Benjamin X. Shi

Foundation interatomic potentials often suffer from inconsistent benchmarks due to infrastructure fragmentation, but QuAcc eliminates this methodological drift by decoupling pure quantum chemistry workflows from the underlying orchestration engines.

Widely used simulations of liquid water are surprisingly inconsistent, with errors exceeding 20% in key properties, but this work provides a robust, reproducible benchmark.

Optimizing the nodal surface in DMC calculations closes the gap with CCSD(T) benchmarks for hydrogen-bonded systems, offering a more accurate and efficient computational approach.

Achieve sub-chemical accuracy in predicting the lattice energies of complex molecular crystals, including drug-like molecules, at DFT-level computational cost, unlocking high-throughput polymorph screening.

Forget generic MLIPs: MACE-POLAR-1 nails long-range electrostatics, slashing errors in protein-ligand binding and crystal energies.

Achieve transition-metal chemical accuracy in DFT calculations for catalytic reactions on metal surfaces by using a new framework for designing non-self-consistent density functional approximations.