Pith. sign in

REVIEW 4 minor 13 references

The paper proves that a homogeneous posynomial that is zero-free on a product of open right half-planes has a concave degree-normalized root, extending Gårding's theorem to real exponents.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

T0 review · deepseek-v4-flash

2026-08-02 07:41 UTC pith:NF42YIMH

load-bearing objection Gårding's theorem for posynomials is real: the paper closes the factor-of-two gap between sector stability and fractional log-concavity, and the proof is clean enough to send to referees.

arxiv 2607.09168 v2 pith:NF42YIMH submitted 2026-07-10 cs.DS math.COmath.PR

G{aa}rding's Theorem for Posynomials

classification cs.DS math.COmath.PR MSC 26B2526C1030C15
keywords posynomialsGårding's theoremsector stabilityfractional log-concavityentropic independencemixing timedeterminantal point processesmatchings
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

This paper proves that a homogeneous posynomial—a positive-coefficient sum of monomials with arbitrary real exponents—that has no zeros on a product of open right half-planes must have a concave degree-normalized root. The result extends Gårding's classical theorem from polynomials to posynomials, where exponents are not required to be integers. Because sector stability with aperture απ implies α-fractional log-concavity, the theorem removes a factor-of-two loss in earlier sampling guarantees. A reader should care because this yields concrete improvements in mixing-time and domain-sparsification bounds for sampling fixed-size matchings and nonsymmetric determinantal point processes.

Core claim

The central claim is that zero-freeness in Γ1 (product of open right half-planes) is enough to make x ↦ p(x)^{1/d} concave on the positive orthant for any homogeneous posynomial of degree d. The proof establishes an angular-contraction estimate—that the argument of p along paths whose coordinate arguments lie in an interval of width φ < π is confined to [0, dφ]—and then uses the representation theorem for complete Bernstein functions to pass from that one-variable angular bound to concavity. The corollary sharpens the connection: Γα-stability implies α-fractional log-concavity, with the α dependence optimal.

What carries the argument

The angular-contraction lemma: for a Γ1-stable homogeneous posynomial p of degree d, if each coordinate of w has argument in [0, φ] with φ < π, then p(w) ≠ 0 and the continuous argument along the straight-line path satisfies 0 ≤ Arg(p(w)) ≤ dφ. This reduces the multivariate stability condition to a one-variable ray monotonicity statement for posynomials with real exponents, which is proved by a generalized Descartes rule; the one-variable statement then feeds a concavity criterion from the complete Bernstein function representation.

Load-bearing premise

The proof leans on the complete representation of a one-variable function that satisfies the angular condition 0 ≤ arg h(z) ≤ arg z; the concavity conclusion requires that the representing measure in that representation be positive, a fact the paper imports without proof.

What would settle it

Exhibit a homogeneous posynomial with positive real coefficients that is zero-free on (open right half-plane)^n but whose degree-normalized root p(x)^{1/d} is not concave on the positive orthant. Since the theorem is concise, the simplest check is numerical: sample two-variable posynomials with non-integer exponents, test zero-freeness on a grid of half-plane points, and test concavity by the midpoint inequality.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

Share X Bluesky LinkedIn Reddit HN

If this is right

  • If p is Γα-stable, then x ↦ log p(x^α) is concave, i.e. α-fractional log-concavity, for all 0 < α ≤ 1.
  • The α dependence is best possible: p(x,y)=x^β+y^β is Γα-stable exactly when αβ ≤ 1, and concavity of (x^{αβ}+y^{αβ})^{1/(αβ)} holds in the same range.
  • For sector-stable generating polynomials of k-subset distributions, the k↔k−r down-up walk with r=⌈1/α⌉ mixes in time O_α(k^{1/α} log(1/μ(S0)) log(1/ε)) and has modified log-Sobolev constant Ω_α(k^{-1/α}).
  • Domain sparsification reduces sampling from such distributions to sparse external fields supported on n^{1−α} poly(k) elements.
  • For fixed-size matchings and nonsymmetric DPPs, the new certificate yields two-site walks with Ω(k^{-2}) log-Sobolev bounds and sparse domains of size n^{1/2} poly(k), improving the previous n^{3/4} poly(k).

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The real-exponent formulation suggests that fractional log-concavity is not an artifact of integer exponents; the same sector-stability certificate might be used directly on generating functions with non-polynomial terms, e.g., partition functions with external fields, without first converting to polynomials.
  • Because Corollary 2 is sharp, any sampling algorithm whose mixing guarantee is governed by fractional log-concavity cannot hope for a better exponent from this certificate alone; further gains must come from additional structure, such as negative dependence or spectral gaps.
  • The angular-contraction technique may generalize to ratios of posynomials or to functions defined by integrals over the positive orthant, provided the one-dimensional ray argument can be extended to a suitable class of functions.
  • An immediate testable extension is to non-homogeneous posynomials: if one normalizes by the total degree and checks Γ1-stability after homogenization, a similar concavity statement may hold; the current proof relies on homogeneity for the perspective step.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

0 major / 4 minor

Summary. The paper proves an analogue of Gårding's theorem for homogeneous posynomials with arbitrary nonnegative real exponents. The main result (Theorem 1) states that if such a posynomial is zero-free on a product of open right half-planes (Γ1-stable), then its degree-normalized root is concave on the positive orthant. Corollary 2 extends this to sector stability: Γα-stability implies α-fractional log-concavity. The proof proceeds through a real-exponent generalized Descartes rule (Lemma 3), a Sendov–Sendov ray-monotonicity theorem (Theorem 4), an angular-contraction lemma (Lemma 5), and a Pick-representation-based concavity lemma (Lemma 6). The applications section derives improved mixing-time and domain-sparsification bounds for fixed-size matchings and nonsymmetric determinantal point processes by removing a previously known factor-of-two loss.

Significance. If the result holds, it is a clean and useful extension of Gårding's theorem to posynomials, with a sharp dependence on the sector aperture α and concrete algorithmic consequences. The proof is essentially self-contained and, notably, has no fitted parameters or post hoc assumptions: the main theorem is derived from first principles with the only external input being the standard Pick/complete-Bernstein representation. The sharpness example shows the exponent range is optimal. The paper also transparently discloses the AI-assisted development of the proof. My reading of the central arguments is that they are correct: the generalized Descartes rule, ray monotonicity, angular contraction, and the perspective argument all check out. The one non-elementary step—the use of the complete-Bernstein characterization in Lemma 6—is correct but terse, and should be explained more explicitly for the intended CS readership.

minor comments (4)
  1. [§3, Lemma 6] The proof asserts, without derivation, that the hypotheses 0≤arg h(z)≤arg z imply that h is a complete Bernstein function, and then cites the Pick representation in [SSV12, Theorem 6.2]. This is the single most load-bearing step in the paper, and although the implication is correct (the sector condition is equivalent to h and -h/z being Pick functions, which is the standard characterization of complete Bernstein functions), the manuscript should either prove this equivalence or give a precise citation to the characterization. As written, a reader cannot tell whether the cited theorem is being used for the representation or for the classification.
  2. [§2, Theorem 4] In the definition of H_β, some coefficients a_j sin(λ_jθ−β) may vanish. The sign-change count should explicitly say that zero coefficients are discarded before applying Lemma 3. This is a minor amendment but avoids ambiguity.
  3. [§3, Lemma 5] The proof of Lemma 5 says that the lift obtained from Theorem 4 is the same branch as the one obtained by moving from 1 to e^{iφ}. This is true because f is zero-free on ℂ\ (−∞,0], which is simply connected, and both paths lie in that set. Stating this explicitly would improve clarity, since Theorem 4 is phrased along a ray of fixed angle, while the desired path is along the unit circle.
  4. [§1, Corollary 2] The sharpness example p(x,y)=x^β+y^β is stated without proof. A one-sentence explanation of why Γα-stability holds exactly when αβ≤1 (using the equation (z1/z2)^β=−1) would be useful.

Circularity Check

0 steps flagged

No significant circularity; the main theorem is derived from first principles with an external Bernstein-function representation theorem.

full rationale

The main derivation chain is Theorem 1 <= Lemma 5 <= Theorem 4 <= Lemma 3; all are proved inside the paper or from the classical (and cited) Pick/complete-Bernstein representation [SSV12, Theorem 6.2] in Lemma 6. None of these steps defines a quantity in terms of the claimed conclusion or fits a parameter and calls it a prediction. Theorem 4 is a strengthened version of Sendov-Sendov ray monotonicity and is proved, not assumed, using the generalized Descartes rule (Lemma 3), which is also proved. Lemma 6 is an external characterization theorem; the paper explicitly states the representation used and the positivity of the measure, so the concavity conclusion follows by integrating concave kernels against a positive measure. This is an input theorem, not a restatement of the target concavity. Corollaries 7 and 8 invoke the author's prior entropic-independence and domain-sparsification theorems as black-box application theorems; these citations are load-bearing only for the sampling consequences, not for the new Gårding-type theorem, and they state independent results rather than assuming the paper's conclusion. No fitted constants, parameter-renaming, or definitional equivalences occur. The only caveat is that the complete-Bernstein characterization is a nontrivial external input; relying on it is a correctness/verification concern, not circularity.

Axiom & Free-Parameter Ledger

0 free parameters · 5 axioms · 0 invented entities

The paper introduces no new free parameters or invented entities. The central derivation depends only on standard analytic background (principal branches, Rolle/Descartes arguments, the Pick representation) and on previously published sampling theorems used for applications.

axioms (5)
  • domain assumption Real powers are defined using the principal logarithm on C\(-∞,0].
    Used throughout the paper; the stability and concavity statements depend on this branch convention.
  • standard math Generalized Descartes rule: a real-exponent exponential sum has at most as many positive zeros as its coefficient sign changes.
    Proved in the paper as Lemma 3 and is the key input to the ray-monotonicity theorem.
  • standard math Complete Bernstein (Pick) representation theorem from [SSV12, Theorem 6.2].
    Used without proof in Lemma 6 to turn the angular bound into concavity on the positive real line.
  • domain assumption Entropic-independence and domain-sparsification theorems from [Ana+21, Theorem 5] and [Ana+22].
    Applied as black boxes in Corollaries 7 and 8; the paper does not re-prove them.
  • domain assumption Sector-stability is preserved by nonzero links and positive external fields.
    Stated in Section 4 without proof or a specific citation; used to extend Γα-stability of generating functions to their links.

pith-pipeline@v1.3.0-alltime-deepseek · 5913 in / 24956 out tokens · 258315 ms · 2026-08-02T07:41:23.319114+00:00 · methodology

0 comments
Cite this review

Pith. "Pith review of G{\aa}rding's Theorem for Posynomials." pith.science (2026). https://pith.science/paper/NF42YIMH

@misc{pith2026260709168,
  author       = {Pith},
  title        = {Pith review of: G\aarding's Theorem for Posynomials},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NF42YIMH}},
  note         = {Machine review of arXiv:2607.09168}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

We extend G{\aa}rding's theorem to homogeneous posynomials: if a finite positive sum of monomials with arbitrary nonnegative real exponents is zero-free on a product of right half-planes, then its degree-normalized root is concave. Consequently, zero-freeness in a sector of aperture $\alpha\pi$ implies $\alpha$-fractional log-concavity. This sharpens generic mixing and domain-sparsification guarantees for fixed-size matchings and nonsymmetric determinantal point processes. The result was developed in an AI-assisted interaction initiated and checked by the author; Codex also assisted with assembling and typesetting the manuscript.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

13 extracted references · 1 linked inside Pith

  1. [1]

    An Inequality for Hyperbolic Polynomials , journal =

    G. An Inequality for Hyperbolic Polynomials , journal =

  2. [2]

    Canadian Journal of Mathematics , volume =

    Sendov, Blagovest and Sendov, Hristo , title =. Canadian Journal of Mathematics , volume =. 2021 , doi =

  3. [3]

    Bernstein Functions: Theory and Applications , edition =

    Schilling, Ren. Bernstein Functions: Theory and Applications , edition =. 2012 , doi =

  4. [4]

    Negative Dependence and the Geometry of Polynomials , journal =

    Borcea, Julius and Br. Negative Dependence and the Geometry of Polynomials , journal =. 2009 , doi =

  5. [5]

    Proceedings of the 29th Conference on Learning Theory , series =

    Anari, Nima and Oveis Gharan, Shayan and Rezaei, Alireza , title =. Proceedings of the 29th Conference on Learning Theory , series =

  6. [6]

    Lorentzian Polynomials , journal =

    Br. Lorentzian Polynomials , journal =. 2020 , doi =

  7. [7]

    Duke Mathematical Journal , volume =

    Anari, Nima and Oveis Gharan, Shayan and Vinzant, Cynthia , title =. Duke Mathematical Journal , volume =. 2021 , doi =

  8. [8]

    53rd Annual ACM SIGACT Symposium on Theory of Computing (STOC) , pages =

    Alimohammadi, Yeganeh and Anari, Nima and Shiragur, Kirankumar and Vuong, Thuy-Duong , title =. 53rd Annual ACM SIGACT Symposium on Theory of Computing (STOC) , pages =. 2021 , doi =

  9. [9]

    54th Annual ACM SIGACT Symposium on Theory of Computing (STOC) , pages =

    Anari, Nima and Jain, Vishesh and Koehler, Frederic and Pham, Huy Tuan and Vuong, Thuy-Duong , title =. 54th Annual ACM SIGACT Symposium on Theory of Computing (STOC) , pages =. 2022 , doi =

  10. [10]

    arXiv preprint arXiv:2106.04105 , year =

    Anari, Nima and Jain, Vishesh and Koehler, Frederic and Pham, Huy Tuan and Vuong, Thuy-Duong , title =. arXiv preprint arXiv:2106.04105 , year =

  11. [11]

    and Vuong, Thuy-Duong , title =

    Anari, Nima and Liu, Yang P. and Vuong, Thuy-Duong , title =. 63rd IEEE Annual Symposium on Foundations of Computer Science (FOCS) , pages =. 2022 , doi =

  12. [12]

    Domain Sparsification of Discrete Distributions Using Entropic Independence , booktitle =

    Anari, Nima and Derezi. Domain Sparsification of Discrete Distributions Using Entropic Independence , booktitle =. 2022 , doi =

  13. [13]

    and Lieb, Elliott H

    Heilmann, Ole J. and Lieb, Elliott H. , title =. Communications in Mathematical Physics , volume =. 1972 , doi =