Seismic velocities, anisotropy, and shear-wave splitting of antigorite serpentinites and tectonic implications for subduction zones


  • Shaocheng Ji,

    Corresponding author
    1. Département des Génies Civil, Géologique et des Mines, École Polytechnique de Montréal, Montréal, Québec, Canada
    • Corresponding author: S. Ji, Département des Génies Civil, Géologique et des Mines, École Polytechnique de Montréal, Montréal, QC H3C 3A7, Canada. (, C. Long, Institute of Geomechanics, Chinese Academy of Geological Sciences, Beijing 100081, China. (

    Search for more papers by this author
  • Awei Li,

    1. Institute of Geomechanics, Chinese Academy of Geological Sciences, Beijing, China
    Search for more papers by this author
  • Qian Wang,

    1. State Key Laboratory of Isotope Geochemistry, Guangzhou Institute of Geochemistry, Chinese Academy of Sciences, Guangzhou, China
    Search for more papers by this author
  • Changxing Long,

    Corresponding author
    1. Institute of Geomechanics, Chinese Academy of Geological Sciences, Beijing, China
    • Corresponding author: S. Ji, Département des Génies Civil, Géologique et des Mines, École Polytechnique de Montréal, Montréal, QC H3C 3A7, Canada. (, C. Long, Institute of Geomechanics, Chinese Academy of Geological Sciences, Beijing 100081, China. (

    Search for more papers by this author
  • Hongcai Wang,

    1. Institute of Geomechanics, Chinese Academy of Geological Sciences, Beijing, China
    Search for more papers by this author
  • Denis Marcotte,

    1. Département des Génies Civil, Géologique et des Mines, École Polytechnique de Montréal, Montréal, Québec, Canada
    Search for more papers by this author
  • Matthew Salisbury

    1. Geological Survey of Canada-Atlantic, Bedford Institute of Oceanography, Dartmouth, Nova Scotia, Canada
    Search for more papers by this author


[1] Antigorite, the high-temperature (HT) form of serpentinite, is believed to play a critical role in various geological processes of subduction zones. We have measured P- and S-wave velocities (Vp and Vs), anisotropy and shear-wave splitting of 17 serpentinite samples containing >90% antigorite at pressures up to 650 MPa. The new results, combined with data for low-temperature (LT) lizardite and/or chrysolite, reveal distinct effects of LT and HT serpentinization on the seismic properties of mantle rocks. At 600 MPa, Vp = 5.10 and 6.68 km/s, Vs = 2.32 and 3.67 km/s, and Vp/Vs = 2.15 and 1.81 for pure LT and HT serpentinites, respectively. Above the crack-closure pressure (~150 MPa), the velocity ratio of antigorite serpentinites displays little dependence on pressure or temperature. Serpentine contents within subduction zones and forearc mantle wedges where temperature is >300°C should be at least twice that of previous estimates based on LT serpentinization. The presence of seismic anisotropy, high-pressure fluids, or partial melt is also needed to interpret HT serpentinized mantle with Vp < 6.68 km/s, Vs < 3.67 km/s, and Vp/Vs > 1.81. The intrinsic anisotropy of the serpentinites (3.8–16.9% with an average value of 10.5% for Vp, and 3.6–18.3% with an average value of 10.4% for Vs) is caused by dislocation creep-induced lattice-preferred orientation of antigorite. Three distinct patterns of seismic anisotropy correspond to three types of antigorite fabrics (S-, L-, and LS-tectonites) formed by three categories of strain geometry (i.e., coaxial flattening, coaxial constriction, and simple shear), respectively. Our results are thought to provide a new explanation for various anisotropic patterns of subduction systems observed worldwide.

1 Introduction

[2] Serpentinites form by hydration or serpentinization of peridotites during hydrothermal alteration of oceanic lithospheric mantle before subduction or percolation of fluids released from the downgoing slab into the overlying mantle wedge [Hyndman and Peacock, 2003]. Serpentinization may play a crucial role in the following geological processes:

  1. [3] Serpentinization can lead to a significant reduction in density, generating both a buoyancy force and a huge volume increase causing high pressure within serpentinizing rocks. If they have sufficiently large volumes, serpentine bodies in subducted slabs or mantle wedges may form diapirs which flow upward, forcing the exhumation of ultrahigh and high-pressure (UHP and HP) metamorphic rocks [Guillot et al., 2009; Pilchin, 2005].

  2. [4] Serpentinite has much lower flow strengths than other categories of ultramafic rocks such as peridotite [e.g., Escartín et al., 1997; Hilairet et al., 2007]. Consequently, the presence of serpentine results in strong localization of strain into weak, narrow shear zones with low flow stresses, and cold thermal structures [e.g., Wada et al., 2008; Uchida et al., 2009]. For the same reason, the presence of serpentine affects frictional stability at the downdip edge of subduction megathrusts and is critical for lubricating lithospheric faults. Thus, the presence or absence of serpentine is believed to control the transition between stick-slip to rate-strengthening, stable-sliding, aseismic behavior [e.g., Dragert et al., 2001; Moore and Lockner, 2007]. The distribution of serpentine thus has an important impact on the distribution of seismicity [e.g., Seno, 2005]. Furthermore, the thickness of a weak serpentine layer may control the mechanical coupling and advection pattern above the subducted slab and further the exhumation of UHP and HP metamorphic rocks [e.g., Gerya et al., 2002]. Serpentinites, when thicker than 2–3 km, may form an efficient flow channel at the top of a subducting slab [e.g., Hilairet and Reynard, 2009].

  3. [5] Serpentine is a group of hydrous phyllosilicates containing ~13 wt% water [Schmidt and Poli, 1998; Ulmer and Trommsdorff, 1995]. Global sea level change may be closely linked with hydration or serpentinization of mantle rocks. According to Rüpke et al. [2004], 5 vol% serpentinization would mean ~500 m sea level change over the past 600 Ma. Serpentine in subduction zones provides an important reservoir of water for partial melting of mantle wedges and producing fore-arc volcanism [e.g., Hyndman and Peacock, 2003]. These processes possibly cause deep focus earthquakes [Dobson et al., 2002]. Furthermore, serpentinization may buffer metamorphic fluids to extremely reducing conditions that are capable of producing hydrogen gas [Sleep et al., 2004].

[6] There are three forms of serpentine—lizardite, chrysotile, and antigorite—all with approximate composition Mg3Si2O5(OH)4 but different structures (flat-layered lizardite, rolled-layered chrysotile, and wavelike structure parallel to the a-axis of antigorite). In altered ultramafic rocks collected from ophiolite suites, the serpentines are lizardite and chrysotile, both of which are stable only at temperatures below ~300°C [Evans, 2004] and result from the hydration of olivine and pyroxene by seawater penetration [Hyndman and Peacock, 2003]. Antigorite is stable over a wide range of P-T conditions [up to 700°C at P = 2.5 GPa, Ulmer and Trommsdorff, 1995; Wunder and Schreyer, 1997]. Experiments of Bromiley and Pawley [2003] suggest that a small amount of aluminum can increase the antigorite stability field to higher temperatures. Thus, antigorite is considered as the main hydrous mineral at depths where arc magmas are generated [90–200 km, Schmidt and Poli, 1998; Hyndman and Peacock, 2003].

[7] The seismic properties (Vp, Vs, and Vp/Vs) of partially serpentinized peridotites have been used to constrain the degree of serpentinization in subduction zones and the forearc mantle wedge [e.g., Kamiya and Kobayashi, 2000; Bostock et al., 2002; DeShon and Schwartz, 2004; Matsubara et al., 2008]. However, the variations of seismic velocities and velocity ratios as a function of the serpentine volume fraction, on which the above applications were based, have been calibrated only for serpentinized peridotites containing lizardite and/or chrysotile [Christensen, 1996; Ji et al., 2002] rather than antigorite. The lizardite and/or chrysotile aggregates, which are the products of alteration below 300°C [O'Hanley, 1996; Evans, 2004], are characterized by their extremely low densities (~2.53) and Vp and Vs values (5.10 km/s and 2.32 km/s, respectively, at 600 MPa) and anomalously high Vp/Vs values (2.151) and Poisson's ratio (0.368) [Christensen, 1996, Ji et al., 2002, Wang et al., 2009]. Based on the seismic properties of low-temperature (LT) serpentinized peridotites, a layer with low velocities and Poisson's ratios larger than 0.29 overlying subducting slabs has been interpreted as extensively serpentinized mantle materials [e.g., Abers, 2005; Kawakatsu and Watada, 2007]. However, it is most certainly wrong to apply the seismic properties of LT serpentinized peridotites [e.g., Christensen, 1966, 1996; Carlson and Miller, 1997] to those in forearc mantle wedges where temperatures are in the range of 350°C to 650°C and antigorite is the only stable serpentine.

[8] The seismic properties of antigorite serpentinites have received little study to date. Birch [1960] investigated, at confining pressures up to 1.0 GPa, a serpentinite sample (ρ = 2.614) from Ludlow, Vermont with an antigorite content of 86 vol%. A sample from Stonyford, California (ρ = 2.665), for which seismic velocities were measured by Christensen [1978] at pressures up to 1.0 GPa, is probably the only serpentinite ever studied consisting of >90% antigorite (95% antigorite, 5% other minerals such as opaque and chlorite). Kern et al. [1997] investigated a serpentinized peridotite (ρ = 2.720) containing 74.9% antigorite, 20.3% olivine, 3.8% opaque minerals, and 1.0% others at pressures up to 600 MPa. Recently, Watanabe et al. [2007] studied, at pressures up to 200 MPa, 13 serpentinized peridotite samples with 42.1–88.8 vol% antigorite, in which other minerals are olivine, tremolite, chlorite, and magnetite. Bezacier et al. [2010] measured the elastic constants of a small (~0.150 mm) antigorite plate (ρ = 2.62) picked from a sample from the Escamray massif (Central Cuba), using Brillouin spectroscopy under ambient conditions (0.1 MPa and 20°C). Using the Hill average, they obtained Vp = 6.76 km/s, Vs = 3.83 km/s, and Vp/Vs = 1.77 for an isotropic aggregate of antigorite. The values are remarkably different from the values of antigorite serpentinite reported by Christensen [2004] even at 200 MPa (Vp = 6.54 km/s, Vs = 3.58 km/s and Vp/Vs = 1.83).

[9] Recently, the seismic velocities (Vp, Vs1, and Vs2) and anisotropy of several serpentinite samples were calculated on the basis of the lattice preferred orientation (LPO), density, volume fraction, and elastic constants of antigorite [e.g., Boudier et al., 2010; Katayama et al., 2009; Bezacier et al., 2010; Hirauchi et al., 2010; Jung, 2011; Nishii et al., 2011]. These calculations considered only the dominant phases present in the rocks while minor and accessory minerals are neglected. While informative, such calculated results cannot replace laboratory measurements of rock seismic properties and should be verified by experiments for the following reasons: (1) A constituent mineral with a complex chemical composition may not have exactly the same elastic constants as the single crystal on which the published elastic constants were determined. (2) Single crystal elastic constants are determined on small crystals of gem quality, free of cracks or inclusions, which are relatively rare in natural rocks. In addition, the effects of grain boundaries and phase interfaces that usually contain retrograde or alteration products are not taken into account in the theoretical calculations. In a rock, the crystalline grains themselves act as perfectly elastic units while the interfaces or contacts between grains often display nonlinear elastic behavior. As a result, the rock may be elastically nonlinear and hysteretic [Ji et al., 2007; Sun et al., 2012]. The traditional theory of linear elasticity and resultant averaging methods thus may not accurately describe the overall elastic properties of these so-called nonlinear mesoscopic elastic materials [e.g., Guyer and Johnson, 1999]. (3) For serpentine minerals, not all grains can be indexed by electron back-scattering diffraction techniques. Hence, the LPOs reported in previous papers may be questionable in statistical representations of the bulk rocks. (4) Theoretically, all of the averaging approaches (e.g., Voigt, Reuss, Hill, and geometric averages) used in the calculation of elastic properties and seismic velocities are applicable only to texture-free rocks in which neither shape preferred orientation (SPO) nor LPO occurs. Both SPO and LPO of serpentine are generally strong in naturally deformed serpentinites. Furthermore, it has not been clear which mixture rule should be used to calculate the average elastic properties of a given polymineralic rock. For example, Crosson and Lin [1971] reported that the Voigt average from room pressure elastic constants offers a better prediction of Vp for the quasi-monomineralic Twin Sisters dunite (94.1% olivine, 4.9% orthopyroxene, and 1% chromite and magnetite) than the other averages at pressures above 200 MPa. However, Christensen and Ramananantoandro [1971] and Babuska [1972] found that the Hill average from room pressure elastic constants provided the best prediction for the elastic wave velocities of dunite and bronzitite measured at 1.0 GPa. Recently, Nishii et al. [2011] studied the antigorite LPO of the Higashi-Akaishi garnet peridotite in the Sanbagawa belt (southern Japan) and noted that the Reuss average yields a seismic anisotropy that is approximately treble the value obtained from the Voigt average.

[10] In this study, we report new results on P- and S-wave velocities and anisotropy of 17 serpentinite samples containing >90 vol% antigorite, measured at room temperature and hydrostatic pressures up to 650 MPa. The experimental results are combined with previous data of serpentinized peridotites in order to distinguish the effects of LT and high-temperature (HT) serpentinization and interpret seismic data in subduction zones and forearc mantle wedges.

2 Samples

[11] The serpentinite samples examined in this study (Tables 1, 2), which consist predominantly of antigorite (ϕ>90 vol%) with generally minor to trace amounts of dolomite, magnesite, talc, lizardite, tremolite, pyrrhotite, brucite, and chlorite, were collected from the famous Yushi (Hadabei Town, Xiuyan County) and Wazigou (Gushan Town, Haicheng City) deposits, Liaoning Province, China. These serpentinites, which are described as “Xiuyan jade” in Chinese culture, have been widely used for fashioning ornamental carvings or gems for ~8200 years [e.g., Cai and Chen, 2000; Zhang, 2002; Cao et al., 1994; Wang and Dong, 2011]. The serpentinites, which occur as elongated lens-shaped bodies (up to 30–210 m long and 1–28 m wide) within magnesite-bearing dolomitic marbles of Paleoproterozoic age (Dashiqiao Group), were formed mainly by the following metasomatic reactions between Si-rich aqueous fluids emanating from felsic plutons and dolomitic rocks at 400–550°C and 300–500 MPa [Wan et al., 2002; Wang and Dong, 2011]:

display math(1)
display math(2)
Table 1. Chemical Composition (wt.%) of Each Sample
  1. a

    Samples contain about 1% porosity.

Table 2. Modal Composition, Parameters of Vp-P Curves and Vp Anisotropy for Antigorite-rich Serpentinite Samples From Liaoning, China
SampleModal compositionaλbDensitycV0S.D.DS.D.B0S.D.kS.D.R2A (%)
(Vol%)(g/cm3)(km/s)(10−4 km/s/MPa)(km/s)(10−2 MPa−1)600 MPa
  1. a

    Abbreviations: Atg, Antigorite; Lz, Lizardite; Dol, Dolomite; Mgs, Magnesite; Tlc, Talc; Chl, Chlorite

  2. b

    Direction of P-wave propagation; c Samples contain about 1% porosity. Parameters defined in text.

YSK1AAtg 97.0, Lz 3.0X2.5826.9580.0161.1920.4050.6200.0452.8390.3430.9919.4
YSK1BAtg 98.0, Lz 2.0X2.5817.0420.0181.2270.4470.3990.0302.9910.5140.98311.7
YSK2AAtg 96.0, Lz 4.0X2.5777.2530.0212.5980.4710.4670.0221.8950.2190.99516.5
YSK2BAtg 94.0, Lz 6.0X2.5787.2180.0291.7470.6980.7510.0672.4810.3970.98612.6
YSK3AAtg 100.0X2.5847.3030.0111.9060.2210.1810.0101.1230.1190.99913.6
YSK3BAtg 98.0, Lz 2.0X2.5826.9740.0132.0130.3120.3700.0192.8040.3450.99311.5
YSK4AAtg 98.0, Lz 2.0X2.5886.5800.0291.2640.7120.9100.0762.7560.3900.9876.1
YSK4BAtg 98.0, Dol 1.0,X2.6056.5540.0351.9740.7810.4120.0351.7150.3490.9856.1
Chl 1.0Z2.5796.0830.0313.2120.6860.6750.0301.6110.1710.996
YSK5AAtg 96.0, Lz 4.0X2.5807.2480.0570.7601.4460.7170.1042.9740.9420.94115.2
YSK5BAtg 95.0, Lz 5.0X2.5807.0170.0401.5900.8970.4440.0391.6720.3600.98213.0
WZG2Atg 91.0, Lz 9.0X2.5777.3410.0241.0870.5680.4690.0282.1390.3200.98816.9
WZG3AAtg 91.0, Lz 9.0X2.5756.9020.0292.0500.5960.3240.0261.2040.2250.9947.1
WZG3BAtg 90.0, Lz 10.0X2.5756.9020.0292.0500.5960.3240.0261.2040.2250.9947.8
WZG4Atg 94.0, Lz 1.0X2.5926.8610.0150.8410.3630.3450.0332.4260.4250.9847.7
Mgs 4.0, Chl 1.0Z2.6206.2900.0251.7590.5710.5430.0402.0390.3150.990
WZG5Atg 99.0, Lz 1.0,X2.5866.8800.0171.0630.4431.1580.1154.0380.4540.9905.5
WZG6Atg 100.0X2.5847.2750.0181.5040.3900.1430.0171.7030.5060.98313.8
WZG8Atg 98.0, Tlc 2.0X2.5856.5520.0161.4560.3930.6070.0543.1540.4220.9893.8

[12] The reaction product “calcite” from equation ((2)) was dissolved in the fluid at large fluid-to-rock ratios and removed from the rock volume [Harlow and Sorensen, 2005]. As a result, the metasomatic reaction formed nearly monomineralic aggregates of antigorite.

[13] The magnesium-rich carbonates of the Dashiqiao Group were shallow marine sediments, as indicated by their oxygen, silicon, and carbon isotopic characteristics (δ18O = +22.6‰, δ30Si = +1.6‰, and δ13C = +0.07‰; Wan et al., 2002; Wang and Dong, 2011). The Si-rich aqueous fluids with δ30Si values between +0.1‰ and +0.3‰ are believed to have been derived from dioritic and granitic magmas which had typical δ30Si values between −0.4‰ and +0.4‰ [Cai and Chen, 2000; Wan et al., 2002]. The metasomatic origin of serpentinites is characterized by high contents of SiO2, moderate contents of MgO, and significantly low contents of Al2O3, FeO + Fe2O3, Cr2O3, and TiO2 compared with most of the serpentinites derived from ultramafic rocks (Figure 1). The loss of volatiles during preparation for whole rock analysis (losses on ignition, LOI) of the studied samples ranges from 11.55% to 13.55% by weight (Figure 2). The so-called “loss on ignition”, which is a test used in the chemical analysis of minerals and rocks, consists of strongly heating (“igniting”) a sample at a specified temperature, allowing volatile substances to escape, until its mass ceases to change. As shown in Figure 2, the loss of volatiles is highly linearly dependent on the volume fraction of serpentine (ϕ). This relationship suggests that the volatile materials are dominantly OH components from serpentine with minor amounts of CO32− from relict dolomite and magnesite.

Figure 1.

Chemical variation diagrams showing the comparison between serpentinites derived from metasomatism of dolomite (solid circles, this study) and serpentinization of peridotite (open circles, Ji et al., 2002; Sun et al., 2012, Wang et al., 2005; Wang and Ji, 2009). Weight percent plots of (a) MgO, (b) Al2O3, (c) and FeO + Fe2O3 versus SiO2, and Al2O3 versus CaO (d).

Figure 2.

Content of volatiles (losses on ignition, LOI) as a function of serpentine volume fraction (%) for the studied samples.

[14] The color of a serpentinite sample qualitatively reflects its mineralogical composition. The pure green “Xiuyan jade” consists of essentially pure antigorite as illustrated by X-ray diffraction [Zhang, 2002; Liu et al., 2009], Raman spectra [Liu et al., 2009], and infrared (IR) spectra [Cao et al., 1994]. The characteristic Raman peaks for the antigorite occur at 231, 378, 684, 1048, 1368, and 1397 cm−1. The IR spectra are characterized by an absorption band of OH (3665–3670 cm−1). The antigorite has a refractive index of between 1.55 and 1.56 [Zhang, 2002]. The antigorite serpentinite has compressive and tensile strengths of 216–245 MPa [Zhang, 2002] and 13–23 MPa [Liu et al., 2009], respectively, at the ambient conditions. A yellow “Xiuyan jade” is an antigorite-lizardite mixture, and the lizardite is regarded as a retrograde product of antigorite produced at temperatures of below ~300°C. In contrast, white pockets, lenses, and stringers in a green sample correspond to relicts of dolomite from HT serpentinization. In addition, the transparency of the Xiuyan serpentinites is controlled by the relative contents of Fe2O3 + FeO and by the Fe2O3/FeO ratios of samples [Zhang, 2002; Cao et al., 1994].

[15] The studied samples display densities ranging from 2.575 to 2.600 under ambient conditions (Table 1), mainly reflecting the relative contents of antigorite (ρ = 2.600), lizardite (ρ = 2.550), dolomite (ρ = 2.860), magnesite (ρ = 2.980), and porosity (~1%). In undeformed serpentinites, antigorite shows generally cross-shaped interpenetrating textures (Figure 3a) with random LPO [Wicks and Whittaker, 1977; Hirauchi et al., 2010]. X-shaped interpenetrating textures are occasionally observed in some weakly deformed serpentinites. In ductile shear zones, however, serpentinites with an original interpenetrating texture have been transformed to LS mylonites that developed both stretching lineation (L) and flattened foliation or schistosity (S). The foliation and stretching lineation represent, respectively, the flow plane and flow direction within the shear zone [Passchier and Trouw, 2005]. These mylonites (e.g., samples WZG2, YSK4B, YSK5A, and YSK5B) display well-developed S-C structures [Berthé et al., 1979] indicating dextral simple shear (Figures 3b–d). S-planes or schistosity planes are defined by a planar fabric caused by the alignment of antigorite flakes while C-planes or cisaillement planes form parallel to the shear zone boundary. The angle between the C and S planes is always acute and defines the shear sense.

Figure 3.

Optical photomicrographs of antigorite serpentinites. (a) Undeformed serpentinites, formed by metasomatic reactions between dolomitic rocks and Si-rich aqueous fluids, display a nonpseudomorphous texture composed mainly of interpenetrating blades of antigorite. (b) Strain gradient across the boundary of a dextral shear zone with S-C structure (sample WZG2). (c–d) LS-mylonite of antigorite with S-C structure (sample WZG2). (e) S-serpentinite characterized by well-developed foliation but no lineation (sample YSK3B). (f) L> > S serpentinite with lineation dominating over foliation (sample YSK2B). All images viewed in cross-polarized light except for Figure 3d in which gypsum λ plate has been inserted.

[16] Deformed antigorite grains display abundant optical evidence of dislocation creep such as undulatory extinction, lattice rotation, folding, kinking, recovery-induced subgrains (Figure 4), and dynamic recrystallization that significantly reduced the grain sizes (Figure 3 and 4). The kink band boundaries (KBB, Figures 4c–d) are perpendicular to the maximum principal stress (σ1). Relict porphyroclasts of antigorite are characterized by their orientation with (001) planes at high angles to the X- or Y-directions and small angles to the Z-direction (Figures 3c–d and Figures 4c–d). The X-direction is parallel to the stretching lineation; the Y-direction is normal to the lineation in the foliation plane, and the Z-direction normal to the foliation plane. The porphyroclasts have generally undergone extensive kinking, lattice bending and/or dynamic recrystallization to rotate their (001) planes progressively towards the bulk shear plane (Figures 4c–d). The recrystallized neograins show lath- or leaf-shaped or flaky habits with their short and long axes parallel to the a- and c-axes, respectively, reflecting their anisotropic growth during ductile deformation in the presence of fluids [Liu et al., 2009]. Petrographic images with an auxiliary gypsum plate show a strong optical fabric (Figure 3d). Scanning electron microscopy-energy dispersive X-ray analyses [Liu et al., 2009] suggested that antigorite crystals in each sample may vary in grain size from site to site but have little variation in chemical composition. S > L (foliation dominates over lineation, samples YSK3A and YSK3B, Figure 3e) and L > S (lineation dominates over foliation, samples YSK2A and YSK2B, Figure 3f) serpentinites were also investigated. A few samples contain some lizardite derived from retrograde metamorphism of antigorite (Table 2).

Figure 4.

Optical evidence for dislocation creep of antigorite: (a) undulatory extinction, (b) recovery-induced subgrains, (c–d) lattice rotation and kinks in relict porphyroclasts, and recrystallized new grains (R). Kink band boundaries (KBB) in Figures 4c–d are approximately normal to σ1. Subgrains indicated by S.

[17] In order to investigate whether any observed seismic anisotropy is caused by compositional heterogeneity or the LPO of constituent minerals, it is necessary to quantify the chemical compositions of three orthogonal minicores from each sample. The data given in Table 1 are the results averaged from these three orthogonal minicores from each sample. The compositional anisotropy of a given chemical component i is defined as: A(i) = (Cmax − Cmin)/Cm × 100%, where Cmax, Cmin, and Cm are, respectively, the maximum, minimum, and mean contents of component i measured for the three orthogonal minicores. All samples except WZG4 show A(SiO2) < 3% (Figure 5a) and all except sample YSK4B display A(MgO) < 2.4% (Figure 5b). Even for LOI, most of the samples show less than 5% compositional anisotropy (Figure 5c).

Figure 5.

Histograms displaying distributions of compositional anisotropy for (a) SiO2, (b) MgO, and (c) LOI (loss on ignition) in weight percent.

3 Experimental Procedure

[18] High-pressure velocity measurements using standard pulse transmission methods [Birch, 1960] were carried out at the GSC/Dalhousie High Pressure Laboratory in Halifax, Nova Scotia [Ji et al., 1993, 2007; Wang and Ji, 2009; Sun et al., 2012]. Three cylindrical minicores, 2.54 cm in diameter and 3–6 cm in length, were cut from each sample in the X-, Y-, and Z-directions. The pressure apparatus is a seven ton, double-walled steel vessel with a 40 cm long by 10 cm diameter working chamber, which can operate to a pressure up to 1.4 GPa. P-waves were generated and received by lead zirconate transducers with a 1 MHz resonance frequency, while S-waves were generated and received using lead zirconate-titanate transducers. In order to prevent the pressure medium from invading the sample during the pressure runs, the minicores were sheathed in impermeable thin copper foil, and the entire sample/transducer/electrode assembly was enclosed in neoprene tubing. Once the sample assembly was sealed in the pressure vessel and the pressure was raised, a high voltage spike from a pulse generator excited the sending transducer and the time of flight to the receiving transducer was measured using a digital oscilloscope. The velocities were not corrected for changes in sample length with pressure. The confining pressure was determined to within 0.1 MPa by direct digital readout from a calibrated strain gauge. The accuracy is estimated to be 0.5% for Vp and 1% for Vs [Christensen, 1985; Ji et al., 1993; Ji and Salisbury, 1993].

4 Results

4.1 Velocity-Pressure Relationship

[19] Typical Vp-P and Vs-P curves measured during depressurization are shown in Figures 6 and 7, respectively. Three Vp measurements (i.e., along the X-, Y-, and Z-directions) and six Vs measurements (i.e., with geometric configurations XY, XZ, YX, YZ, ZX, and ZY, where the first letter refers to the propagation direction and the second to the polarization direction) were performed for each sample using three orthogonal minicores. The curves display a rapid, nonlinear increase in velocity with pressure at low pressures (generally <150 MPa for antigorite serpentinites) and then increase slowly and linearly in velocity at high pressures. The critical pressure at which the transition from the nonlinear regime to the linear regime occurs is inferred as the microcrack-closure pressure [Pc, Christensen, 1978; Ji et al., 1993]. The velocity-pressure curves above Pc can be reproduced in the laboratory, while the curves below this pressure may vary slightly from run to run. This suggests that the rocks become crack-free, compact aggregates in the linear high pressure regime, while in the nonlinear low-pressure regime, seismic velocities are still sensitive to the state of microcracks (e.g., the ratio of crack aperture to length) implying that the rocks still possess discrete memories of their past pressure history in this interval [Ji et al., 2007].

Figure 6.

P-wave velocity (Vp) versus pressure in three orthogonal directions (X, Y, and Z) through samples (a) YSK3B, (b) YSK2B, (c) YSK1A, and (d) YSK5A. Information about the orientation of each minicore is given in Table 2.

Figure 7.

S-wave velocity (Vs) versus propagation and polarization directions in samples (a) YSK1B and (b) YSK2A as a function of pressure. The first letter signifies propagation direction and the second letter the polarization direction. The three minicores are orthogonal, and the information about the orientation of each minicore is given in Table 3.

Table 3. Parameters of Vs-Pressure Curves, Shear-wave Splitting and Anisotropy for Antigorite-rich Serpentinite Samples From Liaoning, China
SampleλaDensitybV0S.D.DS.D.B0S.D.kS.D.R2ΔVs (km/s)As (%)
(g/cm3)(km/s)(10−4 km/s/MPa)(km/s)(10−2 MPa−1)600 MPa
  1. a

    The first letter refers to the propagation direction and the second to the polarization direction.

  2. b

    Samples contain about 1% porosity.


[20] Both P- and S-wave velocities as a function of confining pressure were fitted to the following equation [e.g., Ji et al., 2007]:

display math(3)

where V0 is the reference velocity representing the intrinsic velocity of the nonporous or crack-free rock at zero pressure, which is determined by extrapolating the linear velocity-pressure relationship obtained at high pressures to zero pressure; D is the intrinsic pressure derivative of velocity in the linear elastic regime; B0, which is the ambient velocity drop caused by the presence of pores/microcracks at zero pressure, determines the maximum magnitude of the velocity increases due to the closure of pores and microcracks; and k, which is the decay constant of the velocity drop, controls the shape of the nonlinear segment of the velocity-pressure curve. In equation ((3)), V0 and D are two parameters which describe the intrinsic seismic properties of the microcrack- or pore-free solid matrix, while B0 and k are parameters which describe the extrinsic seismic properties related to the porosity and geometrical shape of pores (e.g., aspect ratio, spatial arrangement, orientation, and size distribution).

[21] Parameters V0, D, B0, and k and their standard deviations were determined for the P- and S-wave velocities of each minicore sample during depressurization using a least square regression method and are given in Tables 2 and 3, respectively. As indicated by the goodness-of-fit coefficients (mostly R2 ≥ 0.95), the pressure-velocity curves for P- and S-waves can be well represented by equation ((3)). Figure 8 summarizes V0, D, B0, and k values for P-wave velocities measured along each direction (i.e., X, Y, and Z) and also shows the arithmetic mean of these three values from the three orthogonal directions in the 17 samples studied. Each of the mean values corresponds to the seismic properties of an equivalent isotropic rock. For the P-wave velocities, V0 varies from 6.912 to 6.338 km/s with an average value of 6.626 km/s, D ranges from 1.055 × 10−4 to 2.437 × 10−4 km/s/MPa with an average value of 1.749 × 10−4 km/s/MPa, B0 lies between 0.247 km/s and 0.845 km/s with a mean value of 0.533 km/s, and k varies from 1.54 × 10−2 and 3.468 × 10−2 MPa−1 with an average of 2.443 × 10−2 MPa−1.

Figure 8.

Statistics of Vp-P curve parameters, defined by equation ((3)), for three structural directions (X, Y, and Z) and the equivalent isotropic aggregate of the antigorite serpentinite samples.

4.2 Mean Velocities (Vp and Vs) and Velocity Ratios (Vp/Vs)

[22] The average seismic velocities of each anisotropic rock, which are equivalent to the properties of its isotropic counterpart, can be computed using the following relations [Ji et al., 2003; Wang and Ji, 2009]:

display math(4)
display math(5)

[23] The P-wave velocities of the 17 equivalent isotropic serpentinites (inline image at 600 MPa), calculated from equation ((4)), range from 6.47 km/s to 7.01 km/s with a mean value of 6.73 km/s. This mean velocity is very close to the value (6.76 km/s) computed from the Hill averages of antigorite elastic constants [Bezacier et al., 2010] but higher than the value (6.54 km/s) given in Christensen [2004]. The S-wave velocities of 11 equivalent isotropic serpentinites (inline image at 600 MPa), calculated from equation ((5)), vary from 3.66 km/s to 3.79 km/s with a mean value of 3.71 km/s, which is 3.1% lower than the value (3.83 km/s) calculated from the Hill averages of antigorite elastic constants [Bezacier et al., 2010] but 3.5% higher than the value (3.58 km/s) provided by Christensen [2004]. It is not surprising that our velocities, which were measured on natural serpentinites, are slightly lower than the theoretical value computed from the elastic constants of antigorite [Bezacier et al., 2010], which were determined on a small crystal of gem quality, free of cracks, inclusions, or altered grain boundaries.

[24] In Figure 9, the arithmetic mean Vp and Vs values of each sample at 200 and 600 MPa are plotted versus the volume fraction (ϕ) of LT (lizardite and chrysotile) and HT (antigorite) serpentines. The least-squares linear fitting of the data from LT and HT serpentine-bearing samples yield the following relations:At 200 MPa,

display math(6)
display math(7)
display math(8)
display math(9)

At 600 MPa,

display math(10)
display math(11)
display math(12)
display math(13)
Figure 9.

Mean P- and S-wave velocities ((a–b)Vp; (c–d) Vs) measured at 200 MPa (a and c) and 600 MPa (b and d) as a function of serpentine volume fraction (ϕ) for low-temperature (LT, open dots, Iturrino et al., 1996; Miller and Christensen, 1997; Ji et al., 2002; Wang et al., 2005; Ji et al., 2007; Watanabe et al., 2007; Wang and Ji, 2009; Sun et al., 2012) and high-temperature (HT, solid squares from this study, and open squares from the references; Birch, 1960; Christensen, 1978; Kern et al., 1997; Watanabe et al., 2007) serpentinization. N: number of samples measured.

[25] The data for LT serpentinized ultramafic samples were taken from Wang et al., 2005, Ji et al. [2007]; Sun [2011] and the Handbook of Seismic Properties of Minerals, Rocks and Ores [Ji et al., 2002]. The data for HT serpentinized ultramafic samples are from Birch [1960], Christensen [1978], Kern [1993], Kern et al. [1997], and Watanabe et al. [2007], and this study. Extrapolation of the linear equations to ϕ = 1 yields Vp = 5.10 and 6.68 km/s for LT and HT serpentinites, respectively, at 600 MPa. Clearly, HT serpentinites and serpentinized peridotites have significantly higher velocities than their LT counterparts, as argued by Christensen [2004], Reynard et al. [2007], Mookherjee and Stixrude [2009], and Bezacier et al. [2010]. The velocity differences (ΔVp) between HT and LT serpentinized peridotites increase linearly with the degree of serpentinization: ΔVp = 1.64ϕ at 200 MPa. The scatter in the data is caused mainly by differences in the contents of constituent minerals (e.g., garnet, clinopyroxene, magnetite, and spinel) other than serpentine or olivine.

[26] The velocity ratio (Vp/Vs), which is a constant for an equivalent isotropic rock at a given temperature and a given pressure, has been regarded as a discriminant of composition for the Earth's interior and provides valuable constraints on its formation and evolutionary processes [e.g., Zandt and Ammon, 1995; Ji et al., 2009; Matsubara et al., 2009; Wang and Ji, 2009]. Typical Vp/Vs-P curves are shown in Figure 10a. For each of the studied serpentinites, the velocity ratio Vp/Vs remains quasi-constant at pressures above the crack-closure pressure (~150 MPa). Our results are consistent with previous investigations [e.g., Christensen, 1996; Wang and Ji, 2009]. At 600 MPa, the velocity ratios of the antigorite serpentinites range from 1.74 to 1.85 with an average value of 1.81, which is higher than the value (1.77) calculated from the elastic constants of an antigorite single crystal from the HP Escambray massif (Central Cuba) using the Hill average, but lower than the value (1.84) suggested by Christensen [2004]. The velocity ratio of 1.81 for antigorite is significantly lower than the value 2.15 shown in Figure 11b or 2.14 [Christensen, 2004] for lizardite at 600 MPa. Our data combined with previous results from Christensen [1978], Kern et al. [1997], and Watanabe et al. [2007] yield a correlation between the velocity ratio (Vp/Vs) and the volume fraction of antigorite (ϕHT): Vp/Vs = 1.78 + 0.01ϕHT at 200 MPa (Figure 11a) and Vp/Vs = 1.77 + 0.04ϕHT at 600 MPa (Figure 11b). For LT serpentinized peridotites, however, Vp/Vs = 1.78 + 0.31ϕLT at 200 MPa (Figure 11a) and Vp/Vs = 1.77 + 0.38ϕLT at 600 MPa (Figure 11b). Clearly, the correlations are remarkably different for the LT and HT serpentinized peridotites.

Figure 10.

(a) Graph of velocity ratio Vp/Vs versus confining pressure for representative antigorite serpentinites (this study). (b) Vp, Vs, and Vp/Vs as a function of temperature for an antigorite-rich serpentinite measured at 600 MPa (Original data from Kern et al., 1997).

Figure 11.

Velocity ratios Vp/Vs measured at (a) 200 MPa and (b) 600 MPa as a function of serpentine volume fraction (ϕ) for low-temperature (LT, open dots) and high-temperature (HT, solid squares from this study, and open squares from references) serpentinization. N: number of samples measured.

4.3 P-wave Velocity Anisotropy

[27] Vp anisotropy (Ap) is defined as: AP = (Vmax − Vmin)/Vm × 100% [Birch, 1961], where Vmax, Vmin, and Vm are, respectively, the maximum, minimum, and mean values of the P-wave velocities measured in a given sample along three orthogonal propagation directions. The intrinsic anisotropy of the samples, calculated from the V0 values measured in three orthogonal directions, ranges from 3.7% (sample WZG8) to 17.9% (sample WZG2) with an average value of 10.9%. The values display almost no difference from those at 600 MPa, which ranges from 3.8% (sample WZG8) to 16.9% (sample WZG2) with an average value of 10.5%.

[28] Three typical patterns have been distinguished for the dependence of Vp anisotropy on pressure (Figure 12):

  • Pattern 1 (Figure 12a): With increasing pressure, the anisotropy decreases rapidly below 100–150 MPa and then slowly decreases or reaches a constant value above this pressure (e.g., samples YSK1A, YSK2A, and YSK3A). This pattern can be attributed to the closure of aligned microcracks which reinforce the anisotropy induced by the serpentine LPO [e.g., Ji et al., 1993].
  • Pattern 2 (Figure 12b): The anisotropy increases first with increasing pressure in the low-pressure range (<100–150 MPa) and then reaches a nearly constant value above this pressure (e.g., samples YSK4A, YSK5A, and WZG8). This pattern can be attributed to the rapid closure at low pressures of aligned microcracks which oppose the LPO-induced anisotropy [Ji et al., 2007].
  • Pattern 3 (Figure 12c): The anisotropy remains almost unchanged with increasing pressure (e.g., samples YSK3B, YSK5B, and WZG3A). This pattern, which is formed by the combination of Patterns 1 and 2, is probably caused by two sets of orthogonal microcracks that have opposite effects on velocity. Thus, the anisotropy is almost directly related to the contribution of serpentine LPO at each pressure.
Figure 12.

Three types of P-wave velocity (Vp) anisotropy versus pressure curves for antigorite serpentinites (see text).

4.4 S-wave Splitting

[29] Vs anisotropy (As) of each sample can also be defined as inline image, where Vmax, Vmin, and Vm are, respectively, the maximum, minimum, and mean values of the S-wave velocities measured along three propagation directions and six polarization directions. A common feature is that As displays a rapid decrease with increasing pressure below ~150 MPa and then reaches a constant value above this pressure (Figure 13a). The intrinsic Vs anisotropy of the serpentinites at 600 MPa ranges from 3.6% (sample WZG5) to 18.3% (sample WZG2) with an average value of 10.4%.

Figure 13.

(a) S-wave velocity (Vs) anisotropy versus pressure curves for antigorite serpentinites. (b) Three types of shear-wave splitting versus pressure curves observed for antigorite serpentinites (see text).

[30] For each propagation direction Λ, shear-wave splitting (ΔVs) Λ is defined as the difference in velocity between fast and slow split, or polarized, S-waves propagating in the same direction Λ. The splitting is both direction and pressure dependent. The variation of shear-wave splitting with pressure can be classified into three patterns (Figure 13b):

  • Pattern 1 displays a pattern in which the splitting increases rapidly, then becomes nearly constant (e.g., X-directions of sample YSK1A and YSK5A). Pattern 1 splitting can be interpreted as destructive interference of the effects of oriented microcracks and LPOs of antigorite. In other words, in the samples showing Pattern 1 splitting, microcracks are preferentially oriented in such a manner that the microcrack-induced splitting opposes the LPO-induced splitting. Microcracks in samples YSK1A and YSK5A (Figure 13b) are inferred to be principally perpendicular to the Y-direction and thus parallel to the XZ plane. The presence of these microcracks causes a decrease in Vs(XY) but has little influence on Vs(XZ). Accordingly, the progressive closure of these microcracks with increasing pressure increases the Vs(XY)- Vs(XZ) splitting in the low-pressure regime. Almost no shear-wave splitting can be observed in the X-direction of sample YSK5A at room pressure because the microcrack- and LPO-induced anisotropies entirely cancel each other (Figure 13b). For the same reason, microcracks in YSK1B (Figure 13b) can be inferred to be preferentially perpendicular to the lineation direction (X). Interestingly, Vs(XZ) is significantly faster than Vs(XY) in sample YSK1A below 30 MPa, indicating that the splitting below this pressure is essentially due to the preferred orientation of microcracks perpendicular to the Y-direction.
  • Pattern 2 displays an initial rapid decrease in splitting with increasing pressure in the low-pressure regime and then approaches a quasi-constant value in the high-pressure regime as in the direction of sample WZG4 or increases slightly with increasing pressure until a constant value is reached at higher pressures (Y-direction of sample WZG2). For this pattern, the quasi-constant splitting above the crack-closure pressure (~150 MPa) should be related to the LPO of serpentine whereas the splitting below the crack-closing pressure is caused by constructive interference between oriented microcracks and the serpentine LPO. In sample WZG2, for example, the rapid decrease in splitting in the low-pressure regime is caused by a progressive increase in Vs(YZ) due to the closure of the foliation-parallel cracks while Vs(YX) is not affected by the presence of these cracks.
  • Pattern 3 shows little variation in splitting with increasing pressure (Figure 13b). This pattern of shear-wave splitting corresponds almost directly to the LPO-induced intrinsic properties of the samples in either the low or the high-pressure regimes. In sample YSK1B, for instance, the microcracks that are inferred to be aligned normal to the Y- and Z-directions are so equally developed that their influences on Vs(XY) and Vs(XZ) are almost equal as well. As a result, the difference between Vs(XY) and Vs(XZ) remains almost constant with increasing pressure (Figure 13b).

[31] Of particular interest is the intrinsic shear-wave splitting above the crack-closing pressure, which is caused by the LPO of antigorite. For the serpentinites deformed dominantly by simple shear (e.g., sample WZG2) or coaxial flattening (e.g., sample YSK3B), significant shear-wave splitting (up to 0.44 km/s at 600 MPa) occurs in the X- and Y-directions with the fast shear wave being polarized parallel to the foliation (Figure 14 and Table 3). In the direction normal to the foliation, however, little splitting (mostly <0.05 km/s) has been detected. In sample YSK1B (Figure 7a), for example, shear-wave splitting for the propagation parallel to the X- and Y-directions is almost the same and equal to 0.40 km/s whereas splitting normal to the foliation is only 0.07 km/s. For serpentinites deformed dominantly by coaxial stretching (e.g., samples YSK2A), the shear-wave splitting is significantly stronger along the Y-direction than the X-direction (Figure 7b and Figure 14).

Figure 14.

Shear-wave splitting along the Y-direction [Vs(YX)- Vs(YZ)] versus shear-wave splitting along the X-direction [Vs(XY)- Vs(XZ)].

5 Discussion

5.1 Interpretation of Slow Regions in the Mantle Wedge

[32] As shown in Figure 9, HT serpentinites and LT serpentinized peridotites display different correlations between seismic velocities and the volume fraction of serpentine (ϕ). At 200 MPa, for example, Vp = 8.00 − 3.11ϕLT (R2 = 0.92, N = 129), Vp = 8.00 − 1.47ϕHT (R2 = 0.32, N = 34), Vs = 4.51 − 2.23ϕLT (R2 = 0.92, N = 109), and Vs = 4.51 − 0.94ϕHT (R2 = 0.45, N = 26). For a given ϕ value, the HT serpentinites and serpentinized peridotites have significantly higher P- and S-wave velocities than their LT counterparts. The velocity differences (ΔVp and ΔVs) between HT and LT serpentinized peridotites increase linearly with the degree of serpentinization: ΔVp = 1.64ϕ and ΔVs = 1.29ϕ at 200 MPa. The above differences occur equally at other pressures or depths. This fact implies that serpentine contents in hydrous slabs and mantle wedges where temperature is >300°C, inferred from the observed seismic velocities, should be at least twice that of previous estimates based on LT serpentinization [e.g., Christensen, 1972, 1996; Horen et al., 1996; Wang et al., 2009].

[33] Slow regions in mantle wedges above subducting slabs have been reported from earthquake tomography (Table 4) and interpreted in terms of the seismic properties of LT serpentinites [e.g., Hyndman and Peacock, 2003; Brocher et al., 2003; Ramachandran et al., 2006]. For example, P-wave velocities as low as 7.16 km/s have been documented in the mantle wedge at depths near 40 km beneath the Cascadia margin [Rondenay et al., 2001]. Antigorite should be dominant in the partially serpentinized peridotites at such depths (>300°C, Table 4). According to our results, this Vp value corresponds to ~66 vol% HT serpentinization, which is about twice the estimate (~31 vol%) based on LT serpentinization [Carlson and Miller, 2003; Hacker et al., 2003]. Based on the seismic properties of LT serpentinized peridotites, Zhao et al. [2001] interpreted the low velocities (7.2 to 7.4 km/s) within the mantle wedge down to 60 km beneath the southeastern end of Vancouver Islands and the Puget Lowland as evidence for 15–20% serpentinization of the mantle wedge. However, our new results suggest 49–63% HT serpentinization in the region where temperatures should be above 300°C (Table 4).

Table 4. Interpretation of Seismic Wave Velocities in the Forearc Mantle Wedge Above Subduction Zones
RegionVelocityDepthaTemperaturebDegree of Serpentinization (vol%)Reference
(km/s)(km)(°C)LT SerpentineHT Serpentinec
  1. aDepth range in which low velocities were observed. bTemperature range estimated for the low-velocity zone according to the reference cited and Peacock [2003] and Syracuse et al. [2010]. cThe volume fraction of antigorite, which is the serpentine stable above 300°C, is estimated according to the results of this study. dSerpentine mud volcanos.

Cascadia marginVp = 7.2–7.635–60300–600 35–63%Brocher et al. [2003]
Cascadia marginVp = 7.2–7.637–45300–500 35–63%Ramachandran et al. [2006]
Cascadia marginVp = 7.2–7.430–60400–600 49–63%Zhao et al. [2001]
Cascadia marginVp = 7.1625–40450–600 66%Rondenay et al. [2001]
Chile marginVp = 7.750–75400–650 28%Graeber and Asch [1999]
Cascadia marginVs = 3.2535–45500–600 100% + fluidBostock et al. [2002]
Izu-Bonin Trench (31°N)Vp = 6.4–7.310–25<20027–57%d Kamimura et al. [2002]
Kanto, Central JapanVp = 6.920–45400–600 84%Kamiya and Kobayashi [2000]
Kanto, Central JapanVp = 6.7–7.230–45400–600 63–99%Seno et al. [2001]
Kii Peninsula, southwestern JapanVp = 7.0–7.530–40350–450 42–77%Matsubara et al. [2009]
Western Shikoku, JapanVp = 7.0–7.530–45400–450 42–77%Matsubara et al. [2009]
Northeast JapanVp = 6.8–7.230–60300–600 63–92%Uchida et al. [2009]
Mariana convergence marginVs = 3.6040–55400–600 100%Tibi et al. [2008]
Nicoya Peninsula, Costa RicaVp = 7.2–7.632–60320–600 35–63%DeShon and Schwartz [2004]
South Andes (36–40°S)Vp = 7.2–7.335–55500–600 56–63%Bohm et al. [2002]

[34] Bostock et al. [2002] found very low S-wave velocities (~3.25 km/s) within the cold forearc mantle in the corner of the mantle wedge located in the southern Cascadia subduction zone at the depths of 35–45 km (Table 4). These velocities are ~10% lower than the lower crustal velocity (~3.6 km/s) immediately above the continental Moho. Based on the seismic properties of LT serpentinized peridotites [Christensen, 1966], they suggested a degree of serpentinization of 50–60%. In the cold forearc mantle below depths of 40 km [<600°C, Hyndman and Peacock, 2003; Syracuse et al., 2010], antigorite is expected to prevail. According to our results, S-wave velocities as low as 3.25 km/s cannot result from HT serpentinization alone because Vs = 3.67 km/s for rock composed of pure antigorite. Thus, the inverted continental Moho velocity reported by Bostock et al. [2002] may be caused by high-pressure fluids rather than serpentinization of the forearc mantle alone.

[35] Graeber and Asch [1999] also reported a 25 km thick zone beneath the Chile margin with P-wave velocities as low as 7.7 km/s and Vp/Vs ratios as high as 1.84. Carlson and Miller [2003] interpreted the properties as results of LT serpentinization with 0–13 vol% lizardite. If the partially serpentinized mantle rocks contain antigorite rather than lizardite at the depths of interest, the properties would be consistent with serpentinite contents of 0 to 28 vol%. Similarly, a decreased Vp [7.2–7.6 km/s between 32 and 60 km depth, 320–600°C, DeShon and Schwartz, 2004] in the forearc mantle wedge along the Nicoya Peninsula, Costa Rica, suggests a HT serpentine (antigorite) volume fraction of 35–63%.

[36] Tibi et al. [2008] documented a 10–25 km thick low-velocity zone (Vs = 3.6 km/s), whose upper boundary occurs at about 40 − 55 km depth, in the forearc mantle wedge beneath the Mariana convergent margin. They interpreted the Vs value as a result of 30–50 vol% lizardite serpentinization. However, antigorite should be the stable serpentine at such depths as temperature is most likely >300°C. Hence the Vs value of 3.6 km/s should infer a pure antigorite serpentinite, corresponding to a chemically bound water content of about 13 wt%.

[37] Kamiya and Kobayashi [2000] observed Vp = 6.9 km/s and Vs = 3.4 km/s in the mantle wedge immediately above the subducting Philippine Sea slab beneath the Kanto district, central Japan. The properties are consistent with lizardite contents of 40–50 vol% if T < 300°C. However, antigorite most likely exists in the mantle wedge where temperatures are estimated to be 400–600°C [Iwamori, 2000]. The P-wave velocity of 6.9 km/s corresponds to an antigorite volume fraction of 84% (Table 4). Furthermore, the existence of high-pressure fluid in the mantle wedge may also contribute to the observed low Vs value (3.4 km/s) which is even lower than that of pure antigorite serpentinite (3.67 km/s).

[38] Based on the seismic properties of LT serpentinized peridotites, Seno et al. [2001] suggested that P-wave velocities of 6.7–7.2 km/s within the subducted slab (30–45 km depth, 400–600°C) beneath the Kanto District of Japan result from 30–47 vol% serpentinization. According to the seismic properties of HT serpentinized peridotites, however, the cited Vp values are consistent with 63–99 vol% HT serpentinization within the subducted slab. In addition, seismic tomographic studies [Zhao et al., 2000] observed a significant low-velocity anomaly (−6%) within the subducted slab beneath the Kii Peninsula. The seismic properties of HT serpentinization suggest an antigorite content of 33–34 vol% (Figure 9) if temperature is higher than 300°C at the depths of interest.

[39] In addition, if peridotites contain >40% lizardite or >84% antigorite, their P-wave velocities become lower than gabbro or mafic gneiss (~7.9 km/s). The Moho, defined by the interface between ultramafic mantle rocks and gabbro or mafic gneiss, will become invisible in seismic reflection as seismic impedance contrasts disappear. If hydration due to fluid trapped at the base of the continental crust occurs in the uppermost part of the mantle, low P- and S-wave velocities of antigorite serpentinites [e.g., Vp(Z) = 6.27 km/s, Vs(ZX) = 3.60 km/s and Vs(ZY) = 3.58 km/s for sample YSK5A] may cause them to be mistaken for felsic crustal rocks, leading to an overestimation of the crustal thickness.

5.2 Velocity Ratios (Vp/Vs)

[40] Christensen [1996] proposed that serpentinites can be distinguished from other rocks by their anomalously high Vp/Vs or Poisson's ratios. As illustrated in Figure 11, the velocity ratio Vp/Vs increases systematically with the volume fraction of LT serpentine (ϕLT): Vp/Vs = 1.78 + 0.31ϕLT at 200 MPa, and Vp/Vs = 1.77 + 0.38ϕLT at 600 MPa. The Vp/Vs ratio at 600 MPa varies from 1.77 for unaltered peridotite to 2.15 for pure LT serpentinite. However, the Vp/Vs ratio shows little increase with increasing antigorite serpentinization (ϕHT): Vp/Vs = 1.78 + 0.01ϕHT at 200 MPa, and Vp/Vs = 1.77 + 0.04ϕHT at 600 MPa. Vp/Vs = 1.81 for pure antigorite aggregates at 600 MPa while the velocity ratios calculated using the Reuss, Voigt, and Hill averages of the antigorite elastic constants determined experimentally by Bezacier et al. [2010] are 1.86, 1.70, and 1.77, respectively. Clearly any pronounced increase in the Vp/Vs ratio can only be due to LT serpentinization (<300°C). Thus, the argument of Christensen [1996] is not applicable to seismic velocities in portions of the mantle wedge with temperatures higher than 300°C.

[41] Figure 11a illustrates that the velocity ratio Vp/Vs remains quasi-constant at confining pressures above the crack-closure pressure (~150 MPa) for most of the serpentinites studied here. Commonly, the Vp and Vs values for crystalline rocks display linear decreases with increasing temperature (Figure 10b) due to thermal effects (e.g., thermal dilatation of mineral lattices, microcracking and grain boundary widening induced by differential thermal expansion) as long as no partial melting, metamorphic reaction, dehydration, or phase transformation takes place during the measurement of seismic velocities. Thus, the variation in Vp/Vs with temperature depends principally on the relative value of dVp/dT with respect to dVs/dT. For an antigorite serpentinite (sample 987a) investigated by Kern [1993], Vp = 7.195−2.277 × 10−4 T (R2 = 0.96), and Vs = 3.635−3.130 × 10−4 T (R2 = 0.99). As a result, Vp/Vs = 1.979 + 1.126 × 10−4 T (R2 = 0.97) at temperatures of 20–500°C and 600 MPa. For sample 755 [Kern et al., 1997], Vp = 7.050−3.385 × 10−4 T (R2 = 0.86), and Vs = 3.666−4.355 × 10−4 T (R2 = 0.86). As a result, Vp/Vs = 1.921 + 1.513 × 10−4 T (R2 = 0.80) at temperatures of 20–700°C and 600 MPa (Figure 10b). It is unclear why the Vp/Vs ratios of the antigorite serpentinite samples investigated by Kern [1993, Vp/Vs = 1.98] and Kern et al. [1997, Vp/Vs = 1.93] are higher than those reported by Christensen [1978, Vp/Vs = 1.84] and our study (Vp/Vs = 1.73–1.84 with a mean value of 1.81) under the same conditions.

[42] As noted earlier, the Voigt, Hill, and Reuss averages yield Vp/Vs values of 1.70, 1.76, and 1.86, respectively, for a pure antigorite aggregate [Bezacier et al., 2010]. The value of Christensen [1978] is close to the Reuss average of Bezacier et al. [2010], which corresponds to the lower bound based on an assumption that the stress is uniform in the aggregate. The mean value obtained from our study is close to the Hill average, which differs very little from the geometric mean. Both the Hill and geometric averages commonly offer the best prediction for the overall mechanical properties of polycrystalline aggregates and multiphase composites [e.g., Berryman, 1995; Ji and Xia, 2002; Mainprice and Ildefonse, 2009]. Clearly, Vp/Vs increases slightly with increasing temperature if dVp/dT is smaller than dVs/dT in serpentinites (Figure 10b). Thus, the Vp/Vs ratio displays only slight dependence on pressure or temperature in the linear elastic regime where microcracks are fully closed (Figure 10).

[43] Considering that the mean Vp/Vs values are 1.77, 1.81, and 2.15 for unaltered peridotite, HT and LT serpentinites, respectively, we suggest that the 25 km thick zone (Vp/Vs =1.75–1.84) beneath the Chile margin [Graeber and Asch, 1999] has undergone inhomogeneous HT rather than LT serpentinization. Matsubara et al. [2009] reported a zone of low Vp (7.0–7.5 km/s), low Vs(3.8–4.1 km/s), and high Vp/Vs(1.80–1.83) at depths of 30–40 km (350–450°C) beneath the Tokai, Shikoku, and Kii regions in southwestern Japan. The data indicate extensive antigorite serpentinization (42–85 vol%) within the mantle wedge of the Eurasian plate owing to the fluids dehydrated from the subducting Philippine Sea plate. The presence of high-pressure fluids may also contribute to the observed high Vp/Vs ratios. The above interpretation is consistent with the occurrence of nonvolcanic tremors and slow-slip events at depths of 30–40 km beneath southwestern Japan.

[44] Deshon and Schwartz [2004] detected a region with Vp/Vs = ~1.85 within the forearc mantle wedge (32–60 km depth) along the Nicoya Peninsula, Costa Rica and interpreted it as a result of serpentinization. The relation between the Vp/Vs ratio and the volume fraction of LT serpentine yields ~21% serpentinization. However, temperature within the forearc mantle wedge (320–600°C) is too high for lizardite or chrysotile to be stable. Based on the seismic properties of HT serpentinized peridotites, however, it is impossible to obtain a Vp/Vs ratio as high as 1.85 unless the wedge mantle is fluid rich or partially molten. Thus, the forearc mantle wedge along the Nicoya Peninsula (Costa Rica) observed by DeShon and Schwartz [2004] can be interpreted as due to the presence of both antigorite and fluid.

5.3 Antigorite-Induced Seismic Anisotropy

[45] The effects of serpentinization on the seismic anisotropy of mantle rocks are still controversial. Kern and Tubia [1993] showed that Vp anisotropy declined from 6–8% in fresh lherzolite to <2% in more serpentinized samples. Horen et al. [1996] reported a decrease in both Vp and Vs anisotropy in peridotites from the Xigaze ophiolite (Tibet) with increasing degrees of serpentinization from 3% to 70%. Schmitt et al. [2007] reported that Vp anisotropy decreases monotonically from over 12% to nearly zero with increasing serpentinization. All of the samples studied above are characterized by the fact that serpentinization took place in the brittle deformation regime at low temperatures (<300°C). The LT serpentinization usually started from grain boundaries and multiple-oriented microfracture networks, forming a “mesh-texture” [Wicks and Whittaker, 1977; Maltman, 1978]. The (001) planes of lizardite are commonly aligned parallel or perpendicular to the original microcracks and grain boundaries [e.g., Dewandel et al., 2003; Boudier et al., 2010], forming an overall complex and random texture with no overall LPO at the scale of the samples [O'Hanley, 1996; Schmitt et al., 2007]. In peridotites, for instance, the commonest fabric pattern is that (010) of olivine and (100) of pyroxenes are preferentially parallel or subparallel to the foliation while [100] of olivine and [001] of pyroxenes are preferentially parallel or subparallel to the lineation [e.g., Ji et al., 1994; Saruwatari et al., 2001; Boudier et al., 2010]. If serpentinization forms a penetrative network composed of three sets of serpentine veins, each perpendicular to X, Y, or Z, and serpentine is aligned with its (001) parallel to the fracture surfaces, lowering both Vp and Vs in the directions normal to the fractures, the serpentinization may not modify the velocity symmetry although the absolute velocity values decrease significantly. If the serpentinization is characterized by only one set of serpentine-filling parallel veins, however, it may markedly increase the seismic anisotropy of peridotites [e.g., Dewandel et al., 2003; Faccenda et al., 2008; Boudier et al., 2010] because single crystal serpentine has much higher seismic anisotropy than olivine and orthopyroxne [Reynard et al., 2007; Bezacier et al., 2010; Mookherjee and Capitani, 2011].

[46] However, the seismic anisotropy in the samples we studied is caused by the LPO of antigorite, which is induced by dislocation creep as indicated by abundant microstructural evidence such as undulatory extinction, lattice rotation, folding, kinking, and dynamic recrystallization (Figures 3 and 4). Thus, seismic anisotropy due to serpentine in-filling along mutually parallel fractures [e.g., Faccenda et al., 2008; Boudier et al., 2010] can be eliminated in the present case.

[47] Serpentine, which develops a layered structure with a pseudo-hexagonal network of linked SiO4 tetrahedra, deforms mainly by intracrystalline slip on systems (001)[010] [Hirauchi et al., 2010; Jung, 2011; Nishii et al., 2011], (001)[100] [Katayama et al., 2009; Moortèle et al., 2010] and possibly (001)[110]. These three slip systems produce the same seismic anisotropy pattern with the fastest and the lowest velocities parallel and perpendicular to the foliation, respectively. However, intracrystalline slip on a single plane certainly cannot produce homogenous deformation because the von Mises criterion for plasticity is not met. According to this criterion, for a crystal to undergo an arbitrary constant volume deformation, at least five independent slip systems are required to be active [e.g., Nicolas and Poirier, 1976]. For this reason, in the serpentinites we examined, there was abundant evidence for other mechanisms (e.g., lattice rotation, folding, kinking, recrystallization, and anisotropic growth) to accommodate intracrystalline slip.

[48] A single crystal of antigorite is characterized by an anisotropy pattern which is very close to that of a single crystal with hexagonal symmetry with fast P-wave velocities (8.0–8.9 km/s) along the a-b plane perpendicular to the c-axis and slow P-wave velocities (5.58 km/s) parallel to the c-axis [Bezacier et al., 2010]. The difference between the velocities along the a- and b-axes is only ~3%. Vp and Vs anisotropy in antigorite can be as high as 46% and 66%, respectively. At high pressures (200–600 MPa) where microcracks are fully closed and seismic anisotropy is intrinsically induced by the LPOs of minerals, the antigorite serpentinites display a prominent Vp anisotropy (at 600 MPa, A = 3.8–16.9% with an average value of 10.5%). It should be noted that the contribution of compositional layering to seismic anisotropy in our samples was minimized because the content of antigorite is higher than 90% in each of the samples. The maximum and minimum Vp anisotropy measured at 600 MPa occur in samples WZG2 and WZG8, respectively. Optical examinations of orthogonally cut thin sections (XZ, XY, and YZ planes) show that sample WZG8, which is only weakly deformed, consists mainly of interpenetrating blades of antigorite that appear to have formed by HT serpentinization under static conditions [Maltman, 1978]. Random LPO is displayed by this texture [Hirauchi et al., 2010]. However, in sample WZG2, in which Vp(X) = 7.41 km/s, Vp(Y) = 6.93 km/s, and Vp(Z) = 6.25 km/s at 600 MPa, (001) cleavages of platy antigorite grains are preferentially aligned parallel to the foliation (XY) and perpendicular to the Y direction but the former are dominant over the latter. Obviously, P-wave velocity anisotropy pattern in antigorite serpentinites is essentially controlled by the preferred orientation of antigorite (001) planes, but displays little dependence on the preferred orientation of [100], [010] or [110] because the P-wave velocities along these directions are almost the same [Bezacier et al., 2010].

[49] The 17 samples can be classified into three categories according to the relative Vp values along the X-, Y-, and Z-directions (Figure 6), and each category can be interpreted according to the antigorite LPO pattern in the samples (Figure 15).

Figure 15.

(a) Flinn-type diagram showing P-wave velocity anisotropy, and (b) pole diagrams showing three end-member types of contoured LPO patterns of antigorite c-axes. The origin of the coordinate axes represents random LPO and thus zero anisotropy. X, lineation; Y, direction normal to lineation in foliation; Z, normal to foliation. Solid dots in Figure 15a indicate the data (600 MPa) from this study, and open dots the data (200 MPa) from references.

[50] (1) Type A: transversely isotropic with Vp(X) ≈ Vp(Y) ≫ Vp(Z). This type of anisotropy (Figure 6a, sample YSK3B), which was also observed in two samples studied by Kern [1993] and Kern et al. [1997] and three samples by Watanabe et al. [2007], is formed by the preferred orientation of antigorite (001) planes parallel to the foliation and the c-axis parallel to the Z-direction (Figure 15b). This type of antigorite LPO is thought to prevail in S- and S > L serpentinites deformed by coaxial flattening. In the S-serpentinites, only S-planes or schistosity developed by the alignment of antigorite flakes without the formation of stretching lineation. In S > L serpentinites, however, foliation (S) is more pronounced than stretching lineation (L). As a result, the c-axes of antigorite cluster around the shortening direction in S- and S > L serpentinites [e.g., Bezacier et al., 2010].

[51] (2) Type B: quasi-transversely isotropic with Vp(X)> > Vp(Y) ≈ Vp(Z). This pattern (Figure 6b, sample YSK2B), which is related to a nearly random concentration of antigorite c-axis in the YZ plane perpendicular to the X-direction (Figure 15b), occurs in L- and L > S serpentinites deformed dominantly by coaxial constriction. In these serpentinites, both the [100] and [010] axes display a girdle or partial girdle distribution within the the XY-plane. This type of antigorite LPO has been reported in Jung [2011], Nishii et al. [2011] and Soda and Takagi [2010].

[52] (3) Type C: orthorhombic symmetry with Vp(X) > Vp(Y) > Vp(Z) (Figures 6c–d). This type of anisotropy, which is common in the serpentinites we examined (e.g., samples YSK1A, YSK5A, YSK5B, and WZG2) and also occurs in a sample of Birch [1960], and two samples (HPS-I and HPS-M) of Watanabe et al. [2007], is characterized by the largest, intermediate, and smallest Vp values in the X-, Y-, and Z-directions, respectively. This type of anisotropy is obviously caused by strong and moderate concentrations of antigorite c-axes in the Z- and Y-directions in the YZ plane, respectively. This pattern of antigorite LPO develops in LS serpentinites deformed in plane strain. Slip on (001) planes contributes mainly to c-axes near the Z-direction while grains aligned initially in unfavorable orientations for slip on (001) planes have their c-axes near the Y- or X-direction. Simple shear strain may progressively rotate the (001) planes aligned initially perpendicular to the X-direction to the bulk shear plane, forming fish-shaped porphyroclasts (Figures 4c–d). Nevertheless, it is difficult for progressive simple shear to eliminate the porphyroclasts whose c-axes cluster in the Y-direction unless extensive dynamic recrystallization has taken place.

[53] Figure 15a, on which the Vp(Y)-Vp(Z) and Vp(X)-Vp(Y) values for antigorite serpentinites are, respectively, plotted on the abscissa and ordinate of a Flinn-type diagram, is constructed to illustrate how Vp anisotropy patterns are related to antigorite LPO. A classic Flinn diagram [Flinn, 1962] plots a value for the principal strain axes Y/Z against X/Y of three-dimensional finite strain ellipsoids which result from deformation of a reference sphere. The data plotted on such a classic Flinn diagram are obtained from analyses of finite strain using strain markers in deformed rocks.

[54] In Figure 15a, the origin of the coordinate axes for the diagram is (0, 0), representing undeformed samples with random LPO which are thus elastically isotropic. Any given Vp anisotropy plots at a particular point on the diagram, and the slope k of the line from the origin (0, 0) to that point is:

display math(14)

[55] The kp value is a coefficient used for classifying the types of LPO-induced Vp anisotropy. The three lines with kp = 0, kp = 1, and kp = ∞ correspond to three end-member types A, C, and B of antigorite c-axis fabrics, respectively. The three lines divide the diagram into two domains: below the kp = 1 line, where the strain is between coaxial flattening and plane strain, and above the kp = 1 line, where strain is transitional between coaxial constriction and plane strain. Thus, the seismic anisotropy patterns may be used to diagnose the orientation and shape of the finite strain ellipsoid in serpentinites.

[56] Plots of Vs(YX)-Vs(YZ) as a function of Vs(XY)-Vs(XZ) are shown in Figure 14. Similarly, we can define the ks coefficient for the serpentinites as

display math(15)

[57] The serpentinites, which deformed dominantly by coaxial stretching (e.g., samples YSK2A) lie in the regime ks> > 1 and are characterized by shear-wave splitting that is significantly stronger along the Y-direction than the X-direction (Figure 7b and Figure 14). In contrast, serpentinites deformed by simple shear (e.g., sample WZG2) are impossible to distinguish from those deformed by coaxial flattening (e.g., sample YSK3B). They all lie directly on or near the line ks = 1 and are characterized by significant shear-wave splitting (up to 0.44 km/s) for the propagation along the X- and Y-directions with the fast shear-wave polarized parallel to the foliation, but with little splitting (mostly <0.05 km/s) in the propagation direction normal to the foliation.

[58] The magnitude of detectable seismic anisotropy in the horizontal plane thus depends on not only the degree of strain but also on the strain geometry of serpentinite within subduction zones. Deformation of antigorite is certainly induced by the movement of the subducting plate with respect to the overriding mantle wedge. In the case of a steeply subducting slab, serpentine is most likely deformed by coaxial flattening, and the serpentine (001) planes are orientated parallel to the plate interface, forming Type A seismic anisotropy with kp ≈ 0 [i.e., Vp(X)≈Vp(Y) ≫ Vp(Z)] and ks ≈ 1 (i.e., the polarization direction of the fast shear-wave component and the largest delay time between the arrivals of the fast and slow shear waves occur when the wave propagation is parallel to the foliation). Consequently, trench-parallel seismic anisotropy and large delay times (δt = 1–2 s) are observed in steep subduction systems such as the Tonga-Kermadec-New Zealand, Aleutian, and Ryukyu trenches [e.g., Long and Silver, 2008, 2009; Wiens et al., 2008; Wirth and Long, 2010; Long and van der Hilst, 2006]. In the case of a shallowly subducting slab, however, antigorite is most likely deformed by simple shear, forming Type C seismic anisotropy with kp ≈ 1 [i.e., Vp(X) > Vp(Y) > Vp(Z)] but also ks ≈ 1. Within the horizontal plane, the P-wave velocities in the trench-normal direction [cosθ Vp(X)], where θ is the dip angle of subducting slab, are smaller, equal to, or even higher than the value in the trench-parallel direction [Vp(Y)]. Taking sample YSK5A [Vp(X) = 7.29 km/s, Vp(Y) = 6.79 km/s] as an example, the trench-parallel Vp could be smaller than the trench-normal value if θ is smaller than 21° [e.g., Cascadia, Currie et al., 2004; South America, Polet et al., 2000; Anderson et al., 2004]. For the same reason, the polarization direction of the fast shear wave could be normal to the trench but the shear-wave splitting could be small [e.g., δt < =0.3 s for the Chile-Argentina subduction zone, Anderson et al., 2004] because near vertical teleseismic rays are close to the Z-direction when θ is smaller than 20°.

6 Conclusions

[59] We have measured P- and S-wave velocities, anisotropy, and shear-wave splitting for 17 antigorite serpentinites at hydrostatic pressures up to 650 MPa. The samples contain 11.55% to 13.55% water by weight. The experimental results combined with previous data yield the following relations between seismic velocities and the volume fraction of LT and HT serpentine (ϕ) within ultramafic rocks: Vp = 8.10 − 3.00ϕLT, Vs = 4.51 − 2.19ϕLT, Vp = 8.10 − 1.42ϕHT, and Vs = 4.51 − 0.84ϕHT. At 600 MPa, Vp = 5.10 and 6.68 km/s, Vs = 2.32 and 3.67 km/s, and Vp/Vs = 2.15 and 1.81 for pure LT and HT serpentinites, respectively.

[60] For a given antigorite serpentinite, the velocity ratio Vp/Vs remains quasi-constant above the crack-closure pressure (~150 MPa) and displays negligible dependence on pressure or temperature. Since the velocity differences (ΔVp and ΔVs) between HT and LT serpentinized peridotites increase linearly with the degree of serpentinization: ΔVp = 1.64ϕ and ΔVs = 1.35ϕ, serpentine contents within the subducting zones and the forearc mantle wedge where temperature is too high (>300°C) for lizardite to be stable should be at least twice as large as previous estimates based on LT serpentinization. However, it is necessary to include the effects of seismic anisotropy and high-pressure fluids in order to interpret the composition of HT serpentinized mantle if Vp < 6.68 km/s, Vs < 3.67 km/s and Vp/Vs > 1.81. The intrinsic anisotropy of the serpentinites studied here (3.8–16.9% with an average value of 10.5% for Vp, and 3.6–18.3% with an average value of 10.4% for Vs) is caused by dislocation creep-induced LPO. Three patterns of seismic anisotropy (Types A, B, and C), which correspond to three types of antigorite LPO (Types S, L, and LS) formed by three categories of strain geometry, have been distinguished using Flinn-type diagrams. Type S serpentinites characterized by well-developed foliation but no lineation; coaxial flattening aligned the c-axes of antigorite parallel to the Z-direction, forming transversely isotropic rocks with Vp(X)≈Vp(Y)> > Vp(Z), almost equal shear-wave splitting in the X- and Y-directions but nearly no splitting in the direction normal to foliation. Type L serpentinites characterized by stretching lineation but no remarkable foliation; coaxial constriction formed a nearly random concentration of antigorite c-axis in the YZ plane normal to the X-direction; thus, Vp(X)> > Vp(Y)≈Vp(Z) and shear-wave splitting is significantly larger along the Y-direction than in the X-direction. Type LS serpentinites developed both foliation and lineation; simple shear aligned the (001) planes of antigorite blades preferentially parallel to the foliation (XY) and perpendicular to the Y-direction but the former are dominant over the latter, causing orthorhombic symmetry with Vp(X) > Vp(Y) > Vp(Z) and significant shear-wave splitting for raypaths in the foliation plane. Thus, combination of strain strength and strain geometry of HT serpentinite within the plate interface and the mantle wedge provides a new explanation for various patterns of seismic anisotropy observed in subduction systems worldwide. In the case of a steeply subducting slab (>300°C), for example, antigorite is most likely deformed by nearly coaxial flattening, causing the trench-parallel seismic velocities and shear-wave splitting to be much larger than trench-normal values. In the case of a shallowly subducting slab (>300°C), however, antigorite is most likely deformed by simple shear. Within the horizontal plane, the seismic velocities and the delay time between the arrivals of fast and slow shear-waves in the trench-normal direction can be smaller, equal to, or even larger than the value in the trench-parallel direction, depending on the dip angle of the subduction zone.


[61] We thank the Chinese Academy of Geological Sciences, the SinoProbe-deep Exploration Project in the Ministry of Land Resources of China for research grants (No. 1212011121274 and SinoProbe-07), and the Natural Sciences and Engineering Research Council of Canada for a discovery grant. R.J. Iuliucci is thanked for technical support. The constructive reviews by M. Savage, Moore, T. Parsons, and the Associate Editor are highly appreciated.