Carolin Sutter-Fella1,Simon Arnold1,2,Corneel Casert1,Edward Barnard1,Stephen Whitelam1
Lawrence Berkeley National Laboratory1,Friedrich-Alexander-Universität Erlangen-Nürnberg2
Carolin Sutter-Fella1,Simon Arnold1,2,Corneel Casert1,Edward Barnard1,Stephen Whitelam1
Lawrence Berkeley National Laboratory1,Friedrich-Alexander-Universität Erlangen-Nürnberg2
Organic-inorganic halide perovskites have excellent optoelectronic properties which enables their application for example in photovoltaics, light emitting diodes, and detectors. The field is plagued by the fact that subtle variations in the synthetic protocol can lead to significant reproducibility challenges. In addition, the parameter space to optimize for desirable optoelectronic properties and stability is very large. Therefore, automation of halide perovskite synthesis at high throughput presents an attractive route to identify synthesis parameter-structure-function properties.<br/>In this talk I will describe our joint effort in setting up a robotic platform, SpinBot One, to synthesize and characterize FAPbI<sub>3</sub> films. SpinBot One is coupled to neural networks which are trained using evolutionary reinforcement learning to realize time-dependent processes that optimize the optical properties of the material under study. SpinBot One fabricates halide perovskite thin films via spin coating and thermal annealing followed by optical characterization (UV Vis, photoluminescence) on the platform and external X-ray diffraction measurements. SpinBot One performed experiments to rationalize the influence of spin speed, spin duration, MACl content, and annealing profile on crystal phase formation and phase stability. The first part of this talk will describe the more technical/ engineering aspects of building our closed loop system while the second part discusses scientific findings enabled by the closed loop.