Answers are grounded in this paper and cite their sources.
Artificial intelligence is transforming perovskite photovoltaics from empirical development toward data-driven and increasingly autonomous research. This review critically examines this transition across multiple scales, connecting materials prediction with device engineering, stability analysis, and experimental decision-making. Machine learning (ML) strategies for bandgap and crystal-structure prediction, synthesizability assessment, and high-throughput materials screening are first evaluated, with particular attention to emerging graph-based models and ML force fields that extend predictive capability toward complex structural and dynamical phenomena. At the device level, recent advances in performance and operational-lifetime prediction, degradation analysis, defect and ion migration modeling, phase stability assessment, and data-driven passivator discovery are systematically discussed. The review further examines the progression from retrospective modeling toward active experimentation through robotic fabrication and characterization, and self-driving laboratories. Particular attention is also given to ML-assisted perovskite tandem photovoltaics, where coupled materials, architectural, spectral, and environmental variables require multidimensional optimization, as well as to AutoML frameworks that reduce the expertise and manual intervention required for model development. Finally, the major challenges limiting the broader application of AI in perovskite photovoltaics are critically discussed, and future research directions toward more reliable, transferable, and experimentally integrated AI frameworks are outlined.
No takes yet. Share an insight, caveat, or question.
Mahiny et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: