BRITE Pivot: Machine Learning Accelerated Optimization of Flash Lamp Processed Thin-films for Flexible Optoelectronic Applications

University Of Texas At Dallas, Richardson TX

Investigators

Abstract

This Boosting Research Ideas for Transformative and Equitable Advances in Engineering (BRITE) Pivot award supports research to develop a new approach to the manufacturing of flexible optoelectronics -- solar cells, light-emitting diodes, and sensors for health and wellness monitoring. The innovative technique uses a flash lamp to deliver short pulses of energy to sinter ultrafine particles, initiate chemical reactions, and form new materials. Replacing traditional heat sources such as ovens with light can greatly reduce time and energy, accelerating manufacturing output and decreasing costs. But finding the processing conditions that optimize materials properties has mostly been done by trial and error, an expensive, inefficient method. By contrast, cutting-edge machine learning approaches guide this project's search for the best processing conditions in fabricating thin films for flexible devices. Today, Pacific Rim nations dominate flexible optoelectronic manufacturing; this project makes the U.S. more competitive. The research involves many technical disciplines, including materials processing and characterization, device fabrication and testing, data analytics and machine learning, and advanced manufacturing. The project provides undergraduates with hands-on research experience. It also develops a diverse Science, Technology, Engineering, and Mathematics workforce by including at all levels women and other under-represented groups. Making high-quality materials often requires a high-temperature annealing process, which takes many iterations to optimize. This project investigates the use of light from a flash lamp instead of furnace heating for thin film processing and adopts machine learning approaches to accelerate process optimization. Photonic curing uses millisecond pulses of intense broadband light to sinter particles, initiate chemical reactions, and transform materials. The energy from the light pulses is preferentially absorbed by the thin film, leading to selective heating, while the underlying substrate remains below its working temperature. Hence, this approach enables processing on plastic substrates which cannot withstand high temperatures, and, therefore, is particularly useful in flexible optoelectronics fabrication. Because photonic curing involves many processing parameters that are intimately coupled with starting material properties, achieving desired quality outcomes is a challenging optimization problem. Traditional varying-one-variable-at-a-time methods are inefficient in exploring the entire parameter space and are thus time-consuming and expensive. The research team’s approach is to collect initial experimental results based on the judicious sampling of input space and apply advanced data analytic techniques that balance between exploring untested phase space and fine-tuning conditions to achieve global optimization. The relationships between input parameters and output improvements for photonic curing are revealed, and physics-based models for thin film processing optimization are developed from the machine learning results. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.