Tensor network algorithm to solve polaron impurity problems

TL;DR

This paper introduces a tensor network and path integral-based method to solve the polaron impurity problem, effectively handling electron-electron and electron-phonon interactions without bath discretization errors.

Contribution

The authors develop a novel tensor network approach that overcomes limitations of existing methods, enabling accurate and flexible solutions for complex polaron impurity problems.

Findings

Accurately reproduces analytic solutions and exact diagonalization results.

Effectively handles non-diagonal hybridization functions.

Enables real-time calculations previously unattainable.

Abstract

The polaron problem is a very old problem in condensed matter physics that dates back to the thirties, but still remain largely unsolved today, especially when electron-electron interaction is taken into consideration. The presence of both electron-electron and electron-phonon interactions in the problem invalidates most existing numerical methods, either computationally too expensive or simply intractable. The continuous time quantum Monte Carlo (CTQMC) methods could tackle this problem, but are only effective in the imaginary-time axis. In this work we present a method based on tensor network and the path integral formalism to solve polaron impurity problems. As both the electron and phonon baths can be integrated out via the Feynman-Vernon influence functional in the path integral formalism, our method is free of bath discretization error. It can also flexibly work on the imaginary, Keldysh, and the L-shaped Kadanoff contour. In addition, our method can naturally resolve several long-existing challenges: (i) non-diagonal hybridization function; (ii) measuring multi-time correlations beyond the single particle Green's functions. We demonstrate the effectiveness and accuracy of our method with extensive numerical examples against analytic solutions, exact diagonalization and CTQMC. We also perform full-fledged real-time calculations that have never been done before to our knowledge, which could be a benchmarking baseline for future method developments.