Toward Generalizable Surrogate Models for Molecular Dynamics via Graph Neural Networks

We present a graph neural network (GNN) based surrogate framework for molecular dynamics simulations that directly predicts atomic displacements and learns the underlying evolution operator of an atomistic system. Unlike conventional molecular dynamics, which relies on repeated force evaluations and numerical time integration, the proposed surrogate model propagates atomic configurations forward in time without explicit force computation and can be applied in an autoregressive manner for multistep temporal evolution. The approach represents atomic environments as graphs and combines message-passing layers with attention mechanisms to capture local coordination and many-body interactions in metallic systems. Trained on classical molecular dynamics trajectories of bulk aluminum, the surrogate achieves sub angstrom level accuracy within the training horizon and exhibits stable behavior during short to midhorizon temporal extrapolation. Structural and dynamical fidelity are validated through agreement with reference radial distribution functions and mean squared displacement trends. The results suggests the model preserves key physical signatures beyond point-wise coordinate accuracy. The results from both the baseline single-step model and the autoregressive model establish that GNN-based surrogate integrators are a promising and computationally efficient complement to traditional molecular dynamics for accelerated atomistic simulations within validated settings.