22-June-2022: Talk by Basita Das
Parameter estimation in Perovskite solar cells using Bayesian Inference.
In a perovskite solar cell, the root cause for underperformance may originate from a large variety of different phenomena such as bulk or interface recombination, transport limitations in the contact layers or bad energy alignment at interfaces. Traditionally, discriminating between these mechanisms has been done by applying a variety of different characterization methods to the samples and then performing some data analysis that often involves fitting the data with analytical equations or numerical models. Such characterizing techniques can be time consuming and often destructive to the device. Furthermore, for the sake of simplicity and to allow fitting the data with a low number of unknowns, the models used for fitting and analyzing the data are often insufficiently complex to really capture all necessary physical phenomena that are relevant to understand the measurement. Ideally one can try to fit more complex device models to the characterization data, but the sheer number of parameters that usually goes into a device simulator, and the correlation between the parameters make the problem intractable. Also, such rudimentary parameter fitting usually gives just one set of parameters that fits the data reasonably well but does not tell us if that combination of parameters is unique or if there are any correlation between the parameters. Information on correlation between parameters is not only important from a device optimization point of view but can also teach us more about the underlying physics controlling the functionality of the device.
In this presentation, we introduce a fast non−destructive method of parameter estimation using Bayesian inference in combination with data current and photoluminescence-voltage measurements taken on perovskite solar cells. From the inferred parameters we identify which region or which layer in our device stack is limiting the performance of our device. Such information is useful to strategize device optimization for better performance.
Bayesian inference methods have been previously used in the field of solar cells for well-studied solar cell technologies like Si solar cells as well as for material systems like SnS solar cells where the number of unknown parameters is small [1–3]. In cases where the number of unknowns are few, parameter estimation using only temperature and illumination dependent current voltage JVTi curves yielded good results. However, this is the first time we are using it for perovskite solar cells which is a bit more complicated given the number of unknown parameters are much larger than in the other solar cell technologies.
 R. E. Brandt, R. C. Kurchin, V. Steinmann, D. Kitchaev, C. Roat, S. Levcenco, G. Ceder, T. Unold, T. Buonassisi, Joule 2017, 1, 843.
 R. Kurchin, G. Romano, T. Buonassisi, Comput. Phys. Commun. 2019, 239, 161.
 R. C. Kurchin, J. R. Poindexter, V. Vähänissi, H. Savin, ¶ Carlos Del Cañizo, T. Buonassisi, How Much Physics Is in a Current-Voltage Curve? Inferring Defect Properties from Photovoltaic Device Measurements, 2019.