Element selection for crystalline inorganic solid discovery guided by unsupervised machine learning of experimentally explored chemistry

Andrij Vasylenko,Benjamin B Duff,Alexandra Morscher,Jacinthe Gamon,Laurence J Hardwick,Ruiyong Chen,Alex R Neale,Luke M Daniels,Michael W Gaultois,Vladimir V Gusev,Marco Zanella,J Felix Shin,John B Claridge,Matthew S Dyer,Paul M Sharp,Matthew J Rosseinsky,Frédéric Blanc

doi:10.1038/s41467-021-25343-7

Andrij Vasylenko, Benjamin B Duff + Show 15 more

Open Access

https://doi.org/10.1038/s41467-021-25343-7

Copy DOI

Journal: Nature Communications	Publication Date: Sep 21, 2021
Citations: 34	License type: open-access

Affiliation: University of Liverpool

Abstract

The selection of the elements to combine delimits the possible outcomes of synthetic chemistry because it determines the range of compositions and structures, and thus properties, that can arise. For example, in the solid state, the elemental components of a phase field will determine the likelihood of finding a new crystalline material. Researchers make these choices based on their understanding of chemical structure and bonding. Extensive data are available on those element combinations that produce synthetically isolable materials, but it is difficult to assimilate the scale of this information to guide selection from the diversity of potential new chemistries. Here, we show that unsupervised machine learning captures the complex patterns of similarity between element combinations that afford reported crystalline inorganic materials. This model guides prioritisation of quaternary phase fields containing two anions for synthetic exploration to identify lithium solid electrolytes in a collaborative workflow that leads to the discovery of Li3.3SnS3.3Cl0.7. The interstitial site occupancy combination in this defect stuffed wurtzite enables a low-barrier ion transport pathway in hexagonal close-packing.

Highlights

The selection of the elements to combine delimits the possible outcomes of synthetic chemistry because it determines the range of compositions and structures, and properties, that can arise
Machine learning (ML) models are powerful tools to study multivariate correlations that exist within large datasets but are hard for humans to identify[16,23]
The collaborative machine learning (ML) workflow[24,25] developed here includes a ML tool trained across all available data at a scale beyond that, which humans can assimilate simultaneously to provide numerical ranking of the likelihood of identifying new phases in the selected chemistries

Summary

Introduction

The selection of the elements to combine delimits the possible outcomes of synthetic chemistry because it determines the range of compositions and structures, and properties, that can arise. We show that unsupervised machine learning captures the complex patterns of similarity between element combinations that afford reported crystalline inorganic materials This model guides prioritisation of quaternary phase fields containing two anions for synthetic exploration to identify lithium solid electrolytes in a collaborative workflow that leads to the discovery of Li3.3SnS3.3Cl0.7. We report a collaborative ML-human expert workflow, in which unsupervised learning addresses the combinatorial problem of the discovery of new materials in the high-level formulation required by synthetic chemists to recognize the patterns at the level of element combinations that define those phase fields known to contain synthetically isolated crystalline compounds This goes beyond the traditional focus at the level of individual materials to support the decisions made in identifying new chemistries to explore. Structural and dynamical analysis of this phase demonstrated a new pathway for lithium transport in hexagonal close-packing (hcp)

Objectives

Methods

Results