Existence of Boutroux curves, $g$-functions and spectral networks from Newton's polygon

Bertrand Eynard; Soufiane Oukassi

REVIEW 3 major objections 4 minor 6 references

Existence of Boutroux curves, $g$-functions and spectral networks from Newton's polygon

T0 review · 3 major / 4 minor · reviewed 2026-08-12 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read An affine space of plane curves with fixed asymptotic behavior always contains a Boutroux curve, whose periods have vanishing real part.

desk verdict The variational strategy and the F=ˇF identity are genuinely nice, but the compactness proof rests on a false zero-count lemma, so the main existence theorem is unproved as written. read the letter →

arxiv 2411.11608 v1 pith:FC6O7BHN submitted 2024-11-18 math-ph math.MP

Bertrand Eynard , Soufiane Oukassi This is my paper · ORCID

classification math-phmath.MP MSC 14H5030F3081Q2060B20

keywords BoutrouxcurvesNewtonpolygong-functionspectralnetworksRiemann-HilbertproblemrandommatrixtheoryregularizedareaStrebelgraphs

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper's central claim is an existence theorem: fix a bivariate polynomial $P$ whose Newton polygon has at least three non-collinear lattice points, a nonempty shifted interior $\overset{\circ}{N}$, and real residues $t_{\alpha,0}$ at its punctures. In the affine space $M = P + \mathbb{C}[\overset{\circ}{N}]$ of curves obtained by varying only the interior coefficients, there is at least one Boutroux curve, meaning $\operatorname{Re}\oint_\gamma Y\,dX = 0$ for every Jordan loop $\gamma$. The proof minimizes a real energy, interpreted as a regularized area of the curve, and shows the minimizer satisfies the Boutroux condition. If correct, this gives a general existence result for the $g$-functions required by Riemann–Hilbert steepest descent, for spectral-network foliations in WKB analysis, and for the equilibrium measures of one- and two-matrix models.

What carries the argument

The central objects are the Newton polygon with its shifted interior $\overset{\circ}{N}$, the affine space $M=P+\mathbb{C}[\overset{\circ}{N}]$, and the regularized-area energy $F$. The Newton data encodes the punctures and their asymptotic “times” $t_{\alpha,k}$; moving inside the affine space leaves all those asymptotics fixed while changing the free interior coefficients. The energy is the area $\frac{1}{2\pi i}\int |Y\,dX|^2$ with divergent puncture terms subtracted, and the proof's main identity is $F=\check F=-\operatorname{Re}\hat F + \pi\,\zeta^t E^{-1}\epsilon$, where $\hat F$ is built from the prepotential $F_0$ and $(\epsilon,\zeta)$ are the real and imaginary parts of the period vector. The positive-definite Hessian of $\check F$ in period coordinates makes the energy strictly convex there, so its minimum satisfies $\zeta=0$, which is exactly the Boutroux condition. The same harmonic function $\phi=\operatorname{Re}\int Y\,dX$ then defines the spectral network as the coincidence set $\{X(p)=X(p'),\ \phi(p)=\phi(p')\}$.

What would settle it

Choose a concrete Newton polygon with at least three non-collinear lattice points and real residues, for example a hyperelliptic $P(x,y)=y^2-R(x)$ whose interior space contains a term of high degree; minimize the regularized area $F$ numerically over $M$ and evaluate $\operatorname{Re}\oint_\gamma Y\,dX$ on a homology basis at the minimizer. A single parameter value for which the real parts are not all zero would disprove Theorem 4.1; if instead the energy level sets appear unbounded while the zero-count bound fails, the gap would be located in the compactness proof of Theorem 3.3.

Watch

Extended reading notes

Core claim

On its own terms, the core discovery is Theorem 4.1: for every $P$ satisfying the Newton-polygon and real-residue hypotheses, the space $M=P+\mathbb{C}[\overset{\circ}{N}]$ contains at least one Boutroux curve, and such curves are isolated in $M$. The minimizing configuration of the energy $F$ is the witness. When $F$ is written in period coordinates it coincides with $\check F = -\operatorname{Re}\hat F + \pi\,\zeta^t E^{-1}\epsilon$, whose Hessian is positive definite because the imaginary part of the Riemann period matrix is positive definite; a minimum therefore sits at $\zeta_i=0$, i.e. at zero real parts of all A- and B-periods, while small loops around punctures have zero real part by the real-residue hypothesis. For a Boutroux curve the differential $Y\,dX$ also defines a single-valued harmonic function $\phi(p)=\operatorname{Re}\int_o^p Y\,dX$, whose level sets organize the curve into the first-kind spectral network; cutting along $\phi=0$ gives the second-kind network used for WKB and random matrices.

Load-bearing premise

The load-bearing premise is that every interior-coefficient polynomial $Q$, pulled back to the curve as $Q(x,Y(x))$, has at most as many zeros as the Newton polygon has lattice points; this zero-count bound is what lets the proof show that energy level sets are compact and that a minimizing curve exists. For high-degree choices of $Q$ the bound can fail, so the compactness argument as written does not cover every case allowed by the theorem.

Editorial extensions

If this is right

For a polynomial one-matrix potential, the Boutroux curve gives the equilibrium measure through $d\mu = \frac{1}{2\pi i}(Y_{\rm left}-Y_{\rm right})\,dx$ on the spectral network, and the energy takes the two-dimensional Coulomb-gas form $\int \operatorname{Re}V\,d\mu - \iint \ln|x-x'|\,d\mu(x)d\mu(x')$.
For the two-matrix model, the construction yields two equilibrium measures $\mu$ and $\tilde\mu$ on two spectral networks, and the $x\leftrightarrow y$ symmetry of the Boutroux condition gives the Matytsin property $X\circ Y=\mathrm{id}$.
The first-kind spectral network gives a canonical decomposition of a Boutroux curve into half-planes, strips, and half-cylinders, with no cylinder faces; this is the atlas used in foliation and moduli-space arguments.
In the hyperelliptic example with prescribed simple poles, the first- and second-kind networks coincide and the graph is the Strebel graph, with faces of prescribed perimeter $2\pi L_\alpha$; the existence theorem thus contains a general construction of such foliations.
Every Boutroux curve is isolated in $M$: near the minimizer the energy is locally strictly convex, so the vanishing-period condition cuts the moduli space transversely rather than along a flat family.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

A natural next test is numerical: for small Newton polygons with large interior coefficients, minimize the regularized energy and compare the minimizer's periods with the prediction $\zeta=0$; this could turn Theorem 4.1 into an effective algorithm for computing $g$-functions.
If the compactness argument can be repaired in the cases where $Q(x,Y(x))$ has many zeros, the same variational scheme should transfer to the proposed generalizations: base curves other than $\mathbb{CP}^1$, logarithmic versions on $\mathbb{C}^*\times\mathbb{C}^*$, and Hitchin spectral curves for arbitrary gauge groups, since only the local puncture regularization changes.
The identification $F=-\operatorname{Re}F_0$ at a Boutroux curve makes the large-$N$ free energy a computable function of periods and times; comparing it with explicit matrix-integral expansions for low degrees would be a sharp test of the whole construction.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

If you're tracking the Boutroux-curve existence problem, this is the Eynard–Oukassi preprint. The variational idea is attractive and the paper is worth knowing about, but the main proof has a hole in it.

What's new: they set up an affine moduli space M = P + C[°N] for plane curves with fixed Newton polygon, define a regularized area F, and show that F coincides with the real part of the prepotential expression ˇF. The computation F=ˇF is clean, and the strict convexity of ˇF on period-coordinate strata is a nice structural result. If a Boutroux curve exists, that part gives uniqueness, and the spectral-network/random-matrix applications in Sections 5–7 are natural.

The problem is Theorem 3.3. Compactness of level sets rests on Lemma 3.2, which asserts that Q(X,Y) has at most #N zeros on the compactified curve. That is false. Take P_N = y^2 + x^N + x^{2N} with N even. The Newton polygon is the triangle (0,2),(N,0),(2N,0), which has 3N/2+3 lattice points total. The exponent (N-2,0) shifts to (N-1,1), which is strictly interior, so Q = x^{N-2} lies in C[°N]. On the normalization, X has two simple zeros over x=0, so Q(X,Y)=X^{N-2} has 2N-4 zeros. For N=50 that is 96 zeros versus 78 lattice points; if #N means the support size 3, the gap is worse. So the pigeonhole step — that among 2#N+1 sample discs, Q can vanish in at most half — has no basis. Without it, the coefficient bound (3.17)–(3.19) does not follow, and Theorem 3.4's minimum is not established. Theorem 4.1 is derived from that minimum, so the central claim is unsupported as written. A second, smaller gap is that the minimizer could sit on a discriminant locus or stratum boundary where the dF=0 condition is not justified.

I do not think this is a crank paper: the F=ˇF identity and the convexity are real, and the intended theorem may well be true. But the current proof does not get there, and the flaw is load-bearing, not cosmetic.

Recommendation: send to a serious referee rather than desk-reject. The problem is important and the gap is identifiable — a repair of the compactness argument would make this a significant paper. I would not cite it as a proof yet.

Referee Report

3 major / 4 minor

Summary. The paper claims to prove a general existence theorem for Boutroux curves: for any fixed bivariate polynomial P whose Newton polygon has nonempty interior, the affine space M = P + C[°N] contains a curve P(x,y)=0 whose periods satisfy Re∮_γ YdX = 0 for all closed loops γ. The strategy is to minimize a regularized area functional F on M, show that its level sets are compact, and then identify the minimum with the Boutroux condition via an identity F = ˇF expressed in period coordinates. The paper also derives structural consequences for spectral networks, Strebel graphs, and random-matrix g-functions. The identity F = ˇF and the local convexity computation are presented in detail, but the central existence proof rests on a compactness argument that I find invalid.

Significance. If the main theorem were correct, it would be a substantial and useful existence result: it would provide a systematic construction of g-functions for steepest-descent arguments, of spectral network foliations, and of Boutroux curves relevant to matrix models and WKB analysis. The paper has genuine strengths: the energy is defined intrinsically from the curve and its periods, the derivation of F = ˇF in Theorem 3.7 is a clear computation, the Hessian computation in Theorem 3.6 is explicit, and the applications to Strebel graphs and matrix models are well motivated. I found no circularity in the use of the prepotential: the equality F = ˇF is invoked as a lemma rather than as an assumption of the existence theorem. However, the main existence theorem is not established by the proof as written, because the compactness of the energy level sets relies on a false zero-count claim.

major comments (3)

[§3, Lemma 3.2] The assertion that 'Q(x,Y(x)) can have at most #N zeros on Σ' is false. Take N even and P_N = y^2 + x^N + x^{2N}. The Newton polygon is the triangle with vertices (0,2), (N,0), (2N,0); the shifted point (N−1,1) is strictly interior, so Q = x^{N−2} belongs to C[°N]. On the normalization of P_N, X has two simple poles over x=∞ and two simple zeros over x=0, so X^{N−2} has 2N−4 zeros on Σ. This exceeds #N = 3, the number of nonzero coefficients of P_N, and for N large it also exceeds the number #\bar N = 3N/2+3 of lattice points in the completed Newton polygon. The pigeonhole step that at most #N of the 2#N+1 sample discs contain a zero therefore has no valid basis, and the boundedness of the evaluations B_l in (3.17)–(3.18), and hence the bound on the coefficients Q_{i,j}, is not established.
[§3, Theorem 3.3, eqs. (3.11)–(3.14)] The proof of the bound on ||Q|| uses a constant K in (3.12) obtained from the fixed curve P and from a domain U that excludes ramification points of P, but the roots Y_i(x) and the denominator P'_y in (3.13)–(3.14) are those of the perturbed curve P+Q. Ramification points and roots vary with Q, and no argument shows that the same U and K control the perturbed roots uniformly over a level set of F. This is a second obstruction to the claimed compactness of level sets.
[§3–§4, Theorems 3.4 and 4.1] Since the existence of a minimum in Theorem 3.4 is proved only through the compactness of level sets, and Theorem 4.1 is derived entirely from that minimum, the main claim of the paper is unsupported as written. The counterexample in the first comment is not an isolated pathology: it lies in the intended domain of the theorem, with C[°N] nontrivial for large N. A substantially different compactness argument would be needed to repair the proof.

minor comments (4)

[Notation, §2.1 and §3] The symbol #N is used ambiguously: it denotes the number of nonzero coefficients of P in (2.1), whereas the proof of Lemma 3.2 needs a bound on the number of zeros in terms of the Newton polygon; the distinction between #N and #\bar N should be made explicit and used consistently.
[Definition 2.2 and Example 2.2] The dimension of C[°N] can be strictly smaller than #°N because of the constraints imposed at zeros of P_d; for instance Example 2.2 has #°N = 3 but dim C[°N] = 0. The proof of Theorem 3.3 should state whether #°N denotes the number of lattice points or the actual dimension of the coefficient space.
[Throughout] There are several typographical and reference errors: 'Harrer-Zagier' should be 'Harer-Zagier', 'Eulcidian' in Theorem 5.5 should be 'Euclidean', and 'Susslin' likely should be 'Suslin'.
[§7.3.5, Theorem 7.3] The proof sketch for the extremal measure relies on classical potential theory and on the already-established Boutroux curve; the text should clarify which parts are conditional on Theorem 4.1, since that theorem is not yet available by the proof in this manuscript.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the Boutroux curve is obtained as an unconstrained minimum of an intrinsically defined energy, and the target periods are not inserted as inputs.

full rationale

The paper's central derivation defines the energy as a regularized area of the curve (Definition 3.1), proves it bounded below and with compact level sets, and then shows at a minimum the derivative of the energy vanishes, which by the explicit formula dF = 2π ζ^T E^{-1} dε forces all real A-periods and, with real residues, all periods to vanish. The Boutroux condition Re∮YdX=0 is therefore a consequence of the minimization, not an assumption built into the energy. The prepotential formalism of [EO07] is used as a lemma to identify F with a period-coordinate expression, but the equality F = F_check is proved within the paper (Theorem 3.7), and the cited prepotential relation dF0 = 2πi Σ η~ dη is an external, checkable integrable-systems result, not a restatement of Boutroux existence. Applications to random matrices invoke the existence theorem rather than fitting the answer from the target data. The proof does contain a serious correctness concern (Lemma 3.2's zero-count claim is false, e.g. Q=x^{N-2} on P_N=y^2+x^N+x^{2N} has ~2N zeros, exceeding #N), but that is an invalid intermediate estimate, not a circular reduction: the zero-count lemma does not define the energy or the Boutroux condition in terms of the conclusion. Accordingly, no step of the derivation reduces by construction to its own input.

Assumptions & free parameters 0 free parameters · 5 assumptions · 0 invented entities

The paper relies on standard results in Riemann surface theory and the prepotential formalism. The only domain-specific assumptions are the real-residue hypothesis and the nonempty interior condition, both explicitly stated. There are no free parameters fitted to external data and no invented entities; the interior coefficients are the variables of the variational problem, not assumed inputs.

assumptions (5)

standard math Riemann's theorem: dim Ω1(Σ) = g and existence of a normalized basis of holomorphic differentials (Theorem 2.1).
Used throughout Sections 2 and 3 to define period coordinates and the prepotential.
standard math Seiberg-Witten prepotential differential identity dF0 = 2πi Σ η̃_i dη_i, cited from [EO07].
Invoked in Section 3.5 to derive dF = dˇF and the convexity of ˇF.
standard math Newton polygon boundary-puncture correspondence (Appendix A).
Used to relate exterior coefficients to times and to define the moduli space M.
domain assumption Real residues hypothesis: tα,0 = Res_α YdX ∈ R for all punctures α (Remark 2.5).
Stated as necessary for the Boutroux condition to be possible.
domain assumption Nonempty interior: C[°N] ≠ 0 (Remark 2.2).
The paper excludes the trivial case where the moduli space is a single point.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Existence of Boutroux curves, $g$-functions and spectral networks from Newton's polygon." pith.science (2026). https://pith.science/paper/FC6O7BHN

@misc{pith2026241111608,
  author       = {Pith},
  title        = {Pith review of: Existence of Boutroux curves, $g$-functions and spectral networks from Newton's polygon},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FC6O7BHN}},
  note         = {Machine review of arXiv:2411.11608}
}

abstract

We prove the existence of an algebraic plane curve of equation $P(x,y)=0$, with prescribed asymptotic behaviors at punctures, and with the Boutroux property, namely, periods have vanishing real part, i.e, $\Re(\int_\gamma y dx)=0$ for every closed loop $\gamma$. This has applications in the Riemann-Hilbert problem, in random matrix theory, in spectral networks, in WKB analysis and Stokes phenomenon, in algebraic and enumerative geometry, and many applications in mathematical physics. From Newton's polygon we can define an affine space such that there exists always a Boutroux curve. This result is applied to random matrix and asymptotic theory, in which a key ingredient is called the $g$-function, the function $g(x)=\int_o^x Y dX$ is a $g$-function precisely if and only if the algebraic plane curve is a Boutroux curve.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

6 extracted references · 3 canonical work pages

[1]

Boutroux curves with external field: equilibrium measures without a minimization problem

[AG97] G. B. Arous and A. Guionnet. “Large deviations for Wigner’s law and Voiculescu’s non-commutative entropy”. Probability theory and related fields 108 (1997), pp. 517–542. ISSN : 1432-2064. DOI: 10.1007/s004400050119. [Ber07] M. Bertola. “Boutroux curves with external field: equilibrium measures without a minimization problem” (2007). arXiv: 0705.306...

work page Pith review arXiv 1997
[820]

[Mum07] D

DOI: 10.1016/0550-3213(94)90471-5 . [Mum07] D. Mumford. Tata Lectures on Theta. Vol. I (no. 28), II (no. 43), III (no. 97). Modern Birkh¨auser Classics. Birkh¨auser Boston, MA,

doi:10.1016/0550-3213(94)90471-5
[1973]

Wall-crossing, Hitchin systems, and the WKB approximation

ISBN : 9780387065175. URL: https://books.google.fr/books?id=-i3vAAAAMAAJ. [GMN13] D. Gaiotto, G. W. Moore, and A. Neitzke. “Wall-crossing, Hitchin systems, and the WKB approximation”. Adv. Math. 234 (2013), pp. 239–403. DOI: 10.1016/j.aim.2012.09.027. arXiv: 0907.3987 [hep-th]. [GZ02] A. Guionnet and O. Zeitouni. “Large Deviations Asymptotics for Spherica...

arXiv 2013
[2007]

Cell decomposition and compactification of Riemann's moduli space in decorated Teichm\"uller theory

[Pen03a] R. C. Penner. “Cell decomposition and compactification of Riemann’s moduli space in decorated Teichm ¨uller theory” (2003). arXiv: math/0306190 [math.GT]. [Pen03b] R. C. Penner. “Decorated Teichm ¨uller Theory of Bordered Surfaces” (2003). arXiv: math/0210326 [math.GT]. [Str84] K. Strebel. Quadratic Differentials. Berlin, Heidelberg: Springer Ber...

work page Pith review arXiv 2003
[2017]

The Geometry of integrable systems. Tau functions and homology of Spectral curves. Perturbative definition

arXiv: 1706.04938 [math-ph]. [Eyn18] B. Eynard. Lectures notes on compact Riemann surfaces

work page Pith review arXiv
[2018]

Invariants of algebraic curves and topological expansion

arXiv: 1805.06405 [math-ph]. [EO07] B. Eynard and N. Orantin. “Invariants of algebraic curves and topological expansion”. Commun. Num. Theor. Phys. 1 (2007), pp. 347–452. DOI: 10.4310/CNTP.2007.v1.n2.a4. arXiv: math-ph/0702045 [math-ph]. [FK12] H. Farkas and I. Kra. Riemann Surfaces. Graduate Texts in Mathematics. Springer New York, 2012.ISBN : 9781468499...

arXiv 2007

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.

[1] [1]

Boutroux curves with external field: equilibrium measures without a minimization problem

[AG97] G. B. Arous and A. Guionnet. “Large deviations for Wigner’s law and Voiculescu’s non-commutative entropy”. Probability theory and related fields 108 (1997), pp. 517–542. ISSN : 1432-2064. DOI: 10.1007/s004400050119. [Ber07] M. Bertola. “Boutroux curves with external field: equilibrium measures without a minimization problem” (2007). arXiv: 0705.306...

work page Pith review arXiv 1997

[2] [820]

[Mum07] D

DOI: 10.1016/0550-3213(94)90471-5 . [Mum07] D. Mumford. Tata Lectures on Theta. Vol. I (no. 28), II (no. 43), III (no. 97). Modern Birkh¨auser Classics. Birkh¨auser Boston, MA,

doi:10.1016/0550-3213(94)90471-5

[3] [1973]

Wall-crossing, Hitchin systems, and the WKB approximation

ISBN : 9780387065175. URL: https://books.google.fr/books?id=-i3vAAAAMAAJ. [GMN13] D. Gaiotto, G. W. Moore, and A. Neitzke. “Wall-crossing, Hitchin systems, and the WKB approximation”. Adv. Math. 234 (2013), pp. 239–403. DOI: 10.1016/j.aim.2012.09.027. arXiv: 0907.3987 [hep-th]. [GZ02] A. Guionnet and O. Zeitouni. “Large Deviations Asymptotics for Spherica...

arXiv 2013

[4] [2007]

Cell decomposition and compactification of Riemann's moduli space in decorated Teichm\"uller theory

[Pen03a] R. C. Penner. “Cell decomposition and compactification of Riemann’s moduli space in decorated Teichm ¨uller theory” (2003). arXiv: math/0306190 [math.GT]. [Pen03b] R. C. Penner. “Decorated Teichm ¨uller Theory of Bordered Surfaces” (2003). arXiv: math/0210326 [math.GT]. [Str84] K. Strebel. Quadratic Differentials. Berlin, Heidelberg: Springer Ber...

work page Pith review arXiv 2003

[5] [2017]

The Geometry of integrable systems. Tau functions and homology of Spectral curves. Perturbative definition

arXiv: 1706.04938 [math-ph]. [Eyn18] B. Eynard. Lectures notes on compact Riemann surfaces

work page Pith review arXiv

[6] [2018]

Invariants of algebraic curves and topological expansion

arXiv: 1805.06405 [math-ph]. [EO07] B. Eynard and N. Orantin. “Invariants of algebraic curves and topological expansion”. Commun. Num. Theor. Phys. 1 (2007), pp. 347–452. DOI: 10.4310/CNTP.2007.v1.n2.a4. arXiv: math-ph/0702045 [math-ph]. [FK12] H. Farkas and I. Kra. Riemann Surfaces. Graduate Texts in Mathematics. Springer New York, 2012.ISBN : 9781468499...

arXiv 2007