💡 New preprint out!
Scaling gas–surface simulations from AIMD to 100k+ trajectories with MLIP.
Efficient sampling + active learning → compact datasets, near ab initio accuracy, detailed scattering physics (energy transfer, rotation, scattering regimes)
arxiv.org/abs/2603.18864
Scaling gas–surface simulations from AIMD to 100k+ trajectories with MLIP.
Efficient sampling + active learning → compact datasets, near ab initio accuracy, detailed scattering physics (energy transfer, rotation, scattering regimes)
arxiv.org/abs/2603.18864
arxiv.org Data-driven construction of machine-learning-based interatomic... Accurate atomistic simulations of gas-surface scattering require potential energy surfaces that remain reliable over broad configurational and energetic ranges while retaining the efficiency... 