NEP

Why couple a GPU MD engine to other codes? Modern simulation workflows rarely live inside a single program. You might want a machine-learned potential for the bulk of a system and a quantum-chemistry (QM) solver for a small reactive region; or an enhanced-sampling library to steer a rare event; or a path-integral or transition-state code wrapped around the dynamics. In every case there is an MD engine — the code that moves the atoms — and a driver — the code that decides where they move or what forces act on them....

The rare-event problem In standard MD, the probability of observing a configuration $\mathbf{R}$ at temperature $T$ follows the Boltzmann distribution: $$P(\mathbf{R}) \propto \exp\left(-\frac{U(\mathbf{R})}{k_B T}\right)$$ where $U(\mathbf{R})$ is the potential energy and $k_B$ is Boltzmann’s constant. Transitions between metastable states separated by free-energy barriers $\Delta G^\dagger \gg k_B T$ are exponentially rare — the mean first-passage time scales as $\tau \sim \tau_0 \exp(\Delta G^\dagger / k_B T)$. For barriers above ~1 eV, these events become effectively inaccessible on MD timescales....