Résumé
We present a dynamical mean field theory (DMFT) solver based on neural quantum states (NQS). NQS are an unbiased variational ground state ansatz, that leverage the power of neural networks to capture long range correlations on complicated graph structures. We showcase this ability by performing a state-of-the-art multi-orbital DMFT calculation, which requires solving notoriously complex quantum impurity problems. Our benchmarks on both the single-orbital Hubbard model and the multi-orbital Hubbard–Kanamori Hamiltonian reveal excellent ground state precision and the capacity to resolve key features associated with Hund’s metals. These promising results open avenues for extending DMFT to more challenging problems.
