Regensburg 2025 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
TT: Fachverband Tiefe Temperaturen
TT 17: Correlated Electrons: Method Development
TT 17.2: Talk
Tuesday, March 18, 2025, 09:45–10:00, H33
Neural-network-supported Configuration Interaction as impurity solver for DMFT — •Alexander Kowalski1, Philipp Hansmann2, Giorgio Sangiovanni1, and Adriana Pálffy1 — 1Institute for Theoretical Physics and Astrophysics, Universität Würzburg, 97074 Würzburg, Germany — 2Department of Physics, Friedrich-Alexander-Universität Erlangen-Nürnberg, 91058 Erlangen, Germany
Solving a strongly correlated lattice model by means of DMFT involves
mapping it to interacting auxiliary Anderson impurity models (AIM)
whose solution consumes the majority of computational resources. For
the solution, algorithms such as QMC, NRG, DMRG or exact
diagonalization can be used, where the latter in particular has the
advantage of being able to compute exact results on the real frequency
axis but is constrained to a small number of bath sites due to the
exponential growth of the Hilbert space. Selected configuration
interaction (CI) based approaches that operate in only a subspace of
the total Hilbert space can greatly alleviate this problem while still
including the most relevant contributions. Recently, a neural network has
been shown to improve basis selection in ground state AIM calculations
[1]. Here we investigate the use of a similar neural-network-supported
CI solver to select the Hilbert space basis for the auxiliary AIM in
DMFT.
[1] P. Bilous, L. Thirion, H. Menke, M. W. Haverkort, A. Pálffy,
P. Hansmann, arXiv:2406.00151
Keywords: dynamical mean-field theory; Anderson impurity model; configuration interaction; neural networks; machine learning