Efficiency versus effort: A better way to compare best photovoltaic research cell efficiencies?

Frequently, trends in record AM1.5 power-conversion efficiencies versus time, such as the NREL efficiency chart, are used to analyze the relative merits of different photovoltaic material technologies. However, this approach belies the effort expended in achieving these levels of performance. We introduce cumulative publi-cations as a proxy for total R&D efforts and find surprisingly that silicon, Cu(In,Ga)Se2 (CIGSe), CdTe, and halide perovskite technologies have each followed essentially the same learning curve of 20-24% efficiency within 10,000 publications and a consistent marginal rate of 5% efficiency increase per factor of 10 in publications. While learning spillover from non-PV technologies, cross-pollination from other PV technologies, and hidden commercial effort are not accounted for by this metric, this analysis still yields useful and novel insights into PV technology trajectories. Trajectories below this learning curve have required more total effort per performance and plateaus of efficiency stagnation at large numbers of publications may indicate (but do not guarantee) the existence of fundamental barriers to commercially relevant performance. Lastly, examples to watch are identified for technologies currently exhibiting higher marginal slopes, including some that appeared dormant by this metric in past years.

Automated summary

Cumulative publications serve as a proxy for total R&D efforts in photovoltaic technologies, showing that silicon, CIGSe, CdTe, and halide perovskite technologies have followed the same learning curve of 20-24% efficiency within 10,000 publications and a consistent 5% efficiency increase per factor of 10 in publications. This analysis provides insights into PV technology trajectories but does not account for learning spillover, cross-pollination, and hidden commercial efforts. Trajectories below the learning curve require more effort per performance, and efficiency plateaus at high publication numbers may indicate fundamental barriers. Technologies with higher marginal slopes are identified as examples to watch.