Hydro-mechanical-chemical modelling of dehydration during serpentinite deformation: Comparing laboratory and numerical experiments

Dehydration reactions play a pivotal role in the dynamics and seismicity at subduction zones and in the deep water cycle. These reactions often occur during rock deformation. The dehydration of antigorite serpentinite is particularly important at subduction zones. This dehydration has been investigated with laboratory experiments of serpentinite deformation. Yet, the reproduction of such laboratory deformation experiments of serpentinite dehydration with mathematical models is still a major challenge. Here, we test a two-dimensional (2D) hydro-mechanical-chemical (HMC) numerical model for serpentinite dehydration by comparing the numerical results with the results of laboratory experiments. The laboratory experiments are performed with a Griggs apparatus. Natural antigorite serpentinites with and without preferred orientation are deformed by vertical compression for a confining pressure of 1.5 GPa and maximum differential stresses between 350 and 700 MPa. For comparison, also experiments with hydrostatic stress are performed. The experimental temperature is between 620 and 650 °C. The applied confining pressure and temperature are in the olivine stability field according to the measured chemical composition of the serpentinite and thermodynamic Perple_X calculations. However, olivine only forms locally in the serpentinite if the serpentinite is deformed under differential stress. Olivine does not form in serpentinite under hydrostatic stress. Hence, we hypothesize that olivine formation is controlled by reaction kinetics and that the kinetics are locally faster in serpentinite that deforms under differential stress. We elaborate a 2D HMC numerical algorithm that can simulate dehydration and olivine generation in a deforming serpentinite [Schmalholz et al., 2023]. The algorithm is based on a staggered finite difference discretization and employs a matrix-free, pseudo-transient iterative solver. Furthermore, the algorithm is programmed in the Julia language, employs the ParallelStencil package, and runs on GPUs. We discuss three major numerical challenges: First, the treatment of large changes in solid density during the generation of olivine by serpentinite dehydration. Second, the treatment of large temporal and spatial gradients in the unknowns, such as porosity and fluid pressure. Third, the treatment of strongly nonlinear relations between unknowns and parameters, such as the relations between density and fluid pressure, porosity and permeability, and porosity and rock viscosity. We implement several mathematical formulations for the reaction kinetics and discuss which formulation can explain the laboratory results best. We further discuss potential numerical benchmarks of such HMC algorithms for modelling the coupling of chemical reactions, fluid flow and rock deformation.        References Schmalholz, S. M., E. Moulas, L. Räss, and O. Müntener (2023), Serpentinite Dehydration and Olivine Vein Formation During Ductile Shearing: Insights From 2D Numerical Modeling on Porosity Generation, Density Variations, and Transient Weakening, Journal of Geophysical Research: Solid Earth, 128(11), e2023JB026985, doi:https://doi.org/10.1029/2023JB026985.