Accelerated design of chalcogenide glasses through interpretable machine learning for composition–property relationships

Abstract

Chalcogenide glasses (ChGs) possess various outstanding properties enabling essential applications, such as optical discs, infrared cameras, and thermal imaging systems. Despite their ubiquitous usage, these materials’ composition–property relationships remain poorly understood, impeding the pace of their discovery. Here, we use a large experimental dataset comprising ∼24 000 glass compositions made of 51 distinct elements from the periodic table to develop machine learning (ML) models for predicting 12 properties, namely, annealing point, bulk modulus, density, Vickers hardness, Littleton point, Young’s modulus, shear modulus, softening point, thermal expansion coefficient, glass transition temperature, liquidus temperature, and refractive index. These models are the largest regarding the compositional space and the number of properties covered for ChGs. Further, we use Shapley additive explanations, a game theory-based algorithm, to explain the properties’ compositional control by quantifying each element’s role toward model predictions. This work provides a powerful tool for interpreting the model’s prediction and designing new ChG compositions with targeted properties. Finally, using the trained ML models, we develop several glass-selection charts that can potentially aid in the rational design of novel ChGs for various applications.