Influence of Fundamental Model Uncertainties on Silicon Solar Cell Efficiency Simulations

Abstract

We determine the uncertainties on simulated efficiencies of silicon solar cells due to uncertainties of the fundamental physical models. To determine these, we refit their parameters to the underlying measurement data. Using a metamodeling and Monte Carlo simulation approach, we then deduce how these propagate to the simulated solar cell efficiency. This is evaluated for 150- $\mu \text{m}$ -thick 1 $\Omega $ cm p-type standard and advanced silicon passivated emitter and rear cells (PERC), as well as for an ideal silicon solar cell. When sticking to the best-known set of physical models, we determine resulting efficiency uncertainties as low as 0.06%abs for usual PERC cells and 0.1%abs for an ideal cell. In a variance-based sensitivity analysis, we find the uncertainties of the bandgap and the hole density-of-states effective mass as well as the Auger recombination in the limiting case to dominate the efficiency uncertainty. In addition to these relatively low uncertainties, larger discrepancies may arise when applying different physical models for otherwise fixed device properties. We determine comparably large efficiency discrepancies of up to 0.6%abs for the two most prominent bandgap narrowing (BGN) models, and up to 0.15%abs for the two most recent parameterizations of free carrier absorption. We also show evidence that for the specific case of a strong space charge region being present at a recombining surface, Schenk’s BGN model fails to replicate experiments. Those discrepancies highlight the necessity of further research on those models. Finally, an error of up to 0.08%abs is observed by simplifying diffused regions on textured surfaces to a planar geometry.