Regularity of Gibbs measures for unbounded spin systems on general graphs

Christoforos Panagiotis; William Veitch

REVIEW 1 major objections 5 minor 1 cited by

Regularity of Gibbs measures for unbounded spin systems on general graphs

T0 review · 1 major / 5 minor · reviewed 2026-08-02 · deepseek-v4-flash

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

Pith's one-line read For unbounded spin systems on arbitrary graphs, boundary conditions can grow double-exponentially without destroying regularity of the Gibbs measure.

desk verdict Strong, self-contained regularity theorem for unbounded spins on arbitrary graphs, a real advance over the Z^d literature, but the advertised extremality of the plus measure is only proved among regular Gibbs measures. read the letter →

arxiv 2603.26319 v3 pith:TC5P5JJT submitted 2026-03-27 math.PR math-phmath.MP

Christoforos Panagiotis , William Veitch This is my paper · ORCID

classification math.PRmath-phmath.MP MSC 60K3582B20

keywords unboundedspinsystemsGibbsmeasuresregularityestimatestightnessplusmeasuresuper-Gaussiantailsbranchingprocessexplorationlong-rangeinteractions

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

This paper proves a regularity estimate for finite-volume Gibbs measures with unbounded real spins whose single-site distribution has super-Gaussian tails, on an arbitrary countable graph with short- or long-range interactions. The estimate controls the Radon–Nikodym derivative of the interacting measure against a non-interacting product measure by an exponential factor built from a function A(x,Λ,ξ,C) that measures the influence of boundary conditions at each site. In the nearest-neighbour case this permits boundary conditions growing double-exponentially in the distance to the boundary, and the paper shows this growth rate is optimal. The result yields infinite-volume 'plus' and 'minus' Gibbs measures that are regular with respect to a product measure, and provides a construction of the plus measure that avoids growing boundary conditions altogether.

What carries the argument

The argument rests on an exploration process that grows a cluster C of sites whose spins exceed thresholds calibrated by A(x,Λ,ξ,C), with thresholds rising as the exploration moves away from the target region. Lemmas 3.1 and 3.2 use Young-type inequalities to absorb the interaction energy of such large spins into the single-site potential, at a cost controlled by the tail exponent n. The cluster size is then dominated by the total progeny of a subcritical branching process whose offspring distribution is given by tail probabilities of the single-site measure; Lemma 3.4 uses this comparison to sum the exploration contributions and close the bound. The admissibility condition (C2) is what keep

What would settle it

Take V=Z with nearest-neighbour interactions, ρ(u)=e^{-a|u|^3}, and boundary conditions ξ_z = exp(exp(3|z|)). Proposition 5.2 predicts these finite-volume measures are not tight. Computing the law of φ_0 and testing whether it has a weak limit would settle the optimality claim.

Watch

Extended reading notes

Core claim

The central claim is Theorem 1.1: under the admissibility condition (C2) — the interactions are symmetric and there is a function f with f(t) ≥ log(1/|t|)^{1/n} near zero such that ∑_y |J_{xy}| f(J_{xy}) is bounded uniformly in x — the density of the finite-volume Gibbs measure restricted to Λ′ is bounded by ∏_{x∈Λ′} e^{C̃ A(x,Λ,ξ,C)^n} dν^0_{Λ′,0,ρ_{a/2},0}. Here A(x,Λ,ξ,C) is the smallest value that keeps boundary contributions along every walk from x within a prescribed decay envelope; it behaves like a non-Gaussian analogue of the harmonic extension of the boundary condition. When A stays bounded in the bulk, the finite-volume measures are tight, and the paper constructs the extremal plu

Load-bearing premise

The load-bearing assumption is the admissibility condition (C2): the interaction weights must admit one function f, growing at least like log(1/|t|)^{1/n} near zero, such that ∑_y |J_{xy}| f(J_{xy}) is bounded uniformly in the vertex x. If the interaction graph has long-range edges that make this sum infinite for every such f, the regularity estimate is not established.

Editorial extensions

If this is right

For nearest-neighbour interactions with n>2, boundary conditions growing at most like K^{(n-1)^{d(o,x)}} are admissible, and this double-exponential threshold is shown to be optimal for non-negative boundary conditions.
Long-range interactions are handled with the same proof; the allowed boundary growth is governed by the decay of the interaction kernel and the admissibility function f.
The plus and minus measures are constructed as limits of finite-volume measures, are regular Gibbs measures, and dominate all regular Gibbs measures in stochastic order.
A second construction of the plus measure uses random boundary conditions from a non-interacting product measure, giving finite-volume measures that are regular up to the boundary and stochastically decreasing in the volume.

Reading between the lines

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

Because the proof never uses translation invariance or amenability, it is plausible that the same scheme extends to graphs of unbounded degree and to non-geometric or random graphs, as long as the uniform summability condition (C2) holds.
The paper notes but does not prove that the argument should adapt to k-body interactions when the tail exponent exceeds k; a natural test is to replace the pairwise Young bounds with a k-variable inequality.
The bound is one-sided. An exact non-Gaussian Cameron–Martin identity would require identifying interactions and boundary conditions for which the inequality becomes an equality, which the paper leaves open.

Share X Bluesky LinkedIn Reddit HN

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

What you should know: Theorem 1.1 is the real product. It gives a Radon-Nikodym bound for finite-volume measures with growing boundary conditions on any countable graph, for both short- and long-range interactions, with the threshold encoded in the A-function. The proof is self-contained (exploration cluster, branching process, FKG) and the constants are explicit. That is a clean advance over Lebowitz–Presutti and Ruelle on Z^d and over [7] for vertex-transitive polynomial-growth graphs. The nearest-neighbour applications are also sharp: double-exponential boundary growth for n>2, exponential for n=2, and Propositions 5.2 and 5.3 show the threshold is optimal for P(phi) models. The alternative construction of the plus measure in Section 5.4 (random boundary conditions, regular up to the boundary, stochastically monotone in volume) is a nice extra and likely useful for later work.

Soft spots. The one that matters is the extremality claim. The abstract and Section 5.3 say the plus measure is extremal, but Proposition 5.8 only shows it is maximal among a-regular Gibbs measures. The proof that nu[F_Lambda^c] -> 0 for arbitrary Gibbs nu uses the regularity bound on nu; if nu is not regular, that step has no support. So 'maximal, hence extremal' is not justified in the standard sense of an extreme point of all Gibbs measures. The main theorem and the construction of nu+ are unaffected; the claim just needs to be qualified to 'maximal among a-regular Gibbs measures' unless the authors can show every Gibbs measure is regular. I don't see an easy fix in the present paper, but it is a presentation issue, not a hole in the core result.

The admissibility condition (C2) is the other thing to check before using the theorem. The f-function is doing real work: it controls the tails of the interactions so that Lemma 3.2 can absorb a neighbouring spin and Lemma 3.4 can keep the branching process subcritical. For long-range interactions with slow decay, you need to know such an f exists. The paper gives reasonable examples (log and power-law type f), so this is an assumption to verify in each application, not a hidden flaw.

I found no sign of circularity or fitted parameters. The proof is long, and I did not machine-check every inequality, but the main line is clear and the lemmas line up. The paper is honest about its limits, mentioning k-body extensions only as expected.

Bottom line: this is a paper for specialists in statistical mechanics and Gibbs measures, and it should go to a serious referee. The extremality overclaim should be fixed, but it is fixable in revision and does not undermine the main regularity theorem.

Referee Report

1 major / 5 minor

Summary. The paper develops regularity estimates for finite-volume Gibbs measures of unbounded spin systems on arbitrary countable graphs with pair interactions satisfying an admissibility condition (C2). The main theorem (Theorem 1.1) bounds the marginal density of the finite-volume measure with boundary conditions ξ by a product of non-interacting single-site measures with super-Gaussian tails, with an exponent controlled by a function A(x,Λ,ξ,C). The proof constructs an exploration cluster of large spins and dominates its size by a subcritical branching process. Applications include tightness for boundary conditions with double-exponential growth in nearest-neighbour models, an optimality result for P(φ) models, a construction of plus/minus infinite-volume Gibbs measures that are regular and maximal among a-regular Gibbs measures, and an alternative finite-volume construction of the plus measure in the nearest-neighbour case via random boundary conditions or vertex-dependent single-site measures.

Significance. If correct, this is a substantial advance over earlier regularity results of Lebowitz–Presutti and Ruelle, which were limited to Z^d and logarithmic boundary growth; the proof is self-contained and works on arbitrary graphs, with explicit control of constants. The exploration/branching argument is a genuinely new tool for this class of models. The paper also gives clean tightness criteria and new constructions of infinite-volume measures. The main regularity theorem is not affected by the overclaim discussed below, but the advertised extremality of the plus measure is presently not proven.

major comments (1)

[Abstract; Section 5.3, Proposition 5.8] The claim that the plus measure is 'extremal' is not supported by the proof. Proposition 5.8 establishes ν^- ⪯ ν ⪯ ν^+ only for a-regular Gibbs measures ν. Maximality in the stochastic order among a-regular Gibbs measures does not imply extremality in the convex set of all Gibbs measures, and the sentence 'maximal, hence extremal' is therefore unjustified. Since the abstract advertises the construction of an 'extremal measure', this is a load-bearing overclaim, though it does not affect Theorem 1.1. Please qualify the abstract (e.g., 'maximal among a-regular Gibbs measures') or add an argument that all Gibbs measures are regular under the stated hypotheses.

minor comments (5)

[Abstract; Section 5.4] The abstract announces 'an alternative construction ... regular up to the boundary' without noting that Section 5.4 is restricted to nearest-neighbour interactions on bounded-degree graphs. Please add the qualification.
[Corollary 1.2] The second stochastic domination ('hence ... ν^0_{Λ,β,˜ρ}') uses monotonicity in β for nonnegative single-site measures, which is not stated at this point. It follows from FKG together with H≥0 on the support of ˜ρ, but should be mentioned to avoid a gap.
[Definition 1.4; Theorem 1.1] The measures ρ_a and ρ_{a/2} are not normalized, so ν^0_{Λ,0,ρ_a,0} is not a probability measure. The normalization convention should be stated explicitly where 'density' is used.
[Section 2.2] Several displayed formulas lose superscripts in the arXiv text, e.g., 'CA(n−1)m' should read 'C A^{(n-1)^m}'. Please proofread the mathematical expressions in the final version.
[Proof of Proposition 5.2] The definition of D_{i,j} is hard to parse; a display with fully parenthesized exponents would improve readability.

Circularity Check

0 steps flagged · score 2.0 of 10

No reductive circularity; main regularity theorem is self-contained. Minor self-citations support only secondary optimality/identification claims.

full rationale

The derivation of Theorem 1.1 does not reduce to its inputs by construction. It is proved from the measure definition (2.1), the admissibility hypotheses (C1)-(C2), Lemmas 3.1-3.4, and the standard branching-process formula (Theorem 2.5). The function A is defined explicitly from ξ, J, f, C (Section 2.2) before the theorem; it is not a fitted parameter, and the theorem asserts a one-sided bound rather than an equality forced by the definition. No quantity is fit to a subset of data and then called a prediction. The only self-citations are secondary: Lemma 5.4 delegates monotonicity of the absolute-value field to [8, Prop. 4.10], and Section 5.3 cites [7, Prop. 2.6] to identify ν+ with an external-field limit. These are used for the optimality corollaries or contextual identification, not for the central regularity theorem, so they are not load-bearing in the reductive sense. The abstract's 'extremal' is stronger than Proposition 5.8's qualified 'maximal among a-regular Gibbs measures'; this is a scope caveat, not circularity. Overall the derivation chain is self-contained against external benchmarks for the main result.

Assumptions & free parameters 0 free parameters · 6 assumptions · 0 invented entities

The central regularity claim rests on the super-Gaussian tail condition and on the f-integrability of interactions. Constants are explicit in the proof rather than fitted to data. Applications add standard monotonicity theorems and volume-growth conditions for the plus measure, but the main regularity theorem is self-contained apart from standard probabilistic tools.

assumptions (6)

domain assumption Single-site measure ρ satisfies ∫ e^{a|u|^n} dρ(u)<∞ for some a>0,n>2 (or (1.2) for n=2).
This is the super-Gaussian tail hypothesis on which Theorem 1.1 and Theorem 4.1 depend; the constants C and C̃ depend on a,n and ρ.
domain assumption Admissible interactions (C1),(C2): J symmetric with sup_x Σ_y |J_{xy}| f(J_{xy}) ≤ M_f for some f with f(t)≥log(|t|^{-1})^{1/n} near 0.
Used in Lemma 3.2 to absorb interaction terms into the single-site potential and in Lemma 3.4 to control the branching process uniformly. Without it the exploration argument does not close.
domain assumption Boundary conditions satisfy Σ_y |J_{xy} ξ_y| < ∞ for all x∈Λ.
Needed for the Hamiltonian in Definition 2.1 to be finite and for the boundary field h_{x,Λ} to be well-defined.
standard math Standard FKG inequality, Griffiths inequality, and monotonicity of the Ising conditional measure.
Used in Proposition 2.3, Lemma 5.4, and Lemma 5.13 for stochastic domination and monotonicity in boundary conditions and interactions.
standard math Total-progeny formula for branching processes, [9, Theorem 3.13].
Used in Lemma 3.4 to sum cluster probabilities and bound the size of the exploration cluster.
domain assumption For the plus-measure construction, the graph has bounded degree (nearest-neighbour cases), interactions are reasonable, and condition (5.10) on f and volume growth holds.
Propositions 5.8, 5.11, and 5.12 need these additional assumptions; they are not part of Theorem 1.1 and are not highlighted in the abstract's 'arbitrary graphs' formulation.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Regularity of Gibbs measures for unbounded spin systems on general graphs." pith.science (2026). https://pith.science/paper/TC5P5JJT

@misc{pith2026260326319,
  author       = {Pith},
  title        = {Pith review of: Regularity of Gibbs measures for unbounded spin systems on general graphs},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TC5P5JJT}},
  note         = {Machine review of arXiv:2603.26319}
}

abstract

We consider a general class of spin systems with potentially unbounded real-valued spins, defined via a single-site potential with super-Gaussian tails on general graphs, allowing for both short- and long-range interactions. This class includes all $P(\varphi)$ models, in particular the well-studied $\varphi^4$ model. We construct an infinite-volume extremal measure called the plus measure as the limit of finite-volume Gibbs measures with weakly growing boundary conditions and show that it is regular, in the sense that it admits a bounded Radon-Nikodym derivative with respect to a product measure of single-site distributions with super-Gaussian tails. Moreover, we provide an alternative construction of the plus measure as the limit of finite-volume Gibbs measures that are regular up to the boundary. As a key intermediate step, we establish regularity and tightness of finite-volume Gibbs measures for a large class of growing boundary conditions $\xi$. Our regularity estimates are encoded in terms of a function $A(\xi)$, which provides precise control on the change of measure induced by boundary perturbations, and can thus be viewed as an analogue of the Cameron-Martin theorem for non-Gaussian fields. In the nearest-neighbour case, this class includes boundary conditions that grow at most double-exponentially in the distance to the boundary when the single-site measure has tails of the form $e^{-a|u|^n}$ for some $n>2$.Our results apply to arbitrary graphs and improve upon earlier results of Lebowitz and Presutti, and Ruelle, which apply in the context of $\mathbb{Z}^d$ and allow only logarithmically growing boundary conditions, as well as subsequent extensions to vertex-transitive graphs of polynomial growth.

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

Subcritical sharpness for real-valued spin models
math.PR 2026-05 unverdicted novelty 7.0 of 10

Exponential decay of correlations is shown for subcritical real-valued spin models on transitive graphs by generalizing the OSSS inequality in the FK percolation representation.

Reference graph

Works this paper leans on

14 extracted references · 3 linked inside Pith · cited by 1 Pith paper

[7]

T. S. Gunaratnam, C. Panagiotis, R. Panis, and F. Severo. Random tangled currents forφ 4: Translation invariant Gibbs measures and continuity of the phase transi- tion.Journal of the European Mathematical Society (to appear), 2025. Available at https://arxiv.org/abs/2211.00319

arXiv 2025
[1]

Albeverio, Y

S. Albeverio, Y. G. Kondratiev, M. R¨ ockner, and T. V. Tsikalenko. A priori esti- mates for symmetrizing measures and their applications to Gibbs states.Journal of Functional Analysis, 171(2):366–400, 2000

2000
[2]

Bellissard and R

J. Bellissard and R. Høegh-Krohn. Compactness and the maximal Gibbs state for ran- dom Gibbs fields on a lattice.Communications in Mathematical Physics, 84(3):297– 327, 1982

1982
[3]

Cassandro, E

M. Cassandro, E. Olivieri, A. Pellegrinotti, and E. Presutti. Existence and uniqueness of DLR measures for unbounded spin systems.Zeitschrift f¨ ur Wahrscheinlichkeits- theorie und verwandte Gebiete, 41(4):313–334, 1978

1978
[4]

Duminil-Copin

H. Duminil-Copin. Lectures on the Ising and Potts models on the hypercubic lattice. InPIMS-CRM Summer School in Probability, pages 35–161. Springer, 2017. Available at https://arxiv.org/abs/1707.00520

arXiv 2017
[5]

Friedli and Y

S. Friedli and Y. Velenik.Statistical Mechanics of Lattice Systems: A Concrete Mathematical Introduction. Cambridge University Press, 2017

2017
[6]

Glimm and A

J. Glimm and A. Jaffe.Quantum Physics : A Functional Integral Point of View. Springer New York, 2nd edition, 1987

1987
[8]

T. S. Gunaratnam, C. Panagiotis, R. Panis, and F. Severo. The supercrit- ical phase of theφ 4 model is well behaved.Preprint, 2025. Available at https://arxiv.org/abs/2501.05353

arXiv 2025

Show all 14 references

[9]

van der Hofstad.Random graphs and complex networks

R. van der Hofstad.Random graphs and complex networks. Volume 1, volume 43 of Cambridge series in statistical and probabilistic mathematics. Cambridge University Press, 2017. 32

2017
[10]

J. L. Lebowitz and E. Presutti. Statistical mechanics of systems of unbounded spins. Communications in Mathematical Physics, 50(3):195–218, 1976

1976
[11]

A. Raoufi. Translation-invariant Gibbs states of the Ising model: general setting.The Annals of Probability, 48(2):760–777, 2020

2020
[12]

D. Ruelle. Superstable interactions in classical statistical mechanics.Communications in Mathematical Physics, 18(2):127–159, 1970

1970
[13]

D. Ruelle. Probability estimates for continuous spin systems.Communications in Mathematical Physics, 50(3):189–194, 1976

1976
[14]

Werner and E

W. Werner and E. Powell.Lecture Notes on the Gaussian Free Field, volume 28 of Cours Sp´ ecialis´ es. Soci´ et´ e Math´ ematique de France, Paris, 2021. 33

2021

Pith tools

Reviewed August 2, 2026 · model on record in the stance chip above.

[1] [7]

T. S. Gunaratnam, C. Panagiotis, R. Panis, and F. Severo. Random tangled currents forφ 4: Translation invariant Gibbs measures and continuity of the phase transi- tion.Journal of the European Mathematical Society (to appear), 2025. Available at https://arxiv.org/abs/2211.00319

arXiv 2025

[2] [1]

Albeverio, Y

S. Albeverio, Y. G. Kondratiev, M. R¨ ockner, and T. V. Tsikalenko. A priori esti- mates for symmetrizing measures and their applications to Gibbs states.Journal of Functional Analysis, 171(2):366–400, 2000

2000

[3] [2]

Bellissard and R

J. Bellissard and R. Høegh-Krohn. Compactness and the maximal Gibbs state for ran- dom Gibbs fields on a lattice.Communications in Mathematical Physics, 84(3):297– 327, 1982

1982

[4] [3]

Cassandro, E

M. Cassandro, E. Olivieri, A. Pellegrinotti, and E. Presutti. Existence and uniqueness of DLR measures for unbounded spin systems.Zeitschrift f¨ ur Wahrscheinlichkeits- theorie und verwandte Gebiete, 41(4):313–334, 1978

1978

[5] [4]

Duminil-Copin

H. Duminil-Copin. Lectures on the Ising and Potts models on the hypercubic lattice. InPIMS-CRM Summer School in Probability, pages 35–161. Springer, 2017. Available at https://arxiv.org/abs/1707.00520

arXiv 2017

[6] [5]

Friedli and Y

S. Friedli and Y. Velenik.Statistical Mechanics of Lattice Systems: A Concrete Mathematical Introduction. Cambridge University Press, 2017

2017

[7] [6]

Glimm and A

J. Glimm and A. Jaffe.Quantum Physics : A Functional Integral Point of View. Springer New York, 2nd edition, 1987

1987

[8] [8]

T. S. Gunaratnam, C. Panagiotis, R. Panis, and F. Severo. The supercrit- ical phase of theφ 4 model is well behaved.Preprint, 2025. Available at https://arxiv.org/abs/2501.05353

arXiv 2025