Time-dependent numerical renormalization group method for multiple quenches: towards exact results for the long time limit of thermodynamic observables and spectral functions
H. T. M. Nghiem, T. A. Costi

TL;DR
This paper introduces an improved time-dependent NRG method for multiple quenches, significantly reducing errors and accurately capturing long-time behavior of thermodynamic and spectral observables in quantum impurity systems.
Contribution
The authors develop a new TDNRG formalism that minimizes numerical errors and enhances accuracy for long-time limits, extending applicability to nonequilibrium steady states.
Findings
Reduced cumulative error in the trace of the density matrix over time
More accurate long-time limits of thermodynamic observables and spectral functions
Formalism can be extended to infinite switch-on times for improved results
Abstract
We develop an alternative time-dependent numerical renormalization group (TDNRG) formalism for multiple quenches and implement it to study the response of a quantum impurity system to a general pulse. Within this approach, we reduce the contribution of the NRG approximation to numerical errors in the time evolution of observables by a formulation that avoids the use of the generalized overlap matrix elements in our previous multiple-quench TDNRG formalism [Nghiem {\em et al.,} Phys. Rev. B {\bf 89}, 075118 (2014); Phys. Rev. B {\bf 90}, 035129 (2014)]. We demonstrate that the formalism yields a smaller cumulative error in the trace of the projected density matrix as a function of time and a smaller discontinuity of local observables between quenches than in our previous approach. Moreover, by increasing the switch-on time, the time between the first and last quench of the discretized…
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
