Generating a FAIR crystal-structure database with the AiiDA informatics infrastructure*

Computer simulations that use powerful electronic-structure techniques are nowadays widely used to characterize or predict materials’ properties. Such efforts rely on databases of measured or calculated data, with structural data being especially useful. Here, we develop and validate a set of protocols to generate a comprehensive structural database of 3D materials abiding to the FAIR data principles. We start from structures taken from three major experimental databases: the Pauling file (MPDS), the inorganic crystal structure database (ICSD), and the crystallography open database (COD). After removal of non-stoichiometric compounds and duplicates, structures are refined with density-functional theory calculations using the open-source SIRIUS accelerated library together with Quantum ESPRESSO. Since calculations are driven by the AiiDA (http://aiida.net) materials’ informatics infrastructure, all curated workflows, the entire provenance of the simulations and the resulting structural data can be shared openly on the Materials Cloud (http://materialscloud.org). We present our protocols and their validation, together with the use of AiiDA's advanced automation and error handling features to create robust workflows for electronic-structure simulations.