Na-ion solid-state electrolytes (Na-SSEs)exhibit high potentialfor electrical energy storage owing to their high energy densitiesand low manufacturing cost. However, their mechanical properties thatare critical to maintaining structural stability at the interfaceare still insufficiently understood. In this study, a machine learning-basedregression model was developed for predicting the mechanical propertiesof Na-SSEs. As a training set, 12,361 materials were obtained froma well-known materials database (Materials Project) and were representedwith their respective chemical and structural descriptors. The developedsurrogate model exhibited remarkable accuracies (R2 score) of 0.72 and 0.87, with mean absolute errors of11.8 and 15.3 GPa for the shear and bulk modulus, respectively. Thismodel was then applied to predict the mechanical properties of 2432Na-SSEs, which have been validated with first-principles calculations.Finally, the optimization process was performed to develop an idealmaterials screening platform by adding the minimized data set, whereinthe prediction uncertainty is reduced. We believe that the platformproposed in this study can accelerate the search for Na-SSEs withideal mechanical properties at minimum cost.
|