Environmental
stability of perovskite solar cells (PSCs) can be improved by a thin layer of
low-dimensional (LD) perovskite sandwiched between the perovskite absorber and the
hole transport layer (HTL). This layer, called ‘capping layer,’ has mostly been
optimized by trial and error. In this study, we present a machine-learning
framework to rationally design and optimize perovskite capping layers. We ‘featurize’
21 organic halide salts, apply them as capping layers onto methylammonium lead
iodide (MAPbI<sub>3</sub>) thin films, age them under accelerated conditions
combining illumination and increased humidity and temperature, and determine
features governing stability using random forest regression and SHAP (SHapley
Additive exPlanations). We find that a low number of hydrogen-bonding donors
and a small topological polar surface area of the organic molecules correlate
with increased MAPbI<sub>3</sub> film stability. The top performing organic
halide salt, phenyltriethylammonium iodide (PTEAI), successfully extends the
MAPbI<sub>3</sub> stability lifetime by 4±2 times over bare MAPbI<sub>3</sub>
and 1.3±0.3 times over state-of-the-art octylammonium bromide (OABr). Through
morphological and synchrotron-based structural characterization, we found that
this capping layer consists of a Ruddlesden-Popper perovskite structure and
stabilizes the photoactive layer by “sealing off” the grain boundaries and
changing the lead surface chemistry, through the suppression of lead (II) iodide (PbI<sub>2</sub>) formation and
methylammonium loss.