Towards Bridging the Data Exchange Gap Between Atomistic Simulation and Larger Scale Models

Abstract

Materials properties are rooted in the atomic scale. Thus, an atomistic understanding of the physics and chemistry is the foundation of computational materials engineering. The MedeA computational environment provides a highly efficient platform for atomistic simulations to predict materials properties from the fundamental interactions effective at the nanoscale. Nevertheless, many interactions and processes occur at much larger time and length scales, that need to be described with microscale and macroscale models, as exemplified by the multiphase field tool MICRESS. The predictive power of these larger scale models can be greatly increased by augmenting them with atomistic simulation data. The notion of per phase-properties including their anisotropies provides e.g., the key for the determination of effective properties of multiphase materials. The key goal of the present work is to generate a common interface between atomistic and larger scale models using a data centric approach, in which the “interface” is provided by means of a standardized data structure based on the hierarchical data format HDF5 . The example HDF5 file created by Schmitz et al., Sci. Technol. Adv. Mater. 17 (2016) 411, describing a three phase Al--Cu microstructure, is taken and extended to include atomistic simulation data of the Al--Cu phases, e.g., heats of formation, elastic properties, interfacial energies etc. This is pursued with special attention on using metadata to increase transparency and reproducibility of the data provided by the atomistic simulation tool MedeA.