Abstract. Anisotropy of upper mantle physical properties results from lattice preferred orientation (LPO) of upper mantle minerals, in particular olivine. We use an anisotropic viscoplastic selfconsistent (VPSC) and an equilibrium-based model to simulate the development of olivine LPO and, hence, of seismic anisotropy during deformation. Comparison of model predictions with olivine LPO of naturally and experimentally deformed peridotites shows that the best fit is obtained for VPSC models with relaxed strain compatibility. Slight differences between modeled and measured LPO may be ascribed to activation of dynamic recrystallization during experimental and natural deformation. In simple shear, for instance, experimental results suggest that dynamic recrystallization results in further reorientation of the LPO leading to parallelism between the main (010)[ 100] slip system and the macroscopic shear. Thus modeled simple shear LPOs are slightly misoriented relative to LPOs measured in natural and experimentally sheared peridotires. This misorientation is higher for equilibrium-based models. Yet seismic properties calculated using LPO simulated using either anisotropic VPSC or equilibrium-based models are similar to those of naturally deformed peridotRes; errors in the prediction of the polarization direction of the fast S wave and of the fast propagation direction for P waves are usually < 15 ø. Moreover, overestimation of LPO intensities in equilibrium-based and VPSC simulations at high strains does not affect seismic anisotropy estimates, because these latter are weakly dependent on the LPO intensity once a distinct LPO pattern has been developed. Thus both methods yield good predictions of development of upper mantle seismic anisotropy in response to plastic flow. Two notes of caution have nevertheless to be observed in using these results: (1) the dilution effect of other upper mantle mineral phases, in particular enstatite, has to be taken into account in quantitative predictions of upper mantle seismic anisotropy, and (2) LPO patterns from a few naturally deformed peridotRes cannot be reproduced in simulations. These abnormal LPOs represent a small percent of the measured natural LPOs, but the present sampling may not be representative of their abundance in the Earth's upper mantle.

No abstract

International audienceThe variation of elastic- wave velocities as a function of the direction of propagation through the Earth's interior is a widely documented phenomenon called seismic anisotropy. The geometry and amount of seismic anisotropy is generally estimated by measuring shearwave splitting, which consists of determining the polarization direction of the fast shear- wave component and the time delay between the fast and slow, orthogonally polarized, waves. In subduction zones, the teleseismic fast shear- wave component is oriented generally parallel to the strike of the trench(1), although a few exceptions have been reported (Cascadia(2) and restricted areas of South America(3,4)). The interpretation of shear- wave splitting above subduction zones has been controversial and none of the inferred models seems to be sufficiently complete to explain the entire range of anisotropic patterns registered worldwide(1). Here we show that the amount and the geometry of seismic anisotropies measured in the forearc regions of subduction zones strongly depend on the preferred orientation of hydrated faults in the subducting oceanic plate. The anisotropy originates from the crystallographic preferred orientation of highly anisotropic hydrous minerals (serpentine and talc) formed along steeply dipping faults and from the larger- scale vertical layering consisting of dry and hydrated crust - mantle sections whose spacing is several times smaller than teleseismic wavelengths. Fault orientations and estimated delay times are consistent with the observed shear- wave splitting patterns in most subduction zones