Four connected directions in molecular computation and predictive design.
Molecular simulations provide detailed atomistic insight into complex molecular processes, but realistic simulations can be expensive in time, energy, and computing resources. We develop data-driven, statistical-mechanical, and geometry-based methods to accelerate molecular simulations and rare-event sampling. Our goal is to make atomistic simulation sufficiently efficient and scalable for high-throughput molecular discovery.
Reliable and efficient energy storage is essential for broader adoption of renewable energy. We combine data-driven methods, molecular simulation, and quantum calculations to understand structure, stability, and atomistic dynamics of energy-storage materials. Our goal is to design new materials for electrical and thermal energy storage with targeted properties.
Designing molecules that bind selectively and strongly to a biological target remains a major bottleneck in drug discovery. We combine physics-based molecular simulation with data-driven models to improve molecular screening, scoring, and design. Our goal is to develop robust computational approaches for efficiently discovering and optimizing new drug molecules.
Spin-active solid-state defects are promising building blocks for quantum technologies, but controlling their formation requires an atomistic understanding of defect dynamics. We develop defect-specific machine-learning interatomic potentials that approach ab initio accuracy at substantially lower computational cost. We use these models to investigate defect formation, migration, stability, and strategies for their selective creation.