▌Prediction of Structural Stability of Layered Oxide Cathode Materials: Combination of Machine Learning and Ab Initio Thermodynamics
層狀氧化物正極材料結構穩定性預測：機器學習與第一原理熱力學的結合
L. T. Wu, M. Zdravković, D. Milosavljević, K. Köster, O. Guillon, J. C. Jiang and P. Kaghazchi*
https://doi.org/10.1002/aenm.202505470
SEED Member: J. C. Jiang
Major Contributions
1. A deep neural network (DNN) combining machine learning with ab initio thermodynamics accurately predicts P2/O3 phase stability for layered oxide Na-ion battery cathodes.
2. Transition metal (TM) ionic potential is identified as the dominant factor controlling Na–TM interactions; sodium concentration and mixing entropy are secondary factors that collectively determine phase outcome.
3. An integrated computational workflow (DFT + Monte Carlo simulation + thermodynamic modeling) provides a transferable rational design strategy for optimizing layered oxide cathode materials.
主要貢獻
1. 結合機器學習與第一原理熱力學的深度神經網路（DNN）模型，準確預測鈉離子電池層狀氧化物正極的 P2/O3 相穩定性。
2. 過渡金屬（TM）離子電位被確定為控制 Na–TM 相互作用的主導因素；鈉濃度與混合熵為次要因素，共同決定相結果。
3. 整合式計算流程（DFT + 蒙地卡羅模擬 + 熱力學模型）提供可遷移的層狀氧化物正極材料理性設計策略。
