Calculation of the piezomagnetic field arising from uniform regional stress in inhomogeneously magnetized crust (II): Limitation in general cases

Abstract

One of the difficulties encountered in modeling the piezomagnetic field is determining the distribution of the initial magnetizations. A previous study showed that, in the case that data are available regarding the magnetic total force anomaly and that the directions of the initial magnetizations are assumed to be uniform, the piezomagnetic field arising from the regional stress is uniquely determined, although the intensities of the initial magnetizations are not uniquely determined. The present study considers a more general situation in which the directions of the initial magnetizations are unknown. The problem addressed here is whether the piezomagnetic field is determined uniquely. To this end, a set of expressions is derived to describe the relation among the initial magnetizations, the magnetic anomalies corresponding to the initial magnetization, and the resultant piezomagnetic field. The expressions show that the correspondence between magnetic anomalies and the piezomagnetic field is not one-to-one. This result suggests that observations of the piezomagnetic field provide quantitative estimations of changes in stress only if we can assume that the directions of the initial magnetizations are uniform.

LETTER Earth Planets Space, 63, 12171220, 2011

Calculation of the piezomagnetic eld arising from uniform regional stress in inhomogeneously magnetized crust (II): Limitation in general cases

Kenichi Yamazaki1.2

1Institute of Seismology and Volcanology, Hokkaido University, N12-W8, Kita-ku, Sapporo, Hokkaido 060-0810, Japan

2Miyazaki Observatory, Research Center for Earthquake Prediction, Disaster Prevention Research Institute, Kyoto University,

3884, Kaeda, Miyazaki 889-2161, Japan

(Received August 17, 2010; Revised June 14, 2011; Accepted August 5, 2011; Online published February 21, 2012)

One of the difculties encountered in modeling the piezomagnetic eld is determining the distribution of the initial magnetizations. A previous study showed that, in the case that data are available regarding the magnetic total force anomaly and that the directions of the initial magnetizations are assumed to be uniform, the piezomagnetic eld arising from the regional stress is uniquely determined, although the intensities of the initial magnetizations are not uniquely determined. The present study considers a more general situation in which the directions of the initial magnetizations are unknown. The problem addressed here is whether the piezomagnetic eld is determined uniquely. To this end, a set of expressions is derived to describe the relation among the initial magnetizations, the magnetic anomalies corresponding to the initial magnetization, and the resultant piezomagnetic eld. The expressions show that the correspondence between magnetic anomalies and the piezomagnetic eld is not one-to-one. This result suggests that observations of the piezomagnetic eld provide quantitative estimations of changes in stress only if we can assume that the directions of the initial magnetizations are uniform.

Key words: Regional stress, piezomagnetic effect, direction of magnetization, double Fourier transform, one-toone correspondence.

1. Introduction

In the Earths crust, changes in stress result in changes in magnetization via the piezomagnetic effect (Nagata, 1970). By observing changes in the magnetic eld arising from the piezomagnetic effect, which are referred to as piezomagnetic elds, it is possible to monitor the stress accumulation caused by various processes such as slab subduction.

However, studies of the piezomagnetic eld encounter difculties relating to the treatment of the magnetization structures in the Earths crust. Changes in magnetization are approximately proportional to the product of applied stress and the initial magnetization with no stress (e.g., Stacey, 1964; Nagata, 1970; Stacey and Johnston, 1972). In the absence of constraints on the spatial distribution of the initial magnetization, it is not possible to perform accurate calculations of the piezomagnetic eld, even in forward problems. Indeed, numerical examples have demonstrated that signicant variations in the magnetic eld are observed near areas having strong contrasts in magnetization (e.g., Oshiman, 1990; Nishida et al., 2007) and that, by ignoring the precise structure of the initial magnetization, we obtain an incorrect estimation of the piezomagnetic eld (e.g., Currenti et al., 2009).

A steady procedure that can be used to calculate the

Copyright c

[circlecopyrt] The Society of Geomagnetism and Earth, Planetary and Space Sci

ences (SGEPSS); The Seismological Society of Japan; The Volcanological Society

of Japan; The Geodetic Society of Japan; The Japanese Society for Planetary Sci

ences; TERRAPUB. doi:10.5047/eps.2011.08.007

piezomagnetic eld is as follows: rst, the spatial distribution of the initial magnetization is assumed by using magnetic survey data; next, the change in magnetization, due to the piezomagnetic effect at each point, is calculated; nally, changes in the magnetic eld generated by different parts of the crust are integrated to obtain the magnitude of the entire piezomagnetic eld (e.g., Oshiman et al., 1997; Nishida et al., 2004). A drawback of this procedure is the non-uniqueness of solutions in the potential theory. Even in the case that the magnetic eld generated by the initial magnetization is completely known, the explicit distribution of the initial magnetizations cannot be determined uniquely. Consequently, the resultant piezomagnetic eld is not determined uniquely. This result means that the observed changes in the magnetic eld are not directly converted to changes in stress.

Yamazaki (2009) proposed a method of calculating the piezomagnetic eld in cases for which (i) the stress eld is uniform in the region of interest, and (ii) the directions of initial magnetizations are parallel to the ambient geomagnetic eld. Taking into account the similarity between this problem and the reduction to the pole calculation (cf. Baranov, 1957), Yamazaki successfully derived a formula to determine the resultant piezomagnetic eld based on magnetic anomaly data. The existence of such a formula implies that the resultant piezomagnetic eld is determined uniquely, regardless of the non-uniqueness of the explicit distribution of the initial magnetization intensities. Therefore, the difculty encountered regarding the

1217

1218 K. YAMAZAKI: LIMITATION IN CALCULATING THE PIEZOMAGNETIC FIELD

non-uniqueness of the initial magnetization is avoided for situations in which conditions (i) and (ii) are met. Condition (i) is valid across a reasonably wide range of situations, including stress accumulation related to plate subduction. However, condition (ii) is somewhat problematic. Although this condition is usually assumed in interpretations of aero-magnetic survey results, the directions of the initial magnetizations are not assured to be uniform, particularly in cases of remanent magnetization.

If the uniqueness of the resultant piezomagnetic eld is still assured without the need to make assumptions regarding the directions of initial magnetization, the efciency of studies on the piezomagnetic eld would be enhanced. In contrast, if the uniqueness of the resultant piezomagnetic eld cannot be assured, we should be cautious in calculating the piezomagnetic eld. Therefore, it is important to consider situations in which the directions of the initial magnetizations are unknown. The aim of the present study is to clarify whether the piezomagnetic eld is uniquely determined in the case that only data of magnetic anomalies arising from initial magnetizations are available and that no constraints are available regarding the directions of the initial magnetizations.

2. Denition of the Problem

It is assumed that changes in magnetization due to the piezomagnetic effect ( Ji) are proportional to the applied stress (i j) and the initial magnetization (Ji). The constitutive law is then summarized as follows (Sasai, 1991):

Ji =

similar expression is obtained for the piezomagnetic eld. Using the constitutive law (Eq. (1)), changes in the total forces of the geomagnetic eld due to the piezomagnetic effect (Fp) are expressed by

Fp(x) =

0 4

3 2li

xi

dx 1

dx 2

H0 dx 3

1 Tjk Jk(x ). (4)

The relations among Ji, Fa, and Fp are more clearly expressed when considered in wavenumber space. To obtain the double Fourier transforms of Eqs. (3) and (4), the following formula is referred to (e.g., Sasai, 1991):

1

2

x j

dr1

dr2

exp[i(k1r1 + k2r2)] r21 + r22 + r23

=

1kH exp(kH|r3|), (5)

where kH = (k21 + k22)1/2. Using this formula, the double

Fourier transforms of Fa and Fp are expressed in the following forms:

Fa(k1, k2) =

04 (li Ki)(K j J#j(k1, k2)), (6)

and

Fp(k1, k2) =

32(li Ki)(K j Tjk J#k(k1, k2)), (7)

where the superscript denotes the double Fourier transform of a function, Ki is a scalar multiplier dened by (K1, K2, K3) = (ik1, ik2, kH), and the superscript # de

notes the integral dened by

f #(k1, k2) =

0 4

32Ti j Jj, (1)

where Ti j is the deviatoric stress tensor given by

Ti j = i j

13(11 + 22 + 33), (2)

and is the stress sensitivity. Although the stress sensitivity largely depends on the mechanical strength of the rocks of the crust (e.g. Hamano, 1983), only the situation in which is constant in the region of interest is considered in the present study. In Eq. (1) and hereafter, summation rules for subscripts are used for simplicity of notation. Note that summation rules are not used in the previous paper (i.e. Yamazaki, 2009). Therefore, attention is required when the equations in this paper are compared with those in the previous paper.

A situation is considered in which anomalies in the geo-magnetic total force values are known at a certain height(i.e., x3). Anomalies in total force values of the geo-magnetic eld (Fa) corresponding to crustal magnetizations (Ji) are approximately expressed by the following formula, which is equivalent to equation (5) in Yamazaki (2009):

Fa(x) =

0 4 li

H0 dx 3 f (k1, k2, x 3) exp(kH x 3). (8)

Based on Eqs. (6) and (7), it is possible to discuss the relation between Fa and Fp. The problem to be solved is whether a given value of Fa uniquely determines the corresponding Fp in the case that the explicit distribution of Ji is unknown.

3. Non-uniqueness of Estimations of the Piezo-magnetic Field

Before considering general cases, a special case is considered in which the directions of the initial magnetizations are uniform so that the magnetization vector is represented as Ji(x ) = m(x )li, where m is a scalar function. In this

case, J#i is expressed as

J#i(k1, k2) = m#(k1, k2)li. (9)

By taking the ratio of Fp to Fa, the factor m# is cancelled. Consequently, Fp is uniquely determined by Fa. The explicit relation is expressed as

Fp(k1, k2) =

3 2

xi

dx 1

dx 2

H0 dx 3

x j

1 Jj(x ), (3)

where li denotes the direction cosines of the ambient geo-magnetic eld and H represents the Curie point depth. A

l j Tjk Kk

li Ki Fa(k1, k2). (10)

K. YAMAZAKI: LIMITATION IN CALCULATING THE PIEZOMAGNETIC FIELD 1219

This formula is equivalent to that derived by Yamazaki (2009), which implies the piezomagnetic eld is uniquely determined for a given magnetic anomaly.

In the general situation, the initial magnetizations are not xed to a single direction. Note that the general situation includes the case of Eq. (9) as a special case. Therefore, if Fp is uniquely determined for a given Fa, the formula of the general situation should also be given by Eq. (10).

In particular, Fa = 0 should yield Fp = 0 if Eq. (10) is

satised.

Unfortunately, a brief inspection of Eqs. (6) and (7) reveals that this expectation is incorrect. For example, consider the following distributions:

J#(k1, k2) = A (k1 k) (k2)

1 0

+i

+A (k1 + k) (k2)

1 0

i

(11)

where k (> 0) and A are constants, and is the Dirac Delta function. One of the explicit forms of J corresponding to Eq. (11) is given by

J(x 1, x 2, x 3) = A

cos kx 1 0

sin kx 1

Fig. 1. Example illustrating the non-uniqueness of the piezomagnetic eld. (a) Spatial distribution of initial magnetizations (J) in the crust. x1 and x3 are the location coordinates, H represents the Curie point depth, and k represents a spatial wavenumber. (b) Changes in the magnetization due to the piezomagnetic effect ( J) corresponding to J. and T represent the stress sensitivity and a deviatoric stress tensor(i.e., Eq. (2)), respectively. In each panel, arrow length is proportional to the intensity of magnetization. The regional stress eld is assumed to be a uniaxial compression with intensity in the x1 direction. The distribution J generates a non-zero magnetic eld above the ground, whereas J does not.

that satises the requirement of Eq. (13) is given by

J#i(k1, k2) =

1

2k2H

, (12)

where A is a constant. With reference to Eqs. (6) and(8), Fa becomes zero for the above distribution of magnetizations. However, Eq. (7) yields non-zero values of the piezomagnetic eld corresponding to this distribution of magnetizations. The initial magnetizations and the changes in magnetizations are illustrated in Fig. 1. This example shows that there are no exact one-to-one correspondences between the magnetic anomalies and the resultant piezo-magnetic elds.

Nevertheless, the possibility is that Fp corresponding to a given Fa can be determined in a statistical sense, and that the counterexample given by Eq. (12) is merely an extreme case. This is an analogy to statistical mechanics (e.g. Landau and Lifshitz, 1980). For a given Fa, we can consider a set of Ji, all members of which generate the given Fa. If Fp corresponding to major members of the set are approximately the same as those corresponding to the initial magnetizations with a uniform direction (i.e. Fp given by Eq. (10)), then the Fp is regarded as a likely estimation corresponding to the given Fa. To conrm whether this expectation is valid or not, it is useful to estimate Fp based on acceptable assumptions regarding Ji, other than the assumption of a uniform direction of Ji (Eq. (9)).

One plausible assumption regarding the initial magnetizations is the minimum length requirement, which is given by

J

#

1 (k1, k2)

40l j K j K iFa(k1, k2), (14)

where (K 1, K 2, K 3) = (ik1, ik2, kH). The correspond

ing piezomagnetic eld is given by

Fp(k1, k2) =

3

4k2H

Ki Ti j K j Fa(k1, k2). (15)

The estimations of piezomagnetic elds given by Eqs. (10) and (15) are different, meaning that estimations of the piezomagnetic eld differ according to the assumed distributions of the initial magnetizations: an assumption of uniform direction yields Eq. (10), while an assumption of minimum magnetizations yields Eq. (15). Therefore, it is not possible to obtain any estimation of piezomagnetic elds unless the assumption regarding the distribution of the initial magnetization can be demonstrated to be valid.

4. Conclusions

In the case of a uniform regional stress, a series of formulae that express the relation between the initial magnetizations, the magnetic anomalies, and the piezomagnetic eld, is derived in wavenumber space. Based on the formulae, it is possible to discuss whether there exists a one-to-one correspondence between the magnetic anomalies and the

2

+ J

2

+ J

2

= min .

#

2 (k1, k2)

#

3 (k1, k2)

(13)

For a given magnetic anomaly Fa, the solution of Eq. (6)

1220 K. YAMAZAKI: LIMITATION IN CALCULATING THE PIEZOMAGNETIC FIELD

piezomagnetic eld. In the case that the direction of the initial magnetizations is assumed to be uniform, the piezomagnetic eld is uniquely determined from the initial magnetic anomalies, regardless of the non-uniqueness of the explicit distributions of the initial magnetizations. In contrast, in cases for which the directions of the initial magnetization are not uniform, the piezomagnetic eld cannot be determined in any sense. This result indicates that the efciency of piezomagnetic studies is diminished in the case that we cannot assure a uniform direction of initial magnetizations within the region of interest.

The above conclusion is somewhat disappointing because it implies that piezomagnetic elds cannot be correctly determined unless the directions of the initial magnetizations are xed. The present analysis was limited to the case in which the regional stress is uniform within the region of interest. However, it is natural to expect that the same conclusion would be obtained for a non-uniform regional stress. Of course, this result does not mean that all efforts to observe piezomagnetic elds are meaningless. Observations of the magnetic eld will continue to be used for the qualitative detection of events, including changes in the rate of stress accumulation. The present results merely highlight the difculty encountered in quantitative interpretations of the observed magnetic eld. In addition, the directions of the initial magnetizations can be assumed to be uniform in many situations. Indeed, this assumption is adopted in many studies that attempt to interpret the results of magnetic surveys. Nevertheless, we must bear in mind that forward calculations of piezomagnetic elds are accurate only if the direction of magnetization is known.

Acknowledgments. The author is grateful to Dr. C. Del Negro and Dr. Y. Sasai for reviewing the manuscript, and Dr. M. Uyeshima for managing the editorial process.

References

Baranov, V., A new method for interpretation of aeromagnetic maps:

Pseudo-gravimetric anomalies, Geophysics, 22, 359383, 1957. Currenti, G., C. Del Negro, A. Di Stefano, and R. Napoli, Numerical simulation of stress induced piezomagnetic elds at Etna volcano, Geophys.J. Int., 179, 14691476, 2009.

Hamano, Y., Experiments on the stress sensitivity of natural remanent magnetization, J. Geomag. Geoelectr., 35, 155172, 1983.

Landau, L. D. and E. M. Lifshitz, Statistical Physics, 3rd Edition, Part 1,544 pp, Butterworth-Heinemann, Oxford, 1980.

Nagata, T., Basic magnetic properties of rocks under mechanical stresses,Tectonophysics, 9, 167195, 1970.

Nishida, Y., Y. Sugisaki, K. Takahashi, M. Utsugi, and H. Oshima, Tectonomagnetic study in the eastern part of Hokkaido, NE Japan: Discrepancy between observed and calculated results, Earth Planets Space, 56, 10491058, 2004.

Nishida, Y., M. Utsugi, and T. Mogi, Tectonomagnetic study in the eastern part of Hokkaido, NE Japan (II): Magnetic elds related with the 2003 Tokachi-oki earthquake and the 2004 Kushiro-oki earthquake, Earth Planets Space, 59, 11811186, 2007.

Oshiman, N., Enhancement of tectonomagnetic change due to non-uniform magnetization in the Earths crustTwo-dimensional case studies, J. Geomag. Geoelectr., 42, 607619, 1990.

Oshiman, N., Y. Sasai, Y. Honkura, Y. Ishikawa, and Y. Tanaka, Geomagnetic changes correlated with crustal movement in the north-eastern part of the Izu Peninsula, Japan, Annali di Geosica, 60, 479494, 1997. Sasai, Y., Tectonomagnetic modeling on the basis of the linear piezomagnetic effect, Bull. Earthq. Res. Inst., Univ. Tokyo, 66, 587722, 1991. Stacey, F. D., The seismomagnetic effect, Pure Appl. Geophys., 58, 522,

1964.

Stacey, F. D. and M. J. S. Johnston, Theory of the piezomagnetic effect in titanomagnetite-bearing rocks, Pure, Appl. Geophys., 97, 146155, 1972.

Yamazaki, K., Calculation of the piezomagnetic eld arising from uniform regional stress in inhomogeneously magnetized crust, Earth Planets Space, 61, 11631168, 2009.

K. Yamazaki (e-mail: [email protected])