Aem1| / |Aem2| versus conductivity at
different frequencies, from which we can
find that for a given frequency |Aem1| / |Aem2|
decreases with the conductivity. In the low-
frequency cases (less than 1 Hz) |Aem1|/|Aem2|
decreases gradually to a constant-like
value with the increase of the
conductivity. For example, at f ¼ 0:1 Hz
jAem1 j=jAem2 j approaches to 1.6 when the
Figure 7. Ratios of the amplitudes |Aem1| / |Aem2| versus the conduc-
tivity at different frequencies. conductivity is higher than 0.04 S/m. But
for the case f > 100 Hz, |Aem1| / |Aem2| is
always larger than 30 in the conductivity
band 10 4 S/m  1.4 S/m. Therefore, we find that in the high-frequency situation Aem1 dominates the EM wave
so that the radiation pattern of the EM wave is mainly controlled by the first two terms of equation (19).
Therefore, by omitting the terms with amplitude Aem2 we have

  iωsem r
eiωsem r  ^ 
^ 0 F þ Aem1 i 1 e 
^ 0  3^r^r B

^ 0 F
Eem ðrÞ ≈ Aem1 B 0 I  ^r^r B  IB
4πr ωsem r 2 ω2 s2em r 3 4π
iωsem r   iωsem r
e i 1 e  
¼ ^^
ðI  r r Þ þ  ^^
ðI  3r r 0 Þ  Aem1 B ^ 0 F
4πr ωsem r 2 ω2 s2em r 3 4π
iωsem r   iωsem r " #
e i 1 e σ 0 B0 1 ^ 0 F (47)
¼ω μ
2
ðI  ^r^r Þ þ  ðI  3^r^r 0 Þ   2  2 ; 2 B
4πr ωsem r 2 ω2 s2em r 3 4π ω G sem  ss
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
p

where the equalities B ^ 0 F and ^r B


^ 0 IF ¼ IB ^ 0 F ¼ ^r B
^ 0 F are used. The term enclosed by the second
bracket which is denoted by p can be taken as an electric dipole since it has a dimension of dipole moment.
Furthermore, we can find that the terms enclosed by the first bracket are identical to the Green’s function of
electric field that is generated by an electrical dipole (see Novotny and Hecht [2006, equation (8.55)]). This
means that the electric field generated by a point force F has approximately the same radiation pattern as
that radiated by an electric dipole, however, whose orientation is not along the force direction but
determined by the cross product of B ^ 0 and F.

4. Comparison With the Electrokinetic Effect


The former sections proved that the motional induction effect is an effective and reasonable mechanism
for the coupling between the seismic wave and EM field. However, it should be noticed that there are
also other candidate mechanisms such as the electrokinetic effect, the piezoelectric effect, and the
piezomagnetic effect. It is possible that two or more than two of these mechanisms simultaneously
contribute to the conversion from seismic wave to the EM field while contribution from each mechanism
may be different. Therefore, it is necessary to study the property of each candidate mechanism and find
out which mechanism dominates the seismic-to-EM conversion under a specific geological condition.
This kind of work will be helpful for distinguishing one mechanism from the other and benefit the
explanation of the EM data observed during actual earthquake. Here we shall compare the motional
induction effect with the electrokinetic effect, which has attracted a lot of attentions and supports [e.g.,
Haartsen and Pride, 1997; Nagao et al., 2000; Karakelian et al., 2002; Gao and Hu, 2010; Hu and Gao,
2011]. The comparison shall be made between the EM responses to an earthquake due to these
two mechanisms.

GAO ET AL. ©2014. American Geophysical Union. All Rights Reserved. 14


Journal of Geophysical Research: Solid Earth 10.1002/2014JB010962

The theoretical basis we use for modeling the electrokinetic phenomenon is Pride’s equations [Pride, 1994]
that are reviewed in Appendix B. According to the work by Gao and Hu [2010], the displacement u and the
electric field E generated by a moment tensor source in a full space are shown below:
Xsem
iωsei ωsr     X sps
iωsei ωsr  T 
α¼ T Fα;s M^r  ^r T M^r ^r þ LFα;s ^r M^r ^r
s¼ss
4πr s¼spf
4πr
Xsem
ei ωsr    
þ T Fα;s 2 trðMÞ^r þ MT ^r þ 2M^r  6 ^r T M^r ^r
s¼ss
4πr
X sps
ei ωsr    
þ LFα;s 2 trðMÞ^r þ MT ^r þ M^r  6 ^r T M^r ^r
s¼spf
4πr
hX  
sem
ei ωsr i 1
þ T Fα;s 
4π ωsr 3 ω2 s2 r 4
X
s¼ss
sps  i
e i ωsr
i 1  
 LFα;s  3trðMÞ^r þ 3MT ^r þ 3M^r  15^r T M^r ^r ; (48)
s¼spf
4π ωsr 3 ω s r
2 2 4

where α = u, E. spf, sps, ss, and sem are the complex slownesses of the fast P wave (Pf), slow P wave (Ps), shear
wave (S), and EM wave, respectively. T Fα;s and LFα;s are the complex amplitudes of the transverse and
longitudinal components [Pride and Haartsen, 1996]. The expressions of the complex slownesses and
amplitudes are detailed in Appendix B. The magnetic field generated by the moment tensor is
(  
1 X sem
ei ωsr ω2 s2 Xsem
ei ωsr iωs 1
H¼ T FE;s ½^r ðM^r Þ þ T FE;s  2 þ 3 ½ðIMÞ  3^r ðM^r Þg; (49)
iωμ s¼ss 4π r s¼ss
4π r r
where
 
I M ¼ δki ek ei Mjk ej ek ¼ δki Mjk ei ej ðek ek Þ ¼ Mji εijl el : (50)

Once the parameters of the porous medium and the source are given, one can calculate the displacement u,
the electric field E, and the magnetic field H at the specified position r from equations (48) and (49).
To get reasonable seismic and EM responses due to the electrokinetic effect, we follow the existing literature
and select proper parameters of the porous medium for numerical simulation. The porous medium is
assumed to be the sandstone that was used in the streaming potential experiment conducted by Jaafar et al.
[2009]. The porosity, permeability, and formation factor are ϕ = 0.19, κ0 = 70 × 10 15 (m2), and F = 17.7,
respectively. The tortuosity α∞ = 3.36 is derived from F = α∞/ϕ. The solid grain modulus and shear modulus
are chosen as Ks = 35.7 GPa and G = 14.7 GPa, respectively. The frame bulk modulus Kb is calculated from the
porosity according to the relations between frame moduli and porosity given by Vernik [1998]. The
parameters that affect the conversion between the seismic waves and EM fields are discussed in detail in
Appendix B. They are, namely, the fluid relative permittivity εr, the fluid viscosity η, the zeta potential ζ , the
rock conductivity σ0 , and the low-frequency electrokinetic coupling coefficient L0, all of which change with
the pore fluid salinity Cf . Their values are either chosen from experiment data or derived from empirical
relations and thus should be realistic. All the parameters of the porous medium are listed in Table 2.
Note that these parameters used for calculating electrokinetic wavefields are for a fluid-saturated porous
medium while the theoretical formulations derived in the present work for modeling the motional induction
effect are based on the elastodynamic theory for nonporous elastic solids. To compute the seismic and EM
wavefields arising from the motional induction effect, we treat the sandstone as a solid rock and then get the
equivalent P velocity (Vp = 4.0 km/s), S velocity (Vs = 2.5 km/s), and density (ρ = 2.3 g/m3). Then we calculate
the elastic moduli H and G from equations (15) and (14), respectively. The conductivity and dielectric
permittivity of the elastic continuum are identical to the bulk conductivity and permittivity of the porous
medium. All the parameters of the equivalent solid rock are listed in Table 1 (solid rock 2).
As shown in Figure 1b a Cartesian coordinates is used with the source being located at the origin point o
and the receiver being located at (x = 40 km, y = 40 km, z =  10 km). The source is a fault with the strike, dip,
and rake being 90∘, 60∘, and 0∘, respectively. The scalar moment is M0 = 1.59 × 1018 N  m corresponding to
an Mw6.1 earthquake. The source time function is also a Ricker wavelet with central frequency of 1 Hz.

GAO ET AL. ©2014. American Geophysical Union. All Rights Reserved. 15


Journal of Geophysical Research: Solid Earth 10.1002/2014JB010962

Figure 8

GAO ET AL. ©2014. American Geophysical Union. All Rights Reserved. 16


Journal of Geophysical Research: Solid Earth 10.1002/2014JB010962

With Cf = 0.001 mol/L and other parameters listed in Table 2, we calculate the displacement uEK, the electric
field EEK, and the magnetic field HEK due to the electrokinetic effect by using the formulas (48) and (49).
With the parameters of the equivalent solid rock listed in Table 1, we compute the wavefields uMI, EMI,
and HMI arising from the motional induction effect by using equations (23), (24), and (36). The displacements
calculated from the two mechanisms are compared in Figure 8. As expected, there is no visual difference
between uEK and uMI. In Figure 8, we also compare the motional induction EM signals EMI and HMI (red lines)
with their electrokinetic counterparts, i.e., EEK and HEK (black lines). Each component of the electric and
magnetic fields is divided into three continuous time periods, i.e., 0  10 s, 10 s  20 s, and 20 s  30 s, in order
to illustrate the EM responses corresponding to the EM wave, the P wave, and the S wave, respectively, which
have different arrival times and amplitude scales.
In the electrograms in Figure 8, we can see three signals that are generated due to the motional induction
effect (red lines). The earliest one EMI
em is the EM wave which arrives immediately after the source originates.
The second one EMI p which has an arrival time of about t = 14 s is the coseismic electric field accompanying
the P wave. The third one which is denoted by EMI s and arrives at t = 23 s is the coseismic electric field
produced the S wave. These three signals are of different amplitudes. EMI em is on the order of 0.0001 μV/m,
smaller than EMI
p which is the coseismic electric field accompanying the P wave and is on the order of
0.01 μV/m, and much smaller than EMI em which is the coseismic electric field accompanying the S wave and
whose amplitude is as large as 20 μV/m. The electric signals resulting from the electrokinetic effect are also
plotted (black lines). We can also identify three signals, namely, EEK
em , Ep , and Es which have similar
EK EK

waveforms, same arrival times but different amplitudes with their motional induction counterparts, i.e., EMI em ,
EMI
p , and E MI
s , respectively. EEK
em is of the order of 0.01 μV/m, 100 times greater than E MI
em (0.0001 μV/m). EEK
p is of
p (0.01 μV/m). This indicates that in the geological condition
the order of 1000 μV/m, much larger than EMI
described by the parameters in Table 2 with Cf = 0.001 mol/L, the electrokinetic effect is the dominant
mechanism for producing the EM wave and the P-converted coseismic electric field. However, EEK s has an
amplitude of about 10 μV/m which is on the same order of EMI
s (20 μV/m), indicating that the two mechanisms
have comparable efficiency in converting the S wave to the electric field.
In the magnetograms in Figure 8, HMIem is the magnetic signal of the EM wave generated by the source due to
the motional induction effect while HEK
em is the one due to the electrokinetic effect. The amplitude of Hem is of
MI

the order of 0.001 pT, 100 times weaker than Hem which is of the order of 0.1 pT, indicating again that the
EK

electrokinetic effect dominates the generation of the EM wave. HMI p with amplitude of 10 pT is the coseismic
magnetic field accompanying the P wave. In contrast, the electrokinetic counterpart HEK
p is zero because the P
wave cannot produce a magnetic field due to the electrokinetic effect [Pride and Haartsen, 1996]. HMI
s and Hs
EK

are the S wave magnetic fields due to the motional induction effect and electrokinetic effect, respectively.
The former (on the order of 1 pT) is 3 orders of magnitude weaker than the latter (on the order of 1 nT),
showing that electrokinetic effect is the dominant mechanism in transforming the S wave energy into the
magnetic signals.
The former comparison between the EM fields arising from the two mechanisms is made under the condition
that the pore fluid salinity is Cf = 0.001 mol/L with the corresponding conductivity being about 7. × 10 4 S/m.
Note that the coupling term σ v × B0 in equation (2) implies that a higher conductivity will give rise to a larger
induction electric current which will generate stronger EM field. On the contrary, theoretical and experiment
works [e.g., Garambois and Dietrich, 2001; Block and Harris, 2006] have indicated that an increase in the
conductivity will cause a decrease in the EM field resulting from the electrokinetic effect. Therefore, it is

Figure 8. Comparisons between the displacements, the electric fields, and the magnetic fields calculated from the elec-
trokinetic formulations given by Gao and Hu [2010] (black lines) and the motional induction formulations derived in this
study (red lines). Each component of the electric and magnetic fields is divided into three continuous time periods, i.e.,
0  10 s, 10 s  20 s, and 20 s  30 s, in order to illustrate the electric and magnetic responses corresponding to the EM
wave, the P wave, and the S wave, respectively, which have different arrival times and amplitude scales. Note that the P
wave cannot produce a magnetic field due to the electrokinetic effect [Pride and Haartsen, 1996] so that we find that
EK EK EK
H p,x = H p,y = H p,z = 0.

GAO ET AL. ©2014. American Geophysical Union. All Rights Reserved. 17


Journal of Geophysical Research: Solid Earth 10.1002/2014JB010962

necessary to consider the influence of the conductivity on each mechanism. In this study, we control the
conductivity by the pore fluid salinity as discussed in Appendix B. Letting the salinity varies from 10 5 mol/L
to 5 mol/L (Note that 5 mol/L is close to the saturated concentration of the NaCl solution at room
temperature), we have that the conductivity changes from 1.13 × 10 4 S/m to 1.4 S/m.
In order to investigate how the conductivity affects the ability of the P and S waves in producing the
coseismic electric and magnetic fields, it is convenient to study the transfer functions E/u and H/u for the
P and S waves arising from both the motional induction effect and the electrokinetic effect. According to
previous studies [Pride and Haartsen, 1996; Garambois and Dietrich, 2001; Gao and Hu, 2010], the transfer
functions E/u and H/u corresponding to the P and S waves arising from the electrokinetic effect are
 
   iωe ρL Hs2p  ρ  
 EK  
Ep  ¼   up ; (51)
 eε Cs2p  ρf 
 
 EK   G s2s  ρ=G
E  ¼ iωμe ρ L jus j; (52)
s  ρf s2s  μeε 
 
 EK   G s2s  ρ=G
H  ¼ iωsse ρ L jus j; (53)
s  ρf s2s  μeε 

where eρ ¼ iη=ðωκ0 Þ. A P wave cannot generate any magnetic field due to the electrokinetic effect so that H/u
vanishes for a P wave [Pride and Haartsen, 1996; Gao and Hu, 2010]. For the motional induction effect, the
transfer functions E/u and H/u are (detailed derivation are given in Appendix C)
 
   μσB  
 MI   0  
Ep  ≈  2 u ; (54)
sp  eεμ p
  2  
 MI     
E  ≈ max 1;  ss   μσB0 jus j; (55)
s eεμ s2  eεμ
s
 
   s σB  
 MI   p 0  
H ≈
 p  2 u ; (56)
sp  eεμ p
 
 MI   ss σB0 
H  ≈  
s s2  eεμjus j: (57)
s

By using equations (51)–(57) and letting |up| = |us| = 0.01 m, we calculate the amplitudes of the coseismic EM
fields produced by the P and S waves due to the electrokinetic effect and the motional induction effect at
different conductivities and four different frequencies, and plot them in Figure 9.
The variation of the coseismic electric field with the conductivity at f = 0.1 Hz is displayed in Figure 9a, from
which we can see that when the conductivity increases the P-converted electric field due to the motional
induction effect EMI
p (red solid line) increases while that due to the electrokinetic effect Ep (blue solid line)
EK

decreases. EEK
p is larger than Ep at conductivities less than the transition point 0.016 S/m for f = 0.1 Hz.
MI

Figures 9b and 9c show that the transition point is augmented to 0.12 S/m and 0.86 S/m when the frequency
is increased to f = 1 Hz and f = 10 Hz, respectively. For f = 100 Hz (Figure 9d), EMI
p is always smaller than Ep ,
EK

indicating that at higher frequency the electrokinetic effect is always the dominant mechanism in converting
the P wave to the electric field in the conductivity region σ0 ≤ 1.4 S/m. Figures 9a–9d also show that for a given
frequency the S wave electric field due to the electrokinetic effect EEKs (blue dotted line) decreases with the
conductivity while that due to the motional induction effect EMI s (red dotted line) almost keeps constant.
Furthermore, EMI
s is always bigger than EEK
s regardless of the conductivity and the frequency, implying that the
motional induction effect always has a stronger ability in transferring the S wave energy into the electric signal.
Figures 9e–9h display the changes of the coseismic magnetic field with the conductivity at four different
frequencies. We find that the magnetic fields produced by the P wave (HMI MI
p ) and the S wave (Hs ) due to the
motional induction effect have comparable amplitudes. This means that the P and S waves have almost the
same efficiency in producing the magnetic field via the motional induction effect. Both HMI
p and Hs increase
MI

with the conductivity while the S wave magnetic field HEK


s resulting from the electrokinetic effect decreases. Hs
EK

GAO ET AL. ©2014. American Geophysical Union. All Rights Reserved. 18


Journal of Geophysical Research: Solid Earth 10.1002/2014JB010962

Figure 9. (a–d) Comparisons between the coseismic electric arising from the motional induction (MI) effect and the
electrokinetic (EK) effect under different conductivities and frequencies. EMI
p (red solid) and Es (red dotted) are the
MI

electric fields accompanying the P and S waves, respectively, due to the motional induction effect. EEK p (blue solid) and Es
EK

(blue dotted) are the ones due to the electrokinetic effect. (e–h) Comparisons between the magnetic fields arising from MI
p (red solid) and Hs (red dotted) are the magnetic
effect and the EK effect under different conductivities and frequencies. HMI MI

fields accompanying the P and S waves due to the motional induction effect. HEK s (blue dotted) is the magnetic field
accompanying the S wave due to the electrokinetic effect. There is no curve of HEKp because the P wave cannot generate any
magnetic field due to the electrokinetic effect [Pride and Haartsen, 1996; Gao and Hu, 2010].

dominates the S-to-magnetic conversion at low conductivities, while HMI s dominates at high conductivities.
The transition conductivity is 0.013 S/m at f = 0.1 Hz (Figure 9e) while it is increased to 0.078 S/m and 0.33 S/m
when the frequency is augmented to 1 Hz (Figure 9f) and 10 Hz (Figure 9g), respectively. Figure 9h shows that
HEK
s is always larger than Hs at f = 100 Hz in the conductivity region σ0 ≤ 1.4 S/m.
MI

In Figure 10a the variations of EMI


p with the conductivity for different frequencies (f = 0.1 Hz, 1 Hz, 10 Hz, and
100 Hz) are compared. For a given frequency, EMI
p rises linearly with conductivity at low conductivities,

GAO ET AL. ©2014. American Geophysical Union. All Rights Reserved. 19


Journal of Geophysical Research: Solid Earth 10.1002/2014JB010962

Figure 10. Variations of the coseismic EM fields arising from the motional induction effect with the conductivity at four
different frequencies, namely, f = 0.1 Hz, 1 Hz, 10 Hz, and 100 Hz. Variations of (a) the electric field (EMI
p ) and (b) the
magnetic field (HMIp ) accompanying the P wave. Variations of (c) the electric field (Es ) and (d) the magnetic field (Hs )
MI MI

accompanying the S wave.

andthen it increases slowly in the high-conductivity range and seems to approach an upper limit. For
example, at f ¼ 0:1 Hz EMI
p rises linearly in the range σ0 < 0.01 S/m and slowly in the range σ0 ≥ 0.01 S/m to an
upper limit of 0.3 μV/m. EMI
p at f = 1 Hz has similar variation but with broader linear part (σ0 < 0.1 S/m) and
seems to approach a larger upper limit (about 3 μV/m). For the cases f = 10 Hz and f ¼ 100 Hz EMI p shows
linear dependence of conductivity over the range of conductivities considered in this article, i.e.,
1.13 × 10 4 S/m–1.4 S/m. At conductivities lower than 0.01 S/m the curves of EMI
p at four different frequencies
coincide with each other, indicating that in this conductivity range EMI
p is independent of the frequency.
From Figures 10b and 10d, we find that the coseismic magnetic fields that accompany the P and S waves, i.e.,
HMI
p and Hs , share similar properties with Ep . Such a dependence of the conductivity and the frequency for
MI MI

the coseismic fields EMIp , Hp , and Hs is similar to the dependence of conductivity for the magnetic field
MI MI

produced by a Rayleigh wave given by Yamazaki [2012] (see Figure 2 in his work). However, the electric field
accompanying the S wave EMI s is different, and for a given frequency, it nearly keeps constant over the range
of the conductivity shown in Figure 10c. If the frequency is augmented by a factor of 10, the value of EMI s is also
7
amplified by the same factor. For example, Es is approximately 3.2 × 10 V/m at f = 0.1 Hz, and it is amplified
MI

to 3.2 × 10 6 V/m, 3.2 × 10 5 V/m, and 3.2 × 10 4 V/m when the frequency is increased to f = 0.1 Hz,
f = 0.1 Hz, and f = 1 Hz, respectively.
The former paragraphs compared the coseismic EM fields arising from the motional induction effect and the
electrokinetic effect. We now investigate the EM waves generated by the moment tensor due to such two
mechanisms. In Figure 8 we find that the electric and magnetic fields of the EM wave caused by the motional
induction effect are on the orders of 0.0001 μV/m and 0.001 pT, respectively, while those caused by the
electrokinetic effect are on the orders of 0.01 μV/m and 0.1 pT, respectively, indicating that the electrokinetic
effect produces a stronger EM wave. This result is obtained under the condition σ0 = 7. × 10 4 S/m. Note that
increase in the conductivity will enhance the mechanoelectric conversion efficiency of the motional
induction effect but weaken that of electrokinetic effect. It is necessary to compare the EM waves caused by
those two mechanisms under other conductivity conditions.
The motional induction EM wave corresponds to the terms with amplitudes Aem1 and Aem2 in the expressions
of the electric and magnetic fields in equations (24) and (36). As is discussed in section 3.3 and illustrated in

GAO ET AL. ©2014. American Geophysical Union. All Rights Reserved. 20


Journal of Geophysical Research: Solid Earth 10.1002/2014JB010962

Figure 7, |Aem1| is always larger than |Aem2| especially for low conductivities and high frequencies, and the EM
wave is mainly controlled by the terms with the amplitude Aem1. Therefore, the electric and magnetic fields of
the EM wave in equations (24) and (36) can be approximately rewritten as
 
iωsem iωsem r 1 eiωsem r 2 eiωsem r i 1 3
em ≈ Aem1 
EMI e Eem1 þ E þ  E ; (58)
4πr 4πr 2 em1
4π ωsem r 3 ω2 s2em r 4 em1

2 2 iωsem r  
1 ω sem e 1 eiωsem r iωsem 1 2
HMI
em ≈ Aem1 Hem1 þ  2 þ 3 Hem1 : (59)
iωμ 4πr 4π r r

According to equations (48) and (49), the electric and magnetic fields of the EM wave resulting from the
electrokinetic effect are
 
iωsem ei ωsr EK;1 eiωsem r EK;2 eiωsem r i 1 EK;3
EEK ¼ T F
 E þ E þ  E ; (60)
em E;sem
4πr em
4πr 2 em 4π ωsem r 3 ω2 s2em r 4 em
2 2 iωsem r  
1 F ω sem e EK;1 eiωsem r iωsem 1 EK;2
HEK
em ¼ T E;sem Hem þ  2 þ 3 Hem ; (61)
iωμ 4πr 4π r r

where

EK;1  
Eem ¼ M^r  ^r T M ^r ^r; (62)

EK;2  
Eem ¼ trðMÞ ^r þ MT ^r þ 2M ^r  6 ^r T M ^r ^r ; (63)

EK;3
Eem ¼ 3trðMÞ^r þ 3MT ^r þ 3M^r  15 ^r T M ^rÞ^r ; (64)

EK;1
Hem ¼ ^rðM ^r Þ; (65)

EK;2
Hem ¼ ðIMÞ  3 ^rðM ^r Þ: (66)

By comparing equations (58) with (60), we find that EMI


em and Eem hold similar expressions although they have
EK

1
different amplitudes in front of their respective brackets and vectors inside. Let us investigate the vector Eem1
EK;1
and its counterpart Eem . By comparing their expressions in equations (25) and (62), respectively, we can
   
 1   EK;1  1 EK;1
approximately have Eem1 ≈Eem  because the amplitudes of Eem1 and Eem should be on the same order or at
        
 2   EK;2   3   EK;3   1 
least they should be closed to each other. Similarly, we can further have Eem1 ≈Eem , Eem1 ≈Eem , Hem1 
     
 EK;1   2   EK;2 
≈Hem , and Hem1 ≈Hem . As a result, dividing equations (58) by (60) and (59) by (61), respectively, we obtain

 MI   MI 
E  H  jAem1 j
 em   em   
EEK  ≈ HEK  ≈  F  : (67)
em em T  E;sem

The ratio between the strengths of the EM waves arising from the two mechanisms is proportional to and
     EK   MI   EK 
 
represented by jAem1 j=T FE;sem . We then plot the variations of EMI       
em = Eem and Hem = Hem with the
 MI   EK 
conductivity in Figure 11. For f = 0.1 Hz, Eem =Eem  < 1 when σ0 < 0.013 S/m, indicating that at low
conductivities the electrokinetic effect produces stronger EM wave than the motional induction effect. If the
frequency is increased to f = 1 Hz and f = 10 Hz, respectively, the dominant range for the electrokinetic effect
is extended to σ0 < 0.08 S/m and σ0 < 0.78 S/m, respectively. If the frequency is augmented to as high as
100 Hz, the electrokinetic effect is always dominant over the range of conductivity considered here. It is
explicitly seen from Figure 11 that under low-conductivity and high-frequency conditions the electrokinetic
effect is the primary mechanism for the generation of the EM wave.

GAO ET AL. ©2014. American Geophysical Union. All Rights Reserved. 21


Journal of Geophysical Research: Solid Earth 10.1002/2014JB010962

5. Discussions and Conclusions


The coupling between the elastic and EM
wavefields arising from the motional
induction effect was studied in this article.
We derived the analytical expressions of the
displacement, the electric field, and the
magnetic field generated by a point force
source and a moment tensor source. Two
types of EM variations can be produced due
to such a motional induction effect. One is
the coseismic EM field accompanying the
seismic wave. The other is an independently
propagating EM wave. With the theoretical
   EK      formulations derived in section 2, we made a
Figure 11. Variations of the ratios of EMI
em
=E  and HMI =HEK  simple estimation of the coseismic EM
em em em
with the conductivity at four different frequencies. signals that were observed in 2008 Mw6.1
Qingchuan earthquake, China. The numerical
results showed that the modeled coseismic electric and magnetic signals can reach the amplitudes of the
observed data, that is, on the orders of 1 μV/m and 0.1 nT, respectively, suggesting that the motional induction
effect is efficient enough to generate observable EM coseismic signals. However, we cannot confirm that these
observed EM signals is caused by the motional induction effect. On the one hand, the agreement between the
amplitudes of the simulated and observed EM signals is obtained by assuming the conductivity to be 0.1 S/m,
but we do not know the exact conductivity near the observation site medium as well as the other parameters.
Therefore, we cannot exclude other mechanisms such as the electrokinetic effect. One the other hand, this
simulation is very rough since only a full-space geological model is considered here. Theoretical signals contain
only the body waves, i.e., the EM, P, and S waves, and they are far behind the observed data which contain
multiple seismic phases resulting from reflections and refractions. If we want to simulate EM signals that
can fit the data better, we should take into account the effect of the boundaries in the real stratum such
as the free surface and the interfaces. This will require a suitable algorithm that allows calculating the
seismoelectromagnetic field in a complex geological model. In addition, the correct parameters of the strata as
well as the source time function that can be obtained via seismic inversion are also needed for the calculation of
the seismoelectromagnetic field.
We also made simulations of the Lorentz fields that were observed in the field experiment conducted by
Haines et al. [2007]. While the electrokinetic effect cannot explain such Lorentz fields, the EM field generated
due to the motional induction effect fits the observed data well (see Figure 5d). This verifies the conclusion
given by Haines et al. [2007]; that is, the motion of the metal hammer plate in the Earth’s magnetic field is
responsible for the Lorentz field.
The expressions of the electric and magnetic fields generated by a point force or a moment tensor source
are somewhat complicated (see equations (19), (20), (24), and (36)). However, we can make some
approximation to the EM wave component. As is discussed in section 3.2 and illustrated in Figure 7, |Aem1| is
always larger than |Aem2|, especially for high frequencies. For example, when the frequency is higher than
100 Hz, |Aem1| / |Aem2| is always larger than 30 in the conductivity band 10 4 S/m–1.4 S/m (black dashed line
in Figure 7). In this situation, terms with the amplitude Aem1 dominate the EM wave component. As a result,
the EM wave generated by a single-point force F is approximately equivalent to an electric dipole (see
^ 0 F rather than the
equation (47)). However, the orientation of the equivalent dipole is along the direction B
force direction. For the excitation of a moment tensor source, we also made approximation to the expressions
of the EM wave component. The electric and magnetic fields are expressed in equations (58) and (59), which
are similar to those arising from the electrokinetic effect as shown in equations (60) and (61).
Comparison between the motional induction effect and the electrokinetic effect was made in the present study.
The electric and magnetic fields generated by an earthquake due to both of the two mechanisms were
calculated, and their waveforms are compared in Figure 8, from which we can see the EM responses that arising
from the two mechanisms have some similarities and dissimilarities. The similarities are embodied in that both

GAO ET AL. ©2014. American Geophysical Union. All Rights Reserved. 22


Journal of Geophysical Research: Solid Earth 10.1002/2014JB010962

of the two mechanisms can result in two kinds of EM signals, i.e., the independently propagating EM wave and
the coseismic EM fields accompanying the P and S waves, which do not simply proportional to the amplitudes
of the seismic waves but depend on the type of the waves. The dissimilarities are that both the EM wave and
the coseismic EM fields arising from the motional induction effect generally differ from their electrokinetic
counterparts in the amplitudes and the dependence on the frequency as well as the conductivity. For the
coseismic EM fields, the transfer functions E/u and H/u corresponding to the P and S waves were derived and
compared with their electrokinetic counterparts. The electric field accompanying the P wave arising from the
motional induction effect EMI
p increases when the conductivity increases, while its electrokinetic counterpart Ep
EK

decreases. As shown in Figures 9a–9d, for low frequencies (f < 100 Hz) the motional induction effect dominates the
P-to-electric conversion at high conductivities. For example, at f ¼ 0:1 Hz EMI
p is larger than Ep when the
EK

conductivity is higher than the transition point 0.016 S/m. The transition point of the conductivity will shift to
0.12 S/m and 0.86 S/m when the frequency is increased to f = 1 Hz and f = 10 Hz, respectively. At frequencies higher
than 100 Hz the electrokinetic effect always dominates in the conductivity range considered in this study
(10 4 S/m–1.4 S/m). The relation between the coseismic magnetic fields accompanying the S wave arising
from the motional induction effect HMI EK
s and its electrokinetic counterpartHs (Figures 9e–9h) is similar to the relation
between Ep and Ep (Figures 9a–9d). However, the relation between the electric field accompanying the S wave
MI EK

arising from the motional induction EMI


s and its counterpart Es is different. For a given frequency Es decreases with
EK EK

the increase of the conductivity, while Es almost keeps unchanged over the range of the conductivity (see
MI

Figures 9a–9d). In addition, EMI


s is always larger than Es under the same condition, indicating that the motional
EK

induction effect is always the primary mechanism in converting the S wave energy to the electric field.
One phenomenon we noticed in Figure 2 that the real data observed during the 2008 Qingchuan earthquake
shows that the P wave-generated coseismic magnetic field when it arrived at the receiver. The electrokinetic
effect seems to have certain obstacle to interpret the coseismic magnetic signals starting with the P arrival,
because it theoretically predicts no magnetic field accompanying the body P wave [Pride and Haartsen, 1996].
Consider that the instruments receiving the seismic and EM signals are usually located close to the free surface,
where the P wave can convert a reflected S wave that can generate significant magnetic field due to the
electrokinetic effect. However, the synthetic magnetic field synchronizing with the P wave has only horizontal
components and no vertical magnetic field can be generated before the S wave arrives [Ren et al., 2012]. This is
because only the SH wave can give rise to the vertical magnetic field while the SV wave only generates
horizontal one [Pride and Haartsen, 1996]. Since in a horizontally layered media the P wave is only coupled with
the SV wave, no vertical magnetic field can be generated unless the SH wave arrives. Nevertheless, we notice
that this conclusion is obtained under the condition that we use a horizontally layered geological model and
assume media are homogeneous and isotropic. Perhaps a complex geological model with heterogeneous or
anisotropic media could result in vertical magnetic field when the P wave arrives. In addition, when the seismic
waves passed the instrument it caused the shaking of the magnetometer, which might give rise to a changing
of the magnetic signal, especially in the vertical component. This needs to be examined in the future studies.
In contrast to the electrokinetic effect, the motional induction effect predicts a significant magnetic field
accompanying the P wave as shown in Figures 3 and 8. The MI effect seems to have no obstacle to interpret
the vertical coseismic magnetic field starting with the P arrival since the P wave can directly generate a
magnetic field with both horizontal and vertical components. As far as this property is concerned, the MI
effect is more likely to be the mechanism for the real EM data than the EK effect if we can confirm that the
vertical magnetic field was not caused by the shaking the of the magnetometer. However, it is still a difficulty
to exclude the electrokinetic effect at present.
Another phenomenon we noticed is that the coseismic EM fields arising from the motional induction effect
is not strictly proportional to the conductivity (Figure 10) although the coupling term σ0v × B0 in equation
(2) implies a direct proportion between the EM fields and the conductivity. In the low range of the
conductivity, EMIp , Hp , and Hs rise linearly with conductivity and do not depend on the frequency. In the
MI MI

high-conductivity range, they increase slowly and seem to approach their respective upper limits. An
increase in the frequency will extend their linear-rising part toward to high conductivities and enhance
their upper limits. This agrees with result obtained by Yamazaki [2012] about the dependence of conductivity
for the magnetic field produced by a Rayleigh wave.

GAO ET AL. ©2014. American Geophysical Union. All Rights Reserved. 23


Journal of Geophysical Research: Solid Earth 10.1002/2014JB010962

The EM waves generated by a fault slip (described by a moment tensor source) due to the two mechanisms
were also compared. The result shows that the motional induction effect can generate a stronger EM wave
than the electrokinetic effect under high-conductivity and low-frequency conditions (Figure 11). Note that
high conductivity gives rise to high attenuation of the EM wave. For example, under the condition that
f = 0.1 Hz and σ0 = 0.1 S/m, the penetration depth of the EM wave is only
sffiffiffiffiffiffiffiffiffiffiffi sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 2
δ¼ ¼ ¼ 5:03 km: (68)
ωμσ0 ð2π0:1 Hz4π107 H=m0:1 S=m

In this situation, the EM generated by the source can only propagate for a very short distance. Although the
EM wave generated due to the motional induction effect is 10 times stronger than that due to the
electrokinetic effect (red dash-dotted line in Figure 11), it may become undetectable at a distant observation
site except that the earthquake is sufficiently large. Nevertheless, there is a possibility that the medium in the
vicinity of the earthquake source is more conductive while that out of the source region is less conductive so
that the EM generated by the source can have strong amplitude and propagate far with a low attenuation out
of the source region. Under this condition, it is reasonable to taken the motional induction effect as the
preferred candidate mechanism for the observed EM waves (e.g., the direct magnetic signals observed by
Okubo et al. [2011]). Otherwise, if the source region is known to have a low conductivity, the electrokinetic
effect should be considered prior to the motional induction effect.

Appendix A: Derivation of Equations (17) and (18)

A1. Displacement
To get the frequency-space response of the displacement, we apply an inverse Fourier transform to equation (12)

∫ueðkÞe
1
uðrÞ ¼ ikr
dk
ð2πÞ3
82 3 9
<   =
1 4  1 1 1
kk5F0 e

ikr
¼  I  2 2 kk þ  eikr dk
ð2πÞ3 :
∫ G k  ω ss
2 2 2 ω ss H k  ω sp sp ω
2 2 2 2 2 ;
h1 1
∫ ∫
1 ikðrr′ Þ 1 1 1
 kkeikðrr Þ dk

¼  e dk I  
G ð2πÞ3 k 2  ω2 s2s Gω2 s2s ð2πÞ3 k 2  ω2 s2s
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
Iu1 Iu2
i

1 1 1
kk eik  ðrr Þ dk  F0 :

þ 
Hω2 s2p ð2πÞ3 k 2  ω2 s2 (A1)
p
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
Iu3

We now solve the first integration in equation (A1). Let r = |r  r′| and ^r ¼ rr
r ; Iu1 can be rewritten as

∫ k
1 1 ⌢

Iu1 ¼  eikr cos θ dk; (A2)


ð2πÞ 3 2
 ω2 s2s

where θ⌢ denotes the angel between k and ^r . Equation (A2) can be solved in a spherical coordinates (r; ⌢
θ ; ϕ):

∞ π 2π

∫ ∫ ∫ k
1 1 ⌢
Iu1 ¼  eikr cos θ k 2 sin ⌢ ⌢ ⌢
θ dkd θ dϕ
ð2πÞ 3
00 0
2
 ω ss
2 2


k2 4π sinðkr Þ
∫
1
¼  dk
ð2πÞ k 3
0
2
 ω ss
2 2 kr

k sinðkr Þ

1
¼   dk
2π2 r 0 k 2  ω2 s2s
1 iωss r
¼ e : (A3)
4πr

GAO ET AL. ©2014. American Geophysical Union. All Rights Reserved. 24


Journal of Geophysical Research: Solid Earth 10.1002/2014JB010962

The final result in equation (A3) was obtained by using the Cauchy’s residual theorem to solve the integration
in the third step. Note that kk in equation (A1) actually correspond to the spatial derivatives. Therefore,
Iu2 can be rewritten as
iωss eiωss r eiωss r ω2 s2s eiωss r
Iu2 ¼ ∇∇Iu1 ¼  2
ðI  3^r^r Þ þ 3 ðI  3^r^r Þ þ ^r^r : (A4)
r r r
The third integration in equation (A1) can be achieved by replacing ss by sp in equation (A4):
iωsp eiωsp r eiωsp r ω2 s2p eiωsp r
Iu3 ¼  ðI  3^r ^
r Þ þ ðI  3^r ^
r Þ þ ^r^r : (A5)
r2 r3 r
With Iu1, Iu2, and Iu3 being determined, we get the displacement in the frequency-space domain:
uðrÞ ¼ GFu ðrj^r ÞF0
n 1 eiωss r 1 eiωsp r
¼ ðI  ^r^r Þ þ ^r^r
G 4πr H 4πr
"   iωss r ! #
1 i 1 e 1 i 1 eiωsp r o
þ  2 2 2   2 2 2 ðI  3^r^r Þ F0 ; (A6)
G ωss r ω ss r 4πr H ωsp r ω sp r 4πr
where GFu ðrj^r Þ corresponds to the terms inside the brace and represents the Green’s function of the displacement.

A2. Electric Field


The electric field in equation (16) can be also transformed into the frequency-space domain by the inverse
Fourier transform


1
e
EðrÞ ¼ EðkÞeikr dk
ð2πÞ3
nh

1 1 1 ′
¼ B0     eikr eikr dkI
ð2πÞ3 G k 2  ω2 s2em k 2  ω2 s2S
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
Ie1


1 1 1 ′
 B0   2  2  kkeikr eikr dk
ð2πÞ 3
Gs 2
s ω 2 k  ω 2 s 2
em k  ω 2 s 2
s
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl
ffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl
ffl}
Ie2


1 1 1 ′
þ B0    kkeikr eikr dk
ð2πÞ3 Hs2p ω2 k 2  ω2 s2em k 2  ω2 s2p
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
Ie3
i o 

1 1 1 1 ikr ikr′
 3 ω 2 s2 G
    kke e dk B0 F0 ω2 μσ0 : (A7)
ð2πÞ em k  ω sem k  ω ss
2 2 2 2 2 2
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl
ffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
Ie4
One can solve the four integrals in equation (A7) in the same way with the derivations of Iu1 and Iu2 and can
easily get that
" #
1 eiωss r 1 eiωsem r
Ie1 ¼  2 2  þ  2 2  I; (A8)
G ω ss  ω2 s2em 4πr G ω sem  ω2 s2s 4πr
  iωsem r
s2em 1 i 1 e 1 eiωsem r
Ie2 ¼   ð I  3 ^
r ^
r Þ  ^
r ^
r
ρ ω2 s2em  ω2 s2s ωsem r 2 ω2 s2em r 3 4π r 4π
  iωss r
1 1 i 1 e 1 eiωss r
  ðI  3^r ^
r Þ  ^
r ^
r ; (A9)
G ω2 s2s  ω2 s2em ωss r 2 ω2 s2s r 3 4π r 4π
  iωsem r
s2 1 i 1 e 1 eiωsem r
Ie3 ¼  em 2 2  ð I  3^r ^
r Þ  ^
r ^
r
ρ ω sem  ω2 s2p ωsem r 2 ω2 s2em r 3 4π r 4π
" ! #
1 1 i 1 eiωsp r 1eiωsp r
  ðI  3^r^r Þ  ^r^r ; (A10)
H ω2 s2p  ω2 s2em ωsp r 2 ω2 s2p r 3 4π r 4π

GAO ET AL. ©2014. American Geophysical Union. All Rights Reserved. 25


Journal of Geophysical Research: Solid Earth 10.1002/2014JB010962

  iωsem r
1 1 i 1 e 1eiωsem r
Ie4 ¼   ð I  3 ^
r ^
r Þ  ^
r ^
r
G ω2 s2em  ω2 s2s ωsem r ω2 s2em r 3 4π
2 r 4π
  iωss r
s2s 1 1 i 1 e 1eiωss r
  ðI  3^
r ^r Þ  ^
r ^
r : (A11)
s2em G ω2 s2s  ω2 s2em ωss r 2 ω2 s2s r 3 4π r 4π

Substituting Ie1, Ie2, Ie3, and Ie4 into equation (A7), we finally obtain the electric field

EðrÞ ¼ GFE ðrj^r ÞF0


n   iωsem r
eiωsem r  ^ 
^ 0 þ Aem1 i 1 e 
^0
^ 0  3^r^r B

¼ Aem1 B 0 I  ^r^r B  IB
4πr ωsem r 2 ω sem r
2 2 3 4π
  iωsem r
e iωsem r
i 1 e  
 Aem2 B^ 0 ^r^r þ Aem2  2 2 3 ^ 0 I  3B
B ^ 0 ^r^r
4πr ωsem r 2 ω sem r 4π
  iωss r
eiωss r i 1 e  
 As1 ^r^r B^ 0 þ As1  IB^ 0  3^r^r B ^0
4πr ωss r 2 ω2 s2s r 3 4π
  iωss r
eiωss r  ^ ^
 i 1 e 
^ 0 I  3B ^ 0 ^r^r

þ As2 ^^
B 0 I  B 0 r r þ As2  B
4πr ωss r 2 ω2 s2s r 3 4π
!
eiωsp r ^ i 1 eiωsp r  ^ 
^ 0 ^r^r gF0 ;
þ Ap B 0 ^r^r  Ap  B 0 I  3B (A12)
4πr ωsp r 2 ω sp r
2 2 3 4π

where B^ 0 ¼ B0 =B0 and B0 = |B0| are the direction vector and amplitude of the ambient Earth’s magnetic field.
Aem1, Aem2, As1, As2, and Ap are the amplitude constants and expressed as follows:

1
Aem1 ¼ ðμσ0 B0 Þ  2 ; (A13)
G sem  s2s
!
1 s2em 1 1
Aem2 ¼ ðμσ0 B0 Þ  ; (A14)
G s2s s2em  s2s s2em  s2p

s2s 1 1
As1 ¼ ðμσ0 B0 Þ ; (A15)
s2em G s2s  s2em
1
As2 ¼ ðμσ0 B0 Þ  2 ; (A16)
G ss  s2em
1 1
Ap ¼ ðμσ 0 B0 Þ : (A17)
H s2p  s2em

In equation (A12), GFE ðrj^r Þ is the Green’s function of the electric field.

Appendix B: Pride’s Equations and Parameters Used in the Text

B1. Pride’s Equations and Complex Amplitudes in Equation (48)


Assuming the time dependence being e iωt, Pride’s equations [Pride, 1994] in an isotropic homogeneous
porous medium can be written as
 
∇H ¼ ½σ0  iωεE þ LðωÞ ∇P þ ω2 ρf u ; (B1)
∇E ¼ iωμH; (B2)
ω2 ðρu þ ρf wÞ ¼ ∇τ þ F; (B3)
 
iωw ¼ L0 E þ κ0 =η ∇P þ ω2 ρf u þ f ; (B4)
    
τ ¼ H  2G ∇u þ C∇w I þ G ∇u þ ∇uT ; (B5)
P ¼ C∇u þ M∇w; (B6)
where H is the magnetic field; E is the electric field; u is the average solid displacement; w is the average
relative fluid-solid displacement; τ is the bulk stress tensor; P is the pore fluid pressure; I is the identity tensor;

GAO ET AL. ©2014. American Geophysical Union. All Rights Reserved. 26


Journal of Geophysical Research: Solid Earth 10.1002/2014JB010962

μ is the magnetic permeability of the porous medium; ρ = (1  ϕ)ρs + ϕρf is the bulk density; ρs is the solid
grain density; ρf is the pore fluid density; ϕ is the porosity; η is the fluid viscosity; ε = ε0[(εr  εs)ϕ/α∞ + εs] is the
dielectric permittivity, where εf and εs are the relative permittivities of the solid and fluid phases, and ε0 is the
vacuum permittivity; and σ0, κ0, and L0 are the bulk conductivity, the permeability, and the electrokinetic
coupling coefficient, respectively. These three variables are actually frequency dependent. However, they
vary so little with frequency in the seismic wave frequency band (less than several hundred hertz) that they
are chosen as static values in this study. The expressions of σ0 and L0 are given in sections B6 and B7,
respectively. F and f are the average force densities exerted on the bulk material and fluid phase, respectively.
H, C, and M are the elastic moduli which are expressed as

H ¼ K b þ 4G=3 þ α2 M; (B7)

C ¼ αM; (B8)

M ¼ K f K s =½ϕK s þ ðα  ϕ ÞK f ; (B9)
where
α ¼ 1  K b =K s : (B10)
Ks and Kf are the bulk moduli of the solid and fluid phases, respectively; Kb and G are the bulk and shear
moduli of the framework, respectively.
According to the plane-wave solution given by Pride and Haartsen [1996], there are four body waves involved
in Pride’s equations, namely, the fast P wave (Pf), the slow P wave (Ps), the S wave (S), and the EM wave. The
fast P wave is actually the P wave in elastodynamics. The slownesses of the Pf and Ps waves are expressed as
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 
4eρρ ρt
2s2pf ;ps ¼ γ∓ γ  þ e
ρ L 2
e
ε ; (B11)
HM  C 2 ρ

where the “” corresponds to the Pf wave slowness spf while the “+” corresponds to the Ps wave sps. γ and ρt
are expressed as
 
ρM þ e
ρH 1 þ e ρL2 =eε  2ρf C
γ¼ ; (B12)
HM  C 2
ρt ¼ ρ  ρ2f =e
ρ: (B13)
The slownesses of the S and EM waves are
  (   2 )1=2
ρt eρL2 ρt e
ρL2 ρ2f L2
2ss;em ¼ þ μeε 1 þ
2
±  μeε 1 þ  4μ ; (B14)
G eε G eε G

where the “+” in front of radical corresponds to the S wave while the “” corresponds to the EM wave. The
complex amplitudes in the expressions of the seismoelectromagnetic wavefields in equations (48) and (49)
have been derived by Pride and Haartsen [1996]. They are
  2  
M spf  e
ρ 1þe ρL2 =eε M
LFu;spf ¼  ; (B15)
HM  C 2 s2  s2 pf ps
   
M s2ps  e
ρ 1þe ρL2 =eε M
LFu;sps ¼  ; (B16)
HM  C 2 s2ps  s2pf
 
s2  μeε 1 þ e ρL2 =eε
T Fu;ss ¼ s  2  ; (B17)
G ss  s2em
 
s2em  μeε 1 þ e ρL2 =eε
T u;sem ¼
F
  ; (B18)
G s2em  s2s
  2
e
ρL M s  ρf =C
LFE;spf ¼ iωLCu;spf ¼ iω pf ; (B19)
eε HM  C 2
s2pf  s2ps

GAO ET AL. ©2014. American Geophysical Union. All Rights Reserved. 27


Journal of Geophysical Research: Solid Earth 10.1002/2014JB010962

  2
e
ρL M s  ρf =C
LFE;sps ¼ iωLCu;sps ¼ iω ps ; (B20)
eε HM  C 2 s2  s2
ps pf

iωμρ L
T FE;ss ¼ iωT Cu;ss ¼  2 f 2  ; (B21)
G ss  sem
iωμρ L
T FE;sem ¼ iωT Cu;sem ¼  2 f 2  : (B22)
G sem  ss

B2. Pore Fluid Conductivity


In this study, we use Sen and Goode [1992] equation to calculate the electrical conductivity of the pore
fluid solution
 
  2:36 þ 0:099 T c 3=2
σf ðT c ; C f Þ ¼ 5:6 þ 0:27T c  1:51104 T 2c C f  pffiffiffiffiffi C f ; (B23)
1 þ 0:214 C f
where Tc = T  273.15 is the Celsius temperature and its unit is oC. T is the absolute temperature in kelvin. Cf is
the salinity of the pore fluid. With the temperature being fixed to Tc = 25∘C (298 K), the pore fluid conductivity
σf varies from 1.224 × 10 4 S/m to 24.72 S/m when Cf changes from 10 5 mol/L to 5 mol/L.

B3. Pore Fluid Electric Permittivity


The permittivity of the pore fluid εr varies with the temperature and the salinity. Here we use Olhoeft’s
empirical equation [Olhoeft, 1980; Revil et al., 1999; Glover et al., 2012] to determine it:

εr ðT; C f Þ ¼ 295:68  1:2283T þ 2:094103 T 2  1:41106 T 3


 13:00 C f þ 1:065 C 2f  0:03006 C 3f : (B24)

Equation (B24) is valid in the temperature range from 273 K to 373 K. With temperature being 298 K the relative
permittivity calculated from equation (B24) varies from 78.24 at Cf = 10 5 mol/L to 36.1 at Cf = 5 mol/L.

B4. Pore Fluid Dynamic Viscosity


The dynamic viscosity of the pore fluid is calculated using Phillips et al. [1978] equation

ηðT c ; ; C f Þ ¼ 4:95166105 þ 6:034658104 expð0:06653081 T c Þ


 
þ 9:703832105 exp 0:1447269 C m
f
 
þ 1:025107103 exp 0:02062455 T c þ 0:1301095C mf : (B25)

In equation (B25) Tc is also the Celsius temperature. C m f is the molality of the pore fluid solution and can be
calculated by C mf ¼ C f =ðρf  AC f =1000Þ, where ρf is the density of the fluid and A is the atomic mass of the
salt in g/mol (in the present study we choose A = 58.44 g/mol for NaCl). With a constant temperature Tc = 25∘C,
the viscosity varies from 1.9 Pa s to 3 Pa s when the salinity changes from 10 5 mol/L to 5 mol/L.

B5. Zeta Potential


The zeta potential ζ is the most important parameter that affects the coupling coefficient L0 significantly. It
cannot be directly measured in the experiment currently but is actually derived from the streaming potential
coefficient Cs via the Helmholtz-Smoluchowski equation [Overbeek, 1952; Hunter, 1981; Morgan et al., 1989;
Wang et al., 2011]

ΔV εr ε0 ζ
Cs ¼ ¼ ; (B26)
ΔP ηðσf þ 2Σs =ΛÞ

where Cs is the ratio of the measured streaming potential ΔV to the applied fluid pressure difference ΔP, σf is
the pore fluid conductivity, Σs is the specific surface conductance, and Λ is a length-scale characteristic of the
pore microstructure. 2Σs/Λ represents the electrical conductivity from surface conduction.

GAO ET AL. ©2014. American Geophysical Union. All Rights Reserved. 28


Journal of Geophysical Research: Solid Earth 10.1002/2014JB010962

Figure B1. (a) Modeled zeta potential as a function of the pore fluid salinity for three pH values with an offset ζ 0 = 0.035 V.
The scattered solid squares represent the experiment data obtained from sandstones by Jaafar et al. [2009]. (b) Modeled
bulk conductivity σ0 of the porous medium with (solid line) and without (dashed line) surface conductance. (c) Variation of
the coupling coefficient L0 with the pore fluid salinity.

Theoretically, calculating the zeta potential is also available. Revil et al. [1999] suggested one model to
calculate the zeta potential [Revil and Glover, 1997, 1998]:
 
χ
ζ ≈ φd exp  ζ ; (B27)
χd
where φd is the Stern potential, χ ζ is the distance from the shear plane to the Stern plane, and χ d is the Debye
Screen length. The Stern potential is given as
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi  !
2k b T 8103 εr ε0 k b TN 10pH þ K Me C f C f þ C a þ C b þ 10pH þ 10pHpK w
φd ¼ ln p ffiffiffi : (B28)
3e 2eΓ0s K ðÞ If

Here kb is the Boltzmann’s constant, T is the absolute temperature, e is the charge on an electron, N is the
Avogadro’s constant, KMe is the binding constant for cation adsorption on quartz, Γ0s is the surface site density,
K() is the disassociation constant for dehydrogenization of silanol surface sites, Ca is the acid concentration,
and Cb is the base concentration. pKw =  log(Kw), where Kw is the disassociation constant of water. If is the
ionic strength. In the present study, we assume that pH = 6 and 10 5 mol/L ≤ Cf ≤ 5 mol/L. Under this
condition, the contributions to the ionic strength from the acid and base are negligible compared with the
dissolved salt (NaCl), so that we can have If ≈ Cf [Revil et al., 1999; Glover et al., 2012]. All the symbols in
equation (B28) are described and listed in Table B1.
The Debye screening length is expressed as
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ε0 εr k b T
χd ¼ : (B29)
2000Ne2 If

The distance from the shear plane to the Stern layer is chosen as χ ζ = 2.4 × 10 10 m, as used by Revil
et al. [1999].

GAO ET AL. ©2014. American Geophysical Union. All Rights Reserved. 29


Journal of Geophysical Research: Solid Earth 10.1002/2014JB010962

Table B1. Parameters Used for Calculating the Zeta Potential and the Surface Conductance
Parameter Symbol Value or Range (Unit) Source
 10
Distance of the shear plane from the Stern plane χζ 2.4 × 10 (m) Revil et al. [1999]
Debye Screen length χd  55 ~ 51.6 (mV) From equation ((A5))
Temperature T 298.15 (K)
 23
Boltzmann’s constant kb 1.38 × 10 (J/K)
 19
Charge on an electron e 1.602 × 10 (C)
23 1
Avogadro’s constant N 6.022 × 10 (mol )
 12
Dielectric permittivity in vacuum ε0 8.854 × 10 (F/m)
pH pH 6 Glover et al. [2012]
 7.5
Binding constant for cation (sodium) adsorption on quartz KMe 10 Glover et al. [2012]
 8.5
Disassociation constant for dehydrogenization of silanol surface sites K() 10 Glover et al. [2012]
10 -2
Surface site density Γ0s 10 (m ) Glover et al. [2012]
5
Salinity of the pore fluid Cf 10 ~ 5 (mol/L)
Ionic strength If If ≈ Cf
 15
Disassociation constant of water Kw 9.22 × 10 Glover et al. [2012]
Acid concentration Ca 0
Base concentration Cb 0
8 2 -1 -1
Ionic mobility of sodium ions βNaþ 5.20 × 10 (m s V ) Crow [1988]
7 2 -1 -1
Ionic mobility of Hydrogen ions β Hþ 3.63 × 10 (m s V ) Crow [1988]
8 2 -1 -1
Ionic mobility of chloride ions βCl 7.90 × 10 (m s V ) Crow [1988]
7 2 -1 -1
Ionic mobility of hydroxyl ions βOH- 2.05 × 10 (m s V ) Crow [1988]
9 2 -1 -1
Ionic surface mobility βs 5 × 10 (m s V ) Revil et al. [1998]
Mean grain diameter d 130 (μm) Glover et al. [2012]

It should be noted that zeta potential ζ calculated from equation (B27) has a negative value at low salinities. It
increases (become less negative) with the increase in the salinity and reaches a positive value at a high
salinity [Glover et al., 2012]. Using the parameters show in Table B1, we find that if Cf > 0.6 mol/L, ζ becomes
positive. However, positive values of ζ have not been observed in actual experiments. Experiment data
[Jaafar et al., 2009; Vinogradov et al., 2010] show that when the salinity increases ζ becomes either constant or
slightly more negative after it reaches its maximum. Glover et al. [2012] introduced a constant zeta potential
offset ζ 0 to equation (B27) and modified the zeta potential as
 
χ
ζ ≈ ζ 0 þ φd exp  ζ : (B30)
χd

Although such an additional constant has not been supported theoretically, it can improve the fit of model to the
experiment data. Figure B1a shows the modeled zeta potential versus salinity and pH with an offset ζ 0 = 0.035 V.
The parameters used for the calculation are listed in Table B1. The scattered solid squares denote the experiment
data obtained from sandstones by Jaafar et al. [2009]. We can see that the theoretical zeta potential curve
with pH = 6 (dashed line) fits the data very well. Therefore, in this study we choose the dashed line to relate the
zeta potential to the salinity, which gives ζ =  0.0899 V at Cf = 10 5 mol/L and ζ =  0.0311 V at Cf = 5 mol/L.

B6. Rock Bulk Conductivity


The bulk conductivity of the porous rock σ0 is also an important parameter that affects the seismoelectric
conversion efficiency. The conductivity arises not only from the pore fluid conductance but also from the
surface conductance. The latter makes an important contribution to the bulk conductivity, especially when
the salinity of the pore fluid is low. The bulk conductivity of the rock is expressed as [Pengra et al., 1999; Revil
and Glover, 1997; Block and Harris, 2006]
 
1 2Σs
σ0 ¼ σf þ ; (B31)
F Λ
where
α∞
F¼ (B32)
ϕ
is the formation factor and Λ is a length-scale characteristic of the pore microstructure. Several models
have been established to relate Λ to the mean grain diameter or to the pore diameter, such as the work by

GAO ET AL. ©2014. American Geophysical Union. All Rights Reserved. 30


Journal of Geophysical Research: Solid Earth 10.1002/2014JB010962

Johnson et al. [1987] and Revil and Cathles [1999]. In this study we choose the relationship given by Revil and
Cathles [1999], that is,
d
Λ¼ ; (B33)
3ðF  1Þ
where d is the mean grain diameter. The surface conductance Σs is given by Revil et al. [1999]
Σs ¼ ΣEDL
s þ ΣProt
s þ ΣStern
s : (B34)
In equation (B34) ΣEDL
s is the contribution from the diffuse layer and is given by Revil and Glover [1998] as
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffih
2000εr ε0 k b TN   
ΣEDL
s ¼ pH
BNaþ C f þ BHþ 10pH Θ1=3  1
C f þ 10
  i
þ BCl C f þ BOH- 10pHpK w Θ1=3  1 ; (B35)

where,
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 ffi 10pH þ K Me C f
Θ¼ 8000εr ε0 k b TN C a þ C b þ C f þ 10pH þ 10pHpK w ; (B36)
2eΓ0s K ðÞ

and Bi = βi + 2εrε0kbT/(ηeZi) is the equivalent ionic mobilities for ion “i.” βi is the ionic mobility of ions “i”
(see Table B1) and Zi is its valence.

ΣProt
s is the proton contribution to the surface conductance, and here we choose the value given by Watillon
and De Backer [1970], that is, ΣProt
s ¼ 2:4109 S.

ΣStern
s is the conductance from the Stern layer

eβs Γ0s K Me C f
ΣStern
s ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2=3 ; (B37)
pH 810 εr ε0 k b TN 
3
10 þ K ðÞ 2eΓ0s K ðÞ
C þ K Me C f

where
  C a þ C b þ C f þ 10pH þ 10pHpK w
C  ¼ 10pH þ K Me C f pffiffiffi : (B38)
If

In equation (B37), βs is the ionic surface mobility, and in this study, we use the value βs = 5 × 10 9 (m2s- 1V- 1)
given by Revil et al. [1998]. With the parameters given in Table B1, the surface conductance calculated from
equation (B34) varies from Σs = 2.43 × 10 9 S at Cf = 10 5 mol/L to Σs = 13.2 × 10 9 S at Cf = 5 mol/L.
Using equations (B23), (B31), and (B33), we calculated the bulk conductivity of the rock σ0 with F = 17.7 and
d = 130 μm and plotted it in Figure B1b. We can find that the significant contribution from the surface
conductance to the bulk conductivity at low salinities. At Cf = 10 5 mol/L, the bulk conductivity calculated
without the surface conductance (dashed line) is only 0.69 × 10 5 S/m, 1 order of magnitude smaller than
that calculated with the surface conductance (solid line), that is, 11.3 × 10 5 S/m. In this article we use the
solid line in Figure B1b to control the conductivity σ0, which gives σ0 = 11.3 × 10 5 S/m at Cf = 10 5 mol/L
and 1.4 S/m at Cf = 5 mol/L.

B7. Electrokinetic Coupling Coefficient


The coupling coefficient L0 is the key parameter that determines the conversion between the mechanical and EM
energies since it links Biot’s equations with Maxwell’s equations and thus leads to the coupling equations, i.e.,
Pride’s equations. According to the work by Pride [1994], the static electrokinetic coupling coefficient is written as
ϕ εr ε0 ζ
L0 ¼  : (B39)
α∞ η
Equation (B39) is derived under the hypothesis that the pore is saturated with only one kind of fluid. In this
study, the porous medium is assumed to be a kind of sandstone which is saturated by water. With ϕ = 0.19,

GAO ET AL. ©2014. American Geophysical Union. All Rights Reserved. 31


Journal of Geophysical Research: Solid Earth 10.1002/2014JB010962

α∞ = 3.36, and other the parameters shown in Table B1 to determine εr, η, and ζ , we calculate the
electrokinetic coupling coefficient L0 and plot it in Figure B1c. The coupling coefficient L0 varies in the range
13.9 × 10 10 sC/kg–1.88 × 10 10 sC/kg when the fluid salinity varies between 10 5 mol/L and 5 mol/L.

Appendix C: Derivation of the Seismoelectric and Seismomagnetic


Transfer Functions
Here, from the viewpoint of the plane-wave theory, we derive the electric field E and magnetic field H as a
transfer function of the displacement u for the P and S waves. Assuming u and E generated by the plane P
and S waves have the following expressions:
^;
u ¼ jujeikr u (C1)

E ¼ jEjeikr e^ ; (C2)

where k ¼ k k^ ¼ ωsk^ is the wave vector with k^ denoting the direction of the wave propagation and s
denoting the slownesses of the P and S waves. |u| and |E| are amplitudes. u^ and e^ are unit vectors. By
substituting equations (C1) and (C2) into the following equation:

∇∇E  ∇2 E  ω2eεμE ¼ ω2 μσuB0 ; (C3)


one can obtain
  
k 2  ω2eεμ I  kk E ¼ ω2 μσuB0 ; (C4)

and further get


 
ω2 μσ kk
E¼ I  ðuB0 Þ: (C5)
k 2  ω2eεμ ω2eεμ

Substituting equations (C1) into (C5), we have



μσB0 2  
E¼ j u j ^
u B^ 0  s k^ k
^ u ^ B^0 ; (C6)
s2  eεμ eεμ
where B^ 0 ¼ B0 =B0 . For the P wave, the directions of the material response and propagation are parallel
 
(giving k^ ¼ u ^ u
^) so that we have k ^ 0 ¼ 0. As a result, the electric field produced by the P wave becomes
^ B
μσB0   ^ 
Ep ¼ ^ B 0 :
up u (C7)
s2p  eεμ

The term u ^ 0 in equation (C7) shows that the direction of Ep is perpendicular to both the directions of
^ B
the displacement u ^ and the geomagnetic field B ^ 0 and also indicates that the amplitude of Ep depends on the
sine of the angle between u ^
^ and B 0. If u ^ 0, Ep becomes zero because u
^ is parallel to B ^ B^ 0 ¼ 0 in this situation. If
 
^ is perpendicular to B
u ^ 0 , we have u ^ B ^ 0  ¼ 1 and then Ep achieves its maximum amplitude. Here assuming
 
u ^ 
^ B 0 ¼ 1 we approximately have the E/u transfer function for the P wave as
 
   μσB0  
Ep ≈  
s2  eεμ up : (C8)

The S wave produces a material vibration in the direction that is perpendicular to the direction of the
propagation. The electric field produced by the S wave can be written as

μσB0 2  
Es ¼ j u s j u^ B^ 0  ss k^ k
^ u ^  ^0 :
B (C9)
s2s  eεμ eεμ
 
Equation (C9) shows that the S wave electric field consists of two terms. If s2s =eεμ≪1, the first term dominates the
2 
^ 0. If s =eεμ≫1, Es is determined by the second
^ B
electric field and in this situation Es is along the direction of u s

GAO ET AL. ©2014. American Geophysical Union. All Rights Reserved. 32


Journal of Geophysical Research: Solid Earth 10.1002/2014JB010962

 
^ i.e., the direction of wave propagation. Similarly, by assuming u
term and its direction is parallel to k, ^0 ¼ 1
^ B
 
 ^  ^ 
and k^ k ^ B 0  ¼ 1 in equation (C9) we approximately get the E/u transfer function for the S wave
u
  2  
 s   μσB0 
jEs j≈ max 1;  s   2 jus j: (C10)
eεμ ss  eεμ

By substituting equation (C6) in to equation (3), we obtain the magnetic field H


1
H¼ ∇E
iωμ
 
1 μσB0 2  
¼ ^
iωsk j u j u^  ^ 0  s k^ k
B ^ u ^ B^0
iωμ s2  eεμ eεμ
h (C11)
sσB0  i
¼ 2 ^ u
juj k ^ B^0
s  eεμ
sσB0 h   i
¼ 2 juj  k^^u B ^ 0 þ k ^B ^0 u ^ :
s  eεμ

For the P wave, k^ ¼ u


^ and the magnetic field becomes
sp σB0  h ^ 
^B
i
^0 u
Hp ¼ 2 up B 0 þ k ^ : (C12)
sp  eεμ

The H/u transfer function for the P wave is approximately


 
   sp σB0  
Hp ≈ u : (C13)
s2p  eεμ p

^ u ¼ 0 and the S wave magnetic field becomes


For the S wave, k^
ss σB0 
Hs ¼ 2 ^B
jus j k ^0 u^: (C14)
ss  eεμ

Equation (C14) means that the direction of Hs is parallel to that of the S wave displacement. Approximately,
the H/u transfer function for the S wave is
 
 ss σB0 
jHs j≈ 2 jus j: (C15)
ss  eεμ

Acknowledgments References
We thank the anonymous reviewers for
their constructive comments that Abdul Azeez, K. K., C. Manoj, K. Veeraswamy, and T. Harinarayana (2009), Co-seismic EM signals in magnetotelluric measurement—A case
improved the manuscript substantially. study during Bhuj earthquake (26th January 2001), India, Earth Planets Space, 61(8), 973–981.
Data supporting the figures in this Aki, K., and P. G. Richards (2002), Quantitative Seismology, Univ. Sci., Calif.
paper are available as in the data set. Block, G. I., and J. G. Harris (2006), Conductivity dependence of seismoelectric wave phenomena in fluid-saturated sediments, J. Geophys.
This work is supported by National Res., 111, B01304, doi:10.1029/2005JB003798.
Natural Science Foundation of China Crow, D. R. (1988), Principles and Applications of Electrochemistry, Chapman and Hall, London, U. K.
(grant 41204039, 41274053, 41090293, Gao, Y., and H. Hu (2010), Seismoelectromagnetic waves radiated by a double couple source in a saturated porous medium, Geophys. J. Int.,
11372091, and 41174110), China 181(2), 873–896, doi:10.1111/j.1365-246X.2010.04526.x.
Postdoctoral Science Foundation (grant Gao, Y. X., H. H. Chen, and J. Zhang (2013a), Early electromagnetic waves from earthquake rupturing: I. Theoretical formulations, Geophys.
2013 T60617), Fundamental Research J. Int., 192(3), 1288–1307.
Funds for the Central Universities, China Gao, Y., X. Chen, H. Hu, and J. Zhang (2013b), Early electromagnetic waves from earthquake rupturing: II. Validation and numerical
(grant WK2080000055), and Chinese experiments, Geophys. J. Int., 192(3), 1308–1323.
Academy of Sciences/State Garambois, S., and M. Dietrich (2001), Seismoelectric wave conversions in porous media: Field measurements and transfer function analysis,
Administration of Foreign Experts Geophysics, 66(5), 1417–1430, doi:10.1190/1.1487087.
Affairs International Partnership Gershenzon, N., and G. Bambakidis (2001), Modeling of seismoelectromagnetic phenomena, Russ. J. Earth Sci., 3, 247–275.
Program for Creative Research Teams. Gershenzon, N. I., M. B. Gokhberg, and S. L. Yunga (1993), On the electromagnetic-field of an earthquake focus, Phys. Earth Planet. Inter.,
77(1–2), 13–19, doi:10.1016/0031-9201(93)90030-D.
Glover, P. W. J., E. Walker, and M. D. Jackson (2012), Streaming-potential coefficient of reservoir rock: A theoretical model, Geophysics, 77(2),
D17–D43.
Haartsen, M. W., and S. R. Pride (1997), Electroseismic waves from point sources in layered media, J. Geophys. Res., 102(B11), 24,745–24,769,
doi:10.1029/97JB02936.
Haines, S. (2004), Seismoelectric imaging of shallow targets, PhD thesis, Stanford Univ., Stanford, Calif.
Haines, S. S., S. R. Pride, S. L. Klemperer, and B. Biondi (2007), Seismoelectric imaging of shallow targets, Geophysics, 72(2), G9–G20,
doi:10.1190/1.2428267.

GAO ET AL. ©2014. American Geophysical Union. All Rights Reserved. 33


Journal of Geophysical Research: Solid Earth 10.1002/2014JB010962

Honkura, Y., H. Satoh, and N. Ujihara (2004), Seismic dynamo effects associated with the M7.1 earthquake of 26 May 2003 off Miyagi
Prefecture and the M6.4 earthquake of 26 July 2003 in northern Miyagi Prefecture, NE Japan, Earth Planets Space, 56(2), 109–114.
Honkura, Y., Y. Ogawa, M. Matsushima, S. Nagaoka, N. Ujihara, and T. Yamawaki (2009), A model for observed circular polarized electric fields
coincident with the passage of large seismic waves, J. Geophys. Res., 114, B10103, doi:10.1029/2008JB006117.
Hu, H., and Y. Gao (2011), Electromagnetic field generated by a finite fault due to electrokinetic effect, J. Geophys. Res., 116, B08302,
doi:10.1029/2010JB007958.
Huang, Q. H. (2002), One possible generation mechanism of co-seismic electric signals, Proc. Jpn. Acad. B Phys. Biol. Sci., 78(7), 173–178.
Hunter, R. J. (1981), Zeta Potential in Colloid Science, Academic Press, London.
Ichihara, H., Y. Hamano, K. Baba, and T. Kasaya (2013), Tsunami source of the 2011 Tohoku earthquake detected by an ocean-bottom
magnetometer, Earth Planet. Sci. Lett., 382, 117–124, doi:10.1016/j.epsl.2013.09.015.
Iyemori, T., T. Kamei, Y. Tanaka, M. Takeda, T. Hashimoto, T. Araki, T. Okamoto, K. Watanabe, N. Sumitomo, and N. Oshiman (1996), Co-seismic
geomagnetic variations observed at the 1995 Hyogoken-Nanbu earthquake, J. Geomagn. Geoelectr., 48(8), 1059–1070.
Jaafar, M. Z., J. Vinogradov, and M. D. Jackson (2009), Measurement of streaming potential coupling coefficient in sandstones saturated with
high salinity NaCl brine, Geophys. Res. Lett., 36, L21306, doi:10.1029/2009GL040549.
Johnson, D. L., J. Koplik, and R. Dashen (1987), Theory of dynamic permeability and tortuosity in fluid-saturated porous-media, J. Fluid Mech.,
176, 379–402, doi:10.1017/S0022112087000727.
Karakelian, D., G. C. Beroza, S. L. Klemperer, and A. C. Fraser-Smith (2002), Analysis of ultralow- frequency electromagnetic field measure-
ments associated with the 1999 M 7.1 Hector Mine, California, earthquake sequence, Bull. Seismol. Soc. Am., 92(4), 1513–1524.
Kuriki, M., M. Matsushima, Y. Ogawa, and Y. Honkura (2011), Spectral peaks in electric field at resonance frequencies for seismically excited
motion of ions in the Earth’s magnetic field, Earth Planets Space, 63(6), 503–507.
Løseth, L. O., H. M. Pedersen, B. Ursin, L. Amundsen, and S. Ellingsrud (2006), Low-frequency electromagnetic fields in applied geophysics:
Waves or diffusion?, Geophysics, 71(4), W29–W40.
Manoj, C., S. Maus, and A. Chulliat (2011), Observation of magnetic fields generated by tsunamis, Eos, 92(2), 13, doi:10.1029/2011EO020002.
Matsushima, M., et al. (2002), Seismoelectromagnetic effect associated with the İzmit earthquake and its aftershocks, Bull. Seism. Soc. Am., 92,
350–360.
Matsushima, M., Y. Honkura, M. Kuriki, and Y. Ogawa (2013), Circularly polarized electric fields associated with seismic waves generated by
blasting, Geophys. J. Int., 194(1), 200–211.
Morgan, F. D., E. R. Williams, and T. R. Madden (1989), Streaming potential properties of westerly granite with applications, J. Geophys. Res.,
94(B9), 12,449–12,461, doi:10.1029/JB094iB09p12449.
Nagao, T., Y. Orihara, T. Yamaguchi, I. Takahashi, K. Hattori, Y. Noda, K. Sayanagi, and S. Uyeda (2000), Co-seismic geoelectric potential
changes observed in Japan, Geophys. Res. Lett., 27(10), 1535–1538, doi:10.1029/1999GL005440.
Novotny, L., and B. Hecht (2006), Principles of Nano-Optics, Cambridge Univ. Press, Cambridge, U. K.
Ogawa, T., and H. Utada (2000a), Electromagnetic signals related to incidence of a teleseismic body wave into a subsurface piezoelectric
body, Earth Planets Space, 52(4), 253–260.
Ogawa, T., and H. Utada (2000b), Coseismic piezoelectric effects due to a dislocation: 1. An analytic far and early-time field solution in a
homogeneous whole space, Phys. Earth Planet. Inter., 121(3–4), 273–288, doi:10.1016/S0031-9201(00)00177-1.
Okubo, K., N. Takeuchi, M. Utsugi, K. Yumoto, and Y. Sasai (2011), Direct magnetic signals from earthquake rupturing: Iwate-Miyagi
earthquake of M 7.2, Japan, Earth Planet. Sci. Lett., 305(1–2), 65–72, doi:10.1016/j.epsl.2011.02.042.
Olhoeft, G. R. (1980), Electrical Conductivity, Paper Presented at Eighth Workshop on Electromagnetic Induction in the Earth and the Moon, Int.
Assoc.o f Geomagn and Aeron, Neuchatel, Switzerland.
Overbeek, J. T. G. (1952), Electrochemistry of the Double Layer, in Colloid Science, 1, Irreversible Systems, Elsevier, New York.
Pengra, D. B., S. X. Li, and P. Z. Wong (1999), Determination of rock properties by low-frequency AC electrokinetics, J. Geophys. Res., 104(B12),
29,485–29,508, doi:10.1029/1999JB900277.
Phillips, S. L., H. Ozbek, and R. J. Otto (1978), Basic Energy Properties of Electrolytic Solutions Database, paper presented at Sixth International
CODATA Conference, CODATA Secr, Santa Flavia, Italy.
Pride, S. (1994), Governing equations for the coupled electromagnetics and acoustics of porous-media, Phys. Rev. B, 50(21),
15,678–15,696.
Pride, S. R., and M. W. Haartsen (1996), Electroseismic wave properties, J. Acoust. Soc. Am., 100(3), 1301–1315.
Ren, H., X. Chen, and Q. Huang (2012), Numerical simulation of coseismic electromagnetic fields associated with seismic waves due to finite
faulting in porous media, Geophys. J. Int., 188(3), 925–944.
Revil, A., and L. M. Cathles (1999), Permeability of shaly sands, Water Resour. Res., 35(3), 651–662, doi:10.1029/98WR02700.
Revil, A., and P. W. J. Glover (1997), Theory of ionic-surface electrical conduction in porous media, Phys. Rev. B, 55(3), 1757–1773,
doi:10.1103/PhysRevB.55.1757.
Revil, A., and P. W. J. Glover (1998), Nature of surface electrical conductivity in natural sands, sandstones, and clays, Geophys. Res. Lett., 25(5),
691–694, doi:10.1029/98GL00296.
Revil, A., L. M. Cathles, S. Losh, and J. A. Nunn (1998), Electrical conductivity in shaly sands with geophysical applications, J. Geophys. Res.,
103(B10), 23,925–23,936, doi:10.1029/98JB02125.
Revil, A., H. Schwaeger, L. M. Cathles, and P. D. Manhardt (1999), Streaming potential in porous media 2. Theory and application to
geothermal systems, J. Geophys. Res., 104(B9), 20,033–20,048, doi:10.1029/1999JB900090.
Sen, P. N., and P. A. Goode (1992), Influence of temperature on electrical-conductivity on shaly sands, Geophysics, 57(1), 89–96.
Stacey, F. D., and M. J. S. Johnston (1972), Theory of the piezomagnetic effect in titanomagnetite-bearing rocks, Pure Appl. Geophys., 97,
146–155.
Tang, J., Y. Zhan, L. Wang, Z. Dong, G. Zhao, and J. Xu (2010), Electromagnetic coseismic effect associated with aftershock of Wenchuan M(s)
8.0 earthquake, Chin. J. Geophys.-Chinese Edition, 53(3), 526–534.
Toh, H., K. Satake, Y. Hamano, Y. Fujii, and T. Goto (2011), Tsunami signals from the 2006 and 2007 Kuril earthquakes detected at a seafloor
geomagnetic observatory, J. Geophys. Res., 116, B02104, doi:10.1029/2010JB007873.
Ujihara, N., Y. Honkura, and Y. Ogawa (2004), Electric and magnetic field variations arising from the seismic dynamo effect for aftershocks of
the M7.1 earthquake of 26 May 2003 off Miyagi Prefecture, NE Japan, Earth Planets Space, 56(2), 115–123.
Utada, H., H. Shimizu, T. Ogawa, T. Maeda, T. Furumura, T. Yamamoto, N. Yamazaki, Y. Yoshitake, and S. Nagamachi (2011), Geomagnetic field
changes in response to the 2011 off the Pacific Coast of Tohoku Earthquake and Tsunami, Earth Planet. Sci. Lett., 311(1–2), 11–27,
doi:10.1016/j.epsl.2011.09.036.
Vernik, L. (1998), Acoustic velocity and porosity systematics in siliciclastics, Log. Anal., 39, 27–35.

GAO ET AL. ©2014. American Geophysical Union. All Rights Reserved. 34


Journal of Geophysical Research: Solid Earth 10.1002/2014JB010962

Vinogradov, J., M. Z. Jaafar, and M. D. Jackson (2010), Measurement of streaming potential coupling coefficient in sandstones saturated with
natural and artificial brines at high salinity, J. Geophys. Res., 115, B12204, doi:10.1029/2010JB007593.
Wang, J., H. S. Hu, G. Yang, X. Li, G. Y. Li, and X. Cheng (2011), Experimental measurements on streaming current and zeta-potential of
core samples under the excitation of low-frequency sinusoidal pressure, Chin. J. Geophys.-Chinese Edition, 54(8), 2169–2176,
doi:10.3969/j.issn.0001-5733.2011.08.025.
Watillon, A., and R. De Backer (1970), Potentiel d’écoulement, courant d’écoulement et conductance de surface à l’interface eau-verre,
J. Electroanal. Chem., Interfacial Electrochem., 25, 181–196.
Yamazaki, K. (2011), Piezomagnetic fields arising from the propagation of teleseismic waves in magnetized crust with finite conductivity,
Geophys. J. Int., 184(2), 626–638, doi:10.1111/j.1365-246X.2010.04883.x.
Yamazaki, K. (2012), Estimation of temporal variations in the magnetic field arising from the motional induction that accompanies seismic
waves at a large distance from the epicentre, Geophys. J. Int., 190(3), 1393–1403, doi:10.1111/j.1365-246X.2012.05586.x.
Yamazaki, K. (2013), Improved models of the piezomagnetic field for the 2011 Mw 9.0 Tohoku-oki earthquake, Earth Planet. Sci. Lett., 363,
9–15.

GAO ET AL. ©2014. American Geophysical Union. All Rights Reserved. 35

You might also like