The use of databases in inorganic chemistry has become increasingly prevalent since 1990 [1]. Consequently, high-throughput (HT) database screenings are today routinely applied to find materials for emerging applications. In order to improve screening success rates accurate descriptors are required. However, accuracy and feasibility typically have to be balanced for a chosen HT approach [2]. Therefore, descriptors using different â??levels of theoryâ? and hierarchical approaches seem to be most promising. In this talk, we use simple structural and more complex energy as well as potential grid-based descriptors to characterize and screen different inorganic materials databases for various applications.
In the first case study, we scan the database of The Materials Project [3], which currently contains >60,000 inorganic crystalline structures obtained from relaxation with electronic density functional theory (DFT) [4,5]. We determine the number of basic structural motifs (e.g., tetrahedra, octahedra) occurring in each structure. A joint probability distribution (e.g., P(Ntet, Noct)) can then be used to construct a fingerprint of a given database. Because of the increasing number of available databases, such a quick visual access to similarities between databases is becoming more and more important.
Second, we analyze zeolite framework-type discoveries with a structural [6] and a thermodynamic [7] descriptor. We find that both descriptors are reliable indicators for the potential of a given zeolite to be deployable in a mature technology [8]. Our analysis also shows that synthesis trends around the new millennium yielded highly distorted, high-energy structures with low potential for technological use, and a similar trend can be currently seen [8].
Finally, we introduce a screening approach for intercalant diffusion in materials that are potentially useful as rechargeable batteries. To this end, we combine an extended version [2] of Voronoi decomposition [9,10] with information from DFT to estimate diffusion barriers obtained from expensive nudged elastic band calculations. The resulting degree of correlation confirms this avenue as a promising path to screen hundreds to thousands of candidate structures.
[1] http://nisseshem.de/blogspot/databases-in-inorganic-chemistry-from-a-publication-statistics-perspective-2.html
[2] Pinheiro et al., CrystEngComm 15, 7531â??7538, 2013
[3] Jain et al., APL Mater. 1, 011002, 2013
[4] Hohenberg and Kohn, Phys. Rev. 136, B864â??B871, 1964
[5] Kohn and Sham, Phys. Rev. 140, A1133â??A1138, 1965
[6] Zimmermann et al., J. Am. Chem. Soc. 137, 13352â??13361, 2015
[7] Kramer et al., J. Am. Chem. Soc. 113, 6435â??6441, 1991
[8] Zimmermann and Haranczyk, Cryst. Growth Des., 10.1021/acs.cgd.6b00272, 2016
[9] Dirichlet, J. Reine Angew. Math., 209â??227, 1850
[10] Voronoi, J. Reine Angew. Math., 97â??178, 1908
