Boundary Free Energies in Disordered Ising Models

Hexiang Wang; Keheng Zhu; Mauris Chueng

Boundary Free Energies in Disordered Ising Models

T0 review · 0 major / 5 minor · reviewed 2026-08-01 · deepseek-v4-flash

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

Pith's one-line read An explicit surface free-energy density exists for bounded disordered Ising models at high temperature; without strong assumptions, no universal surface limit exists.

desk verdict A solid, honest paper: proves a deterministic high-temperature surface free-energy limit for disordered Ising models in the Dobrushin regime, with clean counterexamples showing why none exists in general; deserves a serious referee. read the letter →

arxiv 2607.18326 v1 pith:DEOUKQBX submitted 2026-07-18 math-ph math.MP

Hexiang Wang , Keheng Zhu , Mauris Chueng This is my paper · ORCID

classification math-phmath.MP MSC 82B4482B2060K3582B27

keywords boundaryfreeenergydisorderedIsingmodelsurfacefree-energydensityvanHovesequenceseamflipspinglassself-averagingconcentrationinequality

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 asks whether the boundary correction to the free energy of a disordered Ising magnet has a well-defined thermodynamic limit when divided by the number of boundary bonds. It answers no in general: in one dimension with i.i.d. disorder, different van Hove sequences of boxes give different limiting surface densities, and along intervals the random correction need not converge in probability. It answers yes in the bounded, high-temperature uniqueness regime in dimensions d≥2: for cubic boxes, the normalized free-to-fixed boundary correction converges in expectation, almost surely, and in every L^p to a deterministic constant τ(β), with Gaussian-type surface-order fluctuations. The paper also proves a temperature-uniform variance bound for seam-flip free-energy differences and derives exact Gaussian interpolation identities, while stressing that these do not by themselves settle the low-temperature Gaussian spin-glass surface problem.

What carries the argument

The load-bearing object is the half-space Gibbs state on the half-space H={x∈Z^d: x_1≥1}, with boundary couplings interpolated by a parameter s∈[0,1]. Under the strict high-temperature uniqueness condition (α=2d tanh(βJ*)<1), the single-site influence matrix has row sums bounded by α<1, so an exponential comparison estimate makes this state unique and exponentially close to finite-box Gibbs measures with matching boundary fields; this converts the finite-volume free-energy difference into an integral over s of boundary-spin expectations. Summing squared oscillations of the free energy under resampling individual bonds feeds a bounded-differences concentration inequality, giving the surface-o

What would settle it

The half-space formula (11) makes a sharp prediction: the boundary-spin expectation E[K_{e0}⟨σ_{e1}⟩_{H,s}] is independent of the exhaustion and exterior conditioning used to build the half-space state. If two natural constructions give different values of the integral, the theorem's τ(β) is not well-defined. Alternatively, in the paper's one-dimensional example, sample values ΔF_L/b_L along centered intervals in one fixed i.i.d. environment should fail to converge in probability even though the expectation converges.

Watch

Extended reading notes

Core claim

The central positive claim is Theorem 2: in the strict high-temperature uniqueness regime defined by α=2d tanh(βJ*)<1, with bounded i.i.d. couplings, cubic boxes, and d≥2, the quantity ΔF_L/b_L—the free-to-fixed boundary free-energy difference divided by the number of boundary bonds—converges to a deterministic constant τ(β). The paper identifies τ(β) explicitly as τ(β)=-∫_0^1 E[K_{e0}⟨σ_{e1}⟩_{H,s}] ds, where ⟨σ_{e1}⟩_{H,s} is the boundary-spin expectation in the unique Gibbs state of the half-space lattice with interpolated boundary couplings. Convergence holds in expectation, almost surely, and in every L^p, and the centered correction satisfies the surface-order concentration bound P(|ΔF

Load-bearing premise

The theorem's proof breaks if the strict high-temperature uniqueness condition α=2d tanh(βJ*)<1 fails, or if the couplings are unbounded (e.g., Gaussian at low temperature); without those conditions no convergence or concentration bound is proved, and the paper explicitly leaves the Gaussian low-temperature limit open.

Editorial extensions

If this is right

In the bounded high-temperature regime, two experiments on the same disordered sample—one with free, one with frozen exterior spins—will measure the same bulk free energy and a deterministic surface correction per boundary bond; sample-to-sample surface noise disappears as L→∞.
The normalized correction L(f^fixed_L − f^free_L) tends to 2d τ(β), so the surface free-energy density in this regime is computable from a half-space correlation function.
The surface free-energy density is not independent of box shape: only regular boxes with diverging minimum side are guaranteed to share the same limit, and the paper's one-dimensional example shows that other van Hove sequences can give different values.
Domain-wall (seam-flip) free-energy fluctuations are bounded by 4v|S_L| uniformly in temperature, so a root-mean-square stiffness exponent, if it exists, cannot exceed (d−1)/2.
The Gaussian low-temperature surface limit is not a consequence of these identities; it remains open.

Reading between the lines

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

The explicit half-space formula suggests a practical finite-volume estimator for τ(β): simulate boundary-spin correlations with interpolated boundary couplings and integrate over s; the paper does not propose such an algorithm.
The one-dimensional counterexample operates inside the high-temperature regime, so for bounded couplings the obstruction to a universal surface limit is not low temperature—it is the combination of geometry and randomness; one might expect similar sequence-dependence for anisotropic boundaries in d≥2 outside the regular-box class.
Theorem 3's temperature-uniform bound implies the same variance inequality holds for ground-state energies, so the upper-critical stiffness exponent (d−1)/2 is a zero-temperature statement too; whether the exponent is actually reached or is model-dependent could be tested by exact ground-state enumeration on small boxes.
A possible route to the low-temperature Gaussian problem would be to construct a cube-sequence analogue of the van Hove counterexample, or to prove that the boundary-overlap quantity on the right of the interpolation identity has a unique full-sequence limit; the paper shows neither is immediate.

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

Colleague,

This one is worth your time. The paper proves a real theorem (Theorem 2): for bounded i.i.d. couplings on cubes, in the strict Dobrushin uniqueness regime, the free-to-fixed boundary free-energy difference divided by boundary size converges to an explicit half-space limit, with a.s. and L^p convergence and surface-order concentration. The proof is all there—Dobrushin comparison, the linear interpolation path, McDiarmid via depth-oscillation sums. I checked the constants in (58) and the depth layering; they work, and the stress-test note about the conservative 4 instead of 2 in the oscillation sum is right—it only makes the bound larger.

The negative results are just as good. The nested van Hove exhaustion in Section 6.3 giving two different subsequential surface limits is a clean counterexample, and the one-dimensional failure of convergence in probability (Section 6.4) is the right way to state what goes wrong. Theorem 3's seam variance bound via Efron–Stein is simple and rigorous, with a one-Lipschitz observation that carries to ground states.

What I like most is the honesty about scope. The main theorem is explicitly conditional on α < 1 and bounded couplings; the Gaussian low-temperature problem is stated as open in Section 9; the paper does not oversell. The interpolation identities for Gaussian boundary disorder are nice but are correctly labeled as not sufficient for a low-temperature limit.

Soft spots are minor. The figures look like decorative filler—no reference to them in the text, and their captions read like placeholders. They should be cut. There is also a slight over-broad title; the content is about high-temperature surface limits, not boundary free energies in general. The stochastic-localization discussion in Section 8.3 is more about what does not work than what does, but it is clearly flagged.

I see no circularity or fitting. The finite-volume identities are derived, not assumed, and the half-space representation of τ(β) follows from the comparison estimate. The reader's ACCEPT verdict with moderate confidence seems right to me.

Who should read it: anyone working on surface free energies, spin-glass boundary effects, or rigorous high-temperature Gibbs measures. It is not a grand breakthrough, but it settles a natural question cleanly and leaves the right open problem.

My recommendation: send to a serious referee. It will likely be accepted after removing the decorative figures and tightening a few constants.

Referee Report

0 major / 5 minor

Summary. The paper studies the boundary correction to the free energy of disordered Ising models. It first gives one-dimensional counterexamples showing that the surface free-energy density can depend on the van Hove exhaustion and that the sample correction need not converge in probability. The main positive result, Theorem 2, proves under the Dobrushin uniqueness condition α = 2d tanh(βJ*) < 1 for bounded i.i.d. couplings on cubic boxes in d ≥ 2 that the normalized free-to-fixed boundary free-energy difference converges in expectation, almost surely, and in every L^p to an explicit half-space boundary response τ(β), with an explicit surface-order concentration estimate. A parallel result is given for rectangular boxes with a summability condition for almost-sure convergence. The paper also derives exact Gaussian and independent-copy interpolation identities, proves a temperature-uniform variance bound for seam-flip free-energy differences via Efron–Stein, and discusses why the low-temperature Gaussian Edwards–Anderson limit remains open.

Significance. If the results are correct, the paper establishes the first rigorous deterministic surface free-energy density for high-temperature disordered Ising models with regular boxes, providing an explicit representation as an integral of a half-space boundary response and quantitative concentration with explicit constants. The one-dimensional counterexamples cleanly separate the roles of boundary geometry, normalization, and mode of convergence. The seam-flip variance bound is temperature-uniform and sharp in its scaling, and the honest discussion of the open low-temperature problem clarifies the limits of the method. The proofs are largely self-contained and use standard tools (Dobrushin comparison, Gaussian interpolation, concentration inequalities) with no fitting or post-hoc assumptions. The scope restrictions are stated explicitly and are not hidden.

minor comments (5)

[Theorem 1, item 4] Item 4 lists four inequivalent boundary-condition comparisons, but the proof text only explicitly proves items 1, 2, and 5 (Section 6.3, Eq. (76)). Please either supply a brief proof or reference for item 4 or remove it from the theorem statement.
[Section 6.1 and 6.2] The text refers to 'Theorem 4' in the half-space construction and in the proof of Theorem 2; the cited result is Lemma 4. Please correct the cross-reference.
[Eq. (52)–(53)] The bound in (52) uses Σ_i δ_i(q_e) ≤ 4, but for a single spin observable the sum is 2; the factor is conservative. Similarly, the first term in (53) appears to contain an extra factor of 2 relative to the stated fraction of bad edges. These do not affect convergence, but the constants should be corrected or justified.
[Theorem 2, Eq. (10)] The notation 'set e0 = (e1, 0)' is ambiguous. e0 should be described more explicitly as a boundary edge, e.g., connecting (1,0,...,0) to (0,0,...,0), to avoid confusion with a coordinate vector.
[Eq. (58)] For rectangular boxes the sentence 'At each depth the number of internal edges is at most 2d b_L' mirrors the cube argument, but the analogous vertex-counting statement with b_L defined in (59) is not spelled out. A short clarification would help.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: Theorem 2 is proved from stated hypotheses using standard external tools, and the representation of the limit is derived, not assumed.

full rationale

The paper's central derivation chain is self-contained and does not reduce to its inputs. Theorem 2's limit τ(β) is not fitted or imposed; it is obtained by proving uniform convergence of the normalized boundary average in Eq. (51) via the Dobrushin comparison bound (52)–(53), which relies on the external Dobrushin comparison theorem (Lemma 4, citing Georgii [8]) and on explicit lattice-geometry counting. The half-space Gibbs state in the representation (10)–(11) is constructed and proved unique within the paper under the same α<1 condition; it is not imported as an unproved self-citation. There are no fitted parameters called predictions: the constants C*, J*, vK, bL are defined from the model, and the concentration estimate (13) follows from the explicit oscillation sum (58) and McDiarmid's inequality. Theorem 1 is an exact computation showing genuine non-universality, Theorem 3 is an Efron–Stein variance bound, and the Gaussian interpolation identities (35) and (78) are derived finite-volume identities with explicit bounds. The paper openly states the limits of its regime: the low-temperature Gaussian surface limit is declared open in Section 9, almost-sure convergence for rectangles is conditional on a summability hypothesis, and no claim is made for arbitrary van Hove sequences. No load-bearing argument depends on self-citation: the references are standard external texts and papers. Consequently the derivation is not circular by construction, and the appropriate score is 0.

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

No parameters are fitted; the only inputs are model parameters (β, J*, κ², v). The central claims rely on standard rigorous tools and explicit theorem hypotheses, not on self-cited constructions or newly postulated entities.

assumptions (6)

standard math Dobrushin comparison theorem (Lemma 4) with influence matrix C_{xz} = tanh(βJ*) 1_{x~z} and row sum α = 2d tanh(βJ*) < 1
Used to compare finite-cube and half-space Gibbs expectations in Section 6.2 and to prove uniqueness of the half-space state; the contraction (47) is the engine of Theorem 2.
standard math Gaussian interpolation and Gaussian integration by parts for log-partition functions
Underpins the finite-volume identities in Section 5, especially Eqs. (31), (35), and the endpoint-compatibility discussion.
standard math McDiarmid's bounded differences inequality
Combined with the computed coordinate oscillations (57)-(58) to obtain the exponential concentration estimate (13).
standard math Gaussian concentration / Poincaré inequality
Used for the Gaussian-disorder variance and tail bounds (42)-(43) and for the Gaussian seam tail (17).
domain assumption Coupling field on Z^d realized as a single i.i.d. edge field so different volumes are jointly realized
Needed for almost-sure statements comparing different volumes; stated as part of regimes (D) and (G) in Section 2.2.
domain assumption Uniform boundedness |J_a|, |K_e| ≤ J* and strict Dobrushin condition α = 2d tanh(βJ*) < 1
Explicit hypothesis of Theorem 2 and regime (D); if violated, the Dobrushin comparison and McDiarmid oscillation estimates break.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Boundary Free Energies in Disordered Ising Models." pith.science (2026). https://pith.science/paper/DEOUKQBX

@misc{pith2026260718326,
  author       = {Pith},
  title        = {Pith review of: Boundary Free Energies in Disordered Ising Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DEOUKQBX}},
  note         = {Machine review of arXiv:2607.18326}
}

abstract

The boundary correction to the free energy of a disordered Ising model depends on the boundary condition, the normalization, and the finite-volume sequence. We give a one-dimensional i.i.d. counterexample showing that the surface free-energy density need not be independent of the van Hove sequence and that the corresponding random correction need not converge in probability. For bounded couplings in the Dobrushin uniqueness regime on cubic boxes in dimensions $d\geq2$, we prove convergence of the normalized free-to-fixed boundary free-energy difference in expectation, almost surely, and in every $L^p$, $1\leq p<\infty$. For Gaussian boundary couplings, we derive an exact finite-volume interpolation identity and explain why it does not by itself imply a low-temperature surface limit. For a seam-flip free-energy difference $D_L$ across a set $S_L$ of independent symmetric bonds of variance $v$, we prove $\Var(D_L)\leq4v\abs{S_L}$; one-dimensional examples show that symmetry and finite moments alone do not determine a stiffness exponent, and the low-temperature Gaussian Edwards--Anderson problem remains open.

Figures

Figures reproduced from arXiv: 2607.18326 by the authors.

**Figure 2.** High-resolution numerical rendering of the half-space coupling graph, detailing the boundary layer interactions and spin configurations within the uniform Dobrushin uniqueness regime. Associate an Ising spin σx ∈ {−1, +1} with each x ∈ ΛL. An exterior spin pattern ψ = (ψy) ∈ {−1, +1} is deterministic. For e = (x, y) ∈ BL put qe(σ) = σxψy. 2 [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗

**Figure 4.** High-resolution diagram of the counterexample exhaustion path on Z, demonstrating how alternating geometric densities trigger distinct subsequential limits for the expected surface free-energy term. Multiplication by −1/β gives two distinct free-energy surface limits. The example can be chosen at arbitrarily small x, so it persists inside the one-dimensional Dobrushin regime. This proves that arbitrary van Hove in… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

14 extracted references · 3 canonical work pages

[1]

Aizenman and J

M. Aizenman and J. Wehr, Rounding eﬀects of quenched randomness on ﬁrst-order phase transitions, Comm. Math. Phys. 130 (1990), no. 3, 489–528. DOI: https://doi.org/10. 1007/BF02096933. 17

1990
[2]

Arguin, C

L.-P. Arguin, C. M. Newman, D. L. Stein, and J. Wehr, Fluctuation bounds for interface free energies in spin glasses , J. Stat. Phys. 156 (2014), no. 2, 221–238. DOI: https://doi.org/ 10.1007/s10955-014-1009-8

work page doi:10.1007/s10955-014-1009-8 2014
[3]

Boucheron, G

S. Boucheron, G. Lugosi, and P. Massart, Concentration Inequalities: A Nonasymptotic Theory of Independence , Oxford University Press, Oxford, 2013. DOI: https://doi.org/10.1093/ acprof:oso/9780199535255.001.0001

arXiv 2013
[4]

R. L. Dobrushin, R. Kotecký, and S. B. Shlosman, Wulﬀ Construction: A Global Shape from Local Interaction, Translations of Mathematical Monographs, vol. 104, American Mathematical Society, Providence, RI, 1992. DOI: https://doi.org/10.1090/mmono/104

work page doi:10.1090/mmono/104 1992
[5]

S. F. Edwards and P. W. Anderson, Theory of spin glasses , J. Phys. F: Met. Phys. 5 (1975), no. 5, 965–974. DOI: 10.1088/0305-4608/5/5/017

doi:10.1088/0305-4608/5/5/017 1975
[6]

Eldan, Thin shell implies spectral gap up to polylog via a stochastic localization scheme, Geom

R. Eldan, Thin shell implies spectral gap up to polylog via a stochastic localization scheme, Geom. Funct. Anal. 23 (2013), no. 2, 532–569. DOI: https://doi.org/10.1007/ s00039-013-0214-y

2013
[7]

D. S. Fisher and D. A. Huse, Equilibrium behavior of the spin-glass ordered phase , Phys. Rev. B 38 (1988), no. 1, 386–411. DOI: https://doi.org/10.1103/PhysRevB.38.386

doi:10.1103/physrevb.38.386 1988
[8]

Georgii, Gibbs Measures and Phase Transitions , second ed., de Gruyter Studies in Math- ematics, vol

H.-O. Georgii, Gibbs Measures and Phase Transitions , second ed., de Gruyter Studies in Math- ematics, vol. 9, Walter de Gruyter, Berlin, 2011

2011

Show all 14 references

[9]

Guerra and F

F. Guerra and F. L. Toninelli, The thermodynamic limit in mean ﬁeld spin glass models , Comm. Math. Phys. 230 (2002), no. 1, 71–79. DOI: https://doi.org/10.1007/s00220-002-0699-y

2002 doi
[10]

Ledoux, The Concentration of Measure Phenomenon , Mathematical Surveys and Mono- graphs, vol

M. Ledoux, The Concentration of Measure Phenomenon , Mathematical Surveys and Mono- graphs, vol. 89, American Mathematical Society, Providence, RI, 2001. DOI: https://doi. org/10.1090/surv/089

2001 doi
[11]

C. M. Newman and D. L. Stein, Metastate approach to thermodynamic chaos , Phys. Rev. E 55 (1997), no. 5, 5194–5211. DOI: https://doi.org/10.1103/PhysRevE.55.5194

1997 doi
[12]

C. M. Newman and D. L. Stein, Ordering and broken symmetry in short-ranged spin glasses , J. Phys.: Condens. Matter 15 (2003), no. 32, R1319–R1364. DOI: 10.1088/0953-8984/15/ 32/202

2003 doi
[13]

S. I. Resnick, Extreme Values, Regular Variation, and Point Processes , Applied Prob- ability, vol. 4, Springer-Verlag, New York, 1987. DOI: https://doi.org/10.1007/ 978-0-387-75953-1

1987
[14]

Ruelle, Statistical Mechanics: Rigorous Results , W

D. Ruelle, Statistical Mechanics: Rigorous Results , W. A. Benjamin, New York, 1969. Hexiang W ang, School of Mathematical Sciences, Nankai University, Tianjin, 300071, China Kui6539@outlook.com Keheng Zhu , Academy for Multidisciplinary Studies, School of Mathematics Sciences...

1969

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.

[1] [1]

Aizenman and J

M. Aizenman and J. Wehr, Rounding eﬀects of quenched randomness on ﬁrst-order phase transitions, Comm. Math. Phys. 130 (1990), no. 3, 489–528. DOI: https://doi.org/10. 1007/BF02096933. 17

1990

[2] [2]

Arguin, C

L.-P. Arguin, C. M. Newman, D. L. Stein, and J. Wehr, Fluctuation bounds for interface free energies in spin glasses , J. Stat. Phys. 156 (2014), no. 2, 221–238. DOI: https://doi.org/ 10.1007/s10955-014-1009-8

work page doi:10.1007/s10955-014-1009-8 2014

[3] [3]

Boucheron, G

S. Boucheron, G. Lugosi, and P. Massart, Concentration Inequalities: A Nonasymptotic Theory of Independence , Oxford University Press, Oxford, 2013. DOI: https://doi.org/10.1093/ acprof:oso/9780199535255.001.0001

arXiv 2013

[4] [4]

R. L. Dobrushin, R. Kotecký, and S. B. Shlosman, Wulﬀ Construction: A Global Shape from Local Interaction, Translations of Mathematical Monographs, vol. 104, American Mathematical Society, Providence, RI, 1992. DOI: https://doi.org/10.1090/mmono/104

work page doi:10.1090/mmono/104 1992

[5] [5]

S. F. Edwards and P. W. Anderson, Theory of spin glasses , J. Phys. F: Met. Phys. 5 (1975), no. 5, 965–974. DOI: 10.1088/0305-4608/5/5/017

doi:10.1088/0305-4608/5/5/017 1975

[6] [6]

Eldan, Thin shell implies spectral gap up to polylog via a stochastic localization scheme, Geom

R. Eldan, Thin shell implies spectral gap up to polylog via a stochastic localization scheme, Geom. Funct. Anal. 23 (2013), no. 2, 532–569. DOI: https://doi.org/10.1007/ s00039-013-0214-y

2013

[7] [7]

D. S. Fisher and D. A. Huse, Equilibrium behavior of the spin-glass ordered phase , Phys. Rev. B 38 (1988), no. 1, 386–411. DOI: https://doi.org/10.1103/PhysRevB.38.386

doi:10.1103/physrevb.38.386 1988

[8] [8]

Georgii, Gibbs Measures and Phase Transitions , second ed., de Gruyter Studies in Math- ematics, vol

H.-O. Georgii, Gibbs Measures and Phase Transitions , second ed., de Gruyter Studies in Math- ematics, vol. 9, Walter de Gruyter, Berlin, 2011

2011