Weijie Xu, PhD candidate, Department of Materials Science and Engineering at University of Texas at Dallas

Perovskite solar cells (PSCs) are widely seen as the likely platform for next-generation solar cells due to their high performance and low cost associated with solution processability. However, the critical bottleneck in commercializing PSCs is the processing speed. Minutes or even hours of thermal processing time for perovskite layers are incompatible with high throughput manufacturing methods, e.g., roll-to-roll processes. Additionally, thermal annealing has problems such as low energy efficiency and inducing mechanical failure. Thus, eliminating the thermal annealing step is the key to achieving high-throughput fabrication of flexible PSCs.

In this talk, I will present my work on how to optimize a pulsed light irradiation approach, photonic curing, to replace thermal annealing in PSC fabrication. First, I will demonstrate how to utilize photonic curing to anneal methylammonium lead iodide (MAPbI3) to reduce the processing time for MAPbI3 by 10 times without compromising efficiency. Next, I will present a machine learning framework, Bayesian Optimization (BO), to help ease the optimization of a novel, high-dimensional process like photonic curing. I will compare the performance benchmark between one-variable-at-a-time and BO. I will also demonstrate using BO to optimize photonically cured MAPbI3 on a flexible Willow Glass device. Finally, I will conclude my vision of machine learning in material science research by outlining different examples of integrating BO into our daily research.

Hosted by: Roman Garnett