The high refractive index of perovskite semiconductors at their emission wavelength (n ~ 2.2-2.6) results in a narrow emission escape cone and a large amount of photoluminescence (PL) to be trapped in perovskite thin films. Due to the strong band edge absorption and very small luminescent stokes shift, this leads to considerable photon reabsorption and reemission – an effect known as photon recycling (PR) [1] – and a concurrent red-shift of propagating PL [2]. As recently highlighted, the self-absorption induced redistribution of charge carriers and outcoupling of propagated PL (e.g. due to scattering) can lead to a significant misinterpretation of measurements of the intrinsic bimolecular recombination and charge carrier diffusion coefficients as well as of asymetric PL spectral shapes [3, 4]. For perovskite films with low non-radiative recombination losses, PR can be of great advantage as it allows higher charge carrier densities at steady-state and, thus, increased device open-circuit voltage [5]. However, most current models still start from the simplified assumption that all initially trapped PL will be reabsorbed at some point without the possibility of being outcoupled beforehand, likely overestimating the effect of PR [6].

In this study, we show that in external photoluminescence quantum efficiency (PLQE) measurements employing a typical integrating sphere setup, the PL spectra of polycrystalline perovskite films with layer thicknesses typically employed for thin film solar cells (~150–400 nm), exhibit a notable red-shift and broadening. Such asymetric spectra are often ignored or misinterpreted in the literature, and to our knowledge there is only one recent study that attempts to model them for thin films under the assumption of self-absorption alone [4]. We developed an all-optical model that considers both PR and the outcoupling of initially trapped propagated PL at the crystal surface and grain boundaries due to scattering, which can reproduce several spectral shapes with varying extents of red-shifts. We confirm our assumptions by various control experiments and benchmark the model for perovskite thin films with different grain size, thickness and composition. Our model allows to estimate the escape probability as well as the initially generated internal PL intensity and thus to calculate the internal PLQE. Our approach serves as a new guideline for estimating these important parameters directly from measurements of the external PLQE and a PL spectrum using an integrating sphere setup.