(Invited) Bayesian Optimization of Passivating Contacts for Crystalline Silicon Solar Cells

Abstract

We attempted to utilize Bayesian optimization to realize high-performance crystalline silicon solar cells with passivating contacts, and demonstrates successful application to optimize structural parameter and fabrication process of titanium oxide/silicon oxide/crystalline silicon heterostructure.Bayesian optimization works by constructing a posterior distribution of unknown objective functions based on prior observations. By appropriate choice of the strategy and algorithm, Bayesian optimization could permit reaching the maximum (or minimum) of the unknown objective function by a small number of iterations. This approach is useful when the number of parameters is large and the cost of experiments is high. Formation of titanium oxide/silicon oxide/crystalline silicon heterostructure, which is a promising candidate for passivating contacts, satisfies such a condition since we need to optimize pre-deposition treatment parameters, deposition parameters and post-process parameters.In fact, we optimized three kinds of chemical solution to form silicon oxide interlayer, atomic layer deposition cycle to control the titanium oxide thickness, and process parameters for hydrogen plasma treatment. The parameters include the temperature, time, hydrogen pressure, hydrogen flow rate, RF power and electrode distance. The carrier selectivity, S10, was chosen as the objective function to be optimized. As a consequence, we could reach S10 of 13.63 by only 12 iterations by Bayesian optimization and 10 initial experiments with randomly chosen conditions. These results certifies that Bayesian optimization is useful for accelerating optimization of the practical process conditions in multidimensional parameter space. The optimized conditions could contribute to realization of high efficiency silicon-based heterojunction solar cells.