Accelerated Data‐Driven Discovery of Dual‐Functional Ionic Liquid Passivation for FAPbI ₃ Perovskite Solar Cells Using Graph Neural Network

Achieving efficient and stable formamidinium lead iodide (FAPbI 3 ) perovskite solar cells (PSCs) requires integrated control of crystallization kinetics and defect suppression. While ionic liquids (IL) have shown promise as multifunctional additives, their rational design remains challenging. Here, we develop an attention‐focus graph neural network (GNN) framework that combines the molecular features of IL with device‐level characteristics of FAPbI 3 PSCs. Our model identifies N‐methyl‐N‐butylpyrrolidinium bis(trifluoromethylsulfonyl)imide ([MBPY][TFSI]) as an ideal dual‐functional passivator. The [MBPY] + , acting as a Lewis base, passivates undercoordinated Pb 2+ via Pb‐N coordination bonds, whereas the [TFSI] − anion mitigates interfacial defects via hydrogen bonding with FA + . It is found that the [MBPY] + cation not only suppresses non‐radiative recombination but also enhances the moisture resistance of the perovskite layer due to its hydrophobic alkyl chains. With the synergetic effect of [MBPY] + and [TFSI] − additives, the PSCs achieve a power conversion efficiency (PCE) of 25.03% with an open circuit voltage of 1.182 V, and retain 90.5% of their initial PCE after 1200 h storage at room temperature in air atmosphere (35% relative humidity). This work contributes to ongoing computational and experimental efforts in accelerating the exploration and prediction of potential ionic liquid passivation materials for perovskite solar cells.