ADAPT: Lightweight, Long-Range Machine Learning Force Fields Without Graphs
This work addresses the computational bottleneck in materials science for simulating point defects, offering a more efficient and accurate alternative to existing methods.
The paper tackled the problem of modeling point defects in materials by introducing ADAPT, a machine-learning force field that replaces graph neural networks with a Transformer-based approach, achieving roughly 33% reduction in force and energy prediction errors on a silicon point defect dataset.
Point defects play a central role in driving the properties of materials. First-principles methods are widely used to compute defect energetics and structures, including at scale for high-throughput defect databases. However, these methods are computationally expensive, making machine-learning force fields (MLFFs) an attractive alternative for accelerating structural relaxations. Most existing MLFFs are based on graph neural networks (GNNs), which can suffer from oversmoothing and poor representation of long-range interactions. Both of these issues are especially of concern when modeling point defects. To address these challenges, we introduce the Accelerated Deep Atomic Potential Transformer (ADAPT), an MLFF that replaces graph representations with a direct coordinates-in-space formulation and explicitly considers all pairwise atomic interactions. Atoms are treated as tokens, with a Transformer encoder modeling their interactions. Applied to a dataset of silicon point defects, ADAPT achieves a roughly 33 percent reduction in both force and energy prediction errors relative to a state-of-the-art GNN-based model, while requiring only a fraction of the computational cost.