Title: Accelerated Design of Chalcogenide Glasses through Interpretable Machine Learning for Composition Property Relationships

Abstract: Chalcogenide glasses possess several outstanding properties that enable several ground breaking applications, such as optical discs, infrared cameras, and thermal imaging systems. Despite the ubiquitous usage of these glasses, the composition property relationships in these materials remain poorly understood. Here, we use a large experimental dataset comprising approx 24000 glass compositions made of 51 distinct elements from the periodic table to develop machine learning models for predicting 12 properties, namely, annealing point, bulk modulus, density, Vickers hardness, Littleton point, Youngs modulus, shear modulus, softening point, thermal expansion coefficient, glass transition temperature, liquidus temperature, and refractive index. These models, by far, are the largest for chalcogenide glasses. Further, we use SHAP, a game theory based algorithm, to interpret the output of machine learning algorithms by analyzing the contributions of each element towards the models prediction of a property. This provides a powerful tool for experimentalists to interpret the models prediction and hence design new glass compositions with targeted properties. Finally, using the models, we develop several glass selection charts that can potentially aid in the rational design of novel chalcogenide glasses for various applications.