A central limit theorem for the stochastic cable equation

Soma Nishino

REVIEW 2 major objections 6 minor 10 references

A central limit theorem for the stochastic cable equation

T0 review · 2 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read This paper proves that the spatial average of the solution to a nonlinear stochastic cable equation—the standard linearized model of voltage along a neuron—is asymptotically normal, with a total-variation error of order $L^{-1/2}$ as the…

desk verdict Extends Pu's Malliavin–Stein CLT to the cable equation; the result is plausible and the affine-σ verification is useful, but a load-bearing uniform derivative bound is deferred and there are repairable algebraic slips. read the letter →

arxiv 2506.02755 v2 pith:A6SMAO3U submitted 2025-06-03 math.PR

Soma Nishino This is my paper

classification math.PR MSC 60H1560F0560H07

keywords stochasticcableequationcentrallimittheoremMalliavincalculusStein'smethodtotalvariationdistancefunctionalspace-timewhitenoisespatialaverage

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 establishes a quantitative central limit theorem for the spatial average of the solution to a one-dimensional nonlinear stochastic cable equation driven by multiplicative space-time white noise on an interval of length $L$. The main result, Theorem 1, shows that after centering and normalizing by its standard deviation, this spatial average is close to a standard Gaussian in total variation distance, with error at most $c(t)/\sqrt{L}$ for every fixed time $t$. This matters because the cable equation is the standard linearized model of electrical signal propagation along a neuron, and the spatial average is the natural observable when recording along the whole neuron. The paper also proves a functional version: the centered, length-normalized fluctuation process converges in law in $C([0,T])$ to a Gaussian process with an explicit exponential kernel. A technical assumption behind the functional CLT is verified for affine noise coefficients $\sigma(u)=\sigma_1 u+\sigma_0$, yielding an explicit limiting intensity $f_\sigma(t)$.

What carries the argument

The argument runs through the Malliavin–Stein method. The spatial average is written as a Skorohod integral, $F_L(t)=\delta(v_{L,t})$, with $v_{L,t}(s,y)=L^{-1}1_{(0,t)}(s)1_{[0,L]}(y)\sigma(u(s,y))I_0(t-s,y)$, where $I_0$ is the integral of the cable Green's function. Clark–Ocone's formula and the bound $d_{\mathrm{TV}}(F,N(0,1))\le 2\sqrt{\operatorname{Var}(\langle DF,v\rangle_H)}$ reduce the task to showing this variance decays like $L^{-3}$, which Proposition 4.1 does using a uniform Gaussian-type estimate on the Malliavin derivative of the solution (Lemma 2.2). The functional CLT then follows from tightness via an $L^1$-type increment bound and from covariance convergence via Assumption 1.

What would settle it

Simulate the affine case $\sigma(u)=u$ on a fixed time grid and estimate $\sqrt{L}\,d_{\mathrm{TV}}(F_L(t)/\sqrt{\operatorname{Var}(F_L(t))},N(0,1))$ for $L=10,10^2,10^3$; if the values grow rather than stay bounded, the claimed $O(L^{-1/2})$ rate is false. A direct check of Lemma 2.2 for $\alpha<0$ with large $|\alpha|$ would also settle whether the unproved bound holds.

Watch

Extended reading notes

Core claim

Theorem 1 proves that if $\sigma(1)\neq 0$, then for every fixed $t>0$ there is a constant $c(t)>0$ such that for all $L\ge 1$, $$d_{\mathrm{TV}}\left(\frac{F_L(t)}{\sqrt{\operatorname{Var}(F_L(t))}}, N(0,1)\right)\le \frac{c(t)}{\sqrt{L}},$$ where $F_L(t)$ is the centered spatial average of the mild solution. Theorem 2 shows that under Assumption 1—the convergence of the spatial average of $E[\sigma(u(t,x))^2]$ to an $L^1$ function $f_\sigma(t)$—the process $\sqrt{L}F_L(\cdot)$ converges in law in $C([0,T])$ to $\int_0^t e^{-\alpha(t-s)}\sqrt{f_\sigma(s)}\,dW_s$. Section 6 verifies Assumption 1 for affine $\sigma(u)=\sigma_1 u+\sigma_0$, giving an explicit integrable $f_\sigma(t)$.

Load-bearing premise

The load-bearing premise is Lemma 2.2, an unproved uniform Gaussian-type bound on how strongly the solution at one point responds to earlier noise; if it fails for some admissible $\alpha$, the $L^{-3}$ variance contraction that produces the $L^{-1/2}$ rate collapses.

Editorial extensions

If this is right

At any fixed observation time $t$, the centered and renormalized spatial average is asymptotically standard normal, with total-variation error bounded by an explicit constant $c(t)/\sqrt{L}$ for all $L\ge 1$.
The unnormalized fluctuation process has a Gaussian limit with covariance kernel $\int_0^{t_1\wedge t_2} e^{-\alpha(t_1+t_2-2s)}f_\sigma(s)\,ds$, so the leak rate $\alpha$ and the limiting noise profile $f_\sigma$ fully determine the memory structure of the limit.
For affine noise $\sigma(u)=\sigma_1 u+\sigma_0$, the function $f_\sigma(t)$ is computed explicitly from the Wiener chaos decomposition and is integrable, so the functional CLT applies to this nonlinear case.
Neumann, Dirichlet, and periodic boundary conditions all yield the same qualitative result: the averaging effect over the spatial interval is $L^{-1/2}$, not faster.
The condition $\sigma(1)\neq 0$ is exactly what guarantees the limiting variance is positive, so the normalization by $\sqrt{\operatorname{Var}(F_L(t))}$ is well defined.

Reading between the lines

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

Extending the computation of $f_\sigma$ from affine $\sigma$ to polynomial $\sigma$ of higher degree would likely require higher-order chaos terms, but the same series argument could produce an explicit limit and would be a direct test of Assumption 1.
Proposition 4.1 is stated uniformly in $t,\tau\in[0,T]$, so a strengthened version of Theorem 1 with $c(t)$ uniform over compact time intervals is within reach and would make the functional CLT a direct corollary of the total-variation bound.
In the neuron-cable interpretation, the $L^{-1/2}$ fluctuation scale predicts that longer neurons do not suppress electrical noise; the noise remains visible at a Gaussian scale, and its temporal correlation is shaped by the leakage constant $\alpha$.
A direct numerical check of $\sqrt{L}\,d_{\mathrm{TV}}$ for $\sigma(u)=u$ at increasing $L$ would show whether the $L^{-1/2}$ rate is sharp or whether a faster rate holds for this special case.

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

Hi — quick take on arXiv:2506.02755.

What's new: the paper proves a CLT with L^{-1/2} total-variation rate and an FCLT for the spatial average of the stochastic cable equation on [0,L] with Neumann/Dirichlet/periodic boundary conditions and multiplicative Lipschitz noise. Since the cable Green's function is just e^{-αt} times the heat kernel, this is a natural extension of Pu's heat-equation result; the main addition is the exponential decay factor, and the author handles it cleanly in the covariance asymptotics. The explicit verification of Assumption 1 for affine σ via Wiener chaos is a genuinely useful computation — it gives a closed-form fσ(t) and shows integrability.

Where it stands: the core outline is sound and follows the Huang–Nualart–Viitasaari/Pu framework. The covariance lower bound (Prop 3.1) and the convergence of the limiting covariance (Prop 3.2) are convincing.

Soft spots, in order of importance:

1. Lemma 2.2 — the uniform Gaussian-type bound on the Malliavin derivative — is stated for all α∈R without proof and deferred to Pu [P22]. But Pu only treats α=0. This lemma is the only input that gives the O(L^{-3}) contraction bound in Prop 4.1, which drives the L^{-1/2} TV rate. If the constant were L-dependent or the bound failed for α<0, the rate collapses. I think the lemma is very likely true — the same arguments should carry through with e^{|α|T} factors — but "very likely" is not a proof. The referee should require the proof or a precise statement in the literature.

2. Algebraic slips. In Prop 4.1's periodic case, the Green's function argument in the contraction integral has a typo, and the e^{-α(...)} exponent that comes out is off by an r. In Prop 5.1's Dirichlet case, the prefactor after integrating sin^2 should be 2/L, not 2L; as written the equality is wrong. Both are repairable and the bounds are conservative, but they need fixing before the proof is readable.

These are fixable, not fatal. The central argument holds up in outline, and the affine-σ verification is new and correct.

Bottom line: this paper deserves a serious referee. I'd send it to review with a request to prove or explicitly relocate Lemma 2.2 and to clean up the typos. It's a solid incremental contribution for anyone working on SPDE limit theorems or neuronal cable models.

Referee Report

2 major / 6 minor

Summary. The paper studies the one-dimensional stochastic cable equation ∂u/∂t = (β/2)∂²u/∂x² − αu + σ(u)Ẇ on [0,T]×[0,L] with Neumann, Dirichlet, or periodic boundary conditions, multiplicative space-time white noise, Lipschitz σ, and constant initial condition u₀=1. The main result, Theorem 1, states that for σ(1)≠0 the centered spatial average F_L(t), normalized by its standard deviation, converges to a standard normal in total variation distance with the rate O(L^{-1/2}) uniformly in L≥1. Theorem 2 states that, under a technical Assumption 1 on the L-average of E[σ(u(t,x))²], the process √L F_L(·) converges in law in C([0,T]) to the Wiener integral ∫₀^t e^{-α(t−s)}√fσ(s) dW_s. The paper verifies Assumption 1 in the affine case σ(u)=σ₁u+σ₀ using the Wiener chaos decomposition, obtaining an explicit expression and an L¹ bound for fσ.

Significance. The results, if fully established, are a useful extension of the Malliavin–Stein CLT machinery from the stochastic heat equation [HNV20, P22] to the cable equation with general α∈R and β>0. The total-variation rate uniform in L, the functional CLT under a checkable condition, and the explicit affine verification are all valuable. The derivation is non-circular: Assumption 1 is a verifiable property of the second moment, no parameters are fitted, and the affine computation computes fσ directly from the chaos expansion. The main weakness is that a central estimate, Lemma 2.2, is asserted without proof and is the only route to the O(L^{-3}) contraction in Proposition 4.1; the significance of Theorem 1 is therefore conditional on that estimate being supplied.

major comments (2)

[§2.3, Lemma 2.2] Lemma 2.2 is load-bearing for Theorem 1: Proposition 4.1 uses it to bound both Φ^(1)_L,t,τ and Φ^(2)_L,t,τ, and it is the only estimate producing the O(L^{-3}) contraction that is converted into the O(L^{-1/2}) total-variation rate in (19). However, the lemma is stated without proof for all α∈R, β>0 and all three boundary conditions, with constants uniform in L. The text says only that it 'can similarly be extended' from [P22], which treats the case α=0, β=1. The uniformity in L is exactly what the proof of Proposition 4.1 needs, and the periodic case is not an immediate consequence of [P22] as stated. The authors should provide a complete proof of Lemma 2.2, or a precise reference from which the α, β, and L dependence follows, before Theorem 1 can be considered established.
[Proof of Theorem 2, around (23)–(24)] The finite-dimensional convergence argument normalizes each F_L(t_i) by √Var(F_L(t_i)) and later rescales using the asymptotic variance from (23). This step assumes that Var(F_L(t_i)) ~ c_i/L with c_i>0, i.e., that ∫₀^{t_i} e^{-2α(t_i−s)}fσ(s) ds > 0. Assumption 1 alone does not exclude the degenerate case fσ≡0 on [0,t_i], or fσ vanishing on a set of positive measure for some t_i. In such cases the ratio in (23)–(24) is not justified. The authors should either add a positivity condition on fσ (e.g., fσ>0 a.e. on [0,T], or σ(1)≠0 together with a positivity argument) or handle the degenerate case separately; otherwise the proof of Theorem 2 is incomplete as written.

minor comments (6)

[Proposition 4.1, periodic case] In the periodic-case estimates for Φ^(1) and Φ^(2), the displayed exponent e^{-α(s₁+s₂−r)} should be e^{-α(s₁+s₂−2r)} by the semigroup property; the final bounds remain conservative and the L^{-3} rate is unaffected after adjusting the constant. Also, in the Φ^(2) display the second Green's function is written as G_{s₂−r}(y₂, r); it should be G_{s₂−r}(y₂, z).
[§2.3, Lemma 2.1] Lemma 2.1 is also stated with only a pointer to [P22] and the phrase 'similarly.' The L-uniform moment bound is used throughout the paper; a short proof or a precise reference with the α, β, L dependence would improve the completeness of the manuscript.
[Proposition 6.2] The sentence 'where f_k(t) is the limit in .' is missing the reference; it should read 'the limit in Proposition 6.1.'
[Keywords] The keyword 'Malliain calculus' contains a typo and should read 'Malliavin calculus.'
[Proof of Theorem 2] The notation F is used both for the random vector (F₁,...,F_m) and for the process F_L; renaming the vector, for example F^(L), would avoid confusion.
[Proposition 2.2] Proposition 2.2 is stated for m≥2, but the proof of Theorem 2 also uses the case m=1; this case is covered by Proposition 2.1, but the text should say so explicitly.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the CLT and functional CLT are derived from external Malliavin–Stein bounds and explicit covariance computations; Assumption 1 is a verifiable limit condition, not an encoding of Gaussianity.

full rationale

The paper's derivation chain is self-contained in the relevant sense: Theorem 1 follows from the Malliavin–Stein bound (Proposition 2.1), the stochastic Fubini representation (16), and the variance/covariance lower bounds obtained in Section 3 from the Green's function asymptotics. Proposition 4.1's L^{-3} estimate uses Lemma 2.2 to bound Malliavin derivatives of the solution, and the L^{-3} scaling emerges from integrating Gaussian heat kernels and the spatial domain of size L; it is not an assumption that the spatial average is Gaussian, nor is any fitted parameter renamed as a prediction. The technical Assumption 1 defines f_σ as a spatial limit of second moments; Theorem 2 then computes the covariance limit of sqrt(L)F_L as the integral of e^{-α(t1+t2-2s)} f_σ(s), so f_σ appears in the limit as the derived variance density, not as an input imposing Gaussianity. For the affine case, Section 6 independently verifies Assumption 1 via the Wiener chaos decomposition and explicit series computations, and the resulting f_σ is shown integrable. There are no self-citations that are load-bearing: the cited [P22] is by a different author and provides an external, published moment bound that the paper extends by assertion. Whether Lemma 2.2 really holds for all α∈R and β>0, and whether its constant is uniform in L, is a correctness/completeness risk rather than circularity: the paper's rate would fail if that external bound fails, but the paper does not define its conclusions in terms of its inputs. The algebraic slips noted by a skeptical reader, such as the periodic exponent in Proposition 4.1, do not constitute circular reasoning. Overall, no step in the claimed derivation reduces by construction to the theorem being proved.

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

The central claim rests on the unproved extension of Pu's uniform moment and Malliavin derivative bounds to arbitrary α,β, and on Assumption 1 for the functional version. No free parameters or invented entities appear; all constants are explicit bounds.

assumptions (5)

domain assumption Uniform L^k moment bound for the mild solution: sup_{L≥1} sup_{[0,T]×[0,L]} ||u(t,x)||_k < ∞ (Lemma 2.1)
Stated for α∈R, β>0 without proof, deferred to an analogous result of [P22] for α=0, β=1.
domain assumption Uniform Gaussian-type Malliavin derivative bound (Lemma 2.2)
Stated without proof, claimed to be a similar extension of [P22, Lemma 2.4]; used throughout Propositions 4.1 and 5.1.
domain assumption Assumption 1: existence of fσ(t) as the L→∞ limit of the spatial average of E[σ(u(t,x))²]
Technical condition for Theorem 2; verified for affine σ in Section 6 but not for general Lipschitz σ.
standard math Clark-Ocone formula and Malliavin-Stein inequalities (Propositions 2.1, 2.2)
Imported from [NP12] and [N06].
standard math Existence, uniqueness and L² bounds of the mild solution (Walsh [W86])
Cited in the introduction; defines u and justifies the Picard/chaos representation.

how reviews work

0 comments

Cite this review

Pith. "Pith review of A central limit theorem for the stochastic cable equation." pith.science (2026). https://pith.science/paper/A6SMAO3U

@misc{pith2026250602755,
  author       = {Pith},
  title        = {Pith review of: A central limit theorem for the stochastic cable equation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/A6SMAO3U}},
  note         = {Machine review of arXiv:2506.02755}
}

read the original abstract

We study one-dimensional nonlinear stochastic cable equations driven by a multiplicative space-time white noise. Using the Malliavin-Stein method, we prove a central limit theorem for the spatial average of the solution. The convergence is established in the total variation distance with mild conditions. We also establish a functional central limit theorem with a technical assumption. Furthermore, we show that this assumption holds in a special case.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

10 extracted references · 10 canonical work pages

[1]

Chen: Moments, intermittency, and growth indices for nonlinear stochastic PDE's with rough initial conditions, PhD thesis, No

L. Chen: Moments, intermittency, and growth indices for nonlinear stochastic PDE's with rough initial conditions, PhD thesis, No. 5712 , \'Ecole Polytechnique F\'ed\'erale de Lausanne, ( 2013 ). 342 pp

work page 2013
[2]

Huang, D

J. Huang, D. Nualart, L. Viitasaari: A central limit theorem for the stochastic heat equation, Stochastic Processes and their Applications, 130, 12 , 7170 -- 7184 , ( 2020 )

work page 2020
[3]

Kallianpur, J

G. Kallianpur, J. Xiong: Stochastic differential equations in infinite-dimensional spaces, IMS Lecture Notes Monogr. Ser., 26, ( 1995 )

work page 1995
[4]

Karatzas and S

I. Karatzas and S. E. Shreve: Brownian motion and Stochastic calculus, Springer, Science+Business Media, Inc. in 1998 , 2 nd ed

work page 1998
[5]

Nourdin and G

I. Nourdin and G. Peccati: Normal approximations with Malliavin calculus. From Stein's method to universality, Cambridge Tracts in Mathematics, 192 . Cambridge University Press, Cambridge, ( 2012 ). xiv+ 239 pp

work page 2012
[6]

Nualart: The Malliavin Calculus and Related Topics, Probability and its Applications (New York)

D. Nualart: The Malliavin Calculus and Related Topics, Probability and its Applications (New York). Springer-Verlag, Berlin, ( 2006 ). xiv+ 382 pp

work page 2006
[7]

F. W. J. Olver, D. W. Lozier, R. F. Boisvert, and C. W. Clark (eds.), NIST handbook of mathematical functions, U.S. Department of Commerce, National Institute of Standards and Technology, Washington, DC; Cambridge University Press, Cambridge, ( 2010 )

work page 2010
[8]

Pu: Gaussian fluctuation for spatial average of parabolic Anderson model with Neumann/Dirichlet/periodic boundary conditions, Trans

F. Pu: Gaussian fluctuation for spatial average of parabolic Anderson model with Neumann/Dirichlet/periodic boundary conditions, Trans. Amer. Math. Soc. 375, 2481 -- 2509 , ( 2022 )

work page 2022

Show all 10 references

[9]

J. B. Walsh: A stochastic model of neural response, Advances in Applied Probability, 13( 2 ), 231 -- 281 , ( 1981 )

1981
[10]

J. B. Walsh: An introduction to stochastic partial differential equations, École d'été de probabilités de Saint-Flour, XIV— 1984 pp. 265 -- 439 , Lecture Notes in Mathematics, vol. 1180 , Springer, Berlin, ( 1986 )

1984

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.

[1] [1]

Chen: Moments, intermittency, and growth indices for nonlinear stochastic PDE's with rough initial conditions, PhD thesis, No

L. Chen: Moments, intermittency, and growth indices for nonlinear stochastic PDE's with rough initial conditions, PhD thesis, No. 5712 , \'Ecole Polytechnique F\'ed\'erale de Lausanne, ( 2013 ). 342 pp

work page 2013

[2] [2]

Huang, D

J. Huang, D. Nualart, L. Viitasaari: A central limit theorem for the stochastic heat equation, Stochastic Processes and their Applications, 130, 12 , 7170 -- 7184 , ( 2020 )

work page 2020

[3] [3]

Kallianpur, J

G. Kallianpur, J. Xiong: Stochastic differential equations in infinite-dimensional spaces, IMS Lecture Notes Monogr. Ser., 26, ( 1995 )

work page 1995

[4] [4]

Karatzas and S

I. Karatzas and S. E. Shreve: Brownian motion and Stochastic calculus, Springer, Science+Business Media, Inc. in 1998 , 2 nd ed

work page 1998

[5] [5]

Nourdin and G

I. Nourdin and G. Peccati: Normal approximations with Malliavin calculus. From Stein's method to universality, Cambridge Tracts in Mathematics, 192 . Cambridge University Press, Cambridge, ( 2012 ). xiv+ 239 pp

work page 2012

[6] [6]

Nualart: The Malliavin Calculus and Related Topics, Probability and its Applications (New York)

D. Nualart: The Malliavin Calculus and Related Topics, Probability and its Applications (New York). Springer-Verlag, Berlin, ( 2006 ). xiv+ 382 pp

work page 2006

[7] [7]

F. W. J. Olver, D. W. Lozier, R. F. Boisvert, and C. W. Clark (eds.), NIST handbook of mathematical functions, U.S. Department of Commerce, National Institute of Standards and Technology, Washington, DC; Cambridge University Press, Cambridge, ( 2010 )

work page 2010

[8] [8]

Pu: Gaussian fluctuation for spatial average of parabolic Anderson model with Neumann/Dirichlet/periodic boundary conditions, Trans

F. Pu: Gaussian fluctuation for spatial average of parabolic Anderson model with Neumann/Dirichlet/periodic boundary conditions, Trans. Amer. Math. Soc. 375, 2481 -- 2509 , ( 2022 )

work page 2022