Seismic Anisotropy of Zone Minerals–Contribution of Hydrous Phases

David Mainprice and Benoit Ildefonse

Abstract The seismology is the most effective method to explore the structure of subduction zones to great depth. The distinguishing feature of the in the subduction regions is the presence of hydrated phases, which transport water into the Earth’s interior and release it with dramatic local consequences, triggering earth- quakes and melting. The seismological detection of these hydrous phases and geo- dynamic interpretation of fl ow in the hydrous mantle depend on knowledge of the anisotropic elastic properties and the characteristics of the wave propagation in aniso- tropic media. In this paper we briefl y recall the distinguishing features of anisotropic wave propagation and the observable parameters. We suggest the Vp/Vs ratio is a physically sound parameter than can be observed by seismology in anisotropic regions of the Earth, whereas the Poisson’s ratio, which is often quoted, is not directly observable and does not correspond to the characteristics of wave propagation in an anisotropic or isotropic medium. We report for the fi rst time the ratios of Vp/Vs1 and Vp/Vs2, where Vs1 is the fastest and Vs2 slowest S-wave velocities of an anisotropic media. We present the current knowledge of the anisotropic seismic properties of hydrous minerals in the , , and that are stable along low temperature geotherms associated with subduction, and identify which minerals are likely to infl uence seismological observations because they have very high volume fractions, or very high anisotropies, or both of these. In the upper mantle antigorite and are exceptionally anisotropic (Vp 71%, Vs 68% and Vp 65%, Vs 68%, respectively) and chlorite is also very anisotropic for S-waves (Vp 35%, Vs 76%). Comparatively less anisotropic are hornblende (Vp 27%, Vs 31%), used as proxy for tremolite in subduction zones, and clinohumite (21.8%, Vs 15.9%). Brucite at 4 GPa (Vp 26.5%, Vs 30.9%) and the dense hydrogen magnesium silicate (DHMS) phase A at 9 GPa (Vp 9.3%, Vs 17.6%) are the only hydrous minerals stable in the upper mantle that have had their elastic properties measured at in situ mantle pressures. Except for the phase A, all these minerals are more anisotropic than for at least one parameter (Vp, Vs, Vp/Vs1, or Vp/ Vs2). In the transition zone the major phases hydrous and have moderate to weak anisotropies (Vp 16.3%, Vs 16.5%, and Vp 1.9%, Vs 4.4%, respectively). The DHMS Superhydrous B has a moderate anisotropy (Vp 6.9%, Vs 11.6%). At greater depth the DHMS phase D at 24 GPa (Vp 10.8%, Vs 18.0%)

D. Mainprice Université Montpellier 2, Géosciences Montpellier, CNRS/INSU, UMR 5243, CC 60, Place Eugène Bataillon, 34095 Montpellier, Cedex 5, France, David.Mainprice@gm. univ-montp2.fr

S. Lallemand and F. Funiciello (eds.), Subduction Zone Geodynamics, 63 DOI 10.1007/978-3-540-87974-9, © Springer-Verlag Berlin Heidelberg 2009 64 D. Mainprice, B. Ildefonse

is the only hydrous phase that can transport hydrogen from the transition zone into the lower mantle. The phase D is more anisotropic for S than P waves like many of the hydrous phases. From these data it is clear that hydrous phases are in general very anisotropic. However, pressure can play a strong role in reducing anisotropy. It is the case for brucite and talc, in which increasing pressure from ambient to 4 GPa reduces the anisotropy by about 50% for both P and S waves. In contrast, the anisotropy of the DHMS phase A does not change signifi cantly with pressure. Our picture of the seismic anisotropy of hydrated minerals remains incomplete; the elastic properties of many have not been measured even at ambient conditions (e.g., 10 Å phase and phase E) or not measured in their true elastic symmetry (e.g., clinochlore). The majority of hydrous minerals have not been measured at high pressure, and none related to hydrated mantle have been measured at elevated tem- perature. We have shown in the few cases where hydrous minerals have been mea- sured as a function of pressure, that this variable has an important effect on the velocity distribution and in most cases reduces the degree of anisotropy, hence we would expect seismic anisotropy to play a key role in the determination of the shallow structure of subduction zones in the upper mantle.

Keywords Seismic anisotropy · Elastic properties · Hydrous minerals · Subduction

1 Introduction seismic data in terms of petrology and the anisotropic seismic properties are vital for deduction of fl ow patterns in the mantle. Knowledge of the seismic prop- Subduction zones are regions of extensive recycling of erties of hydrated phases and nominally anhydrous hydrated materials (sediments, hydrated oceanic phases containing hydrogen allows the exploration of and upper mantle) that are transformed by metamor- degree of hydration of the mantle, which is clearly a phic processes into a series of high-pressure mineral at dynamic process in subduction zones with implica- great depths. Many of the minerals postulated to exist tions for seismogenesis (e.g., Seno et al., 2001; Hacker at great depth in subduction zones have only been dis- et al., 2003b; Jung et al., 2004), generation covered in the laboratory at extreme conditions of (e.g., Ulmer and Trommsdorff, 1995) and metamor- pressure and temperature, with no natural occurrences phic transformations (e.g., Hacker et al., 2003a). having been reported. On the other hand minerals pos- In this paper we will investigate the anisotropic elastic tulated to exist on the subduction plane (metamor- properties of hydrated phases and nominally anhy- phosed sediments and oceanic crust) and the hydrated drous phases containing hydrogen and the implications forearc mantle (hydrated peridotites) are exposed on for seismic properties. the Earth’s surface by the slow rise of forearc serpen- Seismology measures the bulk elastic properties tine seamounts, the relatively rapid exhumation pro- of the Earth, with contributions from anelastic and cesses of high pressure metamorphic rocks or the even viscoelastic properties. The bulk elastic properties are more dynamic ascent of volcanic xenoliths. One group controlled to fi rst order by the volumetrically domi- of minerals that are expected to be volumetrically nant elements and hence in this study of seismic prop- important components of the hydrated mantle, which erties of subducted rocks it is natural to consider fi rst is taking part in the recycling of hydrogen at depth, are minerals that have volume fractions of more than 10%. the hydrated minerals, although the contribution of For seismic anisotropy the same reasoning holds, but if anhydrous silicates is now also known to be signifi cant a mineral is exceptionally anisotropic, the classical (Smyth and Jacobsen, 2006; Hirschmann, 2006). example is layered silicates (e.g., biotite) in crustal Knowledge of the seismic properties of minerals at rocks, then even volume factions of less than 10% can depth is obviously essential for the interpretation of Seismic Anisotropy of Subduction Zone Minerals–Contribution of Hydrous Phases 65 have an important effect on the rock properties. The thin veneer of sediments, hydrated oceanic crust, and degree of crystallographic preferred orientation also upper mantle. The temperature of the subducting plate, has a major infl uence on the contribution of anisotro- controlled essentially by its age and subduction rate, pic mineral to overall properties of a rock. In the case mediate the metamorphic processes that produce of mica minerals they have very strong shape and a sequence of transformations of hydrous minerals mechanical anisotropy, and hence they tend to be (e.g., Arcay et al., 2005; Fumagalli and Poli, 2005). strongly oriented in almost all rock types from unde- Theses minerals subsequently release fl uids as the formed sediments to highly deformed metamorphic temperature of the slab rises, that hydrate the surround- rocks. In the context of the strong localised fl ow fi eld ing mantle and trigger fl ux melting (e.g., Ulmer and in subduction zones, it is highly likely that most rocks Trommsdorff, 1995) (Fig. 1), and rising melt reduces are plastically deformed to some degree and the miner- seismic velocities (e.g., Vp −6%) in the als have crystallographic preferred orientation (CPO). (Nakajima et al., 2001). The process of a series of From a point of view the input to a linked transformations of hydrous phases with over- subduction zone can be separated into minerals of the lapping stability fi elds of decreasing water content

Fig. 1 Schematic cross-section of a subduction Double Benioff zones and suduction mineralogy system with the double Benioff zones and their associated mineralogy in cold and warm suduction, simplifi ed after the scenarios of Cold slab & high water contents: antigorite breakdown Fumagalli and Poli (2005) and Poli and Schmidt volcanic front (2002). The Vp seismic velocities are from Serpentine Outer rise Backarc forearc diapirs Earthquakes Nakajima et al. (2001) in NE Japan, considered 0 0 to be a cold subduction. The arrow in the Mantle LIZARDITE 1 Wedge represents the corner fl ow direction Mantle 50 2 Wedge 75 3 Melt 100 Flux Melting Depth (km) Depth 4 (Water Saturated ? ) 5 Water Vp -6 -4 -2 0 ANTIGORITE 150 Velocity perturbation (%) 6

7 200 Pressure (GPa) Pressure Water transfer 10 Å PHASE to the deep mantle 8 250 9 PHASE A 10 = Earthquake 300

Warm slab & low water contents: clinochlore breakdown

volcanic front Serpentine Outer rise Backarc forearc diapirs Earthquakes 0 0 LIZARDITE TALC 1 Mantle 50 2 Wedge Melt 75 3 100 4 (km) Depth Water 5 CLINOCLORE 150 6 Complete dehydration of slab 7 200

Pressure (GPa) Pressure causing a "choke point"

8 250 9 PHASE A 10 = Earthquake 300 66 D. Mainprice, B. Ildefonse with depth can continue to depths of the transition of course its current temperature for the phase stability zone and the lower mantle (e.g., Kawamoto et al., of any given hydrous mineral. 1996). When slab temperatures are suffi ciently low Estimates of the petrology of subducted mate- (e.g., N.E. Japan), when old plates are subducted, anti- rial have been made based on rocks exposed at the gorite is stable over a signifi cant depth range in the surface, samples dredged from the fl oor slab (Fig. 1), and its breakdown will trigger earth- (e.g., Michibayashi et al., 2007) and xenoliths (e.g., quakes in the form of a double Benioff zone (e.g., Michibayashi et al., 2006), thermodynamical calcula- Fumagalli and Poli, 2005; Brudzinski et al., 2007). tions (e.g., Hacker et al., 2003a; Bousquet et al., 2005, At greater depth the dense hydrous magnesium silicate 1997) and laboratory experiments (e.g., Ohtani et al., (DHMS) called the phase D could be preserved to 2004; Iwamori, 2004; Fumagalli and Poli, 2005). 1,200-km depth, being then the only hydrated phase In general some combination of all three approaches is capable of recycling hydrogen into the lower mantle effectively incorporated in these studies to different (Fig. 2). Alternatively, in warmer subduction zones degrees, resulting is some variations in the volume the hydrous phases may be completely de-hydrated at fractions reported by different authors. For the Earth’s some critical depth, called a choke point (Kawamoto mantle we have illustrated the positions of the main et al., 1996), resulting in the breaking of the chain of anhydrous and hydrous phases in pressure-temperature hydrous phases with depth and the failure to transfer space together with the mantle and subduction zone hydrogen to greater depth. Fumagalli and Poli (2005) geotherms in Fig. 2 using the stability fi elds of hydrous suggest this could happen at 200-km depth if the over- phases reported by Schmidt and Poli (1998) and lap in stability fi elds between the 10 Å and phase A is Fumagalli and Poli (2005) for the low pressure region, not maintained by low temperature and water satura- and those of Iwamori (2004) and Ohtani et al. (2004) tion. In this case it is rather the breakdown of clinochlore for high pressures. with a higher temperature stability that causes the The hydrous phases can be divided into 3 main double Benioff zone. Hence the presence of hydrous groups with increasing depth; phases in the deeper part of the subduction zone will 1. Low pressure (P < 5 GPa) minerals, such as antigo- depend on the past temperature of the slab at shallower rite (13.0 wt% H O), clinochlore (13 wt% H O), talc depths necessary to maintain the hydrous chain, and 2 2 (4.8 wt% H2O), Ca-amphibole tremolite or parg-

asite (2.3 wt% H2O) are commonly observed in Temperature (°C) exposed metamorphic rocks. 500Talc 1000 1500 2. Moderate pressure (5–7 GPa) minerals, such as Pargasite Mantle Clino phlogopite (4.8 wt% H2O), 10Å phase (10–13 wt% Antigorite chlore 100 H O), clinohumite (2.8 wt% H O) in hydrated Geotherm 2 2

5 10 Å Phase peridotites, lawsonite (11.5 wt% H O) and zoisite 200 2 Clino (2.0 wt% H O) in hydrated metamorphosed basalts humite 2 Phase A 10 300 (km) Depth and potassium rich phengite (4.6 wt%H2O) in meta- Olivine morphosed sediments. Low T Slab GeothermPhase E HT Slab Geotherm 400 Wadsleyite 3. High pressure (>7 GPa) minerals such as the 15 k-amphibole richterite (2.1 wt% H2O), topaz-OH 500 Pressure (GPa) Pressure (10.0 wt% H2O), phase Egg (11–18 wt% H2O) and 20 Super Ringwoodite the DHMS or alphabet phases, phase A (12 wt% hydrous 600 B H O), phase B (3 wt% H O), phase superhydrous Phase 2 2 Phase D B (2 wt% H O), phase D (10 wt% H O), and phase (Mg,Fe)-Pv 700 2 2 25 + (Mg,Fe)O E (11 wt% H2O), most of which have never observed at the Earth’s surface. Fig. 2 Pressure-Temperature diagram with the main hydrous and anhydrous phases in subduction zones. The stability fi elds The general trend there will be a decrease in the water of hydrous phases reported by Schmidt and Poli (1998) and content in the hydrated slab with depth as the hot Fumagalli and Poli (2005) for the low-pressure region, and those surrounding mantle raises the slab temperature by of Iwamori (2004) and Ohtani et al. (2004) for high-pressures. Geotherms for low and high temperature slabs, and the mantle thermal conduction, causing the breakdown and reduc- are taken from Peacock (1990) tion of the volume faction of hydrous minerals, and Seismic Anisotropy of Subduction Zone Minerals–Contribution of Hydrous Phases 67

fi nally releasing water to the adjacent mantle. As a word shear (or transverse) waves with displacements of caution it should be said that the heterogeneous water perpendicular to the propagation direction. In anisotro- distribution, which is an intrinsic feature of the subduc- pic elastic media there are three types, one compres- tion zone system, implies the presence or the absence of sional and two shear waves with in general three hydrated minerals, as well as the occurrence of both different velocities (Fig. 3). In order to understand the water saturated and water undersaturated conditions. displacements associated with different waves, and The stability of a hydrated mineral implies neither the their relationship to the propagation direction and elas- presence of free water phase nor water saturation (Poli tic anisotropy, it is important to consider the equation and Schmidt, 2002). The relatively low temperature in of propagation of a mechanical disturbance in an subducted slabs favours the formation of hydrous min- anisotropic elastic medium. The solution of elasto- erals that contain large amounts of water as hydroxyl dynamical equation for the displacement of monochro- groups (e.g., antigorite, chlorite, phases E and D), which matic plane wave is given by Christoffel equation as increases the likelihood of water under saturated assem- 2 blages. The appearance of a fl uid is related to the ther- det |Cijkl nj nl - rV dik| = 0 modynamic defi nition of saturation and controlled by assemblages buffering the chemical potential of water where Cijkl is elastic stiffness fourth rank tensor, nj the (e.g., Bolfan-Casanova, 2005); the relationships between propagation unit vector, V are the three phase veloci- ρ δ the amount of water and the presence of free water are ties, is density, and ik is the Kronecker delta. We can not straightforward. Experiments are often conducted in simplify this equation by introducing the Christoffel water-saturated conditions, with free water present (e.g., (Kelvin-Christoffel or acoustic) tensor Tik = Cijkl nj nl Fumagalli and Poli, 2005; Schmidt and Poli, 1998), and three wave moduli M = ρV2. The Christoffel tensor whereas studies of natural samples often infer water is symmetric because of the symmetry of the elastic undersaturated assemblages, particularly at very high constants, and hence the eigenvalues of the 3 × 3 pressure (e.g., Chinner and Dixon, 1973). Christoffel tensor are three positive real values of the In what follows we will develop tensor methods to wave moduli (M) corresponding to ρVp2, ρVs12, ρVs22 assess the contribution of these minerals to seismic of the plane waves propagating in the direction n. The anisotropy and their potential for giving diagnostic three eigenvectors of the Christoffel tensor are the seismic signals to determine the degree of hydration in polarization directions (also called vibration, particle the mantle. movement or displacement vectors) of the three waves. As the Christoffel tensor is symmetric the three eigen- vectors (and polarization) vectors are mutually perpen- 2 The Relationship Between Elastic dicular. In the most general case there are no particular angular relationships between polarization directions and Seismic Anisotropy (pk) and the propagation direction (nj). However, typi- cally the P-wave polarization direction is nearly paral- There are two types of elastic waves, which propagate lel and the two S-waves polarizations are nearly in an isotropic homogeneous elastic medium, the faster perpendicular to the propagation direction, and they compressional (or longitudinal) wave with displace- are termed quasi-P or quasi-S waves to distinguish ments parallel to propagation direction, and the slower them from the P and S of isotropic media (see Fig. 2).

Fig. 3 Relation between Isotropic case : Anisotropic case : polarisation and propagation propagation direction // polarization Vp polarisation of quasi-Vp closest to propagation direction in general Vp > Vs1 = Vs2 in general quasi-Vp > quasi-Vs1> quasi-Vs2 directions of waves in isotropic polarization Vs1 & Vs2 ⊥ propagation direction polarizations are mutually perpendicular and anisotropic elastic media polarization Vs2 polarization quasi-Vs2

propagation direction propagation direction polarization Vp polarization quasi-Vp

polarization Vs1 polarization quasi-Vs1 68 D. Mainprice, B. Ildefonse

In general, the three waves have polarizations that are reason for using Poisson’s ratio in the characteristic perpendicular to one another and propagate in the same seismic properties of anisotropic or even isotropic direction with different velocities, with Vp > Vs1 > media when the more direct observation of Vp/Vs Vs2. Computer methods for calculation seismic wave ratios for an isotropic medium and Vp/Vs1 plus Vp/ velocities in anisotropic are given by Mainprice Vs2 for anisotropic medium have a clear physical (1990). meaning. In the case of subduction zones it has become stan- dard practice to use the isotropic Vp/Vs ratio or Poisson’s ratio to characterise the seismic response 3 Seismic Anisotropy of Hydrated related to hydrated minerals (e.g., serpentine group minerals, Christensen, 1996, 2004), or the presence of Phases dehydration fl uids (e.g., Ito, 1990; Kono et al., 2007), or as yet undetermined minerals with low Vp/Vs (e.g., The anisotropic seismic properties of hydrous phases Rossi et al., 2006). Classically for an isotropic medium shown in Fig. 2 are estimated to have volume fractions there is only one Vp/Vs ratio as the velocity is the same well above 10% and we will use these to illustrate the for all propagations directions and there is only one Vs anisotropic behaviour of hydrous minerals in rocks of velocity. In seismology the Poisson’s ratio (ν) has been peridotite composition in subduction zones. calculated form the classical isotropic relationship

éù(/)VV2 - 2 ν 1 ps = êú2 3.1 Upper Mantle Phases 2 VV ëûêú(/)ps - 1

3.1.1 Above 180 km Depth: Low Temperature However, in anisotropic elastic media there are Serpentine Minerals three velocities (Vp, Vs1, Vs2) and hence two ratios Vp/Vs1 and Vp/Vs2, and of course these ratios vary First, we will consider the case of serpentine poly- with direction. The anisotropic Poisson’s ratio (ν ) ijkl morphs, i.e., lizardite, chrysotile, and antigorite. varies with direction in a complex manner. Poisson’s Although there is no direct information on the amount ratio is defi ned by the elastic strain in two orthogonal of water in the peridotite layer of slabs, it should directions, the longitudinal (or axial) direction (x ) and i contain water due to serpentinization or chloritization. transverse (or lateral) direction (y ). The lateral strain i Serpentine is commonly observed on the ocean fl oor is defi ned by -e y y along y and the longitudinal strain ij i j near fracture zones, in particular at slow-spreading by e x x along x, where e is the infi nitesimal strain ij i j ij ridges (e.g., Lagabrielle et al., 1998; Mével, 2003; tensor. The anisotropic Poisson’s ratio (n ) is given as ijkl Ildefonse et al., 2007). Schmidt and Poli (1998) esti- the ratio of lateral to longitudinal strain (Sirotin and mated that the upper 5 km of the peridotite layer suf- Shakolskaya, 1982) as fers about 10% serpentinization; Ranero and Sallarès ε yyS xxy y (2004) estimated about 17% of serpentinization in the ν = - ij i j = - ijkl i j k l upper 20 km of the peridotite layer of the Nazca plate ijkl ε xxS x xxx kl k l mnop m n o p that is subducted at the North Chile trench. Lizardite is volumetrically the most important serpentine mineral where Sijkl is the elastic compliance fourth rank tensor in serpentinites, especially retrograde metamorphosed in 4 index notation. To compare nijkl to seismic wave (hydrated) peridotites. The low temperature serpen- propagation we could choose the axial and transverse tine minerals chrysotile and lizardite occur widely in strain directions to be parallel to the orthogonal P- and serpentinites, commonly together, and correspond to S-wave polarization directions, respectively. However, inferred temperature conditions ranging from the this analogy is clearly not exact, as the strains associ- surface up to perhaps as high as 400°C (Evans, 2004). ated with seismic waves in an elastic medium require The low temperature serpentine minerals are observed shear strain components not present in the defi nition of to form by hydration on sea fl oor of abyssal peri- Poisson’s ratio. Hence, we can fi nd no sound physical dotites at mid-ocean ridges under static conditions Seismic Anisotropy of Subduction Zone Minerals–Contribution of Hydrous Phases 69

(e.g., Mével, 2003), and are particularly abundant foliation. Watanabe et al. (2007) report a strong in the thick lithosphere of slow spreading ridges foliation, probably indicative of CPO, in 50% of their (e.g., Cannat, 1993; Lagabrielle et al., 1998). Although antigorite samples. There are no single crystal mea- controversial, it appears that chrysotile is meta-stable surements of the elastic tensor of serpentine minerals in the laboratory and eventually transforms to lizard- due to the diffi culty in fi nding natural single crystals of ite (e.g., Grauby et al., 1998; Evans, 2004). The pres- suffi cient size and quality. Recently Auzende et al. ence of the low temperature serpentine minerals in the (2006) and Pellenq et al. (2008) have used transferable hydrothermally altered oceanic crust and lithosphere semi-empirical interatomic potentials to compute the lowers the density and seismic velocities (e.g., Carlson elastic tensor of single crystal lizardite and antigorite, and Miller, 1997; Christensen, 2004). respectively, at ambient pressure. The lizardite elastic In the subducted oceanic lithosphere the major constants of Auzende et al. (2006) predict linear com- release of water in fl uid-saturated systems is caused by pressibility for a- and c-axes of 2.3 × 10−3 and 20.5 × the breakdown of antigorite in cold slabs (Ulmer and 10−3 GPa−1 are in reasonable agreement with the exper- Trommsdorff, 1995), with an associated reduction in imentally measured values of 2.7 × 10−3 and 9.7 × 10−3 P-wave velocity of 5.1% (Kono et al., 2007). Antigorite GPa−1 reported by Hilairet et al. (2006), being good for is expected to be the serpentine mineral stable at rela- the a-axis, but two times too high for the c-axis. The tively higher temperatures and much higher pressures antigorite elastic constants of Pellenq et al. (2008) pre- of the upper mantle (Evans et al., 1976), for example in dict linear compressibility for a- and b-axes of 2.3 × the forearc upper mantle (e.g., Kamiya and Kobayashi, 10−3 and 2.2 × 10−3 GPa−1, which agree well with 2000; Bostock et al., 2002; Brocher et al., 2003; values of 3.7 × 10−3 and 3.3 × 10−3 GPa−1 reported by Hyndman and Peacock, 2003). In particular, the dehy- Hilairet et al. (2006), however the predicted value for dration of antigorite has been associated with the the c-axis 17.9 × 10−3 GPa−1 is again twice the mea- formation of double seismic zones (e.g., Peacock, sured value of 8.2 × 10−3 GPa−1. Reynard et al. (2007) 2001; Hacker et al., 2003b), which are know to be have used ab-initio methods to calculate the elastic global phenomena (Brudzinski et al., 2007). As pointed properties of pseudo-hexagonal lizardite, but compari- out by Christensen (2004) and Watanabe et al. (2007), sons between experimentally measured compressibil- much of the interpretation of wedge mantle seismic ity along the c-axis with the predicted compressibility velocities in terms of the volume fraction of serpentine is fi ve times greater than measured, and hence they are (e.g., Carlson and Miller, 2003) has been made by not acceptable to predict seismic properties. Both the incorrectly using an empirical relationship between model elastic constants of Auzende et al. (2006) and volume fraction of low temperature lizardite-chrysotile Pellenq et al. (2008) predict very strong anisotropy for serpentinites with Vp, Vs and density established by Vp (76.7% and 71.2%) and dVs (81.7 and 67.6%) with Christensen (1966), rather than stable high-pressure high Vp velocities and high shear wave splitting (dVs) phase antigorite. Christensen (1978, 2004), Kern anisotropy in the basal plane for lizardite and antigo- (1993), Kern et al. (1997), and Watanabe et al. (2007) rite, respectively. The polarization of the fastest have all studied samples with clearly identifi ed high S-waves (S1) is parallel to the basal plane in both poly- antigorite volume fractions. These samples show that morphs. All the features in Fig. 4 are very characteris- antigorite has higher Vp, Vs and lower Vp/Vs ratio tic of layer silicates. The Vp/Vs1 and Vp/Vs2 ratios than lizardite-chrysotile serpentinites. In addition Kern for lizardite and antigorite are shown in Fig. 5, they and Tubia (1993), Horen et al. (1996) and Song et al. have the general pattern of high values in the basal (2004) also report ultrasonic measurements of serpen- plane and low values around the c-axis. In detail, the tine samples that have transverse anisotropy, but do not Vp/Vs1 is more complex with small region of high identify which serpentine minerals are present. Both Vp/Vs1 around the c-axis. Overall, lizardite and anti- Christensen (2004) and Watanabe et al. (2007) have gorite are characterised by high Vp/Vs1 and Vp/Vs2 produced new empirical relationships between volume for propagation directions around the basal plane. fraction of high temperature antigorite serpentinites to The velocity ratios have high anisotropies, 36.1–41.2% Vp, Vs and density. Kern (1993) and Kern et al. (1997) for Vp/Vs1 and 92.8–100.4% for Vp/Vs2 for anti- reports a strong crystal preferred orientation of c-axes gorite and lizardite, respectively. The values of the with over three times random densities normal to the velocity ratios vary from 1.25 to 4.30 with direction. 70 D. Mainprice, B. Ildefonse

Fig. 4 Upper hemisphere pole Serpentine Minerals fi gures of P-wave velocity, S-wave splitting anisotropy P velocity (km/s) dVs anisotropy (%) Vs1 polarization (dVs as a percentage) and the [2110] Vp Contours (km/s) AVs Contours (%) polarization direction (white 9.55 81.76 lines) of the fastest S-wave 9.0 70.0 (Vs1) for single crystal lizardite 8.0 50.0 and antigorite, calculated using [0110] 7.0 6.0 30.0 the elastic tensors reported Lizardite by Auzende et al. (2006) and 5.0 10.0 Anzende et al. 06 et al. Anzende upper hemisphere 4.25 0.00 Pellenq et al. (2008) at ambient Max.Velocity = 9.55 Min.Velocity = 4.25 Max.Anisotropy = 81.76 Min.Anisotropy = 0.00 pressure. The c-axis is in the Anisotropy = 76.7 % centre of the pole fi gures, other [100] crystallographic reference 9.14 67.56 directions in the basal plane 8.0 50.0 are marked 7.0 40.0 [010] 30.0

Antigorite 6.0 20.0 5.0 Pellenq et al. 08 et al. Pellenq 4.34 0.84

Max.Velocity = 9.14 Min.Velocity = 4.34 Max.Anisotropy = 67.56 Min.Anisotropy = 0.84 Anisotropy = 71.2 %

Fig. 5 Upper hemisphere pole Serpentine Minerals fi gures of Vp/Vs1, Vp/Vs2 for single crystal lizardite and Vp/Vs1 Vp/Vs2 antigorite, calculated using the [2110] elastic tensors reported by Auzende et al. (2006) and 1.89 4.30 Pellenq et al. (2008) at ambient 1.76 3.5 pressure. The c-axis is in the 1.68 centre of the pole fi gures, other 1.60 3.0 [0110] crystallographic reference 1.52 Lizardite 2.5 directions in the basal plane 1.44 1.36 2.0 are marked 06 et al. Anzende upper hemisphere 1.25 1.42 Max. = 1.89 Min. = 1.25 Max. = 4.30 Min. = 1.42 Anisotropy = 41.2 % Anisotropy = 100.4 % [100] 1.86 3.77

1.76 3.25 1.68 2.75 [010] 1.60 1.52 2.25 Antigorite 1.44 1.75 Pellenq et al. 08 et al. Pellenq

1.29 1.38 Max. = 1.86 Min. = 1.29 Max. = 3.77 Min. = 1.38 Anisotropy = 36.1 % Anisotropy = 92.8 %

The Voigt-Reuss-Hill average isotropic values of Vp/Vs 3.1.2 Above 180 km Depth: Minerals Stable for Auzende et al. (2006) and Pellenq et al. (2008) elas- at Higher Temperature tic constants are 1.77 and 1.73, respectively. For refer- ence, the experimentally measured Vp/Vs values given Several of the minerals present in the fi rst 180-km by Christensen (2004) at 1 GPa pressure for isotropic depth range of subduction zones are stable to higher lizardite and antigorite are 2.17 and 1.87, respectively; temperatures than antigorite. Among the more impor- the Vp/Vs ratio is higher in lizardite than antigorite in tant minerals likely to be stable in warm slabs are talc, both the calculated and experimental values. brucite, clinochlore and tremolite-pargasite. Seismic Anisotropy of Subduction Zone Minerals–Contribution of Hydrous Phases 71

Talc (Mg3Si4O10(OH)2) is of great geological impor- The elastic constants of talc have recently been deter- tance having a widespread occurrence from the surface mined by ab initio methods by Mainprice et al. (2008). of the seafl oor to depths of 150 km in subduction zones. At ambient pressure talc is perhaps the most elastically Perhaps, it is best known to form during the serpentini- anisotropic mineral with P-wave anisotropy of 80% zation of abyssal peridotites at mid-ocean ridges and S-wave anisotropy of 85%. At a representative (e.g., Cannat, 1993; Mével, 2003) being associated pressure of its occurrence in a subduction zone of with alteration of orthopyroxene to serpentine + talc, 0.9 GPa (40-km depth) the P-wave anisotropy is 64.9% where talc can represent up to 41% by volume of the and S-wave anisotropy is 65.1% (Fig. 6). The Vp/Vs1 rock (e.g., Hacker et al., 2003a). At the Mid-Atlantic and Vp/Vs2 ratios have anisotropies of 45.7% and Ridge, talc locally represents 100% of ODP cores 95.2%, respectively. The velocity ratios vary from 1.05 drilled at 14°50′N (Shipboard Scientifi c Party, 2004). to 3.18 with direction.

It also a mineral commonly sampled in detachment Brucite Mg(OH)2 is interesting as it is a model faults associated to oceanic core complexes (e.g., system for understanding dense hydrous magnesium MacLeod et al., 2002; Escartín et al., 2003; Boschi minerals (alphabet phases) under hydrostatic compres- et al., 2006). In subduction zones, it forms in the sion and an important structural unit of many sheet mantle wedge from the breakdown of antigorite to give silicates, such as chlorite, lizardite and talc. It is occa- talc + forsterite + water (e.g., Evans et al., 1976), this sionally a major phase in low temperature and pressure reaction has recently be found to explain the global oceanic serpentinites, as at Hess Deep (e.g., Mével pattern of seismicity that displays double Benioff and Stamoudi, 1996), or at the Mid-Atlantic Ridge zones (Brudzinski et al., 2007). At greater depths, talc (e.g., Bach et al., 2004, 2006). At higher pressures releases water in the dehydration reaction to enstatite brucite commonly occurs in association with antigo- at ca 1.7 GPa 660°C (60-km depth) (e.g., Ulmer and rite (Hostetler et al., 1966) at upper mantle PT condi- Trommsdorff, 1995). Talc should accordingly be pres- tions shown in Fig. 2, and could constitute 13% ent from shallow depths to 60 km. However talc can and 7% by volume of a hydrated harzburgite and lher- still be present at depths of 100 km or more, where it zolite, respectively, according to Hacker et al. (2003a). forms in metamorphosed oceanic crust at ultra-high Brucite is a very anisotropic phase at room pressure, pressure conditions during the hydration of mafi c however at pressures approximating to the in-situ eclogites, even at water under-saturated conditions conditions (P = 4 GPa) in the upper mantle the anisot- lacking a free fl uid phase (Poli and Schmidt, 1997). ropy is reduced, but still very high at 26.5% and 30.9%

Fig. 6 Upper hemisphere pole Talc P= 0.9 GPa fi gures of P-wave velocity, S-wave splitting anisotropy (dVs as a P velocity (km/s) dVs anisotropy (%) Vs1 polarization percentage), the polarization ⊥(100) Vp Contours (km/s) AVs Contours (%) direction (white lines) of the 8.84 65.10 fastest S-wave (Vs1), Vp/Vs1 and 8.0 50.0 7.5 Vp/Vs2 for single crystal talc 40.0 ⊥ 7.0 triclinic (c1) symmetry, calculated (010) 6.5 30.0 6.0 20.0 using the elastic tensor by 5.5 Mainprice et al. (2008) at a upper hemisphere 4.51 0.70 pressure of 0.9 GPa. The c-axis Max.Velocity = 8.84 Min.Velocity = 4.51 Max.Anisotropy = 65.10 Min.Anisotropy = 0.70 is in the centre of the pole fi gures, Anisotropy = 64.9 % other crystallographic reference Vp/Vs1 Vp/Vs2 directions in the basal plane are marked 1.67 3.18

1.52 2.60 1.44 1.36 2.20 1.28 1.80 1.20 1.40

1.05 1.13

Max. = 1.67 Min. = 1.05 Max. = 3.18 Min. = 1.13 Anisotropy = 45.7% Anisotropy = 95.2% 72 D. Mainprice, B. Ildefonse

Fig. 7 Upper hemisphere pole Brucite P= 4.8 GPa fi gures of P-wave velocity, S-wave splitting anisotropy P velocity (km/s) dVs anisotropy (%) Vs1 polarization (dVs as a percentage), the [2110] polarization direction (white 8.42 30.88 30.88 27.0 lines) of the fastest S-wave 8.00 21.0 (Vs1), Vp/Vs1 and Vp/Vs2 for 7.60 [0110] 15.0 single crystal brucite (trigonal 7.20 9.0 symmetry P-3m1), calculated 6.80 using the elastic tensor by Jiang 3.0 upper hemisphere 6.45 0.00 0.00 et al. (2006) at 4.8 GPa pressure. Max.Velocity = 8.42 Min.Velocity = 6.45 Max.Anisotropy = 30.88 Min.Anisotropy = 0.00 The c-axis is in the centre of Anisotropy = 26.5 % the pole fi gures, other crystal- Vp/Vs1 Vp/Vs2 lographic reference directions in the basal plane are marked 1.74 2.27

2.08

1.60 1.92

1.55 1.76 1.50 1.60

1.43 1.47

Max. = 1.74 Min. = 1.43 Max. = 2.27 Min. = 1.47 Anisotropy = 19.3% Anisotropy = 43.1 % for Vp and Vs, respectively (Fig. 7). Brucite like other however we do not illustrate the seismic properties of layer structures has the highest value of Vp/Vs2 in the this mineral, as it is stable at higher temperatures than basal plane of 2.27 and anisotropies of Vp/Vs1 and commonly associated with subduction. In the case of Vs2 of 19.3% and 43.1% that is considerably less clinochlore, they only managed to measure ultrasonic than antigorite at room pressure. For reference the velocities corresponding to 4 elastic constants

Voigt-Reuss-Hill average isotropic of Vp/Vs ratio for (C11,C33,C44 and C66), but as C12 = C11–2C66 we have the

Brucite is 1.68. value for C12. However, the value for C13 has not been

The clinochlore end member of the chlorite series is measured. To estimate the values of C13 we used the a strong candidate to be present in the hydrated upper systematics of all Cij for layer silicates reported by mantle. Experimental evidence reported by Fumagalli Alexandrov and Ryzhova (1961a). A plot of Cij/ and Poli (2005) for the presence of clinochlore in density versus Cij in order of magnitude (C11, C66, C33, hydrated lherzolite and harzburgite at 20–40%wt frac- C12, C13 and C44) for muscovite, phlogopite, biotite and tions in the pressure range up to 6 GPa and tempera- clinochore shows that the values of C13 lie between C12 tures around 700°C make clinochlore a possible and C44 and simple linear interpolation gives C13 = C44 alternative to antigorite to explain mantle wedge seis- + (C12–C44)/2 = 40.2 GPa for clinochlore. The elastic mic anisotropy. Clinochlore is the product of break- constant of clinochlore can be further verifi ed by com- down of tremolite or pargasite in hydrous peridotites paring the experimentally measured linear compress- (Fumagalli and Poli, 2005). The temperature depen- ibility (β) along the a- and c-axes given by Pawley −3 −1 −3 dent overlap of stability fi elds of clinochlore and the et al. (2002), ba = 3.0 × 10 GPa and bc = 5.1 × 10 10 Å phase has major implications for the transfer of GPa−1, with values calculated from Cij of clinochlore −3 −1 −3 −1 H20 to higher pressures phases at depth (Fumagalli of ba = 2.7 × 10 GPa and bc = 8.1 × 10 GPa , and Poli, 2005). However, the elastic constants of the which are very close for the a-axis and slightly too 10 Å phase have not yet been measured, so that seismic high for the c-axis. We calculated values of linear com- interpretation of such a transfer is at present impossi- pressibility from the unit cell data of Welch and ble in the anisotropic case. Alexandrov and Ryzhova Crichton (2002) using ba = (1−a/a0)/P where a is the −4 (1961a) have measured the elastic constants of a num- cell edge at pressure P and a0 is the cell length at 10 −3 −1 ber of layer silicates under the approximation that their GPa pressure, and obtained ba = 3.0 × 10 GPa and −3 −1 elastic properties can be well described by the 5 inde- bc = 5.0 × 10 GPa , very similar to the results of pendent constants corresponding to hexagonal crystal Pawley et al. (2002). To obtain a better match between symmetry. They have measured phlogopite for example, the linear compressibilities would require adjusting Seismic Anisotropy of Subduction Zone Minerals–Contribution of Hydrous Phases 73

Fig. 8 Upper hemisphere pole fi gures of P-wave velocity, Clinochlore S-wave splitting anisotropy (dVs P velocity (km/s) dVs anisotropy (%) Vs1 polarization as a percentage) and the [2110] polarization direction (white 8.09 76.16 lines) of the fastest S-wave (Vs1) 7.50 60.0 for single crystal clinochlorite, 50.0 calculated using the constants [0110] 7.00 40.0 30.0 6.50 20.0 C11, C33, C44, C66 and C12 reported by Alexandrov and Ryzhova 6.00 10.0 upper hemisphere (1961a) and C estimated by the 5.65 0.00 13 Max.Velocity = 8.09 Min.Velocity = 5.65 Max.Anisotropy = 76.16 Min.Anisotropy = 0.00 authors, as described in the text. Anisotropy = 35.5 % The elastic tensor constructed Vp/Vs1 Vp/Vs2 here has the hexagonal 2.90 3.99 symmetry rather than the monoclinic (C2/m) symmetry of 2.60 3.50 2.40 3.25 clinochlore (see text). The c-axis 2.20 3.00 2.75 2.00 is in the centre of the pole 2.50 1.80 fi gures, other crystallographic 2.25 reference directions in the basal 1.45 1.78 Max. = 2.90 Min. = 1.45 Max.= 3.99 Min. = 1.78 plane are marked Anisotropy = 66.9 % Anisotropy = 76.7 %

C13 and other Cij (see Nye, 1957) that are constrained the shallow forearc, but it can be stable to much higher by the ultrasonic data, but further analysis is compli- temperatures than most other hydrous phases. As the cated by the fact that Alexandrov and Ryzhova (1961a) elastic constants of tremolite have never been mea- do not give the composition of their crystal, hence sured we use the ‘generic’ amphibole hornblende to more elaborate constraint on the elastic constants is illustrate the probable anisotropy of this phase. The not justifi ed. Calculated clinochlore seismic properties elastic properties of hornblende have been measured (Fig. 8) have a similar pattern of anisotropy to antigo- by Alexandrov and Ryzhova (1961b). The monoclinic rite, typical of layered silicates, with high Vp and very symmetry of hornblende with a mirror plane normal to strong shear wave splitting in the basal plane, but with [010] is clearly displayed by its seismic properties in a higher symmetry due to hexagonal symmetry of the Fig. 9. Hornblende is very anisotropic for non-layer clinochore elastic constants. The P-wave anisotropy is structure with a P-wave and S-wave anisotropy of signifi cantly lower than antigorite at 35.5%, but the 27.1% and 30.7%, respectively. The Vp/Vs anisotropy shear wave splitting is as strong at 76.2%. The Vp/Vs is also very strong with 24.3% and 39.2% for Vp/Vs1 distribution is similar that the lizardite and antigorite and Vp/Vs2, respectively. in Fig. 5. The anisotropy of Vp/Vs is high at 66.9% and 76.7% for Vp/Vs1 and Vp/Vs2, respectively, and similar in magnitude to antigorite. The highest values 3.1.3 From 180 to 400 km Depth: Minerals of Vp/Vs2 are over 3.50. For comparison the Voigt- Stable at Higher Pressure Reuss-Hill average isotropic values of Vp/Vs are 2.02, which is the highest of the all the hydrous phases in During subduction, dehydration of antigorite from this study (Table 1). about 120 km, depending on the temperature, produces In the upper mantle the amphiboles most likely to forsterite + enstatite ± chlorite + fl uid. Between 120 occur in hydrated peridotites are tremolite-actinolite or and 180 km, hydration of forsterite produces the pargasite (e.g., Fumagalli and Poli, 2005). Hacker hydrous mineral clinohumite (Stalder and Ulmer, et al. (2003a) have estimated volume fractions of trem- 2001). At greater depths it is usually assumed that olite to be between 19–34% for hydrated lherzolite. clinohumite transforms to the phase A. However, The destabilization of tremolite produces clinochlore clinohumite is stable in the presence of hydrogen con- in water saturated peridotites (Fumagalli and Poli, taining forsterite (hydrous olivine) + clinoenstatite ±

2005), hence it is restricted to about 2.5 GPa (80 km) in melt + H20 at 12 GPa (<360 km) (Smyth et al., 2006). Table 1 Isotropic and anisotropic parameters of selected hydrous phases Mineral Density Shear Mmodulus Bulk Mmodulus Poisson’s Vp (km/s) Vs (km/s) Vp/Vs AVp (%) AVs (%) AVp/Vs1 (%) AVp/Vs2 (%) (kg/m3) (GPa) (GPa) ratio Olivine reference 3355 78.01 129.45 0.25 8.34 4.82 1.73 24.3 17.9 25.5 16.4 Mineral Lizardite 2580 33.63 60.90 0.27 6.30 3.56 1.77 76.7 81.7 41.2 100.4 Antigorite 2720 37.15 62.43 0.25 6.36 3.67 1.73 71.2 67.6 36.1 92.8 Talc P = 0.9 GPa 2858 49.21 58.57 0.17 6.56 4.13 1.59 64.9 65.1 45.7 95.2 Clinochlore 2800 29.27 79.25 0.34 6.48 3.20 2.02 35.5 76.2 66.9 76.7 Hornblende 3124 42.84 88.47 0.29 6.83 3.70 1.84 27.1 30.7 24.3 39.2 Clinohumite 3261 73.09 124.72 0.26 8.25 4.73 1.74 21.8 15.9 24.7 16.7 Brucite P = 4 GPa 2605 50.05 74.34 0.22 7.36 4.38 1.68 26.5 30.9 19.3 43.1 Phase A P = 9 GPa 3191 77.43 157.41 0.29 9.04 4.93 1.83 9.3 17.6 10.2 9.1 Superhydrous B 3327 96.99 153.83 0.24 9.23 5.40 1.71 6.9 11.6 18.2 10.5 Phase D P = 24 GPa 3934 169.86 279.28 0.25 11.33 6.57 1.72 10.8 18.0 20.3 25.1 HydroWadsleyite 3395 98.57 149.23 0.23 9.09 5.39 1.69 16.3 16.5 20.9 20.2 HydroRingwoodite 3781 137.43 250.87 0.27 10.71 6.03 1.78 1.9 4.4 4.5 5.8 P = 20 GPa Pressure is room pressure (10−4 GPa) except where indicated. Shear modulus, bulk modulus, Poisson’s ratio, Vp, Vs and Vp/Vs are Hill averages of the Voigt and Reuss isotropic elastic bounds. AVp, AVs, AVp/Vs1, AVp/Vs2 are anisotropic parameters discussed in the text. Marked in bold the mineral names and their anisotropy parameters that are higher than olivine at ambient constants using the elastic constants of Abramson et al. (1997). Seismic Anisotropy of Subduction Zone Minerals–Contribution of Hydrous Phases 75

Fig. 9 Upper hemisphere pole fi gures of P-wave velocity, Hornblende S-wave splitting anisotropy P velocity (km/s) dVs anisotropy (%) Vs1 polarization (dVs as a percentage), the ^(100) polarization direction (white 7.89 30.69 lines) of the fastest S-wave 27.0 7.40 24.0 7.20 21.0 (Vs1), Vp/Vs1 and Vp/Vs2 18.0 7.00 15.0 X2 [010] 6.80 for single crystal hornblende 12.0 6.60 9.0 (monoclinic C2/m symmetry), 6.40 6.0 calculated using the elastic upper hemisphere 6.00 0.80 tensor by Alexandrov and Max.Velocity = 7.89 Min.Velocity = 6.00 Max.Anisotropy = 30.69 Min.Anisotropy = 0.80 Anisotropy = 27.1 % Ryzhova (1961b). The c-axis is Vp/Vs1 Vp/Vs2 in the centre of the pole fi gures, other crystallographic reference 1.97 2.49 directions in the basal plane are 2.32 1.85 2.24 marked 1.80 2.16 1.75 2.08 1.70 2.00 1.65 1.92 1.84 1.60 1.76

1.54 1.68 Max. = 1.97 Min. = 1.54 Max. = 2.49 Min. = 1.68 Anisotropy = 24.3 % Anisotropy = 39.2 %

Fig. 10 Upper hemisphere pole fi gures of P-wave velocity, Clinohumite S-wave splitting anisotropy P velocity (km/s) dVs anisotropy (%) Vs1 polarization (dVs as a percentage), the ⊥(100) polarization direction (white 9.53 15.86 lines) of the fastest S-wave (Vs1), Vp/Vs1 and Vp/Vs2 9.00 12.0 8.80 10.0 for single crystal clinohumite [010] 8.60 8.0 8.40 6.0 monoclinic (C2/m symmetry), 8.20 4.0 calculated using the elastic 8.00 tensor by Fritzel and Bass upper hemisphere 7.65 0.55 Max.Velocity = 9.53 Min.Velocity = 7.65 Max.Anisotropy = 15.86 Min.Anisotropy = 0.55 (1997) at ambient pressure. Anisotropy = 21.8 % The c-axis is in the centre of the pole fi gures, other crystal- Vp/Vs1 Vp/Vs2 lographic reference directions in 2.00 2.03 the basal plane are marked are the pseudo-orthorhombic axes 1.85 used by Fritzel and Bass (1997) 1.80 1.90 1.75 1.85 1.70 1.80 1.65

1.56 1.72

Max. = 2.00 Min. = 1.56 Max. = 2.03 Min. = 1.72 Anisotropy = 24.7 % Anisotropy = 16.7 %

Ti-rich Clinohumite is observed in ultramafi c rocks the unusual setting (a = 100.8°, instead of the conven- exhumed from ultrahigh-pressure conditions in Dabie tional angle b = 100.8° with a = 90°) by Fritzel and Shan, China (Hermann et al., 2007). This body con- Bass (1997) to show the similarity to olivine. In Fig. 10 sists of layered mafi c and ultramafi c rocks ranging one can see many features that resemble olivine, with from harzburgites to orthopyroxenites, associated with high Vp parallel to the a*-axis and high shear wave coesite-eclogites with a peak metamorphic conditions splitting at approximately 45° to a*- and c-axes. of 4 GPa and 700°C, compatible with 120-km depth on Clinohumite has Vp and dVs anisotropies of 21.8% a warm subduction geotherm. Clinohumite is mono- and 15.9%, whereas olivine has values of 24.3% clinic, and its elastic constants have been measured in and 17.9%, respectively, hence very similar to olivine. 76 D. Mainprice, B. Ildefonse

The velocities of Clinohumite are about 1.5% slower 3.2 Transition Zone and Lower than olivine. The anisotropy of Vp/Vs are much lower Mantle Phases than the layer silicates with 24.7 and 16.7% for Vp/ Vs1 and Vp/Vs2, respectively. The phase A is the dense hydrogen magnesium In the transition zone (410–670-km depth) the high- silicate that is stable at the lowest pressures, hence in pressure polymorphs of olivine, wadsleyite and ring- the lower part of upper mantle (Fig. 2). The phase A can woodite, in their hydrous forms have the potential to represent up to 47% and 22% by volume in hydrated store more water as hydroxyl than the on the harzburgite and depleted lherzolite, respectively Earth’s surface (e.g., Jacobsen, 2006). Wadsleyite and (Hacker et al., 2003a). The elastic constants have been ringwoodite can be synthesized with up to 3.1 wt% and 2.8 wt% H O, respectively. The maximum water solu- measured by Sanchez-Valle et al. (2006) at ambient 2 conditions and recently at high pressure (Sanchez-Valle bility in wadsleyite at transition zone conditions (15 GPa, et al., 2008). The anisotropic seismic properties of 1400°C) is about 0.9 wt% (Demouchy et al., 2005). In phase A are shown in Fig. 11 at in-situ upper mantle ringwoodite, water solubility decreases with increasing pressure of 9 GPa. The phase A is hexagonal (P63) and temperature. In subducting slabs, hydrous wadsleyite the velocity distribution is surprisingly different to hex- and ringwoodite can represent volume fractions of agonal and trigonal symmetry elastic constants illus- about 50–60%, resulting in the shallower region of the trated for lizardite, clinochlore and brucite. As illustrated mantle transition zone having a larger water storage in Mainprice (2007) the phase A has a velocity distribu- potential than the deeper region (Ohtani et al., 2004). tion similar to lizardite, clinochlore and brucite at ambi- In the upper part of the transition zone (410–520-km ent pressure, hence the role of pressure is very important depth), anhydrous wadsleyite is replaced by hydrous in DHMS minerals, as also described for the phase D wadsleyite in the hydrated mantle. The elastic constants by Mainprice et al. (2007). Notice the pseudo-tetrago- of hydrous wadsleyite have been determined at ambient nal symmetry, which is most obvious in dVs and Vp/ conditions as a function of water content by Mao et al. (2008). Hydrous wadsleyite with 1.66 wt% H O Vs2 in Fig. 11. The anisotropy of the phase A at in-situ 2 upper mantle pressure is relatively modest compared to has a P-wave and S-wave anisotropy of 16.3% and other layered structures at 9.3%, 17.6%, 10.2%, and 16.5%, respectively (Fig. 12). This compares with 9.1% for P, S, Vp/Vs1 and Vp/Vs2, respectively. anhydrous wadsleyite using the elastic constants of

Fig. 11 Upper hemisphere pole fi gures of P-wave velocity, Phase A P=9 GPa S-wave splitting anisotropy (dVs as a percentage), the P velocity (km/s) dVs anisotropy (%) Vs1 polarization polarization direction (white [2110] lines) of the fastest S-wave 9.55 17.55

(Vs1), Vp/Vs1 and Vp/Vs2 for 9.36 14.0 single crystal phase A (hexago- 9.20 10.0 nal symmetry P63), calculated [0110] 9.04 6.0 using the elastic tensor Sanchez- 8.88 Valle et al. (2008) at 9.0 GPa. 2.0 The c- axis is in the centre of upper hemisphere 8.76 0.00 Max.Velocity = 9.55 Min.Velocity = 8.76 Max.Anisotropy = 17.55 Min.Anisotropy = 0.00 the pole fi gures, other crystal- Anisotropy = 8.7 % lographic reference directions in the basal plane are marked Vp/Vs1 Vp/Vs2

1.85 2.03

2.00

1.80 1.96 1.78 1.76 1.92 1.74 1.72 1.88

1.70 1.85

Max. Ratio = 1.85 Min. Ratio = 1.70 Max. Ratio = 2.03 Min. Ratio = 1.85 Anisotropy = 8.6 % Anisotropy = 9.0 % Seismic Anisotropy of Subduction Zone Minerals–Contribution of Hydrous Phases 77

Fig. 12 Upper hemisphere pole fi gures of P-wave velocity, Hydrous Wadsleyite S-wave splitting anisotropy P velocity (km/s) dVs anisotropy (%) Vs1 polarization (dVs as a percentage), the [100] X1 Vp Contours (km/s) AVs Contours (%) polarization direction (white 9.74 16.52 lines) of the fastest S-wave 14.0 (Vs1), Vp/Vs1 and Vp/Vs2 12.0 9.20 10.0 for single crystal hydrous [010] 9.00 X2 8.0 8.80 6.0 wadsleyite (orthorhombic 8.60 4.0 symmetry Imma), calculated using the elastic tensor Mao upper hemisphere 8.27 0.11 Max.Velocity = 9.74 Min.Velocity = 8.27 Max.Anisotropy = 16.52 Min.Anisotropy = 0.11 et al. (2008) at ambient Anisotropy = 16.3 % pressure. The c-axis is in the centre of the pole fi gures, Vp/Vs1 Vp/Vs2 other crystallographic reference directions in the basal plane are 1.83 1.90 marked 1.70 1.75 1.65 1.70 1.60 1.65 1.55

1.48 1.55

Max. = 1.83 Min. = 1.48 Max. = 1.90 Min. = 1.55 Anisotropy = 20.9 % Anisotropy = 20.2 %

Fig. 13 Upper hemisphere pole fi gures of P-wave velocity, Hydrous Ringwoodite P=20 GPa S-wave splitting anisotropy P velocity (km/s) dVs anisotropy (%) Vs1 polarization (dVs as a percentage), the [100] polarization direction (white 10.80 4.40 lines) of the fastest S-wave 3.5 (Vs1), Vp/Vs1 and Vp/Vs2 for 10.74 2.5 single crystal hydroRingwoodite [010] 10.70 (cubic symmetry Fd3m), 10.66 1.5 calculated using the elastic 10.62 0.5 tensor Wang et al. (2006) at upper hemisphere 10.60 0.00 Max.Velocity = 10.80 Min.Velocity = 10.60 Max.Anisotropy = 4.40 Min.Anisotropy = 0.00 20 GPa. The c-axis is in the Anisotropy = 1.9 % centre of the pole fi gures, other Vp/Vs1 Vp/Vs2 crystallographic reference 1.81 1.83 directions in the basal plane

1.80 are marked 1.78 1.77 1.78 1.76 1.75 1.76 1.74 1.74

1.73 1.73

Max. = 1.81 Min.= 1.73 Max. = 1.83 Min. = 1.73 Anisotropy = 4.5 % Anisotropy = 5.8 %

Zha et al. (1997), with values of 15.4% and 16.8%, hydrated mantle. The elastic constants of hydrous ring- respectively. Hydrous wadsleyite has a Vp/Vs1 and woodite have been measured as a function of pressure Vp/Vs2 of 20.9% and 20.2%, respectively, whereas to 9 GPa by Jacobsen and Smyth (2006) and to 24 GPa anhydrous wadsleyite has values of 22.0% and 19.1%. by Wang et al. (2006). Hydrous ringwoodite at the Clearly it is not possible to distinguish between anhy- in-situ transition zone pressure of 20 GPa has a very drous and hydrous wadsleyite on the basis of anisot- low anisotropy, 1.9% for P-waves and 4.4% S-waves ropy, although there is a linear decrease of velocity with (Fig. 13). Anisotropy for Vp/Vs1 and Vp/Vs2 is water content at room pressure (Mao et al., 2008). slightly stronger at 4.5% and 5.8%, respectively. In the lower part of the transition zone (520–670-km The superhydrous phase B (also called C by some depth), anhydrous ringwoodite is replaced by its authors) is another DHMS mineral with orthorhombic hydrated counterpart hydrous ringwoodite in the (P21 mn) symmetry is stable over the PT conditions of 78 D. Mainprice, B. Ildefonse

Fig. 14 Upper hemisphere pole fi gures of P-wave velocity, Superhydrous B S-wave splitting anisotropy (dVs as a percentage), the P velocity (km/s) dVs anisotropy (%) Vs1 polarization [100] polarization direction (white lines) of the fastest S-wave 9.61 11.57

(Vs1), Vp/Vs1 and Vp/Vs2 for 9.44 8.0 single crystal Superhydrous 9.36 [010] 9.28 6.0 B phase orthorhombic (Pbnm 9.20 4.0 symmetry), calculated using 9.12 the elastic tensor by Pacalo and 8.97 0.19 upper hemisphere Weidner (1996) at ambient Max.Velocity = 9.61 Min.Velocity = 8.97 Max.Anisotropy = 11.57 Min.Anisotropy = 0.19 pressure. The [001] axis is in Anisotropy = 6.9 % the centre of the pole fi gures, other crystallographic reference Vp/Vs1 Vp/Vs2 directions in the basal plane are marked 1.85 1.85

1.80 1.78 1.70 1.76 1.65 1.74 1.72 1.60 1.70

1.54 1.67

Max. = 1.85 Min. = 1.54 Max.= 1.85 Min.= 1.67 Anisotropy = 18.2 % Anisotropy = 10.5 % the transition zone (Fig. 2) constitutes up to 66.5% lower mantle (P = 24 GPa). The velocity distributions and 53.3% of hydrated harzburgite and lherzolite clearly illustrate the trigonal nature of the phase D. The according to Iwamori (2004), whereas Ohtani et al. anisotropy is higher than the super hydrous B phase, (2004) estimated between 10% in the upper part and with P-wave anisotropy of 9.2% and S-wave of 18.0%. 30% in the lower part of the transition zone. Pacalo and Weidner (1996) measured the single crystal elastic constants of the superhydrous phase B at ambient 4 Elastic and Seismic Properties as conditions. The anisotropic seismic properties shown in Fig. 14 reveal that the superhydrous phase B is only a Function of Pressure moderately anisotropic with P-wave anisotropy of 6.9% and S-wave anisotropy of 11.6%. The Vp/Vs1 As already mentioned above, the elastic constants of and Vp/Vs2 anisotropies are also moderate at 18.2% only a few hydrous minerals has been measured, and and 10.5%, respectively. most of these at ambient conditions. A small number The phase D is yet another DHMS mineral and the of minerals have been measured as a function of only known hydrogen bearing mineral stable to lower pressure, and these are illustrated in Fig. 16. To our mantle pressures. Recently, the elastic constants of the knowledge, no measurements of elastic constants of phase D have been calculated using fi rst principle meth- hydrous minerals as a function of temperature have ods by Mainprice et al. (2007) to pressures of 84 GPa. been published. Pressure has a very important effect

At great depth, high temperatures are expected to limit on the seismic anisotropy of brucite Mg(OH)2, which the stability fi eld of the phase D. The percentages of the is considered as an important structural component of phase D present in the transition zone and possibly the many hydrous minerals, and hence one may suspect lower mantle are between 63.5% and 56.3% for harz- that most hydrous minerals will be pressure sensitive. burgite and lherzolite according to Iwamori (2004) and At ambient conditions, brucite is one of the most aniso-

15% for a 2 wt% H2O peridotite according to Ohtani tropic minerals with P wave and S wave anisotropies et al. (2004). As with talc, brucite and phase A the of 58% and 46%, respectively. Anisotropy reduces velocity distribution of the phase D changes with pres- with pressure, to become about half those values at sure. In Fig. 15 the anisotropic seismic properties are upper mantle pressures (Jiang et al., 2006). However, illustrated for a pressure corresponding to the top of the even at upper mantle pressures, brucite remains very Seismic Anisotropy of Subduction Zone Minerals–Contribution of Hydrous Phases 79

Fig. 15 Upper hemisphere pole fi gures of P-wave velocity, Phase D P=24 GPa S-wave splitting anisotropy P velocity (km/s) dVs anisotropy (%) Vs1 polarization (dVs as a percentage), the [2110] polarization direction (white 11.87 18.02 lines) of the fastest S-wave (Vs1), Vp/Vs1 and Vp/Vs2 11.60 14.0 for single crystal phase D [0110] 11.40 10.0 (trigonal symmetry P-31 m), 11.20 6.0 11.00 calculated using the elastic 2.0 tensor Mainprice et al. (2007) upper hemisphere 10.83 0.00 Max.Velocity = 11.87 Min.Velocity = 10.83 Max.Anisotropy = 18.02 Min.Anisotropy = 0.00 at 24 GPa. The c-axis is in the Anisotropy = 9.2 % centre of the pole fi gures, other Vp/Vs1 Vp/Vs2 crystallographic reference 1.88 1.98 directions in the basal plane are marked 1.85 1.75 1.80 1.70 1.75 1.65 1.70 1.65 1.60 1.60

1.53 1.54

Max. = 1.88 Min. = 1.53 Max. = 1.98 Min. = 1.54 Anisotropy = 20.3 % Anisotropy = 25.1 %

Fig. 16 Variation with pressure 12 25 of P- and S-wave anisotropy for 80 10 HydoRw phase D dVs the hydrous phases talc, brucite, 20 Upper Mantle Lower Mantle Lower Upper Mantle

Brucite Vp Zone Transition dVs phase A, ringwoodite, and phase 60 8 Zone Transition Talc Vp 15 D. For reference the anhydrous Talc Vs Rw 6 HydoRw 40 ringwoodite is also plotted 10 Brucite Vs Rw 4 HydoRw dVs Vp phase D Vp 20 phase A dVs Anisotropy of Vp and Vs (%) Vp and of Anisotropy 5 2 Vp Lower Mantle Lower Rw Upper Mantle phase A Vp Zone Transition 0 0 0 0 5 10 15 0 5 10 15 20 25 0 1020304050 Pressure (GPa) Pressure (GPa) Pressure (GPa)

anisotropic compared with most silicates. Talc with hydrous ringwoodite being systematically 1% or

(Mg3Si4O10(OH)2) is probably the most elastically 2% more anisotropic than ringwoodite. The pressure anisotropic mineral reported to-date at room pressure sensitivity of phase D is very different to phase A, the (Mainprice et al., 2008). Like brucite it is a structural anisotropy decreasing signifi cantly for P-wave anisot- element of many hydrous phases, and also like brucite ropy, but remaining nearly constant with pressure its anisotropy is reduced by 50% at 5 GPa pressure, but S-wave. Clearly it is not easy to predict the exact nature even so it is highly anisotropic with Vp and Vs anisot- of the pressure sensitivity of hydrous minerals, as the ropy around 40%. In the same pressure range, the examples presented in Fig. 16 are all different, with DHMS phase A has a radically different pressure sen- exception of brucite and talc that have very similar sitivity for anisotropy. In fact phase A is rather insensi- chemistries. tive to pressure, with the S-wave anisotropy actually increasing with pressure. However, these anisotropy parameters do not reveal all the changes in velocity distribution, which are very important with pressure 5 Discussion for brucite and phase A. In the case of hydrous ringwoodite we can also compare with anhydrous The elastic constants of many hydrous phases have yet ringwoodite. From the data of Wang et al. (2006), the to be measured at ambient conditions, very few have anisotropy of both minerals decreases with pressure, been measured at high pressures and none, with the 80 D. Mainprice, B. Ildefonse

possible exception of lawsonite (CaAl2Si2O7(OH)2 the mantle. The change of anisotropy in hydrogen

H2O, common component of metamorphosed glauco- containing anhydrous major phases (hydrous forms of phane schists) as a function of temperature (Schilling olivine, wadsleyite, ringwoodite) compared to their dry et al., 2003). Hence, in many ways this paper has counterparts is partly answered by data in Mao et al. surveyed the present knowledge of the anisotropic (2008) and Wang et al. (2006) on hydrous wadsleyite seismic properties of hydrous minerals thought to be and ringwoodite (Figs. 12 and 13), with a 1–2% increase present in hydrated mantle. Some of these minerals from the dry phase and the same velocity distribution. have never been observed at the Earth’s surface Some constraint on hydrous olivine can be obtained (e.g., DHMS) and can only be inferred to be present by considering clinohumite to be a ‘super saturated’ from high pressure and high temperature laboratory analogue for olivine, as its velocity distribution is experiments. The lack of pressure and temperature almost identical to olivine when using the crystal set- derivatives of the elastic constants for hydrous miner- ting chosen by Fritzel and Bass (1997). Jacobsen et al. als further hinders direction comparisons between (2006) work on hydrous olivine only gives the isotropic seismic observables and experimentally based petro- elastic parameters. The comparison between dry logical models of the hydrated mantle, which renders olivine and clinohumite show 2.5% and 2.0% reduction inferences about the hydrogen budget of the deep Earth in anisotropy in the hydrous phase. The changes in to be quite speculative. anisotropy of these major mantle phases are quite minor The other aspect, which is almost entirely missing with the incorporation of hydrogen and it seems unlikely in this paper, is the crystal-preferred orientation of that seismic anisotropy could be used as a diagnostic hydrated minerals. There is virtually no CPO data tool to study the degree of hydration of the mantle. apart from the information cited on antigorite form For these minerals the P- and S-wave seismic velocities natural samples and the general observation that lay- are reduced in the hydrogen rich conditions (1 wt% H20, ered structured minerals almost always form strong Smyth and Jacobsen, 2006) by about −0.25 km/s for preferred orientations due to their anisotropic shape Vp for ringwoodite and −0.20 km/s for wadsleyite at and anisotropic mechanical properties. Therefore, the ambient conditions. However, when the P-waves veloc- single crystal seismic anisotropies should be taken as ities are extrapolated to in-situ mantle pressure condi- an upper bound for rock samples due to two majors tions, the velocity difference between wet and dry causes; a) the statistical nature of CPO in polycrystal- phases disappears. For S-waves the effect is slightly line samples implies that crystals are never perfectly stronger, with a velocity reduction in ringwoodite of aligned, b) the presence of other minerals will, in gen- −0.3 km/s for hydrous phase, whereas in wadsleyite eral, have differently aligned elastic anisotropy will there is a velocity increase of about +0.2 km/s at ambi- further reduce the resulting seismic anisotropy of ent conditions. When extrapolated to mantle pressure, multi-phase mantle rocks at depth. the differences in S-wave velocities between dry and It is one of the major objectives of this paper to wet ringwoodite and wadsleyite are reduced from address a fi rst overview of the contribution of hydrated −0.3 to −0.2 km/s, and +0.2 to +0.1 km/s, respectively. phases to seismic anisotropy of the hydrated mantle An alternative seismic approach to studying the associated with subduction, within the limitations cited hydrated mantle, suggested by Van der Meijde et al. above. In making this overview we hope to emphasis (2003), is to measure the displacement of the ‘410’ km their importance and stimulate further work on the elas- seismic discontinuity that is caused by the olivine to tic and mechanical properties of these minerals. Perhaps wadsleyite phase transformation due to water content. the fi rst group of minerals to consider are the ones we The phase transformation depth is affected by the usual know the most about in terms of elasticity and CPO, the considerations of pressure and temperature, plus the nominally anhydrous minerals that dominate the com- effect of different equilibrium concentrations of hydro- position of the pyrolite mantle, i.e., olivine, wadsleyite, gen in the phases, where wadsleyite has been consid- and ringwoodite, leaving aside Mg-perovskite of the ered to have considerably higher concentrations than lower mantle as it is known to have a very low hydro- olivine. The temperature has a very strong effect on the gen concentation (Bolfan-Casanova, 2005). These min- equilibrium hydrogen concentration in wadsleyite. erals are considered to be able to sequestrate signifi cant Increasing the temperature from 900°C to 1400°C volumes of water due to their high volume fractions in reduces the concentration from 2.2 wt% to 0.9 wt% with Seismic Anisotropy of Subduction Zone Minerals–Contribution of Hydrous Phases 81 no signifi cant effect of pressure in the range 13–18 GPa olivine. Hydrous minerals are therefore in general very (Demouchy et al., 2005). Wadsleyite with 0.9 wt% anisotropic; hence the interpretation of subduction water has almost the same concentration as olivine at zone seismology has to take this fact into account. the similar conditions given by recent experiments (Mosenfelder et al., 2006; Litasov et al., 2007) at around 1200°C, raising the temperature of olivine also dramat- ically decreases hydrogen concentration to 0.01wt% at 6 Conclusions 1400°C, well below that of wadsleyite. Hence, recent experimental progress supports the idea of water We have undertaken a survey of the available data on having a signifi cant effect on the depth of the ‘410 km’ the single crystal elastic tensors for hydrous minerals discontinuity. However, the depth of the transition will likely to occur in hydrated mantle of subduction zones. also depend on temperature near 410 km. At tempera- Many hydrous minerals have not had their elastic con- tures of a subducted slab, i.e., 900–1200°C, there is stants measured even at ambient conditions (e.g., 10Å almost no effect, but at 1400–1500°C of a typical man- phase, phase E), some have not been measured in their tle geotherm (Fig. 2), the effect should be present with true symmetry (e.g., clinochlore, phlogopite) and most a ratio of wadsleyite to olivine hydrogen concentration have not been measured at high pressure and none of nearly 100. related to hydrated mantle have measured at high We can see that hydrous minerals have a wide varia- temperature. For serpentine minerals most of the tion in anisotropy (see Table 1, Fig. 16), with layered comparison with seismic observables have been made structures at ambient conditions having exceptionally using wave speed, Vp/Vs or Poisson’s ratio. Available high anisotropy for P- and S-waves (e.g., lizardite, elastic tensors from atomic modelling shows that antigorite, talc, brucite, clinochlore). However the lack lizardite and antigorite are very anisotropic minerals, of good quality single crystal elastic tensors at pres- these minerals are often associated in serpentinites sure is a severe handicap when discussing the seismic with talc and brucite, which are also very anisotropic. properties of antigorite, clinochlore, and to some extent There are several reports of strong CPO in antigorite clinohumite, even though excellent quality elastic con- rocks or equivalently strong anisotropy of ultrasonic stants are available at ambient conditions. Even for velocities. However, the universally used isotropic Vp/ amphiboles, no measurements have been made apart Vs ratio and Poisson’s ratio are clearly not applicable from the 1961 ultrasonic measurements of hornblende to anisotropic rocks composed of oriented anisotropic at ambient conditions (Alexandrov and Ryzhova, hydrous minerals. We have presented an analysis of 1961b). To give some indication of the magnitude of Poisson’s ratio that is clearly not a good parameter to anisotropy in hydrous phases we can use anhydrous describe seismic anisotropy, as the axial and transverse olivine as a reference anisotropic mineral of geody- strains that defi ne Poisson’s ratio are compressional or namic importance, where its anisotropy has had a extensional, whereas the wave propagation involves major impact on interpretation of mantle seismic compressional and shear strains associated with Vp, anisotropy. Olivine has values for the anisotropy Vs1 and Vs2. For an anisotropic medium Vp/Vs has to parameters given in Table 1 as follows; Vp of 24.3%, be replaced by two ratios Vp/Vs1 and Vp/Vs1, but dVs of 17.9%, Vp/Vs1 of 25.5% and Vp/Vs2 of 16.4% until now seismologists have not reported these ratios. using the olivine elastic constants at ambient condi- We have shown in the few cases where hydrous miner- tions of Abramson et al. (1997). The anisotropy of als have been measured as function of a pressure, that olivine is not very sensitive to pressure, whereas the this variable has an important effect on the velocity hydrous minerals seem to be, with exception of the distribution and in most cases reduces the degree of phase A. The minerals that have at least one anisotro- anisotropy, hence we would expect seismic anisotropy pic parameter greater than olivine are lizardite, antigo- to play a key role in the determination of the shallow rite, talc, brucite, clinochore, hornblende, clinohumite, structure of subduction zones in the upper mantle. hydrous wadselyite, and phase D. Phase A is very close to olivine with its S-wave parameter. Only the Acknowledgements DM thanks Carmen Sanchez-Valle (ETH transition zone minerals superhydrous B and hydro- Zurich) for access to her results on the phase A at high pressure ringwoodite are signifi cantly less anisotropic than prior to publication. DM also thanks the Japanese Society for the 82 D. Mainprice, B. Ildefonse promotion of Science (JSPS) for funding a fellowship to visit Bousquet R, Goffé B, Henry P, Le Pichon X, Chopin C (1997) Katsuyoshi Michibayashi at Shizuoka University where this Kinematic thermal and petrological model of the central research seismic properties of subduction zones was initiated in Alps: lepontine metamorphism in the upper crust and eclog- the summer of 2006, and to Wolfgang Friederich and Bernhard itisation of the lower crust. Tectonophysics 273:105–127. Stoeckhert for the invitation to present a keynote on this subject Bousquet R, Goffé B, Le Pichon X, de Capitani C, Chopin C, at the symposium Subduction Dynamics: Bridging the Scales Henry P (2005) Comment on “Subduction factory: 1. in Bochum, Germany, on which this paper is partly based. Theoretical mineralogy, densities, seismic wave speeds, and

We also thank Serge Lallemand and Francesca Funiciello for H2O contents” by Bradley R. Hacker, Geoffrey A. Abers, and organising the Subduction Zone Dynamics conference in Simon M. Peacock. J Geophys Res 110:B02206. Montpellier and editing this volume of Frontiers in Earth Doi:10.1029/2004JB003450. Sciences. We thank Romain Bousquet and the other reviewer for Brocher TM, Parsons T, Tréhu A-M, Snelson C, Fischer MA their constructive comments on the manuscript. This contribu- (2003) Seismic evidence for widespread serpentinized tion is also made possible thanks to the support for computer forearc upper mantle along the Cascadia margin. Geology infrastructure (my_Mac) to DM from INSU-CNRS (France) 31:267–270. and the European Science Foundation (ESF) under the Brudzinski MR, Thurber CH, Hacker BR, Engdahl ER (2007) EUROCORES Programme EuroMinScI, project euroSLAB Global prevalence of double Benioff zones. Science through contract No. ERAS-CT-2003-980409 of the European 316:1472–1474. Commission, DG Research, FP6. Cannat M (1993) Emplacement of mantle-rocks in the seafl oor at mid-ocean ridges. J Geophys Res 98:4163–4172. Carlson RL, Miller DJ (1997) A new assessment of the abun- dance of serpentinite in the oceanic crust. Geophys Res Lett 24:457–460. References Carlson RL, Miller DJ (2003) Mantle wedge water contents estimated from seismic velocities in partially serpentinized peridotites. Geophys Res Lett 30:1250. Doi:101029/ Abramson EH, Brown JM, Slutsky LJ, Zang J (1997) The elastic 2002GL016600. constants of San Carlos olivine to 17 GPa. J Geophys Res Chinner GA, Dixon JE (1973) Some high-pressure parageneses 102:12253–12263. of the Allalin gabbro, Valais, Switerland. J Petrol Alexandrov KS, Ryzhova TV (1961a) Elastic properties of rock- 14:185–202. forming minerals 2 layered silicates. Bull Acad Sci USSR Christensen NI (1966) Elasticity of ultrabasic rocks. J Geophys Geophys Ser 11:871–875. Res 71:5921–5931. Alexandrov KS, Ryzhova TV (1961b) Elastic properties of rock- Christensen NI (1978) Ophiolites seismic velocities and oceanic forming minerals 1 and amphiboles. Bull Acad crustal structure. Tectonophysics 47:131–157. Sci USSR Geophys Ser 9:1165–1168. Christensen NI (1996) Poisson’s ratio and crustal seismology. J Arcay D, Tric E, Doin M-P (2005) Numerical simulations of Geophys Res 101:3139–3156. subduction zones: effect of slab dehydration on the mantle Christensen NI (2004) Serpentine peridotites and seismology. wedge dynamics. Phys Earth Planet Inter 149:133–153. Int Geol Rev 46:795–816. Auzende A-L, Pellenq RJ-M, Devouard B, Baronnet A, Grauby O Demouchy S, Deloule E, Frost DJ, Keppler H (2005) Pressure (2006) Atomistic calculations of structural and elastic prop- and temperature-dependence of water solubility in Fe-free erties of serpentine minerals: the case of lizardite. Phys wadsleyite. Am Mineral 90 1084–1091. Chem Miner 33:266–275. Doi:101007/s00269-006-0078-x. Escartín J, Mével C, MacLeod CJ, McCaig AM (2003) Bach W, Garrido CJ, Paulick H, Harvey J, Rosner M (2004) Constraints on deformation conditions and the origin Seawater-peridotite interactions: fi rst insights from ODP Leg of oceanic detachments: the Mid-Atlantic Ridge core 209, MAR 15°N. Geochem Geophys Geosyst. complex at 15. Geochem Geophys Geosys. Doi:1010.1029/ Doi:10.1029/2004GC000744. 2001GC000278. Bach W, Paulick H, Garrido CJ, Ildefonse B, Meurer WP, Evans BW (2004) The serpentinite multisystem revisited: Humphris SE (2006) Unraveling the sequence of serpentini- chrysotile is metastable. Int Geol Rev 46:479–506. zation reactions: petrography, mineral chemistry and petro- Evans BW, Johannes W, Oterdoom H, Trommsdorf V (1976) physics of serpeninites from MAR 15°N (ODP Leg 209, Site Stability of chrysotile and antigorite in the serpentine multi- 1274). Geophys Res Lett 33, L13306. Doi:10.1029/ system. Schweiz Mineral Petrogr Mitt 56:79–93. 2006GL025681. Fumagalli P, Poli S (2005) Experimentally determined phase Bolfan-Casanova N (2005) Water in the Eath’s mantle. Miner relations in hydrous peridotites to 6–5 GPa and their conse- Mag 63:299–257. quences on the dynamics of subduction zones. J Petrol Boschi C, Früh-Green GL, Delacour A, Karson JA, Kelley DS 46:555–578. (2006) Mass transfer and fl uid fl ow during detachment fault- Fritzel TLB, Bass JD (1997) Sound velocities of clinohumite, ing and development of an oceanic core complex, Atlantis and implications for water in Earth’s upper mantle. Geophys Massif (MAR 30°N). Geochem Geophys Geosyst. Res Lett 24:1023–1026. Doi:10.1029/2005GC001074. Grauby O, Baronnet A, Devouard B, Schoumacker K, Demirdjian Bostock MG, Hyndman RD, Rondenay S, Peacock SM (2002) L (1998) The chrysotile-polygonal serpentinite suite synthe-

An inverted continental Moho and serpentinization of the sized from a 3MgO–2SiO2–excess H2O gel [abs.]: Terra forearc mantle. Nature 417:536–538. Nova Abstract Suppl 1, p. 24. Seismic Anisotropy of Subduction Zone Minerals–Contribution of Hydrous Phases 83

Hacker BR, Abers GA, Peacock SM (2003a) Subduction their relation to the rock fabric. Phys Earth Planet Int 78: factory 1. Theoretical mineralogy densities seismic wave 245–256.

speeds and H2O contents. J Geophys Res 108. Doi:101029/ Kern H, Tubia JM (1993) Pressure and temperature dependence 2001JB001127. of P- and S-wave velocities seismic anisotropy and density Hacker BR, Peacock SM, Abers GA, Holloway SD (2003b) of sheared rocks from Sierra Alpujata massif (Ronda perido- Subduction factory 2. Are intermediate-depth earthquakes in tites Southern Spain). Earth Planet Sci Lett 119:191–205. subducting slabs linked to metamorphic dehydration reac- Kern H, Liu B, Popp T (1997) Relationship between anisotropy tions? J Geophys Res 108. Doi:101029/2001JB001129. of P and S wave velocities and anisotropy attenuation in ser- Hermann J, Fitz Gerald JD, Malaspina N, Berry AJ, Scmbelluri pentinite and amphibolite. J Geophys Res 102:3051–3065.

M (2007) OH bearing planar defects in olivine produced by Kono Y, Ishikawa M, Arima M (2007) Effect of H2O released by the breakdown of Ti-rich humite minerals from Dabie Shan dehydration of serpentine and chlorite on compressional (China). Contrib Mineral Petrol 153:417–428. wave velocities of peridotites at 1 GPa and up to 1000°C Hilairet N, Daniel I, Reynard B (2006) P–V equations of state Phys Earth Planet Int 161:215–223. and the relative stabilities of serpentine varieties. Phys Chem Lagabrielle Y, Bideau D, Cannat M, Karson JA, Mével C (1998) Miner 33:629–637. Ultramafi c-mafi c plutonic rock suites exposed along the Mid-

Hirschmann MM (2006) Water melting and the deep Earth H2O Atlantic Ridge (10°N-30°N); symmetrical-asymmetrical dis- cycle. Annu Rev Earth Planet Sci 34:629–653. tribution and implications for seafl oor spreading processes. Horen H, Zamora M, Dubuisson G (1996) Seismic wave veloci- In: Buck WR, Delaney PT, Karson JA, Lagabrielle Y (Eds) ties and anisotropy in serpentinized peridotites from Xigaze Faulting and magmatism at mid-ocean ridges. Geophysical ophiolite: abundance of serpentine in slow spreading ridge: Monograph, American Geophysical Union, Washington, DC, Geophys Res Lett 23:9–12. United States, pp. 153–176. Hostetler PB, Coleman RG, Evans BW (1966) Brucite in alpine Litasov KD, Ohtani E, Kagi H, Jacobsen SD, Ghosh S (2007) serpentinites. Am Mineral 51:75–98. Temperature dependence and mechanism of hydrogen incor- Hyndman RD, Peacock SM (2003) Serpentinization of the poration in olivine at 12.5–14.0 GPa. Geophys Res Lett forearc mantle. Earth Planet Sci Lett 212:417–432. 34:L16314. Doi:10.1029/2007GL030737. Ildefonse B, Christie DM, Mission Moho Workshop Steering Mainprice D (1990) An effi cient fortran program to calculate Committee (2007) Mission Moho workshop: drilling through seismic anisotropy from the lattice preferred orientation of the oceanic crust to the Mantle. Scientifi c Drilling 4:11–18. minerals. Comput Geosci 16:385–393. Doi:10.2204/iodp.sd.4.02.2007. Mainprice D (2007) Seismic anisotropy of the deep Earth from

Ito K (1990) Effects of H2O on elastic wave velocities in ultraba- a mineral and rock physics perspective. In: Schubert G (Ed) sic rocks at 900°C under 1GPa. Phys Earth Planet Inter Treatise in Geophysics, Volume 2 pp. 437–492. Oxford, UK: 61:260–268. Elsevier.

Iwamori H (2004) Phase relations of peridotites under H2O- Mainprice D, Le Page Y, Rodgers J, Jouanna P (2007) Predicted saturated conditions and ability of subducting plates for elastic properties of the hydrous D Phase at mantle pressures:

transportation of H2O. Earth Planet Sci Lett 227:57–71. implications for the seismic anisotropy of subducted slabs Jacobsen SD (2006) Effect of water on the equation of state of near 670 km discontinuity and in the lower mantle. Earth nominally anhydrous minerals In: H Keppler and JR Smyth Planet Sci Lett 259:283–296. (Eds) Water in Nominally Anhydrous Minerals. Rev Min Mainprice D, Le Page Y, Rodgers J, Jouanna P (2008) Ab initio Geochem 62:321–342. elastic properties of talc from 0 to 12 GPa: interpretation Jacobsen SD, Smyth JR (2006) Effect of water on the sound veloc- of seismic velocities at mantle pressures and prediction ities of ringwoodite in the transition zone. In: SD Jacobsen and of auxetic behaviour at low pressure. Earth Planet Sci Lett S van der Lee (Eds) Earth’s Deep , American 274:327–338. Geophysical Union. Geophys Monogr 168:131–145. Mao Z, Jacobsen SD, Jiang F, Smyth JR, Holl CM, Frost DJ, Jacobsen SD, Jiang F, Smyth JR, Duffy TS, Mao Z, Holl CM, Duffy TS (2008) Single-crystal elasticity of wadsleyites,

Frost DJ (2006) Sound velocities of hydrous olivine and the b-Mg2SiO4,containing 0.37–1.66 wt.% H2O. Earth and Planet effects of water on nominally anhydrous minerals. Eos Trans Sci Lett 266:78–89. AGU 87(52) Fall Meet Suppl Abstract V53F-03. MacLeod CJ, Escartin J, Banerji D, Banks GJ, Gleeson M, Jiang FM, Speziale S, Duffy TS (2006) Single-crystal elasticity Irving DHB, Lilly RM, McCaig AM, Niu Y, Allerton S, of brucite, Mg(OH)(2), to 15 GPa by Brillouin scattering. Smith DK (2002) Direct geological evidence for oceanic Am Mineral 91:1893–1900. detachment faulting: the Mid-Atlantic Ridge, 15 degrees Jung H, Green HW II, Dobrzhinetskaya LF (2004) Intermediate- 45′ N. Geology 30:879–882. depth earthquake faulting by dehydration embrittlement with Mével C (2003) Serpentinization of abyssal peridotites at negative volume change. Nature 428:545–549. mid-ocean ridges C. R. Geoscience 335:825–852. Kamiya S, Kobayashi Y (2000) Seismological evidence for the Mével C, Stamoudi C (1996) Hydrothermal alteration of the existence of serpentinized wedge mantle. Geophys Res Lett upper-mantle section at Hess Deep. In: Mével C, Gillis KM, 27:819–822. Allan JF, Meyer PS (Eds), Proc. ODP, Sci. Results Kawamoto T, Hervig RL, Holloway JR (1996) Experimental 147:College Station, TX (Ocean Drilling Program), pp. 293– evidence for a hydrous transition zone of the early Earth’s 309. Doi:10.2973/odp.proc.sr.147.017.1996. mantle. Earth Planet Sci Lett 142:587–592. Michibayashi K, Abe N, Okamoto A, Satsukawa T, Michikura K Kern H (1993) P- and S-wave anisotropy and shear wave split- (2006) Seismic anisotropy in the uppermost mantle, back-arc ting at pressures and temperature in possible mantle rock and region of the northeast Japan arc: petrophysical analyses of 84 D. Mainprice, B. Ildefonse

Ichinomegata peridotite xenoliths. Geophys Res Lett Sanchez-Valle C, Sinogeikin SV, Smyth JR, Bass JD (2008) 33:L10312. Doi:10.1029/2006GL025812. Sound velocities and single-crystal elasticity of DHMS Michibayashi K, Tasaka M, Ohara Y, Ishii T, Okamoto A, Fryer phase A to high pressure and implications for seismic P (2007) Variable microstructure of peridotite samples from velocities in subducted slabs. Phys Earth Planet Int the southern Mariana Trench: evidence of a complex tectonic 171:229 – 239. evolution. Tectonophysics 444:111–118. Doi:10.1016/j. Schilling FR, Sinogeikin SV, Bass JD (2003) Single-crystal tecto.2007.08.010. elastic properties of lawsonite and their variation with Mosenfelder JL, Deligne NI, Asimow PD, Rossman GR (2006) temperature. Phys Earth Planet Int 136:107–118. Hydrogen incorporation in olivine from 2–12 GPa. Am Schmidt MW, Poli S (1998) Experimentally based water budgets Mineral 91:285–294. for dehydrating slabs and consequences for arc magma Nakajima J, Matsuzawa T, Hasegawa A, Zhao D (2001), Three- generation. Earth Planet Sci Lett163:361–379. dimensional structure of Vp, Vs and Vp/Vs beneath north- Seno T, Zhao D, Kobayashi Y, Nakamura M (2001) Dehydration eastern Japan: implications for arc magmatism and fl uids. J of serpentinized slab mantle: seismic evidence from south- Geophys Res 106:21843–21857. west Japan. Earth Planets Space 53:861–871. Nye JF (1957) Physical properties of crystals–their representa- Shipboard Scientifi c Party (2004) Site 1268. In: Kelemen PB, tion by tensors and matrices. University Press, Oxford. Kikawa E, Miller DJ, et al., Proc. ODP, Init. Repts., 209: Ohtani E, Litasov K, Hosoya T, Kubo T, Kondo T (2004) College Station, TX (Ocean Drilling Program), pp. 1–171. Water transport into the deep mantle and formation of a Doi:10.2973/odp.proc.ir.209.103.2004. hydrous transition zone. Phys Earth Planet Int 143–144: Sirotin Y, Shakolskaya MP (1982) Fundamentals of Crystal 255–269. Physics. Mir, Moscow. Pacalo REG, Weidner DJ (1996) Elasticity of superhydrous. Smyth JR, Frost DJ, Nestola F, Holl CM, Bromiley G (2006) B Phys Chem Miner 23:520–525. Olivine hydration in the deep upper mantle: effects of tem- Pawley AR, Clark SM, Chinnery NJ (2002) Equation of state perature and silica activity. Geophys Res Lett 33:L15301. measurements of chlorite pyrophyllite and talc. Am Mineral Doi:101029/2006GL026194. 87:1172–1182. Smyth JR, Jacobsen SD (2006) Nominally anhydrous minerals Peacock SM (1990) Fluid processes in subduction zones. and Earth’s deep water cycle In: SD Jacobsen and S van der Science 248:329–337. Lee (Eds) Earth’s Deep Water Cycle. Am Geophys Union Peacock SM (2001) Are the lower planes of double seismic Geophys Monogr 168:1–11. zones caused by serpentine dehydration in subducting Song I, Suh M, Woo Y-K, Haob T (2004) Determination of the oceanic mantle? Geology 29:299–302. elastic modulus set of foliated rocks from ultrasonic velocity Pellenq RJ-M, Mainprice D, Ildefonse B, Devouard B, Baronnet measurements. Engin Geol 72:293–308. A, Grauby O (2008) Atomistic calculations of the elastic Stalder R, Ulmer P (2001) Phase relations of a serpentine com- properties of antigorite at upper mantle conditions: applica- position between 5 and 14 GPa: signifi cance of clinohumite tion to the seismic properties in subduction zones. submitted and phase E as water carriers into the transition zone. Contrib to EPSL. Mineral Petrol 140:670–79. Poli S, Schmidt MW (1997) The high-pressure stability of Ulmer P, Trommsdorff V (1995) Serpentine stability to hydrous phases in orogenic belts: an experimental approach mantle depths and subduction-related magmatism. Science on eclogite-forming processes. Tectonophysics 273:169–184. 268:858–861. Poli S, Schmidt MW (2002) Petrology of subducted slabs. Annu Van der Meijde M, Marone F, Glardini D, van der Lee S (2003) Rev Earth Planet Sci 30:207–235. Seismic evidence for water deep in Earth’s upper mantle. Ranero CR, Sallarès V (2004) Geophysical evidence for hydra- Science 300:1556–1558. tion of the crust and mantle of the Nazca Plate during Wang JS, Sinogeikin V, Inoue T, Bass JD (2006) Elastic proper- bending at the north Chile Trench. Geology 32:549–552. ties of hydrous ringwoodite at high-pressure conditions. Reynard B, Hilairet N, Balan E, Lazzeri M (2007) Elasticity Geophys Res Lett 33:L14308. Doi:101029/2006GL026441. of serpentines and extensive serpentinization in subduc- Watanabe T, Kasami H, Ohshima S (2007) Compressional and tion zones. Geophys Res Lett 34:L13307. Doi:101029/ shear wave velocities of serpentinized peridotites up to 2007GL030176. 200 MPa. Earth Planets Space 59:233–244. Rossi G, Abers GA, Ronenay S, Christensen DH (2006) Unusual Welch M, Crichton WA (2002) Compressibility of clinochlore to mantle Poisson’s ratio, subduction, and crustal structure in 8 GPa at 298 K and a comparison with micas. Eur J Mineral central Alaska. J Geophys Res 111:B09311. Doi:10.1029/ 14:561–565. 2005JB003956. Zha CS, Duffy TS, Mao HK, Downs RT, Hemley RJ, Weidner

Sanchez-Valle C, Sinogeikin SV, Smyth JR, Bass JD (2006) DJ (1997) Single-crystal elasticity of beta-Mg2SiO4 to the Single-crystal elastic properties of dense hydrous magne- pressure of the 410 km seismic discontinuity in the Earth’s sium silicate phase A. Am Mineral 91:961–964. mantle. Earth Planet Sci Lett 147:E9–E15. Doi:10.1016/ S0012-821X(97)00010-1.