Development of highly dense Ga-LLZO solid electrolyte pallet for All-Solid-State Battery using Machine Learning Approach

Oxide-based Solid electrolytes require high-temperature sintering ($>$1200 °C) for preferred densification to be used in All-Solid-State Batteries, which invites severe loss of lithium. While its Lower-temperature sintering results in less densification. Thus, its sintering process involves many contradictions and requires extensive trial and error. This study addresses this densification issue using a machine learning (ML) approach. Gallium-doped LLZO (Lithium Lanthanum Zirconium Oxide) pallets were sintered from room temperature to 1200 °C. Data from these sintering processes were used to train ML models with 1000 °C and 1100 °C datapoint using three ML algorithms: Artificial Neural Network (ANN), Support Vector Machine (SVM), and Gaussian Process Regression (GPR). The models were tested for shrinkage prediction accuracy up to 1200 °C. Validation showed all ML models predicted shrinkage with $<$0.018 RMSE, aiding densification improvement. The 1100 °C GPR model had the highest accuracy with a 0.0007 RMSE, outperforming other models.