[RBRC seminar] Worldvolume tempered Lefschetz thimble method as an algorithm towards solving the sign problem

by Dr Nobuyuki Matsumoto

Thursday 18 Nov 2021, 09:30 → 11:00 US/Eastern

The application of Monte Carlo methods to systems with complex action such as finite density QCD is hindered by the sign problem. The "worldvolume tempered Lefschetz thimble method" (WV-TLTM) [1] is proposed towards solving the sign problem. In this algorithm, we make the Hybrid Monte Carlo updates on the continuum set of integration surfaces foliated by the antiholomorphic gradient flow ("the worldvolume of the integration surface"). The WV-TLTM tames the sign and multimodal problems simultaneously as the original TLTM [2], reducing the computational cost significantly compared to the original TLTM. We apply this algorithm to the Stephanov model (a chiral random matrix model), for which the complex Langevin method is known not to work. We also discuss the effect of choosing a specific flow time region on the estimation of observables, especially by analyzing the autocorrelation times and the statistical errors [3].

[1] M. Fukuma and N. Matsumoto, "Worldvolume approach to the tempered Lefschetz thimble method," PTEP 2021, no.2, 023B08 (2021) [arXiv:2012.08468 [hep-lat]].
[2] M. Fukuma and N. Umeda, "Parallel tempering algorithm for integration over Lefschetz thimbles," PTEP 2017, no.7, 073B01 (2017) [arXiv:1703.00861 [hep-lat]].
[3] M. Fukuma, N. Matsumoto, and Y. Namekawa, "Statistical analysis method for the worldvolume hybrid Monte Carlo algorithm," to appear in PTEP [arXiv:2107.06858 [hep-lat]].

 

https://bnl.zoomgov.com/j/1601581422?pwd=dkdxVFgyRTZESUxIbUZDN1RmRURPQT09