Ranging from bio-diagnostics and organic electronics to perovskite solar cells, photoluminescence  (PL) is an ubiquitous technique.

In the characterization of perovskites, PL lineshape is a powerful tool used to obtain information on bandgap, local disorder and phase distribution amongst others. Recently, however, researchers have found that the measured PL spectrum of a perovskite is not merely a product of the material properties. Rather the optical environment, such as the measurement setup [1] and surface morphology [2], plays a non-negligible role when measuring PL which is unbeknownst to many.

Another part of this optical environment is the bulk material itself. For instance, it was shown that, due to the self-absorption of emitted light, MAPbI₃ single crystals give rise to a perceived double-peak where only one peak was emitted [3]. Extending on this, in the related field of organic semiconductors it was found that a low-quality cavity is formed by the organic film on a quartz substrate [4]. The cavity effects enhance emission of certain wavelengths while decreasing others, thereby distorting the intrinsically emitted spectrum. These effects gave rise to a film-thickness dependent measured spectrum.

In our work, we investigate the impact of self-absorption and cavity effects on the PL lineshape of perovskite thin films by varying film thickness. We find for identical perovskite material wildly different PL lineshapes as the thickness is varied. These different lineshapes would typically, and inaccurately, be ascribed to changes in the perovskite material quality.

To account for the cavity effect and self-absorption, a modelling approach is developed taking film thickness, refractive index, extinction coefficient and carrier diffusion length as input parameters. By application of this model we are able to successfully extract the intrinsic PL spectrum of a given perovskite material. Inversely applying the same model, we were able to correctly predict the spectrum measured for another film thickness. Finally, the applicability of the model is extended to multilayer systems, emphasizing its significance for complex and device-relevant stacks.

Taken together, our results show the need for careful consideration of the optics when characterizing perovskite films using PL since the measured spectrum can be very different eventhough the material is identical. For accurate comparison between publications (and experimental reproducibility), film characteristics and measurement conditions should be reported alongside the PL spectrum.