JARVIS-FF: Evaluation and comparison of classical interatomic potentials
JARVIS-FF is a database designed to help Molecular Dynamics and/or Monte Carlo users to choose the classical empirical potential that is right for their simulational needs. At the moment the database consists of 3128 entries among energetics and elastic property calculations, and is currently being expanded to include surface and defect-related properties as well as phonon-related data. All these properties are computed for all empirical potentials available to LAMMPS users or deposited into the NIST Interatomic Potentail Repository (IPR), and are compared to density functional theory (DFT) as well as to experimental data, where available. The most notable characteristic of this database is its easy-to-use web-interface that has been designed expressly to facilitate direct comparison between potentials. This is accomplished through clear look-up tables that allow to compare at a glance the results from different potentials for the same or similar systems. Lastly, both the complete database and the software coding used in the process have been released for public use online (presently at Github) in a user-friendly way designed to enable further material design and discovery.
Code Availability: The code is written in python and it is publicly available on Github .
Reproducibility and Technical Validation: All the data in the database is intended to be reproduced very easily, as they were obtained by running the publicly available code LAMMPS and all the used input files are available for download from the zipped folders stored on the web for each calculation. To validate the FF applicability, the elastic constants and cohesive energy data of each structure are compared with those obtained using DFT. These DFT data were taken from the Materials Project (MP).
Derived Quantities: Thanks to the high-throughput nature of Jarvis, the user has a multitude of data available for the material/structure of interest. Interesting derived quantities can then be obtained, as, for instance, convex-hull plots that are used to know the phase stability of a material on the basis of composition and energetics information. The scripts for plotting muti-component convex-hull plot is provided at the Github page.
Data analytics: Data analytics tools provide the opportunity to determine correlations in data sets that can illuminate the accuracy of predictions and the flexibility of underlying models. As an example, to investigate if there are any trends in the elastic properties predicted by the various potentials, we used the data in Jarvis to perform a principal-component analysis on the relative errors of the predicted elastic constants compared to experimental data experimental data for the four FCC elements with the largest number of available potentials, i.e. Al, Cu, Ni, and Ag. The research paper for JARVIS-FF has been submitted to journal and will be available soon.
Outreach: Currently listed among the "Pre/Post Processing Tools for use with LAMMPS" LAMMPS Tools