This paper presents PerovskiteVision, a machine vision tool for quantifying the homogeneity of perovskite thin films over large areas. The tool uses adapted machine vision algorithms to spatially quantify substrate coverage, film thickness, and defect density from images. Combined with an optical model, it predicts photovoltaic cell and module current density. This allows for a posteriori identification of optimal process conditions for large-area perovskite deposition via gas-knife assisted slot-die coating, maximizing both performance and throughput. The study demonstrates how machine vision accelerates characterization for multi-objective optimization of thin-film deposition processes.

Nina Taherimakhsousi, Mathilde Fievez, Benjamin P. MacLeod, Edward P. Booker, Emmanuelle Fayard, Muriel Matheron, Matthieu Manceau, Stéphane Cros, Solenn Berson, Curtis P. Berlinguette