Researchers at Helmholtz-Zentrum Berlin (HZB), Humboldt-Universität zu Berlin, the Fritz Haber Institute of the Max Planck Society and Technical University Berlin have identified why co-evaporated perovskite films tend to grow unevenly on textured silicon in perovskite-silicon tandem cells, and fixed it with a thin caesium chloride (CsCl) seed layer. This method enabled a power conversion efficiency of 30.3% (29.7% certified), among the highest reported for a fully evaporation-based tandem device.
Co-evaporation is an attractive way to deposit the perovskite top cell industrially, since it's solvent-free, scalable, and can conformally coat the pyramid-textured surface used on silicon bottom cells to improve light absorption. However, getting the perovskite to grow evenly on that texture has been a persistent obstacle. The team used two nanoscale synchrotron characterization techniques at HZB's BESSY II facility - infrared scattering-type scanning near-field optical microscopy (IR-s-SNOM) and X-ray photoemission electron microscopy (XPEEM) - to trace the problem to the hole-transport layer underneath the perovskite. That layer, a carbazole-based self-assembled molecule called MeO-2PACz, is meant to form a single molecular layer. In practice, it pools unevenly: it collects in the valleys of the silicon's pyramid texture and leaves the pyramid facets more thinly covered.
That uneven coverage carried through into the perovskite layer itself. Where the hole-transport layer was thin, the perovskite struggled to incorporate its organic precursor during evaporation and instead left behind residual lead iodide at the buried interface - a defect-prone byproduct that traps charge, blocks extraction, and can photodecompose over time into metallic lead and iodine.
Adding a 20-nanometer caesium chloride seed layer between the hole-transport layer and the perovskite fixed this: it promoted organic precursor incorporation even over the thinly covered regions, suppressing lead iodide formation and producing larger, more uniform perovskite grains across the textured surface, in both the pyramid valleys and facets. Applied to full tandem devices on industrially relevant textured silicon (roughly 2-micron pyramids), the CsCl-treated cells reached higher efficiencies with a narrower device-to-device spread than untreated reference cells, driven mainly by improved fill factor and open-circuit voltage, along with a tighter, better-centered perovskite band gap (around 1.68 eV) versus the wide 1.65–1.75 eV spread seen without the seed layer. The best device reached 30.3% in-house and 29.7% certified - said to be the highest confirmed efficiency for a perovskite-silicon tandem with a fully vacuum-evaporated perovskite absorber.
Unencapsulated cells with the CsCl seed layer showed a T80 operational lifetime of 416 ± 37 hours at 25°C under continuous maximum-power-point tracking. Because the co-evaporation process uses no solvents, the researchers note it may also help address device stability, one of the main obstacles to commercializing tandem cells. "This enables the uniform growth of co-evaporated perovskite layers on textured silicon and prevents the formation of interface defects caused by uneven coverage of the hole transport layer," said Viktor Škorjanc, the study's lead author. "This represents a real step forward in translating record efficiencies from the laboratory into reliable, industrially manufacturable tandem solar technologies," added Marcel Roß, who leads HZB's evaporated perovskite solar cell team.
