A machine-learning model has been trained to discover Heusler compounds, which are intermetallics exhibiting diverse physical properties attractive for applications in thermoelectric and spintronic materials. Improving these properties requires knowledge of crystal structures, which occur in three subtle variations (Heusler, inverse Heusler, and CsCl-type structures) that are difficult, and at times impossible, to distinguish by diffraction techniques. Compared to alternative approaches, this Heusler discovery engine performs exceptionally well, making fast and reliable predictions of the occurrence of Heusler vs non-Heusler compounds for an arbitrary combination of elements with no structural input on over 400,000 candidates. The model has a true positive rate of 0.94 (and false positive rate of 0.01). It is also valuable for data sanitizing, by flagging questionable entries in crystallographic databases. It was applied to screen candidates with the formula AB 2 C and predict the existence of 12 novel gallides MRu 2 Ga and RuM 2 Ga (M = Ti-Co) as Heusler compounds, which were confirmed experimentally. One member, TiRu 2 Ga, exhibited diagnostic superstructure peaks that confirm the adoption of an ordered Heusler as opposed to a disordered CsCl-type structure.

In the pursuit of materials with exceptional mechanical properties, a machine-learning model is developed to direct the synthetic efforts toward compounds with high hardness by predicting the elastic moduli as a proxy. This approach screens 118 287 compounds compiled in crystal structure databases for the materials with the highest bulk and shear moduli determined by support vector machine regression. Following these models, a ternary rhenium tungsten carbide and a quaternary molybdenum tungsten borocarbide are selected and synthesized at ambient pressure. High-pressure diamond anvil cell measurements corroborate the machine-learning prediction of the bulk modulus with less than 10% error, as well as confirm the ultraincompressible nature of both compounds. Subsequent Vickers microhardness measurements reveal that each compound also has an extremely high hardness exceeding the superhard threshold of 40 GPa at low loads (0.49 N). These results show the effectiveness of materials development through state-of-the-art machine-learning techniques by identifying functional inorganic materials.

This Editorial is intended for materials scientists interested in performing machine learning-centered research. We cover broad guidelines and best practices regarding the obtaining and treatment of data, feature engineering, model training, validation, evaluation and comparison, popular repositories for materials data and benchmarking datasets, model and architecture sharing, and finally publication.In addition, we include interactive Jupyter notebooks with example Python code to demonstrate some of the concepts, workflows, and best practices discussed. Overall, the data-driven methods and machine learning workflows and considerations are presented in a simple way, allowing interested readers to more intelligently guide their machine learning research using the suggested references, best practices, and their own materials domain expertise. File list (2) download file view on ChemRxiv BestPractices_submitted.pdf (2.22 MiB) download file view on ChemRxiv BestPractices paper-SI.pdf (3.00 MiB)

Most discoveries in materials science have been made empirically, typically through one-variable-at-a-time (Edisonian) experimentation. The characteristics of materials-based systems are, however, neither simple nor uncorrelated. In a device such as an organic photovoltaic, for example, the level of complexity is high due to the sheer number of components and processing conditions, and thus, changing one variable can have multiple unforeseen effects due to their interconnectivity. Design of Experiments (DoE) is ideally suited for such multivariable analyses: by planning one's experiments as per the principles of DoE, one can test and optimize several variables simultaneously, thus accelerating the process of discovery and optimization while saving time and precious laboratory resources. When combined with machine learning, the consideration of one's data in this manner provides a different perspective for optimization and discovery, akin to climbing out of a narrow valley of serial (one-variable-at-a-time) experimentation, to a mountain ridge with a 360° view in all directions.