Automatic Materials Design Through Atomistic Simulations and Machine Learning

Event Sponsor: 
Data Science and Learning Seminar
Start Date: 
Jul 11 2019 - 1:00pm
Building 240/Room 1406-1407
Argonne National Laboratory
Rafael Gomez-Bombarelli
Speaker(s) Title: 
Massachusetts Institute of Technology
Ian Foster
Machine learning tools combined with theoretical simulations can effectively accelerate the design of novel materials. Data-driven approaches can access the information embedded in years of experiments, perform rapid optimization of high-dimensional experimental conditions and design parameters, increase the accuracy and speed of physics-based simulations, or design new molecules and crystals automatically. By deploying automated atomistic simulations (molecular dynamics, electronic structure) to create bottom-up representations of materials, and by using those as inputs to machine learning models, we can build effective and accurate predictors. Here, we will describe recent results and ongoing work in using machine learning as the connector between multiple scales of simulation and experiment in materials design. These include (i) high-throughput screening of molecular materials such as organic light emitting-diodes or small molecule battery electrolytes using electronic structure simulations; (ii) inverse design tools based on deep generative models for automatic chemical discovery; (iii) automated learning of all-atom and coarse-grained potentials for discovery of soft materials like ion-conducting polymers; (iv) graph-based representations that accurately predict and rationalize polymorphism in nanoporous zeolite materials. 

Rafael (Rafa) Gomez-Bombarelli received his BSc, MSc (2006) and PhD (2011) in Chemistry from Universidad de Salamanca (Spain). After postdoctoral work at Heriot-Watt University (Edinburg, UK 2012-2014) and Harvard’s Department of Chemistry and Chemical Biology (2014-2016), Rafa co-founded Calculario, a materials discovery startup and joined Kyulux North America Inc. an OLED startup focused in thermally-assisted delayed fluorescence materials for display. Since January 2018 Rafa is an Assistant professor at MIT DMSE. There, he leads a 10-people group working at the interface between deep learning and atomistic simulations for materials design, with a strong focus on molecular and nanoporous materials, inverse design and dimensionality reduction, and predicting supramolecular recognition. He was recently awarded the Google Faculty Research award (2019).