Boston Chapter of the IEEE Photonics Society

Abstract:  The chemical space of organic optoelectronic materials is extremely vast. This allows exquisite fine-tuning of molecular designs to achieve desired properties, but hinders the systematic exploration of structure and property space. Although physics-based simulations can screen candidates much faster than chemical synthesis and device fabrication, autonomous chemical design is still a challenge. The discrete, graph-like nature of molecules presents a difficult optimization challenge; simulations may not capture all the experimental design and performance parameters and often ignore the vast amounts of pre-existing data.

Recent machine learning advances have allowed progress in many of these issues. Here, I will describe an ML-accelerated design cycle of optically active compounds that combines (i) data extraction from the literature: (ii) unsupervised and semisupervised deep learning to generate discrete molecules from continuous vectors so that numerical optimization methods can be applied to chemical design; (iii) ML-based calibration of theoretical results with respect to experiment; (iv) neural simulators that replace expensive atomistic simulations. The domains of application range from organic light emitting diodes, to photodiodes or optical switches.

Biography:  Rafael Gomez-Bombarelli is the Toyota Assistant Professor in the Department of Materials Science and Engineering (DMSE). Rafa  joined the MIT faculty in January 2018. He received a B.S., M.S., and Ph.D. in Chemistry from Universidad de Salamanca in Spain, followed by postdoctoral work at Heriot-Watt University and Harvard University after which he was a senior researcher at Kyulux NA applying Harvard-licensed technology to create real-life commercial organic light-emitting diode (OLED) products. At MIT, his research focus on the interplay between atomistic simulations and machine learning for materials design.