Dr. Amir Barati Farimani, 

Stanford University




Room 1043, ERF


In the first part of my talk, I will discuss our deep learning methodology that can directly learn the physics of transport phenomena based only on data. Transport phenomena includes the exchange of energy, mass, momentum, and charge between systems, encompassing fields as diverse as continuum mechanics and thermodynamics, and is heavily used throughout all engineering disciplines. Here, we show that modern deep learning models, such as generative adversarial networks, can be used for rapid simulation of transport phenomena without knowledge of the underlying constitutive equations, developing generative inference based models for steady state and time-dependent heat conduction and incompressible fluid flow problems with arbitrary geometric domains and boundary conditions. In contrast to conventional procedure, the deep learning models learn to generate realistic solutions in a data-driven approach and achieve state-of-the-art computational performance while retaining high accuracy. Deep learning models for physical inference can be applied to any phenomena in cyber physical systems, given observed or simulated data, and can be used to learn and predict directly from experiments where the underlying physical model is complicated or unknown. In the second part of my talk, I will discuss the deep learning technology we developed for materials discovery. We applied molecular deep learning cheminformatics technology for virtual screening of thousands of candidates for organic photovoltaic (OPV) material and processing conditions and built a predictive model for OPV performance. We first collected and combined experimental datasets from literature sources. From these datasets, we extracted chemical species and processing conditions and used them to seed exploration of the parameter space. We then identified the most important chemical and experimental condition descriptors to lead fundamental research toward the engineering of molecular-scale features that contribute to higher power conversion efficiencies (PCE) in OPVs. By examining the feature space of the model, we then determined which inputs lead to the largest changes in performance metrics. This feature space might include the chemical identities of the active molecules (i.e., donor and acceptor), processing method (e.g., spin-coating, inkjet printing, roll-to-roll printing), and processing conditions (e.g., solvent identity, annealing temperature and time, additive content). We also generated thousands of hypothetical acceptors and donors and screened them using our model. Our model was able to closely predict the performance of the newly synthesized organic molecules.

Dr. Barati Farimani is currently a post-doctoral fellow in Chemistry at Stanford. He received his PhD from the University of Illinois at Urbana-Champaign in 2015 in Mechanical Engineering. During his PhD, he used molecular and atomistic simulations to characterize thin nanoporous membranes for bio-sensing applications and water desalination. He received the Stanley I. Weiss best thesis award from the University of Illinois in 2016 and was recognized as an Outstanding Graduate Student in 2015. During his post-doctoral fellowship at Stanford, Dr. Barati Farimani has developed data-driven, deep learning techniques for inferring, modeling, and simulating the physics of transport phenomena and for materials discovery for energy harvesting applications.

Host: Dr. Kenneth Brezinsky
For more information, please contact Prof. Kenneth Brezinsky at kenbrez@uic.edu