Superdiffusive Scaling Limits for the Symmetric Exclusion Process with Slow Bonds

Dirk Erhard; Tertuliano Franco; Tiecheng Xu

REVIEW 1 major objections 5 minor 5 references

Superdiffusive Scaling Limits for the Symmetric Exclusion Process with Slow Bonds

T0 review · 1 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read This paper proves that when the number of equally spaced slow bonds grows with the system size, the symmetric exclusion process, run at the critical superdiffusive speed $k^2 n^{1+\beta}$, converges to the homogeneous heat equation on the…

desk verdict Three genuinely new superdiffusive scaling limits for SSEP with slow bonds, with a sound relative-entropy proof modulo a sign typo in Lemma 4.8 that the stress-test over-read. read the letter →

arxiv 2412.04396 v1 pith:6UMDXT54 submitted 2024-12-05 math.PR

Dirk Erhard , Tertuliano Franco , Tiecheng Xu This is my paper · ORCID

classification math.PR MSC 60K35

keywords symmetricexclusionprocessslowbondssuperdiffusivescalinghydrodynamiclimitrelativeentropymethodheatequationtorus

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 studies the symmetric exclusion process on a discrete torus with $nk$ sites in which $k$ equally spaced bonds, one at the end of each block of $n$ sites, are slowed by a factor $n^{-\beta}$, with $\beta>1$ fixed. It asks what happens macroscopically when the process is run faster than the usual diffusive scaling, so that each block equilibrates internally before mass can cross a slow bond. The answer is a family of three superdiffusive limits: at a subcritical speed the density freezes at the initial mass of each block; at the critical speed $k^2 n^{1+\beta}$ the block densities obey the discrete heat equation when $k$ is fixed; and when $k$ also grows to infinity the averaged block densities obey the continuous heat equation on the torus, with no boundary conditions. The last result is the paper's central claim: a growing number of weak barriers leaves no trace in the continuum limit, and the evolution is the same as for the homogeneous exclusion process.

What carries the argument

The central object is the averaged empirical measure $\pi^n_t$, which assigns to each macroscopic point $i/k$ the average occupation of the whole block of $n$ sites between two slow bonds; this is the quantity whose limit the paper characterises. The proof for $k\to\infty$ is carried by a time-dependent reference measure: a Bernoulli product measure $\nu^n_t$ whose parameter is $\rho_t(i/k)$ on block $i$, together with a refined relative-entropy inequality that bounds the derivative of $H(\mu^n_t|\nu^n_t)$ by the entropy itself plus negligible terms. The estimates rest on subgaussian bounds and $\ell$-dependence for the centred occupation variables $w_t(x)=(\eta_t(x)-\rho_t(i/k))/(\rho_t(i/k)(1-\rho_t(i/k)))$, which control the replacement errors inside blocks and at slow bonds. For fixed $k$, an entropy method with a replacement lemma identifies the limit points, and tightness comes from a martingale whose quadratic variation vanishes.

What would settle it

Simulate the process with $\alpha=1$, $\beta=2$, $k=\lfloor n^{0.6}\rfloor$, torus size $nk$, initial profile $\gamma(u)=\tfrac12+\tfrac14\sin(2\pi u)$, and acceleration $k^2 n^3$; at a fixed time measure the gap $|\langle\pi^n_t,G\rangle-\int_{\mathbb{T}}G(u)\rho_t(u)\,du|$ for a smooth test function $G$. If this gap does not converge to $0$ as $n\to\infty$, Theorem 2.4 would be false. A sharper test of the proof's boundary is to repeat with $\gamma(u)=u$, which touches $0$ and $1$: the paper's entropy bound no longer follows, so convergence or failure here isolates the role of the non-degeneracy assumption.

Watch

Extended reading notes

Core claim

On the torus with $nk$ sites and $k$ equally spaced slow bonds of strength $\alpha n^{-\beta}$ with $\beta>1$, consider the averaged empirical measure that replaces each block of $n$ sites by a single atom at $i/k$ carrying the block's average occupation. The paper proves three scaling limits under superdiffusive accelerations of the form $k^2 n^{2+\theta}$. For $k$ fixed and $0<\theta<\beta-1$, the measure converges in probability to the frozen profile $\sum_{i=0}^{k-1} \bar\gamma(i)\delta_{i/k}$, with $\bar\gamma(i)=k\int_{i/k}^{(i+1)/k}\gamma(u)\,du$, so each block keeps its initial mass and there is no time evolution. For $k$ fixed and $\theta=\beta-1$, meaning the time scale is $k^2 n^{1+\beta}$, the limit is $\sum_{i=0}^{k-1}\rho_t(i)\delta_{i/k}$, where $\rho_t$ solves the discrete heat equation $\partial_t\rho_t=\alpha\Delta_k\rho_t$ with initial datum $\bar\gamma(i)$. Finally, if $k=k(n)\uparrow\infty$ at the same critical time scale, the averaged empirical measure converges to the strong solution $\rho_t$ of the continuous heat equation $\partial_t\rho_t=\alpha\Delta\rho_t$ on the torus with initial datum $\gamma$, i.e. the slow bonds become macroscopically invisible. The proof uses a relative-entropy method with a time-dependent Bernoulli reference measure whose parameter at site $x$ is $\rho_t(i/k)$ inside block $i$.

Load-bearing premise

The initial density profile must be smooth and bounded strictly away from $0$ and $1$; if $\gamma$ touches the empty or full state, the relative-entropy bound $H(\mu^n|\nu^n_0)\le Cn$ used throughout the proof breaks down, so Theorem 2.4 is not established for such profiles.

Editorial extensions

If this is right

For fixed $k$, the critical speed $k^2 n^{1+\beta}$ makes each slow-bond rate exactly comparable to the macroscopic clock, so the $k$ block masses undergo a discrete heat equation with diffusivity $\alpha$; slower speeds leave those masses frozen.
When $k$ grows without bound at the same speed, the discrete Laplacian on $k$ sites converges to the continuous Laplacian on the torus, and the limiting evolution is the homogeneous heat equation with no boundary conditions.
At any superdiffusive acceleration $k^2 n^{2+\theta}$ with $\theta>0$, the system equilibrates instantly inside each block, so the only macroscopic degrees of freedom are the $k$ block averages.
The replacement lemma for the fixed-$k$ method holds as long as $k=o(n^{(\beta-1)/2})$, while the refined relative-entropy method used for Theorem 2.4 works for any $k\to\infty$, so the two regimes are proved by different arguments.
The result identifies a full phase diagram in the time-scale exponent and the number of boxes: subcritical freezing, critical discrete heat flow, and the $k\to\infty$ homogeneous continuum limit.

Reading between the lines

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

Read as a homogenisation statement, the $k\to\infty$ limit says that weakly slowed bonds of strength $n^{-\beta}$ and density $1/k$, with $\beta>1$, do not lower the macroscopic diffusivity: the effective equation is the bare heat equation with coefficient $\alpha$.
The freezing regime suggests a sharp transition at $\theta=\beta-1$, where inter-box flux first appears; this paper establishes the critical case, and one might expect the subcritical and critical regimes to match continuously as $\theta\uparrow\beta-1$, though that matching is not proved here.
The non-degeneracy assumption $\varepsilon_0<\gamma<1-\varepsilon_0$ is likely an artefact of the entropy method; a natural test is whether profiles touching $0$ or $1$ obey the same hydrodynamic limit, possibly requiring different control terms.
Because the discrete-to-continuum step only requires $k\to\infty$ and the heat-equation solution is strong, the result should extend to more general initial profiles by approximation whenever the entropy bound can be replaced, but this extension is not proved in the paper.

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

Read the paper. The three scaling limits are real and new: for fixed k, subcritical superdiffusive speed gives frozen box averages; at k^2 n^{1+β} you get the discrete heat equation; and as k→∞ the averaged empirical measure converges to the continuous heat equation with no boundary conditions, so the slow bonds become macroscopically invisible. The main technical contribution is using the Jara–Menezes refined relative entropy method for k→∞, which removes the replacement-lemma restriction k=o(n^{(β−1)/2}) from the earlier entropy approach. That is a genuine improvement and the proof is largely convincing.

I checked the structure of Proposition 4.2 and Corollary 4.3. The bound H(µ^n_0|ν^n_0)≤Cn gives H(µ_t|ν_t)=o(kn) once k→∞, and the error terms in Lemmas 4.6 and 4.7 are indeed o(kn) with the choice ℓ=n√k. The subgaussian estimates are standard from Appendix F of Jara–Menezes.

The soft spots are minor but worth fixing. First, the abstract silently reparametrizes the time scale: the theorem statements use k^2 n^{2+θ} with θ∈(0,β−1), while the abstract writes k^2 n^θ with θ∈(2,1+β). That will confuse readers. Second, there is a typo in the proof of Lemma 4.8. The display (4.23) shows a k^2 n^{β−1} prefactor multiplying {H+kn log3/ℓ}. If that were the actual bound, the Gronwall step in Proposition 4.2 would indeed not close for arbitrary k→∞. But plugging the stated choice δ=k^2 n^{β−1}ℓ^{−1}/8 into (4.24) gives k^{-2} n^{-(β−1)} instead, which is harmless and consistent with the lemma's statement. So the stress-test concern does not survive contact with the algebra; the authors just need to correct the exponent in (4.23). Also note that Theorem 2.4 assumes the initial profile is uniformly separated from 0 and 1; that ellipticity is used in Corollary 4.3 and in the subgaussian bounds, and it should be flagged in the statement.

Verdict: this deserves a serious referee. The results are a natural completion of the Franco–Gonçalves–Neumann phase diagram at superdiffusive scales, and the proof is sound modulo the typo and a clarity issue. I would bring it to a reading group and cite it if I worked on slow-bond systems. Recommendation: send it to a good probability journal; the authors should fix the sign typo and align the abstract with the theorem statements.

Referee Report

1 major / 5 minor

Summary. The paper studies the symmetric simple exclusion process on a discrete torus of size nk with k equally spaced slow bonds of strength n^{-β}, for β>1, and establishes hydrodynamic limits at superdiffusive time scales. The main results are: for fixed k and time scale k^2 n^{2+θ} with θ<β−1, the averaged empirical measure converges to a time-independent box-average profile (Theorem 2.2); for fixed k and time scale k^2 n^{1+β}, it converges to the solution of the discrete heat equation on the k boxes (Theorem 2.3); and for k→∞ at the critical scale k^2 n^{1+β}, it converges to the solution of the continuous heat equation on the torus (Theorem 2.4). The fixed-k results are proved by Varadhan's entropy method and a replacement lemma, while the k→∞ result is proved by the refined relative entropy method of Jara and Menezes.

Significance. If the results hold, the paper provides a complete phase diagram for the superdiffusive scaling limits of the slow-bond exclusion process, including the striking conclusion that when the number of slow bonds diverges, the slow bonds become macroscopically invisible and the effective evolution is the homogeneous heat equation. The paper is clearly organized, the fixed-k proofs are detailed and follow established techniques, and the use of the Jara–Menezes relative entropy method is appropriate and nontrivial. The main theorems are plausible and of genuine interest to the hydrodynamic-limit community. However, one load-bearing estimate in the proof of Theorem 2.4 (Lemma 4.8) is misstated in a way that currently prevents the Gronwall argument from closing; the issue appears to be a typo, but it must be corrected before the theorem can be considered proven.

major comments (1)

[§4.3, Lemma 4.8 (Eq. (4.23))] The proof of Lemma 4.8 derives the intermediate estimate (4.23) with the prefactor C(ε0)||G||^2_∞ k^2 n^{β−1} multiplying {H(µ^n_t|ν^n_t) + kn log3/ℓ}. With the choice ℓ = n√k, adopted in Lemma 4.7, this term is of order k^2 n^{β−1}H + k^{3/2} n^{β−1}, which is not o(kn) for arbitrary k→∞; for example, k = log n, β > 1, and H of order n give k^2 n^β, whose ratio to kn diverges. The printed estimate therefore does not imply the lemma's stated bound (4.18), and the Gronwall step in Proposition 4.2, which requires an o(kn) remainder, is not justified by the text. The derivation from (4.24) with δ = k^2 n^{β−1}ℓ^{-1}/8 actually yields the prefactor k^{-2} n^{-(β−1)} on the H-term, suggesting a sign error in the exponent in (4.23); the authors should correct this estimate and confirm that the final entropy bound closes.

minor comments (5)

[Abstract and Section 1] The abstract describes the time scale as k^2 n^θ with θ∈(2,1+β), while Theorem 2.2 and the introduction use k^2 n^{2+θ} with θ∈(0,β−1); the relation between the two parametrizations should be stated explicitly or the notation unified.
[§4.1, after Eq. (4.6)] The sentence 'where we have used above that ρ is the solution of the heat equation given in (2.4)' should refer to equation (2.7), the continuous heat equation.
[Lemma 4.12] The statement says the constant is C(ε0, k), but the proof concludes with C(ε0, κ); the dependence on k is presumably a typo and should be corrected.
[Theorem 2.4] The assumptions that γ is smooth and satisfies ε0<γ<1−ε0 are stated only in the paragraph before the theorem; they should be included in the theorem statement, since the proof relies on them for the subgaussian estimate (5.1) and the initial entropy bound (4.10).
[Proof of Proposition 2.1] The proposition explicitly allows k→∞, but the proof in Section 3.1 is written as if k is fixed; the authors should indicate briefly why the same estimates hold for k=k(n)↑∞.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the superdiffusive heat-equation limit is derived from the generator via relative entropy, not imposed by construction; self-citations are contextual only.

full rationale

Theorem 2.4 is proved by the standard relative-entropy strategy: a time-dependent reference measure νn_t is built from ρt, the solution of the heat equation (2.7), and the proof establishes H(µn_t|νn_t)=o(kn) (Corollary 4.3) using Lemma 4.1, which is cited from the external work of Jara and Menezes [4]. The limit ρt is a candidate profile, not a fitted parameter; the nontrivial content is the entropy-production estimate, which is computed from the generator Ln and the conductances ξ (equations (4.3)-(4.9)) and controlled by the Dirichlet form and subgaussian estimates. No step redefines the conclusion as an assumption. The self-citations to [2] and [3] appear only in the introduction as background on the diffusive phase transition and are not used in the proofs of Theorems 2.2-2.4, whose martingale-problem and relative-entropy arguments are self-contained apart from the external tools in [4] and [5]. The appendix's subgaussian lemmas are also imported from [4], an external source, so there is no self-citation chain. The skeptic's concern about the prefactor in (4.23) is a potential correctness gap (an incomplete estimate), not a circular reduction: an erroneous prefactor would make the proof not close, but it would not make the claimed limit equal to the input by definition. Accordingly, no circular step is identified.

Assumptions & free parameters 0 free parameters · 3 assumptions · 0 invented entities

No fitted parameters, no invented entities. α, β, k, n, θ are model inputs set by the problem, not tuned to make the proof work. The imported lemmas from [4] are the only unproved external inputs.

assumptions (3)

domain assumption Refined Yao's inequality (Lemma 4.1, quoted from Jara-Menezes [4, Lemma A.1])
Bounds ∂_t H(µ^n_t|ν^n_t) by a Dirichlet form plus an explicit functional; this is the starting point of Proposition 4.2 and is used as a black box.
domain assumption Subgaussian and ℓ-dependence lemmas (Hoeffding, Lemma 5.2, Lemma 5.3, Lemma 5.4 from Appendix F of [4])
Control the fluctuation terms (4.7)-(4.9) in Lemma 4.8 and Lemma 4.11; these estimates are quoted without proof.
standard math Maximum principle and uniform smoothness of solutions to ∂_tρ=αΔρ on the torus
Guarantees ε0<ρ_t<1-ε0, which makes the random variables w_t(x) uniformly subgaussian and the initial entropy H(µ^n|ν^n_0) of order n; invoked after (2.7) and in Corollary 4.3.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Superdiffusive Scaling Limits for the Symmetric Exclusion Process with Slow Bonds." pith.science (2026). https://pith.science/paper/6UMDXT54

@misc{pith2026241204396,
  author       = {Pith},
  title        = {Pith review of: Superdiffusive Scaling Limits for the Symmetric Exclusion Process with Slow Bonds},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6UMDXT54}},
  note         = {Machine review of arXiv:2412.04396}
}

abstract

In \cite{fgn1}, the hydrodynamic limit in the diffusive scaling of the symmetric simple exclusion process with a finite number of slow bonds of strength $n^{-\beta}$ has been studied. Here $n$ is the scaling parameter and $\beta>0$ is fixed. As shown in \cite{fgn1}, when $\beta>1$, such a limit is given by the heat equation with Neumann boundary conditions. In this work, we find more non-trivial super-diffusive scaling limits for this dynamics. Assume that there are $k$ equally spaced slow bonds in the system. If $k$ is fixed and the time scale is $k^2n^\theta$, with $\theta\in (2,1+\beta)$, the density is asymptotically constant in each of the $k$ boxes, and equal to the initial expected mass in that box, i.e., there is no time evolution. If $k$ is fixed and the time scale is $k^2n^{1+\beta}$, then the density is also spatially constant in each box, but evolves in time according to the discrete heat equation. Finally, if the time scale is $k^2n^{1+\beta}$ and, additionally, the number of boxes $k$ increases to infinity, then the system converges to the continuous heat equation on the torus, with no boundary conditions.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

5 extracted references · 4 canonical work pages

[1]

Boucheron, G

S. Boucheron, G. Lugosi, and P. Massart. Concentration inequalities. Oxford University Press, Oxford, 2013. A nonasymptotic theory of independence, With a foreword by Mi chel Ledoux

work page 2013
[2]

Franco, P

T. Franco, P. Gonçalves, and A. Neumann. Hydrodynamical behavior of symmetric exclusion with slow bonds. Ann. Inst. H. Poincaré Probab. Statist. , 49(2):402–427, 05 2013

work page 2013
[3]

Franco, P

T. Franco, P. Gonçalves, and A. Neumann. Phase transitio n of a heat equation with Robin’s boundary condi- tions and exclusion process. Trans. Amer . Math. Soc., 367(9):6131–6158, 2015

work page 2015
[4]

Jara and O

M. Jara and O. Menezes. Non-equilibrium ﬂuctuations of i nteracting particle systems. ArXiv, 2018. https://arxiv.org/abs/1810.09526

arXiv 2018
[5]

Kipnis and C

C. Kipnis and C. Landim. Scaling limits of interacting particle systems , volume 320 of Grundlehren der mathe- matischen Wissenschaften. Springer-Verlag Berlin Heidelberg, 1st edition, 1999. UFBA, I NSTITUTO DE MATEMÁTICA , C AMPUS DE ONDINA , A V. A DHEMAR DE BARROS , S/N. CEP 40170-110, SALVADOR , B RAZIL Email address: erharddirk@gmail.com UFBA, I NSTI...

work page 1999

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.

[1] [1]

Boucheron, G

S. Boucheron, G. Lugosi, and P. Massart. Concentration inequalities. Oxford University Press, Oxford, 2013. A nonasymptotic theory of independence, With a foreword by Mi chel Ledoux

work page 2013

[2] [2]

Franco, P

T. Franco, P. Gonçalves, and A. Neumann. Hydrodynamical behavior of symmetric exclusion with slow bonds. Ann. Inst. H. Poincaré Probab. Statist. , 49(2):402–427, 05 2013

work page 2013

[3] [3]

Franco, P

T. Franco, P. Gonçalves, and A. Neumann. Phase transitio n of a heat equation with Robin’s boundary condi- tions and exclusion process. Trans. Amer . Math. Soc., 367(9):6131–6158, 2015

work page 2015

[4] [4]

Jara and O

M. Jara and O. Menezes. Non-equilibrium ﬂuctuations of i nteracting particle systems. ArXiv, 2018. https://arxiv.org/abs/1810.09526

arXiv 2018

[5] [5]

Kipnis and C

C. Kipnis and C. Landim. Scaling limits of interacting particle systems , volume 320 of Grundlehren der mathe- matischen Wissenschaften. Springer-Verlag Berlin Heidelberg, 1st edition, 1999. UFBA, I NSTITUTO DE MATEMÁTICA , C AMPUS DE ONDINA , A V. A DHEMAR DE BARROS , S/N. CEP 40170-110, SALVADOR , B RAZIL Email address: erharddirk@gmail.com UFBA, I NSTI...

work page 1999