Machine Learning Force Fields for Organic and Inorganic Systems [Materials Provided]

A message-passing machine learning force field with DFT-level accuracy accelerates organic, inorganic, and hybrid material simulations with high precision.

Machine learning force fields, also known as "machine learning interatomic potentials," have emerged as an important tool for achieving cost-effective atomic-level simulations of diverse chemical systems, often achieving accuracy comparable to density functional theory (DFT) at significantly lower computational costs. Recent advancements in message-passing networks have overcome the challenge faced by traditional MLFFs of being limited in the types of elements they can accommodate. Furthermore, the introduction of atomic charges and electrostatic interactions using charge equilibration methods allows for precise reproduction of multiple charge states, ionic systems, and electronic response characteristics, achieving even higher accuracy by explicitly considering long-range interactions. Our MLFF architecture, "MPNICE," incorporates explicit electrostatics for accurate charge representation. We also provide a set of pre-trained models trained on materials covering the entire periodic table (89 elements). MPNICE emphasizes high throughput performance, enabling long-duration and large-scale atomic-level simulations that were difficult to achieve with traditional methods while maintaining high accuracy.

simulatorEmbedded OSEAI/ETL/WEB application server