DOI: https://doi.org/10.36347/sjet.2026.v14i07.002

Pages: 374-385

The development of third-generation photovoltaics perovskite solar cells (PSCs), organic solar cells (OSCs), and quantum dot solar cells (QDSCs) is constrained by the impracticality of exhaustively exploring multidimensional chemical and processing parameter spaces using conventional trial-and-error experimentation. The discovery of stable, non-toxic, and high-efficiency nanomaterials requires navigation of combinatorial spaces spanning elemental compositions, molecular architectures, synthesis routes, and processing conditions that far exceed experimental throughput capabilities. Artificial intelligence (AI) and machine learning (ML) offer a paradigm shift by enabling data-driven prediction, optimization, and generation of materials prior to laboratory synthesis. Through training on curated experimental datasets, high-throughput density functional theory (DFT) calculations, and multi-scale simulations, ML models can approximate complex structure–property relationships, identify optimal synthesis windows, and guide experimental efforts with quantifiable uncertainty. This review provides a systematic evaluation of ML methodologies applied to photovoltaic nanomaterial design, covering data acquisition and representation, supervised learning, deep learning, generative models, active learning, and self-driving laboratories. Key challenges including data scarcity, interpretability, domain transfer, and reproducibility are critically assessed.