High Performance Computations of Subsurface Reactive Transport Processes at the Pore Scale

Field applications such as carbon sequestration drive the geochemistry of porous media far from equilibrium in relatively short time scales. In these short time frames, feedback processes between flow and geochemical reactions (e.g., mineral dissolution-precipitation) that take place at the pore scale are key to understanding the discrepancy between lab-derived reaction rates and the continuum scale reaction rates that are typically used in reservoir scale models. In the DOE Energy Frontier Research Center for Nanoscale Control of Geologic Carbon (NCGC), pore scale modeling is being used to gain insight into the scale dependence of parameters such as reactive surface area or reaction rates as they affect CO2 sequestration, with an objective of upscaling these parameters to continuum scale models. Under the SciDACe program we have partnered with the NCGC EFRC to develop a new capability for direct numerical simulation of reactive transport processes associated with CO2 sequestration at the pore scale. Building on high performance computations of microscale flows in complex geometries developed in the Chombo framework at LBL as part of the SciDAC APDEC project, we use operator splitting to couple a new flow and scalar transport solver in Chombo with the geochemical code CrunchFlow. The framework makes use of higher-order algorithms based on adaptive mesh refinement and finite volume methods. In current work we are focusing on experimental validation of flow and reactive transport in 3D packed bed systems. Here we discuss performance optimizations that enable high resolution calculations of flow and reactive transport in tightly packed capillary tubes using the new Chombo-CrunchFlow framework. We demonstrate resolution of less than 4 microns for problems involving calcite dissolution and precipitation.