Determining Parameters of Metal-Halide Perovskites Using Photoluminescence with Bayesian Inference

In this work, we demonstrate that time-resolved photoluminescence data of metal halide perovskites can be effectively evaluated by combining Bayesian inference with a Markov-chain Monte-Carlo algorithm and a physical model. This approach enables us to infer a high number of parameters that govern the performance of metal halide perovskite-based devices, alongside the probability distributions of those parameters, as well as correlations among all parameters. Via studying a set of halfstacks, comprising electron- and hole-transport materials contacting perovskite thin films, we determine surface recombination velocities at these interfaces with high precision. From the probability distributions of all inferred parameters, we can simulate intensity-dependent photoluminescence quantum efficiency and compare it to experimental data. Finally, we estimate mobility values for vertical charge-carrier transport, which is perpendicular to the plane of the substrate, for all samples using our approach. Since this mobility estimation is derived from charge-carrier diffusion over the length scale of the film thickness and in the vertical direction, it is highly relevant for transport in photovoltaic and light-emitting devices. Our approach of coupling spectroscopic measurements with advanced computational analysis will help speed up scientific research in the field of optoelectronic materials and devices and exemplifies how carefully constructed computational algorithms can derive valuable plurality of information from simple datasets. We expect that our approach can be expanded to a variety of other analysis techniques and that our method will be applicable to other semiconductors.