[S-IT16] Deep Earth Sciences

[SIT16-09] Coble Creep May Control Lower Mantle Rheology

*Jac van Driel¹, David Dobson¹, John Brodholt¹ (1.University College London)

Keywords:Mantle, Rheology, Grain Boundaries, Machine Learning, Diffusion, Bridgmanite

Constraining the dynamics of Earth’s silicate-rich lower mantle informs our understanding of the entire planet. Of the lower mantle’s set of phases, bridgmanite is by far the most influential. Occupying up to 70% of the Earth’s lower mantle volume, the ubiquity of this mineral suggests that it controls much of the lower mantle’s thermal and chemical transport. The dominant deformation mechanism of Earth’s lower mantle has long been speculated. Until recently, the lack of seismic anisotropy throughout much of the lower mantle has led many to suggest that diffusion creep is preferable to dislocation creep, in which the latter typically leaves a seismically observable signature because of the crystal-preferred orientation generated by specific crystallographic slip systems. Nevertheless, this stance has recently been questioned due to the possibility of dislocation climb, which, unlike dislocation glide, generates no seismic anisotropy. The distinct advantage of dislocation creep for geodynamic models is the lack of dependence on the grain size, a much-contested parameter. Furthermore, dislocation creep adheres to a linear dependence on the system’s shear stress, unlike diffusion creep. This property facilitates the propensity for dislocation creep to generate low viscosities under high-stress conditions. Nevertheless, pure climb creep and Nabarro-Herring creep are intrinsically rate-limited by the slowest diffusing species within the lattice. Thus, models suggest both Nabarro-Herring and pure-climb require relatively high vacancy concentrations.

Another possible explanation is that materials could be deforming via Coble creep. This diffusion mechanism facilitates the propagation of vacancies and atoms along grain boundaries rather than through the lattice interior. However, to date, relatively poor constraints have been placed on the essential parameters of grain boundary diffusivity and width, not to mention the mean grain size of the lower mantle. Computational methods have modelled the kinetics of processes such as diffusion, dislocations, and, more recently, grain boundaries. Such approaches have led to an enhanced understanding of the different pathways through which atoms migrate and the relative strength of slip systems within lower mantle minerals. Grain boundaries have been shown to influence the rheology of polycrystalline aggregates through the nucleation of dislocations and the migration of interfaces. Nevertheless, the role of grain boundaries in Earth materials is most important through the process of Coble creep. Coble, like its lattice counterpart, Nabarro-Herring, deforms a material through the diffusional flux of atoms and vacancies between different surfaces. However, unlike Nabarro-Herring, Coble creep is an order of magnitude more sensitive to the grain size of the medium. Furthermore, grain boundaries typically facilitate faster diffusion along the interface when compared to the lattice interior, a process that further controls more expansive properties such as ionic conductivity.

This work utilises a developed Machine Learning Potential to establish low-energy grain boundary structures for bridgmanite at six orientations. Following the analysis of grain boundary properties such as width, energy and structure, the diffusivity of the grain boundary region is obtained through large-scale molecular dynamics calculations and the mean squared displacement analysis. The properties of grain boundary width and diffusivity are then used to place constraints on the grain-size-dependent viscosity of the lower mantle via the Coble Creep formula.