Predicting viscoplastic anisotropy in the upper mantle: a comparison between experiments and polycrystal plasticity models

Viscoplastic deformation of mantle rocks produces crystal preferred orientations (CPO or texture) of olivine and, by consequence, an anisotropic mechanical response for subsequent deformation events. Olivine polycrystals deformed experimentally in torsion and then in extension normal to the previous shear plane have extensional strengths ∼2 times higher than the torsional ones. Implementation of this viscoplastic anisotropy in geodynamic codes depends nevertheless on the predictions of polycrystal plasticity models, since the full range of orientation relations between applied stresses and the broad range of texture types observed in upper mantle rocks may only be tested numerically. Here, we compare instantaneous viscoplastic responses of olivine polycrystals predicted by tangent and second-order viscoplastic self-consistent (VPSC) and stress equilibrium-based models to experimental data. These polycrystal plasticity models, in which only dislocation glide is considered to accommodate strain, reproduce qualitatively the viscoplastic anisotropy observed in the laboratory, but overestimate its magnitude for rocks with strong textures. The lower bound model better approaches the laboratory results, suggesting that the discrepancy between laboratory and numerical experiments arises from the models forcing strain compatibility in olivine polycrystals that can only deform by dislocation creep. The experimental data is indeed well reproduced by the second-order VPSC model modified to include additional isotropic deformation mechanisms, which accommodate the deformation components that cannot be produced by dislocation glide but do not produce plastic spin. These results corroborate that other processes in addition to dislocation glide contribute significantly to viscoplastic deformation under laboratory conditions, reducing the magnitude of anisotropy. In nature, coarser grain sizes probably result in lower activity of these additional deformation processes. We therefore propose that standard and modified second-order VPSC models provide a good prediction of the possible range of texture-induced viscoplastic anisotropy in upper mantle domains deforming by dislocation creep.