Potential of Close-Space Sublimation for Scalability of Perovskite Solar Cells

Vacuum techniques for perovskite photovoltaics (PV) are promising in their potential for scalability but are rarely studied with techniques that are readily adaptable for industry. Co-sublimation is one such technique, whose potential as a dry, additive technique for highly uniform and large-area depositions makes it attractive as an industrial candidate for perovskites. However, low-deposition rates, and its dependence on high-vacuum and in-situ rate monitoring make integration on large scales difficult to imagine. Furthermore, the organic sources used (e.g. FAI and MAI) tend to be unstable, decomposing over time. This is severe enough that the organic sources are not reused, instead discarded after each evaporation. In this work, we study the use of close-space sublimation (CSS) for making perovskite solar cells, a technique that has already seen wide-spread use in industry, including in PV and benefits from high material-transfer and low working pressures. We show that organic sources of FAI used in these systems can be cycled multiple times (>30 depositions) and estimate they can be continuously sublimated for months before needing replacement. We show the conversion of inorganic perovskite precursor layers in a 2-step process using CSS for the organic, FAI source. We show that a rough vacuum process (10 mbar) can be used for the sequential conversion of evaporated PbI₂/PbCl₂/CsI thin-films. The FA_0.9Cs_0.1PbI₃:Cl perovskites converted in this work have a bandgap at 810 nm (1.53 eV), with large grains (>400 nm). Devices were then thermally stressed at 85 ^oC and exhibited a stable photoconversion efficiency (PCE) for >650 hours. This thermal stressing was found to be necessary for high -performing devices, who experienced an increase in average PCE from 12.2 – 17.5% after 1 week annealing. We report a champion cell of 18.7% PCE. To explain this drastic increase in PCE, we collect the JV characteristics of the devices at varying light intensities (dark, 0.1, 0.5, and 1 sun) and use drift-diffusion simulations to explain the transformation. These drift-diffusion simulations are powered by SIMsalabim, and by fitting multiple JV curves at different light intensities, we are able to ensure a unique fit of the data. In this way, a reduction in trap density of 5 orders of magnitude was observed after 1 week of thermal stressing.