Atomistic simulations are playing an increasingly important role in realistic scientifi c and industrial applications in many areas including advanced materials design, drug design, renewable energy, and nanotechnology. The predictive capability of these approaches hing es on the accuracy of the model used to describe atomic interactions. Modern models are optimized to reproduce experimental values and electronic structure estimates for the forces and energies of representative atomic confi gurations deemed important for the problem of interest. However, no standardized approach exists for estimating the accuracy of interatomic models. In addition, a lack of standardization in the programming interface of interatomic models and the lack of a systematic infrastructure for archiving and retrieving them make it diffi cult to use existing models for new applications and to reproduce published results. These limitations prevent the fi eld of atomistic modeling from realizing its true scientifi c and technological potential.These limitations are being addressed by a four-year project initiated in October 2009 as part of the National Science Foundation Cyber-Enabled Discovery and Innovation program, led by the authors of this article, called the Knowledgebase of Interatomic Models (KIM). The project consists of two stages, each of which would constitute a signifi cant contribution to atomistic computational materials science. In the fi rst stage, an open-source online infrastructure consisting of a web portal, repository, and processing pipeline is being developed at http://openKIM .org. The repository will contain interatomic models, simulation codes that test the predictions of these models (referred to as "tests" for short), and reference data obtained from fi rst principles calculations and experiments. Users will be able to freely upload and download models and tests and to search the content of the site. When a new model is uploaded, a processing pipeline will automatically couple it with all tests already stored in the repository, generating new predictions that will be stored in the repository. Similarly new tests will be coupled with all models on the site to generate new predictions. In this way a great deal of information on the accuracy of interatomic models will be gathered over time. Development of the infrastructure described above will also involve the establishment, in collaboration with the National Institute of Standards and Technology and the research community, of terminology dictionaries, testing/benchmarking procedures, and application programming interface (API) standards for atomistic simulations. In particular, the KIM API will allow any interatomic model to work seamlessly with any atomistic simulation code.The second stage of the KIM project will focus on developing a framework to evaluate the transferability and precision of interatomic models. Transferability refers to the ability of a model to provide accurate predictions for properties that it was not fi tted to reproduce. Precision refe...
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.