Photovoltaics and Solar Energy

Established absorbers like crystalline silicon and GaAs are approaching their theoretical efficiency limits. Next-generation technologies (perovskites, organic-inorganic hybrids, multi-junction tandems) offer higher ceilings but face a bottleneck: the space of possible absorber compositions, contact layers, and heterostructure configurations is enormous, and stability plus manufacturability constraints eliminate most candidates. Perovskite solar cells illustrate the tension well. Thousands of A-site, B-site, and halide combinations exist, but only a fraction combine high efficiency with the operational stability needed for commercial deployment. Sequential optimization of efficiency, then stability, then processability wastes effort on candidates that fail downstream.

Photovoltaic material discovery typically starts from Shockley-Queisser analysis to identify optimal bandgaps, then screens databases or runs combinatorial synthesis for materials near those targets. This treats the absorber in isolation from the full device stack, ignoring interface recombination, contact resistance, and encapsulation compatibility that govern real-world performance. Machine-learning models trained on reported efficiencies overrepresent silicon and a few perovskite compositions while offering little guidance for novel chemistries. Compositional screening of perovskite variants can generate candidates but cannot enforce the stability constraints that disqualify most compositions in practice.