Machine Learning-Assisted Design of Na-Ion-Conducting Glasses

Asan alternative to liquid electrolytes, all-solid-state sodium-ionbatteries are receiving significant attention due to their potentialfor improved safety and efficiency. Here, we propose a combined experimentaland machine learning (ML) approach for discovering glass electrolyteswhile also providing insights into the role of different glass components.Specifically, we experimentally prepare and measure the ionic conductivityof 27 glass compositions of the sodium aluminophosphate glass family.Further, we train ML models on this dataset to predict the ionic conductivity,which exhibits excellent agreement with the experimental results.We interpret the composition-conductivity relationship learnedby the ML model using Shapely additive explanations (SHAP), whichreveals the role played by the glass components in governing the conductivity.Employing these observations, glass compositions with improved conductivityvalues are predicted and experimentally validated. The results corroboratethe insights from SHAP analysis and enable optimized glass formulationsin real-world experiments. This demonstrates how ML tools can significantlyaccelerate the discovery of Na-ion-conducting glass electrolytes.

Automated summary

As an alternative to liquid electrolytes, all-solid-state sodium-ion batteries are receiving significant attention for their potential safety and efficiency improvements. In this study, an experimental and machine learning approach is proposed to discover glass electrolytes and gain insights into the role of different glass components. Experimental measurements and machine learning models are used to predict and understand the ionic conductivity of different glass compositions. The results show excellent agreement between the predictions and experimental measurements, demonstrating the potential of machine learning tools for accelerating the discovery of Na-ion-conducting glass electrolytes.