Intelligently screening passivation materials is critical for improving the power conversion efficiency (PCE) values of perovskite solar cells (PSCs), which are still lacking. Herein, machine learning is employed to map the correlations between the PCE and interface passivation material at the atomic level, enabling intelligent material screening. The dataset includes around 100 interface materials used at the perovskite/hole transport layer interface. The random forest model best predicts the PCE, with a root mean square error of 0.7%. High-throughput predictions are further made and rationalized using density functional theory calculations. It is revealed that a material with a high binding energy with the [PbI₄]²⁻ surface promotes strong passivation effects, and small organic cations with an NH₃⁺ terminal have high potential. Experimental validation using methylammonium iodide and phenethylammonium iodide as the interface materials reveals the reliability of the predictions. Our work enables the high-throughput and rapid screening/design of interface materials for highly efficient PSCs, and it also provides general screening rules for interface materials at the atomic level.

中文翻译：

---

通过机器学习智能筛选界面钝化材料用于高效钙钛矿太阳能电池

智能筛选钝化材料对于提高钙钛矿太阳能电池 (PSC) 的功率转换效率 (PCE) 值至关重要，而这方面仍然缺乏。在此，机器学习用于在原子水平上映射 PCE 和界面钝化材料之间的相关性，从而实现智能材料筛选。该数据集包括大约 100 种用于钙钛矿/空穴传输层界面的界面材料。随机森林模型最好地预测 PCE，均方根误差为 0.7%。使用密度泛函理论计算进一步做出和合理化高通量预测。结果表明，具有高结合能的材料与 [PbI ₄ ] ^2-_{表面促进强钝化作用，具有NH 3} ⁺末端的小有机阳离子具有高电位。使用甲基碘化铵和苯乙基碘化铵作为界面材料的实验验证揭示了预测的可靠性。我们的工作使高效PSC的界面材料的高通量和快速筛选/设计成为可能，并且它还为原子水平的界面材料提供了一般筛选规则。