In a recent Letter [Dabelow and Reimann, Phys. Rev. Lett. 124, 120602 (2020)], a perturbative relaxation theory is applied to the Bose-Hubbard model and compared to data from a quantum simulator experiment [Trotzky et al. Nature Phys 8, 325 (2012)]. In this Comment, I argue that the discrepancy found is due to the quasi-local observable measured in the experiment being affected by a harmonic trapping potential that is unaccounted for in the analytic calculation. I support my claim with quasi-exact numerics computed using a parallel time-evolving block decimation algorithm (TEBD), and show that the relaxation theory gives accurate results in the correct limit when compared to an appropriate local observable. Finally, I note that my numerics disagree with the t-DMRG calculations of Trotzky et al. but give better qualitative agreement with their experimental results.
The perturbation theory based on typicality introduced by Dabelow and Reimann [Phys. Rev. Lett. 124, 120602 (2020)0031-900710.1103/PhysRevLett.124.120602] and further refined by Dabelow et al. [Phys. Rev. Res. 2, 033210 (2020)2643-156410.1103/PhysRevResearch.2.033210; J. Stat. Mech. (2021) 0131061742-546810.1088/1742-5468/abd026] provides a powerful tool since it is intended to be applicable to a wide range of scenarios while relying on only a few parameters. Even though the authors present various examples to demonstrate the effectiveness of the theory, the conditions used in its derivation are often not thoroughly checked. It is argued that this is justified (without analytical reasoning) by the robustness of the theory. In the paper at hand, said perturbation theory is tested on three spin-based models. The following criteria are taken into focus: the fulfillment of the conditions, the accuracy of the predicted dynamics, and the relevance of the results with respect to a mesoscopic case.