Passivation Engineering Using Ultrahydrophobic Donor–π–Acceptor Organic Dye with Machine Learning Insights for Efficient and Stable Perovskite Solar Cells

Abstract

To prevent the degradation of perovskite solar cells (PSCs) and optimize the solar energy conversion process, a donor–π–acceptor (D–π–A) organic blue dye as a passivation layer and as a hole‐transporting layer is introduced. The terminal chains of D–π–A dye confer the ultrahydrophobic character (contact angle > 100°) of the interface layer, protecting the perovskite from ambient moisture while mitigating ionic diffusion in the device. The dye interlayer primarily improves the perovskite by reducing grain boundary defects. The perovskite/D–π–A architecture enhances the interfacial hole extraction, suppressing nonradiative carrier recombination and enabling power conversion efficiency (PCE) reaching 20.90%, outperforming by 2.05% the PCE of control cells. Unsealed PSCs retain 84% and 62% of their efficiency after photovoltaic operation for 1000 and 3000 h, respectively. Statistical correlation of bivariant and multivariant analyses of photovoltaic parameters is performed and Pearson's correlation identifies underlying patterns in experimental data collections. Machine learning (ML) of regression algorithms is used to predict the minimum errors and the coefficient of determination, which confirm the analysis quality. The linear regression ML model suggests the importance of photovoltaic parameters (Rs > Vmpp > Jsc > Voc > fill factor > Jmpp > Rsh) toward higher PCE. An efficient online prediction model is also developed to support the estimation of PCEs with high accuracy.