CCQ Seminar: Kieron Burke

Date & Time

Title: Making density functionals with machine learning
Kieron Burke
Departments of Chemistry and of Physics, UC Irvine
The quantum mechanics of electrons determines the glittering diversity of chemistry and materials. Electronic structure calculations have become crucial for many aspects of modern science, from drug discovery, to finding new materials, to understanding planets.   They are crucial because they yield quantitative predictions of energies needed to make and break chemical bonds, and of reaction rates.   Density functional theory provides the approximations that are used in most simulations.  About 50,000+ scientific papers each year report results of these DFT calculations.
Our present approximations yield useful accuracy, but all fields using DFT would be revolutionized by significant improvements (factor of 3) in accuracy, while retaining reliability and efficiency.  I will describe some of our (and others) fledgling attempts to apply the power of modern machine learning techniques to transform this field.
Advancing Research in Basic Science and MathematicsSubscribe to Flatiron Institute announcements and other foundation updates