SiO2 Fracture: Chemomechanics with a Machine Learning Hybrid QM/MM Scheme

PI James Kermode, King's College London
SiO2 Fracture: Chemomechanics with a Machine Learning Hybrid QM/MM Scheme
Project Description

According to the World Business Council for Sustainable Development, about five percent of total human energy consumption is currently expended by crushing and grinding rocks. Doing this more efficiently would eliminate millions of tons of carbon dioxide emissions per year. However, the underlying “stress corrosion” fracture processes in rocks are poorly understood.

This INCITE project is pioneering simulation methodologies for predictive modelling in these kinds of systems with the goal of modeling failure processes in oxides, which is relevant not only for the mining industry, but for structural glasses, photovoltaic devices, and biomedical implants, as well. Insights garnered from these studies could help rationalize and guide future materials design and processing developments.

Until recently, the theoretical study of these processes has been difficult because of the high cost of experiments and the tight chemo-mechanical coupling of the chemistry and elastic fields, which creates an inextricably multiscale problem.

To solve this, the research team is applying a hybrid multiscale simulation program that combines various levels of theories to help describe the fracturing of silicon dioxide in a wet environment. In this quantum mechanical/molecular mechanical (QM/MM) scheme, higher level theories account for the breaking of chemical bonds; the less expensive levels enable the inclusion of thermal fluctuations that link the microscopic behavior of these breaking bonds to the macroscopic stress that drives crack propagation.

As the INCITE project develops, the QM/MM approach will be coupled with a machine learning method, Learn on the Fly, that allows for quantum mechanical accuracy on a large model system.

Allocations