Anisotropy tuning with the Wilson flow

Szabolcs Borsányi, Stefan Krieg, Simon Mages, Stephan Dürr, Andreas Schäfer, Kálman Szabó, Thorsten Kurth, Zoltán Fodor, S. D. KatzarXiv 2012 · DOI · arXiv

Abstract

We use the Wilson flow to define the gauge anisotropy at a given physical scale. We demonstrate the use of the anisotropic flow by performing the tuning of the bare gauge anisotropy in the tree-level Symanzik action for several lattice spacings and target anisotropies. We use this method to tune the anisotropy parameters in full QCD, where we also exploit the diminishing effect of a well chosen smearing on the renormalization of the fermion anisotropy.

Introduction

In Lattice Quantum Chromodynamics (QCD), simulations typically utilize a uniform grid where spatial ($a_s$) and temporal ($a_t$) spacings are equal. However, specific physical problems—such as the study of excited particle spectra or high-temperature matter—require significantly higher resolution in the temporal direction. This necessitates the use of anisotropic lattices where the ratio $\xi = a_s / a_t > 1$.

The primary challenge lies in the "renormalization" of the grid by quantum fluctuations. The input "bare" anisotropy ($\xi_0$) does not directly correspond to the physical "renormalized" anisotropy ($\xi$) observed in the simulation. Traditional tuning methods, such as Klassen’s method using Wilson loops, are computationally expensive, sensitive to short-distance noise, and often require ambiguous fitting procedures. This work introduces a precise, automated alternative using the Wilson flow.

Methodology: The Anisotropic Wilson Flow

The Wilson flow (or gradient flow) is a process that systematically smooths gauge fields, effectively removing high-frequency quantum noise. As the flow progresses through a "flow time" $\tau$, it probes the physics at a specific length scale $\sqrt{8\tau}$.

Key Innovations

  1. Anisotropic Flow Equations: The authors adapt the standard flow equations for anisotropic grids by introducing specific smearing coefficients that account for the lattice geometry.
  2. The $R_E$ Ratio: A critical metric is defined based on the ratio of the spatial and temporal components of the action density derivative. When the flow's internal anisotropy ($\xi_w$) matches the lattice's physical anisotropy, this ratio reaches a predictable value (unity in the appropriate normalization).
  3. Self-Consistent Tuning: By executing the flow across a range of test anisotropies, the exact physical anisotropy $\xi_g$ can be determined through interpolation.
The tuning process for gauge anisotropy. The intersection in the bottom panel identifies the physical anisotropy $\xi_g$ where the ratio of action density derivatives is unity.

Results and Discussion

The method was validated across quenched (pure glue) and dynamical (full QCD with $N_f = 2+1$ flavors) ensembles.

Precision and Automation

The Wilson flow approach achieves sub-percent precision in determining the gauge anisotropy $\xi_g$. Unlike traditional methods, the continuous nature of the flow equations allows for an automated tuning process, eliminating the need for manual selection of fit ranges. The results across various lattice spacings are summarized in the table below.

Summary of tuned parameters ($\xi_g$, $w_0$) for the tree-level Symanzik action across different lattice spacings ($\beta$).

Fermion Alignment through Smearing

A significant bottleneck in anisotropic QCD is the separate tuning required for glue and quark (fermion) fields. The authors demonstrate that applying "stout smearing" to the quark links causes the fermion anisotropy to naturally align with the gauge anisotropy. In dynamical tests with four steps of stout smearing, the fermion anisotropy matched the gauge anisotropy within 2% without dedicated fermion tuning.

Conclusion

The use of the Wilson flow provides a robust, high-precision framework for tuning anisotropic lattices. By defining anisotropy at a continuous physical scale ($w_0$), the method offers cleaner results and better scalability than Wilson-loop-based techniques. Future work will explore extending this to a broader range of lattice spacings and diverse quark actions to further generalize these findings.

Summary updated

Abstract

We use the Wilson flow to define the gauge anisotropy at a given physical scale. We demonstrate the use of the anisotropic flow by performing the tuning of the bare gauge anisotropy in the tree-level Symanzik action for several lattice spacings and target anisotropies. We use this method to tune the anisotropy parameters in full QCD, where we also exploit the diminishing effect of a well chosen smearing on the renormalization of the fermion anisotropy.
BibTeX
@misc{borsanyi2012anisotropy,
  author = {Szabolcs Borsányi and Stefan Krieg and Simon Mages and Stephan Dürr and Andreas Schäfer and Kálman Szabó and Thorsten Kurth and Zoltán Fodor and S. D. Katz},
  title = {{Anisotropy tuning with the Wilson flow}},
  year = {2012},
  eprint = {1205.0781},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1205.0781},
}
Figure