Growing demand in applications like portable electronics and electric vehicles calls for cost-effective, safe, and high-performance energy storage systems. Development of solid-state electrolytes with Li\(^{+}\) ionic conductivities comparable to those of the current liquid chemistries is an important step towards meeting these needs. Unfortunately, one of the most promising solid electrolytes known to date, lithium lanthanum zirconium oxide (LLZO) garnets, exhibits far from ideal ionic conductivity. Thus, significant efforts, often through aliovalent substitution, have been devoted to increasing their ionic conductivity. Given the high-dimensional design space involved and the time required for synthesis, processing, and characterization of new materials, brute force approaches are not ideal to identify optimal compositions. We assess whether machine learning tools can be used to effectively explore the design space of LLZO garnets and potentially reduce the number of experiments involved in their development. We collected, curated, and filtered all the experimental results of Li\(^{+}\) ionic conductivity in LLZOs published in the scientific literature. Exploration of this data provides insights into the mechanisms that govern ionic transport in these oxides. Furthermore, we show that active learning with predictive models based on random forests can effectively be used with current data for the design of experiments. Our results indicate that the current highest Li\(^{+}\) ionic conductivity garnet LLZO could have been discovered with only 30% of the experimental studies conducted to date. All data and models are available online and can be used to drive future investigations.

便携式电子产品和电动汽车等应用的需求不断增长，这就要求具有成本效益，安全和高性能的储能系统。用Li \（^ {+} \）开发固态电解质离子电导率可与目前的液体化学相比，这是满足这些需求的重要一步。不幸的是，迄今为止已知的最有前景的固体电解质之一，氧化镧镧锆锆（LLZO）石榴石，远没有达到理想的离子电导率。因此，人们通常致力于通过异价取代来增加其离子电导率。考虑到所涉及的高维设计空间以及合成，加工和表征新材料所需的时间，蛮力方法对于确定最佳成分并不是理想的选择。我们评估了机器学习工具是否可用于有效探索LLZO石榴石的设计空间，并可能减少参与其开发的实验数量。我们收集，策划，科学文献中发表的LLZO中的离子电导率（^ {+} \）。对这些数据的探索提供了对控制这些氧化物中离子迁移的机理的见解。此外，我们表明，基于随机森林的预测模型进行的主动学习可以有效地与当前数据一起用于实验设计。我们的结果表明，迄今为止只有30％的实验研究发现了当前最高的Li （（^ {+} \））离子电导率石榴石LLZO。所有数据和模型均可在线获得，并可用于推动未来的调查。