Machine learning unveils composition-property relationships in chalcogenide glasses

Due to their unique optical and electronic functionalities, chalcogenide glasses are materials of choice for numerous microelectronic and photonic devices. However, to extend the range of compositions and applications, profound knowledge about composition-property relationships is necessary. To this end, we collected a large quantity of composition-property data on chalcogenide glasses from the SciGlass database regarding glass transition temperature (T-g), coefficient of thermal expansion (CTE), and refractive index (n(D)). With these data, we induced predictive models using four machine learning algorithms: Random Forest, K-nearest Neighbors, Neural Network (Multilayer Perceptron), and Classification and Regression Trees. Finally, the induced models were interpreted by computing the SHapley Additive exPlanations (SHAP) values of the chemical features, which revealed the key elements that significantly impacted the tested properties and quantified their impact. For instance, Ge and Ga increase T-g and decrease CTE (two properties that depend on bond strength), whereas Se has the opposite effect. Te, As, Tl, and Sb increase n(D) (which strongly depends on polarizability), whereas S, Ge, and P diminish it. The SHAP interaction analysis indicated two-element pairs that are likely to exhibit the mixed-former effect: arsenic-germanium and sulfur-selenium. Knowledge about the role of each element on the glass properties is precious for semi-empirical compositional development trials or simulation-driven formulations. The induced models can be used to design novel chalcogenide glasses with the required combinations of properties. (c) 2022 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.