Austin Flick1,Mathilde Fievez1,Thomas Colburn1,Qiaohao Liang2,Alexander Siemenn2,Tonio Buonassisi2,Reinhold Dauskardt1
Stanford University1,Massachusetts Institute of Technology2
Austin Flick1,Mathilde Fievez1,Thomas Colburn1,Qiaohao Liang2,Alexander Siemenn2,Tonio Buonassisi2,Reinhold Dauskardt1
Stanford University1,Massachusetts Institute of Technology2
Rapid spray plasma processing (RSPP) has been demonstrated as a successful method to deposit and rapidly cure homogeneous metal halide perovskite thin films over large areas in open-air conditions.<sup>[1]</sup> An ultrasonic spray coater accompanied by an atmospheric pressure nitrogen blown-arc discharge plasma dynamically moves over the substrate to rapidly form a compact perovskite film in under a second. One of the key challenges towards optimizing the processing space of the RSPP system is the high-dimensional processing space and the complex relationships between input variables (coating and curing distances, speed, etc.) and output film morphology and optoelectronic properties. Furthermore, developing a pathway for commercialization of perovskite technologies using the RSPP system requires a multi-objective approach to achieve both high power conversion efficiency and long-term stability.<br/><br/>In previous work, a sequential learning framework using Bayesian optimization was adopted to evaluate the initial performance of RSPP perovskite films in a p-i-n photovoltaic device architecture (glass/ITO/NiO/CsFAPbI<sub>3</sub>/C<sub>60</sub>/BCP/Ag). This optimization study focused on six individual processing parameters: substrate temperature, linear process speed, spray flow rate, plasma gas flow rate, plasma height, and plasma duty cycle.<sup>[2]</sup> Probabilistic knowledge constraints built from prior knowledge with the RSPP system informed a regression-based model to refine the high-dimensional processing space towards an optimal regime; however, this initial optimization lacked the complementary stability characterization for a multi-objective approach.<br/><br/>In this work, we advanced the multi-objective optimization through a complementary evaluation of the long-term stability of the previously explored six-dimensional processing space. Over 400 unencapsulated RSPP perovskite devices spanning a broad range of initial power conversion efficiencies were aged for over 400 days in inert atmospheric conditions and evaluated for their performance retention. Within the previously refined processing space, optimal device stabilities were found to retain >90% initial power conversion efficiency, up to 100% performance retention in devices with initial efficiencies of >15%. Furthermore, continued exploration within the refined processing space, incorporating optimization studies of previously fixed processing parameters, enabled further increases in device efficiency >19%. The long-term stability and elevated performances of the RSPP perovskite devices demonstrate a strong foundation towards developing a pathway for commercialization of RSPP perovskite technologies.<br/><br/>[1] Rolston, N., Scheideler, W.J., Flick, A.C., Chen, J.P., Elmaraghi, H., Sleugh, A., Zhao, O., Woodhouse, M., and Dauskardt, R.H. “Rapid Open-Air Fabrication of Perovskite Solar Modules.” <i>Joule</i>, <b>2020</b>.<br/>[2] Liu, Z., Rolston, N., Flick, A.C., Colburn, T.W., Ren, Z., Dauskardt, R.H., and Buonassisi, T. “Machine Learning with Knowledge Constraints for Process Optimization of Open-Air Perovskite Solar Cell Manufacturing.” <i>Joule</i>, <b>2022</b>.