Machine Learning Potentials in AMS

Within the Amsterdam Modeling Suite, several pre-trained machine learning potentials are available, including ANI, AIMNet2, M3GNet, eSEN, MACE, and UMA. AMS also supports external ML potentials through the ASE engine, including models such as CHGNet, FairChem, MatterSim, ORB, and many more.

All integrate with AMS workflows, the graphical user interface, and PLAMS scripting, so you can use ML potentials for setup, simulation, analysis, and automation in the same environment as the rest of AMS.

Training or refining your own machine learning potentials is possible with ParAMS and Simple Active Learning, including custom M3GNet and MACE models.

Your own (proprietary) in-house ML potentials can be used via the ASE engine, offering instant access to all AMS driver functionality such as our advanced Molecular Dynamics or potential energy surface exploration methods.

AMS includes and connects to machine learning potentials for a wide range of molecular and materials simulations:

• ANI-1x, ANI-1ccx, ANI-2x: fast conformers and thermochemistry for organic molecules
• AIMNet2: charged molecules, atomic charges, dipoles, and IR intensities
• eSEN-S-Con-OMol: organic, bio-relevant, electrochemical, and inorganic chemistry
• M3GNet-UP-2022: materials, cohesive energies, and periodic systems
• MACE-MP-0 and MACE-MPA-0: foundation models for inorganic materials
• UMA-S-1.2 variants: surfaces, MOFs, electrolytes, molecular crystals, molecules, and materials
• External ML potentials through ASE: CHGNet, FairChem, MatterSim, ORB, and more
• AMS Driver workflows: optimizations, MD, NEB, PES scans, phonons, and thermochemistry
• Active learning and ParAMS training for M3GNet and MACE
• GPU support and package-manager installation

Parametrization, Active Learning and Battery applications

Webinar recording on training Machine Learning Potentials in ParAMS and using Active Learning for applications in Battery Modeling.

Fast and accurate Li intercalation potentials in layered cathodes

Using the machine learning potential M3GNET Li intercalation potentials of layered cathode materials are screened, fast and accurate.

Concentration dependent Li migration barriers in LiTiS2

Calculate Li migration barriers inside a crystal structure with the fast and accurate M3GNET machine learning potential.

Fast and accurate prediction of Kevlar’s mechanical properties

Learn how to use the ANI-2x machine learning potential to calculate the Elastic Tensor and mechanical properties of Kevlar.