REVIEW 3 major objections 5 minor 25 references
Scale setting of $\mathrm{SU}(N)$ Yang-Mills theories via Twisted Gradient Flow
T0 review · 3 major / 5 minor · reviewed 2026-08-09 · deepseek-v4-flash
Pith's one-line read The gradient-flow scale $\sqrt{t_0}$ for SU(5) Yang-Mills is positively biased by topological freezing at finite volume, but the bias vanishes in the infinite-volume limit when the PTBC algorithm is used.
desk verdict A solid preliminary SU(5) scale-setting measurement from the TGF+PTBC program; the central vanishing-bias claim is credible but rests on data whose mixing is demonstrated at only one lattice spacing. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The central machinery is PTBC combined with twisted gradient flow. PTBC runs several replicas of the lattice that differ only in the boundary conditions on a small spatial defect, interpolating between periodic and open boundaries; configurations are swapped between neighbouring replicas with a Metropolis step tuned to about 20% acceptance, so the topological charge can change through the nearly-open replica. The scale $t_0$ is defined from the flowed energy density by $N/(N^2-1)\langle t^2 E(t)\rangle = 0.1125$, and the restricted scale $t_0^{(0)}$ uses only $Q=0$ configurations to quantify the bias a frozen algorithm would produce.
What would settle it
At the finest lattice spacing, run the standard local algorithm until it is clearly frozen (integrated autocorrelation time of $Q^2$ above $10^5$ sweeps) on the same volumes used for PTBC, and measure $t_0$ and $t_0^{(0)}$. The claim predicts that the frozen $t_0$ lies above the PTBC $t_0$ by a correction that vanishes as the volume grows; if the two infinite-volume extrapolations separate by more than the combined statistical error, the central claim fails.
Extended reading notes
Core claim
The central claim is that a completely frozen topology gives a positively biased $t_0$ compared with the value obtained when the topological charge is properly sampled. The paper checks this by comparing $t_0$ with the restricted scale $t_0^{(0)}$, defined by projecting onto configurations with $Q=0$: at finite volume $t_0^{(0)} \ge t_0$, and the difference shrinks as the volume grows. For the three lattice spacings simulated in SU(5), the infinite-volume extrapolation of $t_0^{(0)}$ is compatible with $t_0$, indicating that the bias drops out, as expected from theoretical arguments about fixed-topology ensembles.
Load-bearing premise
The PTBC replica chain, with swap acceptance tuned to about 20%, samples the target distribution efficiently enough at every simulated beta and volume that the physical-replica ensembles are unbiased; the paper demonstrates this efficiency at only one SU(5) point.
Editorial extensions
If this is right
- A standard algorithm frozen in the $Q=0$ sector will overestimate $t_0$ at finite volume; the overestimate behaves as a finite-volume correction and does not affect the infinite-volume scale.
- With PTBC, the infinite-volume $t_0$ in the twisted gradient flow scheme is unbiased for SU(5), so it can serve as the reference scale for the step-scaling determination of $\Lambda\sqrt{8t_0}$.
- The same comparison between $t_0$ and $t_0^{(0)}$ provides a practical diagnostic: agreement between the two extrapolations signals that topological freezing is not biasing the scale.
- The 0.2% precision on $\sqrt{8t_0}/a$ at the finest lattice spacing suggests the planned $\Lambda$-parameter determination will not be limited by scale-setting statistics.
Reading between the lines
- Because the bias originates only from the correlation between flowed energy density and topological sector, the same pattern should appear for SU(3) and SU(8); a PTBC-based scale setting with the same replica tuning should also produce unbiased $t_0$.
- The $1/V$ dependence of $t_0^{(0)}$ could be used as a correction scheme: extrapolate $t_0$ measured with an ordinary algorithm from several volumes instead of running PTBC, provided the autocorrelation time is manageable.
- A testable extension is to check whether the infinite-volume agreement between $t_0$ and $t_0^{(0)}$ survives at finer lattice spacings than those in Table 1, where freezing is more severe and PTBC efficiency must be re-established.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. This proceedings article reports a preliminary scale-setting study for SU(5) Yang–Mills theory using twisted boundary conditions and the gradient-flow scale t0. To avoid topological freezing, the authors use the Parallel Tempering on Boundary Conditions (PTBC) algorithm. They define two scales: the standard t0 from Eq. (4) and a restricted scale t0^(0) from Eq. (5) obtained by projecting onto the Q=0 topological sector. They compute both scales at nine (beta,L) points grouped into three lattice spacings, and they argue that t0^(0) is larger than t0 at finite volume and that the difference vanishes in the infinite-volume limit, so a frozen algorithm would produce a positive bias in t0 that disappears at large volumes. The paper is explicitly preliminary and states that a fuller analysis will appear in a forthcoming publication.
Significance. If the central claim holds, the paper provides useful evidence that topological freezing biases the gradient-flow scale t0 upward and that this bias is a finite-volume effect controlled by the Q=0 sector, as expected from fixed-topology arguments. The use of PTBC for SU(5) and the data in Table 1 also constitute a useful methodological data point for the ongoing TGF step-scaling programme toward the SU(N) Lambda-parameter. The manuscript is honest about its preliminary status and cites prior work for the PTBC scheme and algorithm. However, the significance is moderated by two gaps: the algorithmic mixing of PTBC is demonstrated at only one simulation point, and the infinite-volume extrapolation that carries the main compatibility claim is not documented with a fit.
major comments (3)
- [Sec. 4, Fig. 3] The central claim that the Q=0 scale t0^(0) extrapolates compatibly with t0 at beta=19.34158 is not supported by the reported analysis because the finite-volume extrapolation is not documented. The text states that the infinite-volume extrapolation is compatible, and Fig. 3 shows three points on an axis labelled 1/L^4, but no fit form, fit range, fit parameters, chi-squared, or number of degrees of freedom are given. Since this compatibility is the main physical conclusion, the manuscript must report the extrapolation details, including whether the L=40 point is included or excluded and how the final error is obtained. Without these details the compatibility claim is an assertion rather than a result.
- [Sec. 3, Fig. 1 and Table 1] The reliability of all measured t0 and t0^(0) values depends on the PTBC physical replica efficiently sampling all topological sectors, but mixing is demonstrated at only one SU(5) point, beta=17.98526, L=20, with tau_Q2=2.5(3)x10^2 sweeps. No autocorrelation information, topological-charge history, or round-trip diagnostic is reported for the finer ensembles in Table 1, in particular for beta=19.34158, where the compatibility in Fig. 3 is drawn. Tuning the swap acceptance to 20% by increasing N_r controls adjacent-replica overlap but does not by itself guarantee decorrelation of Q^2 on the accumulated statistics. If the finest-beta ensembles were biased toward Q=0, the measured t0 would be inflated toward t0^(0), the residual difference at L=60 would understate the freezing bias, and the apparent 1/V drop-out would be an algorithmic artifact. The authors should provide per-ensemble autocorrelation times for Q^2 or otherwise justify that the mixing established at the coarsest point transfers to finer lattice spacings.
- [Sec. 4, Table 1] The statement that all determinations of t0 at beta=19.34158 are compatible within 0.2% accuracy is ambiguous and not directly supported by the quoted errors. The three central values are 25.349(82), 25.366(68) and 25.302(85), whose relative errors are around 0.3%, so the spread of central values being about 0.2% does not establish a 0.2% accuracy for the finite-volume dependence. Please state explicitly what quantity is claimed to be flat to 0.2% and how that number is computed, for example the maximum relative deviation of the central values or a confidence interval from a common-value fit.
minor comments (5)
- [Sec. 4, Fig. 3] The text says the Q=0 correction scales as the inverse of the volume, while the horizontal axis of Fig. 3 is labelled with 1/L^4; since the physical volume of this lattice is L^4, please state explicitly that 1/V = 1/L^4 is used and make the axis label unambiguous.
- [Table 1] The column header appears as "L mu N_r L_d L t0/a^2 ..." and is hard to parse; the quantity L_mu used to define the renormalization scale should be typeset clearly, for instance as L_\mu, and separated from the column for L.
- [Sec. 2, Eqs. (4) and (5)] The notation with four vertical bars in Eqs. (4) and (5) is typesetting-heavy; using the standard convention of evaluating the equation at t=t0 would improve readability without changing the content.
- [Sec. 1] The sentence "In physical units this corresponds to sqrt(8 t0) ~ 0.5 fm" is stated after the SU(3) definition in Eq. (2) but could be misread as applying to the SU(N) convention in Eq. (4); please clarify that the physical value is the conventional one used for SU(3) t0.
- [References] Reference [5] is cited as a Ph.D. thesis for the values of Lambda/mu_had in SU(N); if a peer-reviewed publication with the same results is available, it should be cited instead or additionally.
Circularity Check
No significant circularity: the t0 and t0^(0) determinations are conventional scale definitions measured from independent flow data, not outputs recycled as inputs.
full rationale
The paper's central claims are that t0 can be biased by topological freezing and that the bias disappears at large volumes. The quantities t0 and t0^(0) are defined by conventional conditions (Eqs. (4) and (5)), with the constants 0.1125 and c = 0.3 taken from the literature; they are not fitted parameters, and neither definition presupposes the values that are later reported. The inequality in Eq. (18) makes t0^(0) >= t0 a mathematical consequence for each ensemble, but this is an analytic relation between the conditioned and unconditioned energy densities, not a circular reuse of the target result. The PTBC algorithm is inherited from the authors' previous work (Ref. [18]), and the paper cites that work for the finding that PTBC reduces topological-charge autocorrelations; however, the SU(5) scale-setting measurements in Table 1 and the infinite-volume comparison in Fig. 3 are new, independent numerical determinations and do not reduce to the cited algorithm paper as their conclusion. The lack of autocorrelation data at the finer beta values is a robustness concern, not circularity, because the potential failure of mixing would make the quoted t0 values unreliable rather than force a predetermined answer. No equation in the paper is equivalent by construction to another, and no fitted parameter is renamed as a prediction. Accordingly, the circularity score is 0.
Assumptions & free parameters
free parameters (4)
- t0 reference value =
0.1125
- topological charge smoothing radius c =
0.3
- twist parameter k =
k=1,2,3 for N=3,5,8
- PTBC replica parameters (N_r, L_d, intermediate c(r)) =
N_r=13,21,32,44; L_d=3,4,5 in Tab. 1; intermediate c(r) not reported
assumptions (4)
- domain assumption Wilson plaquette action with twisted boundary conditions (Eq. (6)-(8)) is a valid lattice regularization of SU(N) Yang-Mills.
- standard math The gradient-flow scale t0 has a well-defined continuum limit and can be used to set the scale.
- domain assumption The lattice topological charge Q_clov at sqrt(8t)=c l, with c=0.3, is close to an integer and the projection |Q|<0.5 selects the zero-topology sector.
- standard math PTBC replica swaps satisfy detailed balance, so the physical replica samples the untwisted, periodic-boundary target distribution.
Cite this review
Pith. "Pith review of Scale setting of $\mathrm{SU}(N)$ Yang-Mills theories via Twisted Gradient Flow." pith.science (2026). https://pith.science/paper/BW55IKSF
@misc{pith2026250118449,
author = {Pith},
title = {Pith review of: Scale setting of $\mathrmSU(N)$ Yang-Mills theories via Twisted Gradient Flow},
year = {2026},
howpublished = {\url{https://pith.science/paper/BW55IKSF}},
note = {Machine review of arXiv:2501.18449}
}
abstract
We present preliminary results for the scale setting of $\mathrm{SU}(N)$ Yang-Mills theories using twisted boundary conditions and the gradient-flow scale $\sqrt{t_0}$. The end goal of this study is to determine the $\mathrm{SU(N)}$ $\Lambda$-parameter through the step-scaling method. The scale $\sqrt{t_0}$, being defined from the flowed action density of the gauge fields, is correlated with their topological charge and thus could be affected by topological freezing. We deal with this problem with the Parallel Tempering on Boundary Conditions algorithm, which we found to be effective for the same numerical setup in a previous work.
Figures
Reference graph
Works this paper leans on
-
[1]
A. Ramos,The gradient flow running coupling with twisted boundary conditions,JHEP11 (2014) 101 [1409.1445]
arXiv 2014
-
[2]
E.I. Bribián and M. García Pérez,The twisted gradient flow coupling at one loop,JHEP03 (2019) 200 [1903.08029]
arXiv 2019
-
[3]
E.I. Bribián, J.L.D. Golán, M. García Pérez and A. Ramos,Memory efficient finite volume schemes with twisted boundary conditions,Eur. Phys. J. C81(2021) 951 [2107.03747]
arXiv 2021
-
[4]
M. Lüscher, P. Weisz and U. Wolff,A Numerical method to compute the running coupling in asymptotically free theories, Nucl. Phys. B359 (1991) 221
work page 1991
-
[5]
Dasilva Golán,Topology and volume dependence in SU(N) Yang-Mills theories, Ph.D
J.L. Dasilva Golán,Topology and volume dependence in SU(N) Yang-Mills theories, Ph.D. thesis, Universidad Autónoma de Madrid, 2023
work page 2023
- [6]
-
[7]
L. Del Debbio, G.M. Manca and E. Vicari,Critical slowing down of topological modes, Phys. Lett. B594 (2004) 315 [hep-lat/0403001]
arXiv 2004
-
[8]
ALPHA collaboration,Critical slowing down and error analysis in lattice QCD simulations, Nucl. Phys. B845(2011) 93 [1009.5228]
arXiv 2011
Show all 25 references
-
[9]
Hasenbusch,Fighting topological freezing in the two-dimensional𝐶𝑃 𝑁−1 model, Phys
M. Hasenbusch,Fighting topological freezing in the two-dimensional𝐶𝑃 𝑁−1 model, Phys. Rev. D96(2017) 054504 [1706.04443]
2017 arXiv
-
[10]
Bonanno, C
C. Bonanno, C. Bonati and M. D’Elia,Large-𝑁 𝑆𝑈(𝑁) Yang-Mills theories with milder topological freezing,JHEP 03(2021) 111 [2012.14000]
2021 arXiv
-
[11]
Bonanno, G
C. Bonanno, G. Clemente, M. D’Elia, L. Maio and L. Parente,Full QCD with milder topological freezing,JHEP 08(2024) 236 [2404.14151]
2024 arXiv
-
[12]
Narayanan and H
R. Narayanan and H. Neuberger,Infinite N phase transitions in continuum Wilson loop operators, JHEP03(2006) 064 [hep-th/0601210]
2006 arXiv
-
[13]
Lohmayer and H
R. Lohmayer and H. Neuberger,Continuous smearing of Wilson Loops, PoSLATTICE2011 (2011) 249 [1110.3522]. 8 SU(N) scale setting via TGF Andrea Giorgieri
2011 arXiv
-
[14]
Lüscher,Trivializing maps, the Wilson flow and the HMC algorithm, Commun
M. Lüscher,Trivializing maps, the Wilson flow and the HMC algorithm, Commun. Math. Phys.293 (2010) 899 [0907.5491]
2010 arXiv
-
[15]
Lüscher,Properties and uses of the Wilson flow in lattice QCD, JHEP08(2010) 071 [1006.4518]
M. Lüscher,Properties and uses of the Wilson flow in lattice QCD, JHEP08(2010) 071 [1006.4518]
2010 arXiv
-
[16]
M. Cè, M. García Vera, L. Giusti and S. Schaefer,The topological susceptibility in the large-𝑁 limit of SU(𝑁) Yang-Mills theory, Phys. Lett. B762 (2016) 232 [1607.05939]
2016 arXiv
-
[17]
Brower, S
R. Brower, S. Chandrasekharan, J.W. Negele and U.J. Wiese,QCD at fixed topology, Phys. Lett. B560 (2003) 64 [hep-lat/0302005]
2003 arXiv
-
[18]
Bonanno, J.L
C. Bonanno, J.L. Dasilva Golán, M. D’Elia, M. García Pérez and A. Giorgieri,The SU(3) twisted gradient flow strong coupling without topological freezing,Eur. Phys. J. C84 (2024) 916 [2403.13607]
2024 arXiv
-
[19]
’t Hooft,A Property of Electric and Magnetic Flux in Nonabelian Gauge Theories, Nucl
G. ’t Hooft,A Property of Electric and Magnetic Flux in Nonabelian Gauge Theories, Nucl. Phys. B153 (1979) 141
1979
-
[20]
González-Arroyo and M
A. González-Arroyo and M. Okawa,The Twisted Eguchi-Kawai Model: A Reduced Model for Large N Lattice Gauge Theory,Phys. Rev. D27(1983) 2397
1983
-
[21]
F.ChamizoandA.González-Arroyo, Tachyonicinstabilitiesin2+1dimensionalYang–Mills theory and its connection to number theory, J. Phys. A50 (2017) 265401 [1610.07972]
2017 arXiv
-
[22]
Creutz,Monte Carlo Study of Quantized𝑆𝑈(2) Gauge Theory,Phys
M. Creutz,Monte Carlo Study of Quantized𝑆𝑈(2) Gauge Theory,Phys. Rev. D21 (1980) 2308
1980
-
[23]
Kennedy and B.J
A.D. Kennedy and B.J. Pendleton,Improved Heat Bath Method for Monte Carlo Calculations in Lattice Gauge Theories, Phys. Lett. B156(1985) 393
1985
-
[24]
Creutz,Overrelaxation and Monte Carlo Simulation, Phys
M. Creutz,Overrelaxation and Monte Carlo Simulation, Phys. Rev. D36(1987) 515
1987
-
[25]
Fritzsch and A
P. Fritzsch and A. Ramos,The gradient flow coupling in the Schrödinger functional, JHEP 10 (2013) 008 [1301.4388]. 9
2013 arXiv
Reviewed August 9, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.