Molecular Noninteracting Kinetic Energy by Machine Learning
In this project the candidate will learn cutting-edge techniques from the fields of convex analysis and machine learning to develop new approaches to the kinetic energy of electronic systems — the key to unlocking the full potential of Density-functional theory.
Density-functional theory (DFT) is the standard approach to quantum-mechanical simulations of molecules, striking a good balance between accuracy and efficiency by calculating the energy from the density r rather than the many-body wave function. Today, this goal is only partially achieved, however, since a noninteracting wave function is introduced to calculate the noninteracting kinetic energy T[r] accurately. To unlock the full potential of DFT, also the noninteracting wave function must be avoided by calculating the kinetic energy for a given density in a different manner.
Taking advantage of the convexity of the T[r] in r, we propose to obtain it by convex conjugation of the noninteracting energy E[v] as a function of the external potential. Since the noninteracting energy is much simpler than its interacting counterpart, it can be readily calculated accurately for any potential v, providing the large datasets needed for machine learning. The DFT kinetic energy T[r] may then be obtained accurately and quickly by convex conjugation of machine-learned energies.
- The candidate should have a MSc in Physics, Chemistry, Materials Sciences, or close subjects; preferably in the fields of condensed matter, statistical or computational physics, as well as in the fields of theoretical, computational, or physical chemistry.
- Candidates with documented experience in scientific programming and/or multiscale modelling will be prioritised.
Call 2: Project start autumn 2022
This project is in call 2, starting autumn 2022.