A High-Throughput Computational Study Driven by the AiiDA Materials Informatics Framework and the PAULING FILE as Reference Database

Confronted with the explosion of computing power and the vast quantities of materials data thus produced, Gray proposed in 2009 the “Fourth Paradigm of Science: Data-Intensive Discovery through Data Exploration (eScience)" [Gray, J. (2009). The Fourth Paradigm, Data-Intensive Scientific Discovery (ed. T. Hey, S. Tansley and K. Tolle), xvii-xxxi. Redmond, Washington: Microsoft Corporation], defined by the unification of experiment, theory and computation. Within computational materials science, the third (computational) and fourth (data) paradigms can become the supporting pillars for a fifth one, of database-driven and database-filling research. There, our goal is to map out the missing pieces of all possible combinations of two elements in arbitrary stoichiometries and for any structure-a search simple in its definition, but that even for very simple model potentials (e.g. a binary Lennard-Jones) can give rise to millions of different structures for just one binary compound. In addition the number of potential chemical element combinations (equal to all potential inorganic solids) forces us to develop approaches that are able to reduce it to a realistic subset of the most likely ones, to be the first targets for a simulation and then to be experimentally investigated, in the search for novel inorganic solids. Here, we outline how a trustworthy database (comprehensively spanning the published experimental inorganic solids) linked to a database of high-throughput density functional theory (DFT) calculations developed from a curated reference database of experimental results (namely, the PAULING FILE-Binaries Edition (http://www.paulingfile.com)) can bridge the fourth and fifth paradigms in the case of materials science.