Predictivestatmech.org shows off predictive models for new physics and chemistry that appear when moving up from the atomic to the nano and micro-scale. To support this goal, we are developing the thermodynamics of far-from equilibrium systems, building functional data structures for supercomputing and applying Bayesian inference to mine simulation data. Work in these topics builds on recent advances in fundamental computer science, applied statistics, and nonequilibrium physics and chemistry. Together, new developments in these fields will allow unprecedented access to electron through device-level simulations and analysis for materials design grounded in fundamental physics.
Two research areas collectively have the potential to greatly reduce the time and effort building, running, and analyzing molecular and continuum simulations for modern high-performance computing platforms. The first expands the theory and techniques of statistical mechanics for probabilistic simulation of energy conversion devices. The second applies advances in domain-specific languages to eliminate the lag between defining a physical, Hamiltonian model and carrying out dynamics and other computations on its potential energy landscape.