Electrical conductivity of serpentinite with various amounts of magnetite was measured during shear deformation at high pressure and temperatures (P = 1.0 GPa, T = 750 K) corresponding to mantle wedge conditions to evaluate the contribution of aligned magnetite to the bulk conductivity of serpentinite. Under hydrostatic conditions, the sample conductivity considerably increases when the magnetite volume fraction exceeds 25% in volume, suggesting the presence of the percolation threshold for magnetite interconnection. During shear deformation, the conductivity for the samples with less than 25 vol.% magnetite increased by an order of magnitude or higher with increasing shear strain up to 9, which is likely a result of the clustering or realignment of magnetite grains in the serpentinites. However, activation enthalpy was nearly constant before and after deformation experiments, suggesting that shear deformation is unlikely to enhance establishment of interconnection of magnetite. Consequently, more than 25 vol.% magnetite is needed to establish connectivity of magnetite in serpentinite. On the other hand, the conductivity of serpentinite with low volume fraction of magnetite (5%), which is typical concentration of natural serpentinites, is almost similar to that of magnetite-free serpentinites. The present results show that the interconnection of magnetite in serpentinites by shear deformation is not expected as an origin of the high conductivity anomaly occasionally observed at the slab interface in the mantle wedge. The origin of high conductivity, therefore, indicates the presence of aqueous fluid with high salinity rather than the magnetite interconnection.
 In the mantle wedge of the subduction zones, magnetotelluric surveys occasionally reveal high electrical conductivity anomaly [e.g., Matsuno et al., 2010; Soyer and Unsworth, 2006; Yamaguchi et al., 2009]. The conductivity values observed in the mantle wedge vary but are occasionally very high up to 0.1 S/m in the hydrated wedge of the hot Ryukyu and Kyushu subduction zones [Ichiki et al., 2009]. The conductivity of olivine, which is the main upper mantle mineral, is less than 10−4 S/m at temperature and pressure expected at the slab interface, even if olivine is hydrated [e.g., Poe et al., 2010; Yang, 2012; Yoshino et al., 2006]. Therefore, a presence of highly conductive material capable of enhancing the conductivity at low temperatures is expected in the mantle wedge.
 There are two appropriate candidates accounting for such high conductivity anomaly observed in the subduction zone; aqueous fluid and serpentinite. In the absence of highly conductive minerals such as graphite, magnetite and sulphides, conductive fluid can be regarded as the most preferable candidate for explaining high conductivities [Hyndman and Shearer, 1989]. On the other hand, serpentinization can be also considered as an attractive candidate to enhance conductivity in the mantle wedge. The subducting plate releases most of the water into the mantle wedge by the dehydration reactions, and the expelled water reacts with mantle rocks, forming serpentinite at the plate interface [e.g., Hyndman and Peacock, 2003; Rüpke et al., 2004]. The existence of a hydrous layer, including serpentinites, has been detected in several subduction zones, evident as low-velocity seismic wave anomalies with a high Poisson's ratio [e.g.,Kamiya and Kobayashi, 2000; Brocher et al., 2003].
 Previous measurements of electrical conductivity of natural serpentinites of the Indian Oceanic ridge has shown an extremely high conductivity compared to the unaltered peridotites [Stesky and Brace, 1973]. In addition, the conductivity measurement of lizardite, which is a low T serpentine [Zhu et al., 2001], yielded conductivities (10−3–10−1 S/m) close to those measured in the cold mantle wedge of active subduction. However, recent electrical conductivity measurements have shown that antigorite, the more relevant serpentine polymorph in the hot subduction zone conditions, reveals low electrical conductivities similar to the dry mantle minerals [Guo et al., 2011; Reynard et al., 2011], suggesting that serpentine alone cannot explain the high conductivity anomaly observed above the descending slab.
 Serpentinites are commonly associated with magnetite; therefore the high conductivity anomaly can be explained if magnetite is interconnected in serpentinite. The mantle rocks at the plate interface are subjected to noncoaxial stress and widely develop a strong alignment of constituent minerals, as seen in the strong seismic anisotropy [e.g., Katayama et al., 2009; Bezacier et al., 2010]. Such extensive deformation at the plate interface could generate interconnection and reorientation of magnetite at the mantle shear zones, which may be the cause of the high electrical conductivity anomaly. In this study, we measured the electrical conductivity of synthetic serpentinites with various amounts of magnetite during shear deformation to examine the influence of magnetite interconnection on electrical conductivity in the deformed samples.
2. Experimental Methods
 Starting materials were prepared from the powder mixture of antigorite with various amounts of magnetite (5, 10, 20, 30, 50 vol.%). Natural Fe-free antigorite from China [Osako et al., 2010] and magnetite were crushed and ground in agate mortar. The mixed samples enclosed in a platinum capsule were then sintered in a piston cylinder apparatus at P = 1.0 GPa and T = 773 K within the stability field of antigorite. The recovered samples showed nearly homogeneous distribution of magnetite in antigorite matrix (Figure S1a in the auxiliary material).
 We developed a cell assembly for in-situ electrical conductivity measurement during shear deformation as shown in Figure S2 in theauxiliary material. The sintered starting samples with 2–3 mm length, 0.3 mm thick and 0.8 mm2of cross-sectional area were prepared, and were sandwiched between two alumina pistons cut at 45° from the maximum compression direction. Ni electrodes for the electrical conductivity measurement, which control an oxygen fugacity assuring magnetite stability, were place at each edge of sample along the shear direction. Two lead lines with a composition of W97Re3 were connected to each electrode, and were insulated from the graphite heater by Al2O3insulator. The assembly included a pyrophyllite cubic pressure medium with a 21-mm cubicle edge length, ZrO2 thermal insulator and cylindrical graphite furnace.
 In-situ conductivity measurements were performed using a DIA-type cubic anvil apparatus with deformation facility. Impedance spectroscopic measurements were carried out using a Solartron 1260 impedance gain-phase analyzer combined with a Solartron 1296 interface. Impedance spectra were obtained in the frequency range of 10−1–106 Hz. The electrical conductivity measurements of serpentinites with various amounts of magnetite were performed at a constant pressure P = 1.0 GPa. We measured electrical conductivity during both heating and cooling cycles (500–750 K) and both before and after deformation Temperature was calibrated from applied power using the previously determined relation between input power and temperature in this cell assembly. The upper and lower bounds of sample resistance were checked for the cell assembly using a magnetite-free sample and magnetite, respectively. After the conductivity had stabilized at 750 K, the piston was advanced by 200 μm/h (strain rate of ∼10−4 s−1) was fixed for 8 hours in the shear deformation experiment. The electrical conductivity of serpentinites along the shear direction was measured during shear deformation every 10 minute.
 The total shear strain for each run was estimated from the offset of the alumina piston observed in the recovered samples. The time-strain relationship was calculated based on the assumption that strain rate was constant during movement of the piston. The electrical conductivity was also corrected for the variation of sample dimension, which was determined by measuring the length between two electrodes after deformation. Experimental conditions at 1 GPa and 750 K are summarized inTable 1.
The conductivity measured at 750 K before shear deformation.
The conductivity measured at 750 K after shear deformation.
The activation enthalpy determined from measurement before shear deformation.
The activation enthalpy determined from measurement after shear deformation. Numbers in parentheses are 1σ deviation from the mean.
Shear strain inferred from the recovered samples.
Conduction mechanism of interconnected phase inferred from activation enthalpy for electrical conductivity of bulk samples. P denotes proton conduction in antigorite. H indicates hopping (small polaron) conduction in magnetite.
3. Results and Discussion
3.1. Electrical Conductivity Measurements Before Deformation
 The results of conductivity measurements before deformation showed a wide range of conductivity, 10−4 to 102 S/m, depending on the amounts of magnetite in the synthesized serpentinites (Figure 1a). In all experiments, conductivity increases with increasing temperature. This relationship is expressed by the Arrhenius formulation,
where σ0is a pre-exponential factor, ΔH is activation enthalpy, k is the Boltzmann constant and T is temperature. Two distinct trends were observed in terms of activation enthalpy; ΔH = 0.65–0.87 eV for samples with less than 20 vol.% magnetite, which is similar to that of antigorite (0.68–0.74 eV [Guo et al., 2011]), whereas relatively lower activation enthalpy, ΔH = 0.01–0.25 eV, for samples with higher abundance (>25 vol.%) of magnetite (Table 1) is much closer to that of magnetite (0.06–0.07 eV) [Tannhauser, 1962; Yamanaka et al., 2001]). Although Reynard et al. considered that the dominant conduction mechanism in antigorite is small polaron conduction due to the electron hopping between ferric and ferrous iron sites, it is unlikely that small polaron conduction is dominant for the Fe-free antigorite used by the present study. This indicates that the proton conduction in serpentinite can be a dominant mechanism for the samples with a relatively lower amount of magnetite, whereas the hopping conduction between ferrous and ferric iron sites in magnetite is a dominant mechanism for the magnetite-rich serpentinites. The activation enthalpy of conductivity of a bulk rock should correspond to that of a conductive phase if the conductive phase establishes interconnection in the resistive matrix [Yoshino et al., 2003, 2004, 2008; Watson et al., 2010; Watson and Roberts, 2011; Yoshino and Noritake, 2011]. It is concluded that the percolation threshold for magnetite in the antigorite matrix under hydrostatic conditions is between 20 and 25 vol.%.
 The effect of the magnetite volume fraction on electrical conductivity was evaluated by the Hashin and Shtrikman  bounds, in which the upper and lower bounds are expressed as,
where σmgt and σserp are the single phase conductivity of magnetite and serpentine at T = 750 K and P = 1.0 GPa, ϕ is the volume fraction of magnetite, and δσ is the conductivity difference between magnetite and serpentine. The experimental results do not follow the inferred upper or lower bound relation, but show an abrupt increase of conductivity at the magnetite fraction ranging 20–30 vol.% (Figure 1b). This implies that the equivalent circuit abruptly changes from series to parallel by changing volume fraction of magnetite. The critical volume fraction for connectivity of magnetite predicted from a considerable change of the conductivity is consistent with that for a change of activation enthalpy.
3.2. Effect of Deformation
 After the measurements of electrical conductivity under static conditions, the uniaxial piston was advanced and the samples were deformed in a simple shear geometry (Figure S2 in the auxiliary material). Although it is difficult to monitor differential stress due to a large friction in our assembly, the recovered samples show highly deformed textures, including the stretched antigorite and alignments of magnetite toward the shear direction (Figures S1b–S1d in the auxiliary material).
 During shear deformation, the electrical conductivity of serpentinites increased with increasing shear strain. The conductivity of magnetite-free serpentinite was also increased by the realignment of antigorite because of conductivity anisotropy of antigorite [Guo et al., 2011] (Figure 2a). The sample with 5 vol.% magnetite showed a small increase of conductivity during deformation, and the conductivity values were close to those of the magnetite-free samples. In the samples with 10–25 vol.% magnetite, the results showed nearly one order of magnitude increase during deformation, whereas the electrical conductivities were not largely changed in the samples with high magnetite fraction (>30 vol.%). The increase in conductivity is likely to be a result of the elongation of magnetite parallel to the shear direction in the serpentinites. If magnetite is completely interconnected during deformation, an abrupt increase in conductivity is expected; however, the results showed a gradual increase in conductivity and did not reach the value close to the magnetite conductivity. Microstructures of the recovered sample show thin disconnected layers of magnetite in antigorite aggregates (Figure S1 in theauxiliary material), which is consistent with the incomplete connectivity of magnetite during deformation.
 After the deformation experiments, the samples were measured at various temperatures (Figure 2b). The results showed almost same activation enthalpy before and after deformation. The serpentinites with low magnetite fraction (<20 vol.%) have larger activation enthalpy, whereas the relatively small activation enthalpy was observed in the samples with high volume fraction of magnetite (>25 vol.%). These observations suggest that an interconnection of magnetite for the samples with less than 25 vol.% of magnetite was not established by shear deformation (γ < 9). Therefore, it is concluded that the shear deformation did not contribute to the development of interconnection in the measured range of shear strain, because magnetite grains were elongated parallel to shear direction and simultaneously destroyed by development of pinch and swell structure (microboudin).
 Dynamic wetting has been experimentally reported in some geological materials such as molten Fe alloy in olivine matrix [Bruhn et al., 2000] and partial molten peridotite [Jin et al., 1994] when the conductive phase is distinctly softer than the matrix one. However, the present study did not demonstrate dynamic wetting by shear deformation. It implies that magnetite is harder than antigorite. Thus, in the case of a rock with a small amount of harder conductive phase, shear deformation can enhance conductivity due to the development of thin conductive layer parallel to the shear direction. However, shear deformation cannot contribute to establishing long-range connectivity in a millimeter order over several hundreds of grains, because interconnection formed by a thin hard conductive layer is readily destroyed by deformation of the softer matrix under further shear deformation.
4. Origin of the Conductivity Anomaly at the Slab Interface
 The present experiments showed that electrical conductivity of serpentinites increases during deformation due to the elongation of magnetite grains parallel to shear direction; however, the conductivity does not increase largely for the sample with low magnetite fraction (5 vol.%). The magnetite fraction is commonly less than 5 vol.% in natural serpentinites; therefore the significant increase in electrical conductivity is not expected during shear deformation at the plate interface. Although variable range of conductivity was reported in the serpentinized peridotites [Stesky and Brace, 1973; Toft et al., 1990], they may reflect the heterogeneous distribution of magnetite, and such local clustering may not be detected in the long-period magnetotelluric profiles. Consequently, other mechanisms are necessary to explain the high conductivity anomaly seen in the subduction zones.
 One possible candidate is a presence of free aqueous fluid with high salinity. Recent experimental data have shown that electrical conductivity in fluid-bearing quartzite is not enough to explain the conductivity anomaly [Shimojuku et al., 2012]. If interstitial fluids have the conductivity as high as seawater (∼5 S/m), the upper bound of Hashin and Shtrikman's  relation implies a fluid fraction of 0.5–10 vol.% in serpentinite to explain the electrical conductivity observed in the subduction zones (Figure 3). Increase of salinity in aqueous fluid due to progressive serpentinization can lead to significant enhancement of the conductivity even if the fluid fraction is very low, less than 1 vol. % [Reynard et al., 2011]. The high conductivity regions are characterized by low seismic velocities and high Poisson's rations [e.g., Bostock et al., 2002; Matsubara et al., 2008], which also suggest the presence of excess water in the serpentinized mantle wedge. Water released from the subducting slab tends to migrate along the plate interface and is likely accumulated at the mantle wedge corner due to the permeability barrier [Kawano et al., 2011], resulting in the high electrical conductivity and low velocity anomalies in the mantle wedge corner.
 We thank D. Yamazaki, E. Ito, A. Yoneda and A. Shimojuku for useful discussions and for technical assistance. We also acknowledge M. Osako for providing samples. Comments by two anonymous reviewers helped to improve the manuscript. This work was supported by a Grant-in-Aid for Scientific Research on Innovative Areas (Research in a Proposed Research Area), “Geofluids: Nature and Dynamics of Fluids in Subduction Zones” from the Japan Society for Promotion of Science (21109003).
 The Editor thanks the anonymous reviewer for assistance in evaluating this paper.