31-Jan-2026
Happy new year! We hope you had a pleasant winter holiday and wish you a productive 2026. We have some exciting updates to share, especially for users interested in machine learning potentials.
Announcements
- A new generalized embedded-atom method (GEAM) model driver with various parameterizations is now available in OpenKIM. Originally developed by Amit Samanta and coworkers at LLNL, GEAM is a data-driven interatomic potential aimed at modeling primarily BCC metals with accuracy comparable to machine learning potentials. See the announcement here.
- The universal MACE-MP-0-a medium foundation model is now available as a KIM Portable Model using the TorchML Model Driver, enabling plug-and-play compatibility with any KIM API compliant simulator code for this venerable interatomic potential. See the announcement here.
- Because providing the machine learning requirements for the TorchML Model Driver can be difficult on bare metal, we have developed a version of the KIM Developer Platform Docker image with all prerequisites installed. It can be used with Docker or other container software such as Singularity/Apptainer or Podman to run TorchML Portable Models out of the box, including support for utilizing the host GPU. See the documentation here.
- We have significantly improved the reproducibility of computation results reported in OpenKIM by consolidating the containers that our calculations are run in. The "profiling.container-id" key reported with each result contains a version number that corresponds to a release of the KIM Developer Platform packaging the exact environment used for the computation. This key can be found by inspecting the "pipelinespec.edn" file provided with each result (see example here), or by querying the OpenKIM database.
- A new Test Driver for computing monovacancies in arbitrary crystals under the Crystal Genome (XtalG) framework is now available. It computes the relaxed and unrelaxed formation energies and relaxation volumes for each symmetrically distinct site in any crystal, as well as relaxed and unrelaxed effective formation energies for a composition-preserving distribution of vacancies. Like all XtalG Test Drivers, it is available as a standalone Python class as part of the kimvv package for computations on users' own models and structures. See the announcement here.
Ongoing Developments
- Additional Crystal Genome (XtalG) Test Drivers are in active development, including phonons and finite temperature properties such as Gibbs free energy, heat capacity, and anisotropic thermal expansion coefficients.
How to stay up-to-date