The performance of perovskite solar cells/modules (PSC/Ms) has improved through optimisation of the perovskite and charge transport layers (CTLs). However, little attention has been paid to the substrate. Almost all thin film photovoltaics rely on soda lime glass (SLG) which contains alkali ions. Na⁺ is the most likely to diffuse into the active layers due to its small radius and low charge. This is especially true in PSMs which use the P1-P2-P3 monolithic interconnection because in the P1 lines, the bottom CTL is in contact with SLG. Previous studies on Na doping of PSCs have found beneficial effects but also detrimental ones when the Na concentration is too high. Therefore, studying inadvertent Na diffusion from SLG in PSMs is of interest.

Here, we used spectroscopy and microscopy techniques to study Na diffusion in PSMs with a Cs_0.05(CH₃NH₃)_0.14(CH(NH₂)₂)_0.81PbI_2.7Br_0.3 perovskite. We used XRD to compare the crystallography of perovskite deposited on SLG and quartz and found that the perovskite peaks of SLG samples are shifted to lower angles, indicating wider lattice plane spacings. We examined the perovskite grain morphology with SEM imaging and observed Brownian tree-shaped areas (trees) growing perpendicularly from the edges of P1 lines up to ~250 μm into the active area. While the bulk of the film contains plate-shaped grains on top of the perovskite grains, very few of these plates are found inside the trees. AFM, KPFM, and cathodoluminescence (CL) mapping showed that these plates are excess PbI₂ from the precursor solution. CL also shows that near the P1 lines, the perovskite’s luminescence is redshifted by 19 meV and stronger by 6-7x, indicating less non-radiative recombination. The Brownian tree shape of these PbI₂-less areas suggests that their formation was controlled by diffusion from the P1 lines, with Na being the most likely diffusant. We performed cross-sectional elemental mapping with STEM-EDX and SIMS to confirm that Na diffused from SLG to CTL/perovskite inside P1 lines and then from P1 lines into active areas. We found strong Na-Br correlation, indicating NaBr formation inside the perovskite. A mechanism can thus be proposed: annealing of the CTL/perovskite provided enough energy for Na to diffuse from SLG. In the perovskite layer close to P1 lines, Na bonds with Br, leaving the perovskite precursor Br-poor. To compensate, more PbI₂ reacted to form I-rich, Br-poor perovskite which explains the XRD and CL peak shifts. NaBr then boosts the perovskite’s local emission by passivating defect sites, as previously observed with KBr.