Combining Materials Modelling and Informatics Techniques for Efficient Search of Fast Lithium Ionic Conductors for All-Solid-State Battery Application

摘要

There is now a rapidly growing interest to efficiently and systematically search/screen for new functional materials using high-throughput materials modelling. Oftentimes, the workhorse tool employed for this purpose is based on first-principles density functional theory (DFT) approach because of good predictive accuracy with respect to experimental results. However, for search/screening of fast lithium ionic conductors, the heavy computational cost for evaluating Li ionic conductivity has made it impractical to use the latter as a search/screening criterion and it also impose a severe restriction on search space size. Machine learning can aid in addressing this problem by building probabilistic models for computation-intensive material properties such as ionic conductivity. In here, we present our recent efforts aimed towards the development of an effective and efficient search/screening workflow for fast lithium ionic conductors. We highlight our developed general representation schemes (accounting for cell periodicity, atom ordering, and number of atoms) for inorganic compounds or crystalline solids, demonstrating their predictive power for various DFT-calculated material properties such as cohesive energy, material density, electronic band gap energy, and thermodynamic decomposition energy[1]. We also present our informatics-aided DFT-based screening of a chemical search space with olivine- and tavorite-type structures and with ion migration energy and thermodynamic stability as criteria[2,3,4].References[1] Jalem et al., Sci. Tech. Adv. Mater. 2018, in press.[2] Jalem et al., J. Chem. Inf. Model. 2015, 55, 1158-1168.[3] Jalem et al., J. Mater. Chem. A 2014, 2, 720-734.[4] Jalem et al., Chem. Mater. 2012, 24, 1357-1364.