REVIEW 4 major objections 4 minor 9 references
On analytic continuation from imaginary to real chemical potential in Lattice QCD
T0 review · 4 major / 4 minor · reviewed 2026-08-09 · deepseek-v4-flash
Pith's one-line read A new analytic-continuation method for lattice QCD works by inverting the Cauchy formula, reproducing Padé and Taylor results up to mu_B/T ~ 2.
desk verdict Candid status report of a promising inverse-Cauchy continuation method; idea is real, evidence is preliminary. 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 object is the inverse Cauchy problem. The standard Cauchy integral formula $f(z_0) = \frac{1}{2\pi i}\oint_C \frac{f(z)}{z-z_0}dz$, written for a circle of radius $R$ centered at the origin, is discretized with Gauss-Legendre quadrature into the linear system $A x = b$: the matrix elements are $A_{ik} = \frac{1}{2\pi} w_k \frac{R e^{i\theta_k}}{R e^{i\theta_k} - z_i}$, the unknowns $x_k$ are the boundary values $f(R e^{i\theta_k})$ at the quadrature nodes, and the data $b_i$ are the known interior values $y_i$ on the imaginary axis. Solving the system determines the boundary values; then the same quadrature formula, applied directly, continues the function to any point inside the circle, in particular the real axis. The machinery relies on treating the discretized formula as exact (the paper writes 'we trust the formula as exact!') and on the analyticity of $f$ inside the circle.
What would settle it
Use the same inverse-problem procedure on a known function that has a simple pole inside the circle, e.g. $f(z)=1/(z-a)$ with $|a|<R$, feeding exact values on the imaginary axis: if the real-axis continuation still achieves $\mathcal{O}(10^{-8})$ accuracy, the analyticity requirement is not the operative constraint; if it fails, the method breaks when a singularity crosses the contour. A lattice-QCD version would be to fix a temperature and choose a circle radius $R$ that encloses the Lee-Yang singularity known from the Padé analysis, then check whether the inverse-problem continuation diverges from the Taylor-series results.
Extended reading notes
Core claim
The central claim is that a numerically performed analytic continuation from imaginary to real chemical potential can be obtained by solving the inverse Cauchy problem: given values of a function at $n$ points on the imaginary axis, the function's values at the $n$ Gauss-Legendre quadrature nodes on a circle of radius $R$ are the solution of the linear system $A x = b$ constructed from the discretized Cauchy formula (Eqs. 9--10). Once these boundary values are known, the same quadrature formula is used in the forward direction to compute the function at real values of the chemical potential. The procedure is demonstrated on $\sin(z)$: with $n=40$--$50$ nodes and $R=1$, real-axis values are reproduced to $\mathcal{O}(10^{-8})$, and the residue-related Laurent coefficient $C_{-1}$ comes out at $\mathcal{O}(10^{-15})$. Applied to lattice QCD data for the net baryon number density at $T \sim 155$ MeV, the method gives real-axis results that agree with multi-point Padé and eighth-order Taylor series up to $\mu_B/T \sim 1.5$--$2.0$, beyond which the results become strongly dependent on the input. The paper also reports that the recovered boundary values $\hat{f}_k$ looked like nonsense, which it interprets as evidence of an effective quadrature at work.
Load-bearing premise
The method assumes the unknown function (the net baryon number density) is analytic on and inside the chosen circle of radius $R$ centered at zero, so that the discretized Cauchy formula can be trusted as exact and the resulting linear system can be inverted despite ill-conditioning.
Editorial extensions
If this is right
- The inverse Cauchy method offers a parameterization-free route from imaginary to real chemical potential, standing alongside multi-point Padé and Taylor series as a third way to perform the analytic continuation.
- In the tested range up to $\mu_B/T \sim 1.5$--$2.0$, the method agrees with the Padé and Taylor results, indicating that all three approaches are limited by the same nearby singularity (the Lee-Yang edge).
- Beyond that threshold, the results become strongly dependent on which input data are used, and the Taylor series errors grow large, marking the practical limit of any analytic continuation in this setup.
- The same inverse-problem machinery extends naturally to Laplace (anti)transforms, which would give a new tool for reconstructing spectral functions from lattice QCD.
Reading between the lines
- The paper's observation that the recovered boundary values $\hat{f}_k$ look like nonsense while the real-axis results are accurate suggests the method is not performing a genuine inversion of the Cauchy formula; a direct test would be to plug the recovered $\hat{f}_k$ back into the forward quadrature and check whether the original imaginary-axis inputs are reproduced, which would separate a real i
- Because the paper states the quadrature also handles functions with pole singularities via the residue theorem, the analyticity requirement may be replaceable by meromorphy as long as the contour avoids the poles; this would make the method applicable when Lee-Yang singularities lie on the real axis outside the chosen circle.
- From a numerical-analysis perspective, the linear system $A x = b$ is a Fredholm integral equation of the first kind, and its ill-conditioning could be mitigated by standard regularization; a regularized version might stabilize the boundary values and push the reliable continuation range beyond the current $\mu_B/T \sim 2$ threshold.
- The precision reached on $\sin(z)$ ($\mathcal{O}(10^{-8})$ with 40--50 nodes) offers a clean benchmark against which other analytic-continuation methods in lattice QCD could be compared, independent of the physical uncertainties of the gauge data.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper, a Lattice 2024 proceedings contribution from the Parma group, reports on two approaches to analytically continuing lattice QCD results from imaginary to real baryon chemical potential: multi-point Padé approximants, previously used by the Bielefeld-Parma Collaboration, and a new method based on solving the Cauchy integral formula as an inverse problem. In the inverse-Cauchy method, values of the net baryon number density on the imaginary axis are used as the right-hand side of a discretized Cauchy formula, the linear system in Eqs. (9)-(10) is inverted to obtain boundary values on a circle of radius R, and the same quadrature is then used to evaluate the function on the real axis. The paper demonstrates the method on sin(z), reports application to lattice QCD number density, and compares the results with Padé and Taylor continuations. The authors state that the reconstructed boundary values initially 'looked like non-sense', attribute the eventual success to an 'effective quadrature', and announce that substantial progress will appear in a future publication.
Significance. If validated, the inverse-Cauchy method would be a conceptually interesting addition to the toolkit for analytic continuation in finite-density lattice QCD. The paper is commendably honest about the fact that the boundary values produced by the inversion are not meaningful and about the dependence of the results on the input data. The forward test on sin(z) and on Laurent coefficients is a useful sanity check, and the comparison with Padé and Taylor methods addresses the right question. However, the central claim that the inverse problem provides a reliable numerical analytic continuation is currently supported only by a benign test on an entire function and by a visually plausible lattice result without error bars. The necessary stability analysis is absent, and the self-reported 'non-sense' boundary values indicate an uncontrolled step in the method. The significance of the paper therefore depends entirely on whether the missing numerical analysis can be supplied; in its present form the evidence is not sufficient to establish the method.
major comments (4)
- [Section 3, Eqs. (8)-(10)] The transition from the approximate quadrature formula in Eq. (8) to the exact linear system in Eq. (9), where the authors write 'notice that now we trust the formula as exact!', is the load-bearing step of the entire method. The quadrature error of Eq. (8) is discarded without any estimate, and because the matrix A in Eq. (10) is inverted, there is no guarantee that this error, or the Monte Carlo noise in the vector b, remains under control. The paper reports that the reconstructed boundary values f_k 'looked like non-sense', which is exactly what one expects from an unstable inversion. The subsequent 'effective quadrature' explanation in Section 3 is a qualitative assertion, not a derived property. To support the central claim, the authors should provide a condition-number study of A as a function of n and R, a convergence test in n, and a noise-injection test in which synthetic data with controlled statistical errors are propagated through the inversion.
- [Section 3, Fig. 4 (left panel)] The sin(z) test is not a demanding test for the inverse Cauchy problem. Since sin(z) is entire, the Gauss-Legendre quadrature error in Eq. (8) is far below machine precision, and the test cannot expose the ill-conditioning that occurs when the function has singularities near or inside the contour. The situation relevant to QCD, where Lee-Yang singularities restrict the radius of analyticity, is not probed by this example. A more informative test would use a meromorphic function with a known singularity, for example f(z)=1/(z-z_s) or a function with a branch cut, with z_s placed at varying distances from the contour, and would report the error of the continued values as a function of that distance. Without such a test, the good result for sin(z) cannot be taken as evidence that the method works for the lattice number density.
- [Section 3, Fig. 4 (right panel); Section 2] The lattice QCD results are presented without error bars, and the two inverse-Cauchy curves (blue circles and black diamonds) differ from each other depending on the input data used. The agreement with the Padé continuation is therefore not quantitatively assessable. The authors should specify exactly which input data were used for each curve (number of derivatives, imaginary-chemical-potential interval, number of quadrature nodes n, and radius R), propagate the Monte Carlo errors through the linear system in Eq. (10), and show the sensitivity of the real-axis result to n and R. The claim that the inverse-Cauchy method works up to mu_B/T ~ 1.5-2.0 is currently supported only by a visual coincidence with the Padé curve.
- [Section 3, analyticity assumption] The method assumes that f(z) is analytic on and inside the circle of radius R centered at the origin. For the net baryon number density, this requires R to be smaller than the distance to the nearest Lee-Yang singularity. The paper does not check whether the chosen R satisfies this condition, nor does it discuss what happens if a singularity lies inside the contour. Since the Cauchy representation and the inverse problem are only valid in a singularity-free disk, the authors should either verify the singularity-free radius using the Lee-Yang results from Ref. [7] or test the method on synthetic data with a known singularity inside the contour. This is not a minor point: if the contour encloses a singularity, the system in Eq. (10) is not a discretization of a well-posed problem.
minor comments (4)
- [Whole manuscript] The manuscript contains repeated 'Working Notes on the Inverse Cauchy problem' pages with handwritten annotations and an equation numbering that differs from the main text (Eqs. (1)-(12) in the notes versus Eqs. (6)-(11) in the body). This material must be removed or integrated into the text before publication.
- [Figures] Figure 1 is used for both the complex-mu plane sketch and the Cauchy contour cartoon, and Figure 3 similarly mixes a sketch of the complex plane with a pictorial explanation of the inverse problem. The figures should be renumbered and referenced consistently.
- [Abstract and Section 1] The collaboration name is written as 'Bielefeld Parma Collaboration' in the abstract and as 'Bielefeld-Parma Collaboration' in the body; please use one consistent form.
- [Working Notes, Section 1] The inserted note contains the typo 'olomorphic' for 'holomorphic'; this is part of the handwritten material that should be cleaned up.
Circularity Check
No circularity: the inverse Cauchy continuation inverts a well-defined integral representation and is validated against data that are independent Monte Carlo measurements, not fitted targets.
full rationale
The paper's derivation chain is self-contained and non-circular. The inverse Cauchy method starts from the Cauchy integral formula (Eq. 6), specializes to a circular contour (Eq. 7), approximates the integral by Gauss-Legendre quadrature (Eq. 8), and then promotes the quadrature sum to an exact linear system (Eq. 9) whose solution gives boundary values f-hat_k, which are subsequently used in the forward quadrature formula to evaluate the function on the real axis. The real-axis results are not fitted to real-axis data: the linear system A x = b is solved using only the imaginary-axis input values y_i, and the same inputs are not also used as targets of the continuation. The lattice QCD data come from the collaboration's earlier Monte Carlo simulations [6,7], which are independent numerical measurements, and the comparison with multi-point Pade and Taylor series [8] is a consistency check, not an input to the inversion. The self-citations are for data generation and context, not for a load-bearing uniqueness claim. The admitted ill-conditioning, the 'non-sense' boundary values, and the absence of error bars are genuine numerical-validation concerns, but they are not circularity: they concern whether the method is stable and reliable, not whether the claimed prediction is equivalent by construction to its inputs. No equation or fitted parameter is defined in terms of the target real-axis quantity, and no cited prior result is used to forbid alternatives. Therefore the circularity score is 0.
Assumptions & free parameters
free parameters (4)
- Cauchy contour radius R
- Number of Gauss-Legendre nodes n =
n=40,50 for sin tests; unspecified for lattice data
- Input data selection (derivatives and imaginary-mu interval)
- Padé degrees (m,n)
assumptions (4)
- domain assumption The number density is analytic on and inside the chosen circle in the complex chemical potential plane.
- domain assumption The Gauss-Legendre quadrature discretization of the Cauchy integral is trusted as exact.
- ad hoc to paper The linear system A x = b is solvable and its solution gives meaningful boundary values.
- standard math The Cauchy integral formula and its derivative version (Eqs. 6-7 and 11) apply as standard theorems of complex analysis.
Cite this review
Pith. "Pith review of On analytic continuation from imaginary to real chemical potential in Lattice QCD." pith.science (2026). https://pith.science/paper/VRKINRN7
@misc{pith2026250203392,
author = {Pith},
title = {Pith review of: On analytic continuation from imaginary to real chemical potential in Lattice QCD},
year = {2026},
howpublished = {\url{https://pith.science/paper/VRKINRN7}},
note = {Machine review of arXiv:2502.03392}
}
abstract
Imaginary baryon number chemical potential simulations are a popular workaround for the (in)famous sign problem plaguing finite density QCD studies on the lattice. One is necessarily left with the problem of analytically continuing results to real values of $\mu_B$. In the framework of the Bielefeld Parma Collaboration, we have in recent years studied a multi-point Pad\'e description of the net baryon number density computed as a function of imaginary baryon number chemical potential. While our main emphasis has till now been on the determination of Lee-Yang singularities, the method is per se a natural tool to analytically continue results. We report on the status of our projects with this respect, comparing the Pad\'e approach to analytic continuation to another, new strategy, which is an application of the Cauchy integral formula in the sense of an inverse problem.
Figures
Reference graph
Works this paper leans on
- [7]
-
[1]
P. de Forcrand, Simulating QCD at finite density , https://doi.org/10.22323/1.091.0010 PoS LAT2009 (2009) 010 [ https://arxiv.org/abs/1005.0539 1005.0539 ]
arXiv 2009
-
[2]
C.R. Allton, S. Ejiri, S.J. Hands, O. Kaczmarek, F. Karsch, E. Laermann et al., The QCD thermal phase transition in the presence of a small chemical potential , https://doi.org/10.1103/PhysRevD.66.074507 Phys. Rev. D 66 (2002) 074507 [ https://arxiv.org/abs/hep-lat/0204010 hep-lat/0204010 ]
arXiv 2002
-
[3]
R.V. Gavai and S. Gupta, Pressure and nonlinear susceptibilities in QCD at finite chemical potentials , https://doi.org/10.1103/PhysRevD.68.034506 Phys. Rev. D 68 (2003) 034506 [ https://arxiv.org/abs/hep-lat/0303013 hep-lat/0303013 ]
arXiv 2003
-
[4]
P. de Forcrand and O. Philipsen, The QCD phase diagram for small densities from imaginary chemical potential , https://doi.org/10.1016/S0550-3213(02)00626-0 Nucl. Phys. B 642 (2002) 290 [ https://arxiv.org/abs/hep-lat/0205016 hep-lat/0205016 ]
arXiv 2002
-
[5]
M. D'Elia and M.-P. Lombardo, Finite density QCD via imaginary chemical potential , https://doi.org/10.1103/PhysRevD.67.014505 Phys. Rev. D 67 (2003) 014505 [ https://arxiv.org/abs/hep-lat/0209146 hep-lat/0209146 ]
arXiv 2003
-
[6]
P. Dimopoulos, L. Dini, F. Di Renzo, J. Goswami, G. Nicotra, C. Schmidt et al., Contribution to understanding the phase structure of strong interaction matter: Lee-Yang edge singularities from lattice QCD , https://doi.org/10.1103/PhysRevD.105.034513 Phys. Rev. D 105 (2022) 034513 [ https://arxiv.org/abs/2110.15933 2110.15933 ]
arXiv 2022
-
[8]
HotQCD collaboration, Taylor expansions and Pad\'e approximants for cumulants of conserved charge fluctuations at nonvanishing chemical potentials , https://doi.org/10.1103/PhysRevD.105.074511 Phys. Rev. D 105 (2022) 074511 [ https://arxiv.org/abs/2202.09184 2202.09184 ]
arXiv 2022
Show all 9 references
-
[9]
write newline
" write newline "" before.all 'output.state := FUNCTION blank.sep after.quote 'output.state := FUNCTION fin.entry output.state after.quoted.block = 'skip 'add.period if write newline FUNCTION new.block output.state before.all = 'skip output.state after.quote = after.quoted.blo...
Reviewed August 9, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.