Identification of the dominant recombination process for perovskite solar cells based on machine learning

Over the past decade, perovskite solar cells have become one of the major research interests of the photovoltaic community, and they are now on the brink of catching up with the classical inorganic solar cells, with efficiency now reaching up to 25%. However, significant improvements are still achievable by reducing recombination losses. The aim of this work is to develop a fast and easy-to-use tool to pinpoint the main losses in perovskite solar cells. We use large-scale drift-diffusion simulations to get a better understanding of the light intensity dependence of the open-circuit voltage and how it correlates to the dominant recombination process. We introduce an automated identification tool using machine learning methods to pinpoint the dominant loss using the light intensity-dependent performances as an input. The machine learning was trained using >2 million simulations and gives an accuracy of the prediction up to 82%. Le Corre et al. demonstrate the application of machine learning methods to identify the dominant recombination process in perovskite solar cells with 82% accuracy. The machine learning algorithms are trained and tested using large-scale drift-diffusion simulations, and their applicability on real solar cells is also demonstrated on devices previously reported