Data refinement for enhanced ionic conductivity prediction in garnet-type solid-state electrolytes

Abstract

The demand for advanced energy storage drives an urgency to accelerate material discovery in solid-state electrolytes. In pursuit of this aim, this study presents an innovative methodology that integrates materials science insights with machine learning techniques to improve the ionic conductivity prediction in garnet-based solid electrolytes. Utilizing an expanded dataset comprising 362 data points, and exploiting easily obtainable pre-synthesis inputs, our approach incorporates rigorous data preprocessing inspired by materials science and machine learning methodologies. Through systematic feature selection and hyperparameter tuning, the model achieved an improved R-squared value of 0.85. This study highlights the efficacy of the proposed approach and underscores the potential of machine learning in streamlining materials discovery and design for next-generation solid-state batteries.