Advancing Machine-Learning Augmented Free-Energy Density Functionals for Fast and Accurate Quantum Simulations of Warm Dense Plasmas

University Of Rochester, Rochester NY

Investigators

Abstract

Density-functional theory (DFT), one of the most successful methods in many-body physics, has been one of the main tools for understanding the physics and chemistry of nature as well as for improving our daily life. Examples of DFT applications range from guiding experimentalists to discover near-room-temperature superconductors and other functional materials, to controlling chemical reactions for better products, and to designing drugs to cure diseases. The success of DFT relies on the accuracy of approximations describing how electrons in a material interact with each other, the so-called exchange-correlation (XC) free-energy density functional and the non-interacting free-energy functional for orbital-free DFT. In this research project, finite-temperature XC-functionals and non-interacting free-energy functionals, advanced by machine learning, will be developed to significantly improve the predictive capability of DFT for both plasma physics and materials research. The outcome of this research project is expected to make significant impact in a variety of scientific fields and applications such as planetary science, astrophysics, fusion energy and national defense, as well as make a positive impact on the society through delivering tools to speed up discoveries of novel materials. Warm-dense matter, at pressures ranging from millions to hundreds of billions of atmospheres, exists vastly in the universe -- from planetary cores and astrophysical objects such as brown and white dwarfs, to diamond-anvil-cell compression, to shocks and inertial confinement fusion implosions created in a laboratory. Reliably predicting the static, transport and optical properties of matter at such extreme conditions depends on the accuracy of first-principles methods such as DFT. This project establishes a research program to improve the accuracy and speed of DFT for quantum simulations of extreme materials. The objectives of this project include: (1) developing fully thermalized and numerically efficient XC free-energy functionals for ab-initio molecular-dynamics simulations; (2) eliminating the prohibitively expensive bottleneck in the orbital-based Mermin-Kohn-Sham (MKS) scheme at elevated temperature by developing a novel class of orbital-free (OF) non-interacting free-energy functionals; (3) taking these developments, augmented with machine-learning techniques, to enable an efficient OF-DFT implementation for accelerating the electronic structure calculations that preserve the MKS level of accuracy; and (4) applying these advanced tools to answer key questions in high energy density plasma physics and in extreme materials science. This award is jointly supported by the Plasma Physics program in the Division of Physics and the Condensed Matter and Materials Theory program in the Division of Materials Research. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.