Experimental and numerical investigation of dislocation-based transient creep mechanisms in the upper mantle

Transient creep of olivine in the upper mantle plays an important role in large-scale Earth processes such as glacial isostatic adjustment and postseismic creep, as well as (exo-)planetary tidal heating and orbital dynamics. Yet, an experimentally confirmed microphysical understanding of transient creep across all timescales relevant to Earth processes remains elusive. An increasing body of laboratory and geodetic work suggests that nonlinear, dislocation-based dissipation mechanisms may play a more important role than previously thought. In response, several dislocation-based transient creep mechanisms have been proposed to explain transient creep in the upper mantle, including intergranular plastic anisotropy and the build-up of backstresses arising from long-range dislocation interactions.    The time-dependent dissipation of strain energy during transient creep manifests as attenuation, Q-1, in the frequency domain. Therefore, the constitutive equations of the proposed mechanisms should be able to predict the attenuation in polycrystalline olivine subjected to forced oscillations, providing an independent test of their applicability. Here we present numerical investigation of the nonlinear constitutive equations of these models in the frequency domain and comparisons thereof to the mechanical results of a set of high-stress, forced-oscillation experiments on polycrystalline olivine performed in a deformation-DIA coupled with synchrotron analysis techniques. Key microstructural variables needed to inform these comparisons, such as grain size, plastic anisotropy, and dislocation density, were obtained from electron backscatter diffraction and dislocation decoration.   The experiments demonstrate amplitude-dependent attenuation, which is characteristic of dislocation-based dissipation. In addition, we find that Q-1 depends on the maximum stress amplitude experienced by the sample. Dislocation-density piezometry indicates that this history effect can be linked to dislocation density evolution as post-experiment dislocation densities reflect the highest stresses obtained in the experiment rather than the stresses obtained near the end of the experiment. Numerical analysis of the constitutive equations yields high Q-1 values, up to ~5, which is similar to the experimental observations. We find that the experimental observations are consistent with predictions from the backstress model for the grain sizes and dislocation densities of our samples. When extrapolated to lower stress amplitudes, the backstress mechanism produces approximately linear behavior and behaves as a Burgers model in frequency space, suggesting that dislocation interactions may contribute to seismic wave attenuation as well.