Data‐Driven Materials Exploration for Li‐Ion Conductive Ceramics by Exhaustive and Informatics‐Aided Computations

Abstract

Interest in all‐solid‐state Li‐ion batteries (LIBs) using non‐flammable Li‐conducting ceramics as solid electrolytes has increased, as safe and robust batteries are urgently desired as power sources for (hybrid) electric vehicles. However, the low Li‐ion conductivities of ceramics have hindered all‐solid‐state LIB commercialization; many researchers have attempted to develop fast Li‐ion conductors. We introduce two efficient high‐throughput computational approaches for materials exploration: (i) exhaustive search and (ii) informatics‐aided prediction. For demonstration, ∼400 Li‐ and Zn‐containing oxide (Li−Zn−X−O) compounds of varied crystal structures are extracted from Materials Project datasets. We calculate the migration energies for Li‐ion conduction and the phase stabilities (decomposition energies) of these materials by simulation and apply Bayesian optimization to determine the material with the highest ionic conductivity. The results show much greater efficiency than a random search algorithm.

A Critical Review of Machine Learning of Energy Materials

“…Using less expensive computational techniques allows the generation of a larger quantity of training data. For example, Nakayama et al trained PLS and GBR models on the Li migration energy in 400 Li‐containing compounds computed using bond‐valence force fields (BVFFs). While statistically more robust, it is unclear whether the source of the training data—BVFF calculations—are sufficiently accurate to yield useful predictions, i.e., data quantity may be sufficient, but data quality is in question.…”

Shaping the Future of Solid‐State Electrolytes through Computational Modeling

“…Data resources such as the Materials Project [ 34 ] can be used to assist in the detection of materials that can be considered as SSEs through machine learning. [ 35,36,37 ] Machine learning is applied to filter and select the materials based on their properties. Computational methods such as DFT and AIMD simulations can then be used to determine further properties and the diffusion coefficient of the filtered materials.…”

“…Computational methods such as DFT and AIMD simulations can then be used to determine further properties and the diffusion coefficient of the filtered materials. [ 35 ] The need for access to a large amount of meaningful, high‐quality data is the main problem of studies based on machine learning; [ 36 ] however, this will improve in the future as the potential of this approach is realized.…”

Bayesian-optimization-guided experimental search of NASICON-type solid electrolytes for all-solid-state Li-ion batteries

“…3,4 Solid electrolytes have lower ionic conductivities than liquid electrolytes; to address this, methods to prepare solid electrolyte materials with high ionic conductivities are actively explored. [5][6][7] Recently, NASICON-type oxide-based solid electrolytes attracted considerable interest due to their high Li-ion conductivity. 8,9 Li 1.3 Al 0.3 Ti 1.7 (PO 4 ) 3 (LATP), prepared based on LiTi 2 (PO 4 ) 3 (LTP), 10,11 is a well-known NASICON-type solid electrolyte that builds on the advantages of LTP, such as high conductivity and stability under ambient conditions.…”