This was part of Machine Learning in Electronic-Structure Theory

Neural Quantum States For Ab Initio Many-Body Electronic Problems

Giuseppe Carleo, EPFL (Ecole Polytechnique Fédérale de Lausanne)

Wednesday, March 27, 2024



Abstract: I will describe applications of Neural Quantum States (NQS) to the study of strongly correlated electronic systems. I will specifically discuss recent work on the Electron Gas and on the Fermi Gas, showing methodological advances based on a message-passing parameterization of the backflow transformation, as well as the use of a pairing wave function based on the Pfaffian.  I will then address the important problem of solving for the dynamics of electrons, solving the time-dependent Schroedinger equation variationally. In the context, I will show new results on the dynamics of quantum dots in two dimensions, with unprecedented accuracy.