Deciphering the facet-dependent degradation mechanism of hybrid perovskites by machine learning potentials

Abstract

To elucidate the microscopic mechanisms underlying moisture-induced degradation in perovskite materials, we developed a machine learning potential capable of describing the interactions between various facets of formamidinium lead iodide (FAPbI3) and water with a density functional theory level accuracy. Among the studied (100), (110), (111), and (210) facets, we find that the (100) facet with a PbI2-rich termination exhibits superior intrinsic moisture resistance. This stability arises from two critical mechanisms: (i) our newly introduced structural descriptor, the molecular orientation index, reveals that FA molecules on PbI2-rich (100) show enhanced resistance to water perturbation and (ii) this surface forms significantly fewer hydrogen bonds with water molecules compared to other facets. Furthermore, the (100) facet undergoes a unique layer-by-layer degradation process, a phenomenon not observed in other orientations. Notably, the unstable FAI-rich (100) surface becomes water-resistant when exposed to water vapor rather than liquid water, highlighting the collective behavior of water molecules in the degradation process. Our findings provide critical mechanistic insights into facet-dependent degradation pathways in FAPbI3 and offer promising strategies for enhancing perovskite stability through facet engineering.