2019 was a fantastic year for the KIM project, and 2020 promises to be even better with many exciting projects in the works. Here are some highlights from 2019; click on the [Find out more] link at the end of each item for more details.
KIM is a member organization of DataCite as of January 2019 and issues DOIs to all interatomic potentials archived on openkim.org. This allows potentials to be cited in publications, and for their usage to be tracked through Clarivate Analytics to recognize the contribution of code developers. [Find out more]
Version 2.0 of the KIM API was released comprising a major rewrite, many new features, portable CMake based build, and extensive documentation. [Find out more]
KIM API 2.0 supports two classes of interatomic potentials: Portable Models (PMs) that work with all simulators, and Simulator Models (SMs) that only work with a particular simulator (like LAMMPS). This makes it possible to archive and test cutting edge and high-performance potentials (e.g. GPU-based). [Find out more]
Binaries for the KIM API and all interatomic potentials archived on openkim.org are now available for macOS (via Homebrew) and all major Linux platforms. Support for Windows is planned. [Find out more]
Full support for KIM API 2.0 is now available for the following major simulation platforms: Asap, ASE, DL_POLY, GULP, LAMMPS, MDStressLab, Potfit, Pyiron, Quasicontinuum. [Find out more]
Examples of how to use KIM models in different simulation codes have been added to the documentation. [Find out more]
A web-based query mechanism has been created making it possible to programmatically obtain the predictions of a given potential archived in openkim.org for a specified material property through a web call. This can be used in simulation setup and post-processing. [Find out more]
KIM has been tightly integrated with LAMMPS:
All prebuilt LAMMPS binaries for Linux now automatically include the KIM API as a dependency making it easy to use interatomic potentials from openkim.org with LAMMPS. [Find out more]
Between PMs and SMs, all pair_style potentials supported by LAMMPS can be easily uploaded to openkim.org by simply filling in an online form and providing the parameter files. [Find out more]
Commands have been added to LAMMPS to select a KIM model (SM or PM) (kim_init and kim_interactions), perform a property query to openkim.org and store the result in a standard LAMMPS variable (kim_query), access and/or change the parameters of PMs (kim_param), automatically handle unit conversion, and provide citation information for KIM models used in a simulation in BibTex format. [Find out more]
The KIM-based Learning-Integrated Fitting Framework (KLIFF) was released. KLIFF can be used to fit both physics-based potentials and machine learning potentials, such as neural network potentials. KLIFF is open source, compatible with the KIM API, and has excellent documentation. [Find out more]
The openkim.org website has been redesigned with improved navigation and up-to-date extensive documentation. [Find out more]
Many other developments are in the works and will be announced soon.
The KIM Development Team