Designing Quantum Chemistry Methods with and Beyond Chemical Intuition
Quantum chemistry simulations are essential for connecting electron-level behaviors to macroscopic chemical phenomena, such as chemical reactions, spectroscopy, and electromagnetic properties. However, balancing computational cost and accuracy remains challenging due to the inherently complex nature of electronic structure. In this talk, I will discuss how this complexity impedes universal, scalable solutions and propose guidelines for designing efficient quantum chemistry methods using chemical intuition. I will illustrate these guidelines through two computational algorithms I developed. First, I will introduce a new multi-reference approach that connects strong and weak electron correlations via a non-orthogonal configuration interaction (NOCI) formulation, showing how capturing different types of electron correlations enhances accuracy. Next, I will present density matrix embedding theory (DMET) and its finite-temperature extension, demonstrating how understanding the entanglement structure enables a balance between accuracy and computational efficiency. Moving beyond chemical intuition, I will then explore the possibility of a universal quantum chemistry method. Our recent development of a neural network quantum state (NNQS) with a generative model based on normalizing flows illuminates the path toward universal solutions in quantum chemistry. Finally, I will share my perspective on the evolving role of quantum chemistry: from making accurate predictions to enabling active chemical discovery.
Hosted by Professor Richard Remsing