DMSE Special Seminar: Dr. Rafael...
Thursday, Mar 23, 2017
12:00 pm - 1:00 pm
Title: Computer-Driven Design of Molecular Materials: Combining First Principles Simulation and Machine Learning Many of the challenges our society now faces are related to finding optimal materials and chemicals for new and old tasks: energy production and storage, sustainable use of resources, climate change, healthcare, agriculture . . . Thanks to continuous advances in hardware and in algorithms, computation offers a way to accelerate and improve how researchers design and deploy new materials. Machine learning tackles larger than ever datasets to find unexpected patterns, to classify and regress, to generate novel ideas or to design optimal strategies in experimental design. First principles simulations have never been faster or more accurate and materials scientists are now equipped to conduct wide predictive virtual searches over large spaces of candidate materials. Machine learning has been combined with quantum chemical simulations with experimental intuition and some healthy pragmatism to discover novel useful molecular materials. This talk will cover progress in virtual screening of organic light-emitting diodes for displays and lighting, organic electrolytes for electrical energy storage, and dyes for novel solar cells. In addition, this talk will examine how deep learning techniques have allowed researchers to encode molecular graphs as continuous vectors that can be optimized numerically and decoded back into optimal molecules.
Dr. Rafael Gomez-Bombarelli