Modeling the Limiting Efficiency of Perovskite/Lead-Sulfide Quantum Dot Tandem Solar Cells

Multi-junction (â??tandemâ?) solar cells have consistently achieved the top power-conversion efficiencies in the world for nearly three decades. In combination with solar concentrators and high-performance materials like GaAs, these devices routinely break the thermodynamic efficiency limit derived by Shockley and Queisser for single-junction devices. At the same time, quantum dots (QD) and Perovskites have become highly attractive as low-cost active materials, the former for their wide tunable band gap range and the latter for their high power-conversion efficiencies (PCE).

Here we model the limiting efficiency of a two-junction tandem solar cell with active layers of Perovskite material CH₃NH₃PbI₃ (MAPbI₃) and lead sulfide quantum dots (PbS QD). Starting from the ideal single-cell case (32.47% maximum PCE), we then consider a version of the PVMirror system proposed by Yu et al., with a long-pass dichroic mirror directing short and long-wavelength light onto the wide and narrow-band gap junctions, respectively. An idealized thermodynamic model was used to optimize the band gap of each junction and the splitting wavelength of the dichroic mirror. When the two junctions are wired independently, PCE is maximized with a splitting wavelength of 700 nm and band gaps of 1.77 eV and 0.94 eV for the Perovskite and QD junctions, respectively. This optimal system achieved a theoretical maximum PCE of 46.12%, with the Perovskite and QD junctions converting 27.80% and 18.32% of the AM1.5 spectrum, respectively. The predictive accuracy of the model was tested by comparing the theoretical and experimental performance of a tandem solar cell comprising 1.59 eV MAPbI₃, 1.08 eV PbS QD, and an 800 nm dichroic mirror in both independent and series wiring configurations. In the independent configuration, the tandem cell achieved a maximum experimental PCE of 18.08%, but this number was reduced to 6.45% when wired in series, indicating significant losses due to circuit resistances. By contrast, the ideal PCEs predicted by our model for this particular system in the independent and series configurations were 42.99% and 33.96%, respectively. This discrepancy between predicted and measured PCE calls for further refinements to our model to account for losses due to realistic absorbance profiles and nonradiative recombination. Additionally, efforts should be made to improve electrical contacts between the two junctions in future experiments in order to ascertain the additional loss mechanism present in the series circuit.