Data-driven discovery of green solvents for sustainable, high-performance perovskite solar cells

Eco-friendly solvents are essential for the mass production of solution-processed perovskite solar cells (PSCs). However, identifying greener alternatives to toxic solvents such as N,N-dimethylformamide (DMF) remains challenging. Here, we present a data-driven workflow integrating machine-learning (ML) predictions, density-functional-theory (DFT) calculations, and experimental validation. This approach enables rapid screening of a vast chemical space (~169 million PubChem molecules) using ML-predicted physicochemical and environmental properties, followed by DFT evaluation of solvent-precursor interactions and device-level testing. This workflow identified dimethyl malonate (DMM) as a CHEM21-Recommended, low-toxicity solvent. A 1:4 mixture of DMM and dimethyl sulfoxide (DMSO) yielded highquality FAPbI₃ films and PSCs with 25.0% PCE (V OC = 1.19 V, J SC = 24.7 mA cm⁻², FF = 0.85), comparable to DMF-based devices. To our knowledge, this is the first demonstration of DMM for perovskite synthesis. By integrating ML-based discovery with DFT, this study establishes a generalizable pathway toward sustainable PSC manufacturing.