Monolithic perovskite/silicon tandem solar cells increasingly rely on self-assembled monolayer (SAM) hole-selective contacts, often combined with a NiOx interlayer, to achieve low interfacial losses. On pyramid-textured silicon bottom cells, however, uniform coverage is difficult to achieve: in our devices, NiOx nanoparticles accumulate in the pyramid valleys while tips and facets remain thinly covered, leaving regions where the SAM (Me-4PACz) does not fully cover the NiOx. These SAM-uncovered sites act as recombination centres and shunt paths at the buried interface, and their density is expected to be higher on textured than on flat substrates.

Here, we use area-selective atomic layer deposition (AS-ALD) of Al2O3 to passivate these sites without disturbing the SAM-covered surface. Thermal ALD (trimethylaluminium/H2O, 100 °C) is applied directly onto NiOx/Me-4PACz. Because the SAM-terminated surface lacks hydroxyl groups for precursor chemisorption, Al2O3 nucleation is delayed on SAM-covered areas and proceeds preferentially on exposed NiOx, so that at low cycle numbers the passivation layer is confined to the defective regions. This concept was recently demonstrated on flat single-junction devices [1], while reactions between trimethylaluminium and phosphonic acid anchors have also been reported [2]; we therefore examine how the selectivity window depends on cycle number.

In planar wide-bandgap single-junction cells, 1–5 ALD cycles increased the average PCE from 19.4% to 20.7%, driven mainly by a fill-factor gain (76.0% to 79.8%) and higher shunt resistance, with only a small increase in open-circuit voltage (~10 mV). Beyond 7 cycles, the series resistance rose and the fill factor dropped to 70.2% at 9 cycles, consistent with Al2O3 growth extending onto the SAM-covered surface. Contact-angle measurements on NiOx/SAM likewise show a non-monotonic dependence on cycle number. We then transfer the process to textured silicon heterojunction bottom cells, starting from a reference monolithic tandem of 29.5% (1 cm²), and compare the optimum cycle window on textured versus planar substrates. Surface-potential mapping and cross-sectional TEM are used to correlate the location of Al2O3 with the pyramid morphology. This work positions AS-ALD as a self-aligned, defect-selective passivation route for textured tandem architectures.