Discovery of superionic conductors by ensemble-scope descriptor

Abstract

Machine learning accelerates virtual screening in which material candidates are selected from existing databases, facilitating materials discovery in a broad chemical search space. Machine learning models quickly predict a target property from explanatory material features called descriptors. However, a major bottleneck of the machine learning model is an insufficient amount of training data in materials science, especially data with non-equilibrium properties. Here, we develop an alternative virtual-screening process via ensemble-based machine learning with one handcrafted and two generic descriptors to maximize the inference ability even using a small training dataset. A joint representation with the three descriptors translates the physical and chemical properties of a material as well as its underlying short-and long-range atomic structures to describe a multifaceted perspective of the material. As an application, the ensemble-scope descriptor learning model was trained with only 29 entries in the training dataset, and it selected potential oxygen-ion conductors from 13,384 oxides in the inorganic crystal structure database. The experiments confirmed that we successfully discovered five compounds that have not been reported, to the best of our knowledge, as oxygen-ion conductors.

Structure–property correlation in oxide-ion and proton conductors for clean energy applications: recent experimental and computational advancements

“…153 As mentioned above, one of the main limitations in the application of ML methods to ion conductors to identify potential candidates, from hundreds of thousands of compounds in a database as a rst-pass screening, is related to the fact that "materials data are not big data". 154 As already used in ref. 153, some further descriptorsrelated to ion conductivity can be employed to compensate for the relatively low number of experimental training data.…”

“…154 still requires to be improved, for example, regarding the defect chemistry and doping strategies in the identied stoichiometric compounds, which are oen the key source of ion conductivity, the concept of ensemble-scope descriptor learning provided a signicant advance in virtual screening, in particular for its efficient search capability using only a few tens of training datasets. 154 Lee and co-workers performed a rst-principles screening on 90 types of garnet-type oxides (A 3 B 2 C 3 O 12 ) to consider their potential as oxygen ion conductors based on the recent computation work on Ca 3 Fe 2 Ge 3 O 12 which indicated this composition as a possible novel conductor characterized by an oxygen interstitial (O i ) diffusion mechanism. 156,157 In detail, the authors generated 90 combinations of A 3 B 2 C 3 O 12 with typical valences of cations A, B, and C of 2+, 3+, and 4+, respectively, to search for energetically and dynamically stable garnet-type oxides with low E mig for the O i on the migration path.…”

Machine Learning Boosts the Design and Discovery of Nanomaterials

“…(d) Three descriptors of the virtual-screening process: a handcrafted descriptor and two state-of-the-art generic descriptors, i.e., the smooth overlap of atomic positions (SOAP) and the reciprocal three-dimensional voxel space (R3DVS). Oxygen-ion conductivity (σ) is the target property (reprinted from an open access article, ref ).…”

“…ML algorithms combined with high-throughput screening can search for novel functional materials, such as high-performance two-dimensional photovoltaic (2DPV) materials and high Curie point two-dimensional ferromagnetic materials (Figure c). , Via a ML strategy, a series of novel Co-based superalloys with good high-temperature oxidation resistance at 1000 °C were successfully screened out from over 210,000 candidates . Kajita et al developed an alternative virtual-screening process via ensemble-based ML with one handcrafted and two generic descriptors (Figure d) to maximize the inference ability even when using a small training data set with only 29 entries . This model selected potential oxygen-ion conductors from 13,384 oxides in an inorganic crystal structure database, five of which had not been reported.…”

Search for high-capacity oxygen storage materials by materials informatics

“…1 . This method has been used in a variety of applications, such as in the search for oxide ion conductors, 10 organic EL materials, 11 cathode materials for lithium-ion batteries, 12 and lithium ion conductors. 13,14 While there are many researches that have reported to able to accelerate material developments by MI, there are still scarce reports that achieved experimental validations.…”