Optical thin film structures: two practical applications for photonics and optoelectronics, and AI-assisted design
Guest Speaker: Prof. L. Jay Guo, Emmett Leith Collegiate Professor of Electrical and Computer Engineering, IEEE Fellow, Optica Fellow，University of Michigan，USA
Date & Time: 9:30-11:00, July. 30, 2025
Location: Yiucheng Lecture Hall （500），Xu Zuyao Building
Inviter:Assco.Prof. Kehang Cui
主讲嘉宾介绍/ Biography
L. Jay Guo is the Emmett Leith Collegiate Professor of Electrical and Computer Engineering at the University of Michigan. He currently serves as the director of the Macromolecular Science and Engineering program, and holds courtesy appointment in Mechanical Engineering, and Applied Physics. He is a fellow of IEEE and a fellow of Optica. Professor Guo’s lab is involved in interdisciplinary research, with activities ranging from polymer-based photonic devices and sensor applications, flexible transparent conductors, nanophotonics, structural colors and AI assisted design, hybrid photovoltaics and photodetectors, to nanomanufacturing technologies. Professor Guo has close to 300 journal publications; with citation more than 35,000 times, and an H-index of 93 (by google scholar). His professional service includes Editor-in-Chief of Optics and Photonics Research, co-Editor-in-Chief of Micro and Nano Manufacturing, and member of the Editorial Advisory Board of Advanced Optical Materials, and Opto-electric Science. His entrepreneur activities include co-founding two startup companies to commercialize technologies from his lab.
Light interacting with metallic and dielectric structures can produce various interesting optical effects. Optical multilayer thin film structure is one of the most important photonic structures widely used in many applications, including color filters, optical cavities or resonators, radiative cooling, and reflectors for extreme UV lithography. But the design of multilayer structures to satisfy the desired optical response is not an easy task, requiring experience/knowledge and intuition. We report a neural network model, OptoGPT, to tackle the challenge of inverse design. We introduce “structure tokens” to denote the material and thickness information and serialize a multilayer structure into a token sequence. In this way, for arbitrary optical response, the OptoGPT model can output any type of multilayer structure, with different number of layers, different materials, and thicknesses. This model provides a unified inverse design for different applications, and also works when designing different incident angles and polarization states. The design process can finish within 0.1 seconds, therefore significantly simplify the inverse design process.
On the experiment side, I will introduce two types of thin film structures that we explored and have found practical applications. The first is to seek ITO replacement with an ultrathin and stable silver film as the flexible transparent electrode, which was used in OPV, and in OLED, with which the waveguide mode can be completely eliminated, leading to enhanced light extraction efficiency. Such transparent conductors have been commercialized by mature industrial sputtering process and applicable to a wide range of applications, e.g. photovoltaics, writable touch screens, electrochromic windows, transparent EM shielding and transparent antenna. It also provides design flexibility than traditional ITO due to its multilayer structure and the choices of dielectric layers. I will also describe recent progress in multi-layer thin film based structural colors, as a more environmentally sustainable solution than the traditional dye-based pigments. Such structures can be mass-produced for commercial applications by physical vapor depositions. We applied Reinforcement Learning algorithm to treat the multilayer structures as a sequence generation process to design the structures satisfying the targe optical spectrum. As an example, a coating that can mimic the chrome-appearance is designed and fabricated and replace the traditional toxic chrome-plating process. Designs with RF transparency can also be realized, which is not possible with traditional chrome-finishes.