Gaussian deconvolution on $\mathbb R^d$ with application to self-repellent Brownian motion

Yucheng Liu

Gaussian deconvolution on $\mathbb R^d$ with application to self-repellent Brownian motion

T0 review · 0 major / 4 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read A general deconvolution theorem on R^d yields the Laplacian Green function decay |x|^{-(d-2)}, and the result pins down the critical two-point function of self-repellent Brownian motion in d>4.

desk verdict Clean R^d deconvolution theorem with a real application; one proof step in Lemma 3.1 is under-justified but patchable. read the letter →

arxiv 2411.16058 v1 pith:ZVKIYIKH submitted 2024-11-25 math.PR math-phmath.MP

Yucheng Liu This is my paper · ORCID

classification math.PRmath-phmath.MP MSC 60K3542B1060J6582B41

keywords Gaussiandeconvolutioncriticaltwo-pointfunctionself-repellentBrownianmotionlaceexpansioninfraredboundFouriertransformanisotropicGreendecay

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 proves a deconvolution theorem for the continuum equation $(\delta-J)*G=g$ on $\mathbb R^d$, $d>2$. If $J$ and $g$ are even, decay fast enough, and the Fourier transform of $J$ satisfies an infrared bound with $\hat J(0)=1$, then the solution $H=G-g$ decays at infinity like $(x\cdot\Sigma^{-1}x)^{-(d-2)/2}$ up to an explicit constant, with $\Sigma$ the second-moment matrix of $J$. This is the same power law as the Green function of the Laplace operator, extended from the lattice to the continuum and to anisotropic kernels. The paper applies the theorem to self-repellent Brownian motion in $d>4$, upgrading a previously known upper bound to the asymptotic formula $G_{\alpha,\lambda_c(\alpha)}(x)\sim c_d|x|^{-(d-2)}$ with $c_d=a_d(1+O(\alpha))$.

What carries the argument

The argument compares the true solution with the critical two-point function $C$ of a Gaussian random walk. Let $D$ be the centered Gaussian density with covariance matrix $\Sigma$; the covariance matrix is chosen to match the second moments of $J$, and the decomposition $H=\hat g(0)C+f$ isolates the leading decay. In Fourier space the remainder is $\hat f=\hat E/((1-\hat D)(1-\hat J))$, with $E=(g*J-g*J*D)-\hat g(0)(D*D-D*D*J)$; matching $\Sigma$ to $J$'s second moments makes the second derivatives of $\hat E$ vanish at the origin, softening the singularity of $\hat f$ compared with $\hat C$. Infrared bounds on $\hat D$ and $\hat J$ control the denominators, and weak derivatives combined with a standard integrability inequality show that $\hat f$ is $d-2$ times weakly differentiable and integrable, so the Riemann-Lebesgue lemma forces $f(x)=o(|x|^{-(d-2)})$.

What would settle it

Choose $d=3$, an even compactly supported $J$ with $\hat J(0)=1$, the infrared bound (1.7), and nondegenerate covariance matrix $\Sigma$, and an even Schwartz $g$ with $\int g=1$; evaluate the Fourier integral (1.10) along a ray. If $|x|H(x)$ does not converge to $a_3/\sqrt{\det\Sigma}$, the asymptotic (1.11) fails, and one such pair would refute the theorem.

Watch

Extended reading notes

Core claim

The central claim is Theorem 1.3: under Assumption 1.1, the Fourier-integral solution $H$ to $(\delta-J)*H=J*g$ obeys $H(x)=a_d(\int g)(\sqrt{\det\Sigma})^{-1}(x\cdot\Sigma^{-1}x)^{-(d-2)/2}+o(|x|^{-(d-2)})$, where $a_d=\Gamma((d-2)/2)/(2\pi^{d/2})$ and $\Sigma=\mathrm{diag}(\int x_i^2J(x)\,dx)$. If $g$ itself is $o(|x|^{-(d-2)})$, the solution $G=H+g$ of the original equation shares the same asymptotics. The theorem is then applied to self-repellent Brownian motion: in $d>4$ and for sufficiently small repulsion $\alpha$, the critical two-point function satisfies $G_{\alpha,\lambda_c(\alpha)}(x)\sim c_d|x|^{-(d-2)}$ with $c_d=a_d(1+O(\alpha))$, matching the Laplace Green function. The proof extends a recent lattice deconvolution theorem to the continuum, requiring only even symmetry rather than lattice symmetry, and formulates hypotheses as moment conditions.

Load-bearing premise

The entire proof rests on the kernel $J$ having Fourier transform equal to $1$ at zero and dipping quadratically near zero; in the Brownian-motion application this property is established only when the repulsion strength $\alpha$ is sufficiently small.

Editorial extensions

If this is right

Any continuum model whose lace-expansion kernel satisfies Assumption 1.1 inherits $|x|^{-(d-2)}$ decay of its critical two-point function, with anisotropy encoded in $\Sigma$.
For self-repellent Brownian motion in $d>4$ with $\alpha$ small, the critical two-point function is asymptotic to $a_d(1+O(\alpha))|x|^{-(d-2)}$, upgrading the Gaussian-domination upper bound to a two-sided exact asymptote.
Because the assumptions allow signed $J$, the theorem covers lace-expansion kernels that are not probability distributions.
If also $g=o(|x|^{-(d-2)})$, the solution $G$ of the original convolution equation inherits the same asymptotic as the auxiliary solution $H$.
The moment-based hypotheses are weaker and often easier to verify than the polynomial-decay hypotheses used in earlier lattice deconvolution theorems.

Reading between the lines

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

A natural next target is the subcritical case $\hat J(0)<1$, where the same Fourier estimates should produce exponential decay with a rate depending on $J$; this would matter for finite-size scaling of continuum models on a torus.
The anisotropic conclusion suggests that for the random connection model with a connection function lacking lattice symmetry, the critical connection probability should decay as $(x\cdot\Sigma^{-1}x)^{-(d-2)/2}$ with the anisotropy carried by $\Sigma$; a direct comparison with simulations would test this.
The proof hints at a bootstrap: starting from a rough upper bound and iterating the deconvolution statement could reach the asymptotic without relying on an external Gaussian-domination bound, making the smallness condition on $\alpha$ quantitative.
Because only evenness and moment conditions are used, any candidate continuum model reduces to a check of the infrared bound (1.7); that single inequality is the practical bottleneck for new applications.

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

The paper proves a genuinely useful deconvolution theorem on R^d: under moment and infrared-bound conditions, the solution to (δ−J)*G = g decays like the Green's function, with an anisotropic diagonal covariance Σ. This extends the Liu–Slade Z^d result to the continuum, weakens decay to moments, and handles anisotropy. The application to self-repellent Brownian motion in d>4 gives the first sharp asymptotic for the critical two-point function, |x|^{-(d-2)}. That application is the real payoff and looks solid, since the inputs from Bolthausen–Koenig–Mukherjee provide far stronger decay than Assumption 1.1 requires.

I agree with your accept-leaning read. The Fourier strategy is standard and the proof is detailed. The main soft spot is exactly what your stress-test flagged: Lemma 3.1. For the case 3 ≤ |γ| ≤ d−2, the paper needs |x|^{d−2}E ∈ L^p with p < d/4 to get the strict improvement in Lemma 2.2, then the ε' in the denominator. It says this follows from Assumption 1.1 and Young, but Young alone, applied to the given p near d/4, can give a convolution exponent r > d/4 (e.g., d=5, p=1.2 gives r=1.5). That is a genuine gap in the written proof.

That said, the gap is easily closed: L^p moment bounds imply the same bound in L^{p'} for any p' ≤ p, and one may choose p' small enough that 2/p' − 1 > 4/d, so the convolution lands in L^r with r < d/4. The paper should spell this out. As written, the proof of Theorem 1.3 is incomplete for the full range of Assumption 1.1, but the missing step is short, not a counterexample, and Theorem 1.4 is unaffected because the application uses (4.3).

Minor note: the paper leans on a very recent preprint [2] for the application, but that is appropriate and not a circularity.

Who is this for? Anyone working with lace expansion on continuous space—random connection models, self-repellent Brownian motion, possibly torus models. It deserves a serious referee. I would send it out and ask for a revision that fixes Lemma 3.1's crowding step. The central idea is sound and the application is a nice step forward.

Referee Report

0 major / 4 minor

Summary. The paper proves a continuum, anisotropic analogue of the authors' lattice deconvolution theorem. Under Assumption 1.1 (evenness, L^1∩L^2 and moment conditions on J and g, plus an infrared lower bound on 1−Ĵ), the Fourier-integral solution H of (δ−J)∗H=J∗g is shown to satisfy the asymptotic H(x)=a_d(∫g)/√detΣ (x·Σ^{-1}x)^{-(d-2)/2}+o(|x|^{-(d-2)}), where Σ is the diagonal matrix of second moments of J. The proof decomposes Ĥ as ĝ(0) times the Gaussian random-walk two-point function plus a remainder, and uses weak-derivative estimates in Fourier space to show the remainder decays faster. The paper then verifies Assumption 1.1 for the self-repellent Brownian motion in d>4 using the lace-expansion bounds of [2], yielding G_{α,λ_c(α)}(x)∼c_d|x|^{-(d-2)} with c_d=a_d(1+O(α)).

Significance. This is a clean and useful extension of deconvolution theory from Z^d to R^d. It removes lattice symmetry, permits anisotropic decay through a J-dependent covariance matrix, and states hypotheses in moment form, which is convenient for lace-expansion applications. The proof is elementary Fourier analysis and is mostly self-contained; no parameter is fitted to the target decay, and the asymptotic constant is explicit. The application to self-repellent Brownian motion is honest: it relies on the external model-specific bounds of [2], not on the theorem being proved. I also checked the delicate moment estimate for E in Lemma 3.1: the concern that Young's inequality alone gives too weak an exponent does not land, because using the unweighted L^1 factor in each convolution gives |x|^{d−2}(f∗g)∈L^p whenever |x|^{d−2}f and |x|^{d−2}g lie in L^p and f,g∈L^1. The proof would nonetheless benefit from spelling this out.

minor comments (4)

[Section 3, Lemma 3.1, case 3≤|γ|≤d−2] The sentence 'The required moment conditions on E follow from Assumption 1.1 and Young's convolution inequality' is terse: a naive Young bound with both factors weighted gives |x|^{d−2}(g∗J)∈L^r with 1/r=2/p−1, which can exceed d/4 within Assumption 1.1. Please add the standard weighted-convolution argument using an unweighted L^1 factor, i.e. |x|^A(f∗g) is controlled by (|x|^A f)∗g + f∗(|x|^A g). This is a clarity issue, not a mathematical error.
[Section 3, proof of Proposition 1.6, sentence before (3.13)] The sentence 'we always have at least two factors of derivatives of ĝ, Ĵ, or D̂' is not literally true: if all of α_2 falls on ĝ, the expanded term has only one differentiated factor, with the other factors undifferentiated. The subsequent use of Lemma 2.3 remains valid with zero-derivative indices, so please rephrase to avoid confusion.
[Section 3, Lemma 3.1] There is a typo in the lemma statement: 'Assumpton 1.1' should be 'Assumption 1.1'.
[Section 4, around (4.7)] To help the reader, spell out the cancellation from det Σ=σ^{2d} and (x·Σ^{-1}x)^{-(d-2)/2}=σ^{d-2}|x|^{-(d-2)} that yields the constant a_d/σ^2 in (4.7).

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: Theorem 1.3 is a forward Fourier proof from explicit hypotheses, and Theorem 1.4 is a verification of those hypotheses using external results from [2].

full rationale

The derivation is self-contained in the sense that the target decay is not inserted by definition or by fitting. In (1.25), the Fourier solution is decomposed as H_hat = g_hat(0) D_hat^2/(1-D_hat) plus a remainder f_hat, with the Gaussian kernel D chosen by (1.27) so that the second moments of E vanish. The leading constant a_d integral(g)/sqrt(det Sigma) and the exponent (d-2) come from the explicit Gaussian random-walk computation in Lemma 1.5, not from matching the conclusion. Proposition 1.6, which makes the remainder o(|x|^{-(d-2)}), is proved from Assumption 1.1 using the regularity lemmas of Section 2; the citations to [22] are for standard weak-derivative calculus and do not carry the theorem's conclusion. The application in Theorem 1.4 verifies Assumption 1.1 for J = g = lambda_c phi_1 + Pi_lambda_c using the lace expansion and bounds of [2], which is external prior work, and then invokes Theorem 1.3. The infrared bound (1.7) is an explicit hypothesis on J, not a renamed version of the target decay, and the constant c_d = a_d(1+O(alpha)) is obtained from Sigma = (1+O(alpha)) Id in (4.7)-(4.8). I found no step where a fitted parameter is relabelled as a prediction or where a self-citation chain forces the result. The terse assertion in Lemma 3.1 that the moment conditions on E follow from Assumption 1.1 and Young's inequality is a possible proof-completeness point, but it is not circularity: the inequality |x|^{d-2} <= C(|x-y|^{d-2}+|y|^{d-2}) makes the Young step valid, and in any case a terse proof is not a circular reduction. Overall, the central claim stands independently of its inputs.

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

The paper introduces no fitted parameters and no invented entities. The machinery is Fourier analysis plus the Gaussian random-walk comparison function C(x). The central theorem's assumptions are explicit, and the application leans on previously established lace-expansion results for the self-repellent Brownian motion.

assumptions (3)

domain assumption Assumption 1.1 conditions on J and g: h ∈ L^1 ∩ L^2, |x|^2 h ∈ L^1 ∩ L^2, |x|^{2+ε} h ∈ L^1, and for d>4, |x|^{d-2} h ∈ L^p ∩ L^2 with 1 ≤ p < d/4; \hat J(0)=1; infrared bound \hat J(0)-\hat J(k) ≥ K_IR(|k|^2 ∧ 1).
These hypotheses are introduced in Section 1.2 and used throughout the proof, especially the infrared bound to invert (1-\hat J) and to obtain the |k|^{-2} singularity.
standard math Standard Fourier analysis: L^p Fourier transform bounds (1≤p≤2), Riemann-Lebesgue lemma, weak derivatives and Hölder's inequality.
Used in Sections 2 and 3 via references [22, Lemma A.4] and [6, Chapter 5] to control derivatives of Fourier transforms.
domain assumption External results from [2] for self-repellent Brownian motion: lace expansion equation (4.1), Gaussian domination bound (1.16), and bound (4.2) on Π_λ.
These results are cited from Bolthausen-Koenig-Mukherjee (2024) and are used in Section 4 to verify Assumption 1.1 and to identify the L^2 solution.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Gaussian deconvolution on $\mathbb R^d$ with application to self-repellent Brownian motion." pith.science (2026). https://pith.science/paper/ZVKIYIKH

@misc{pith2026241116058,
  author       = {Pith},
  title        = {Pith review of: Gaussian deconvolution on $\mathbb R^d$ with application to self-repellent Brownian motion},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZVKIYIKH}},
  note         = {Machine review of arXiv:2411.16058}
}

abstract

We consider the convolution equation $(\delta - J) * G = g$ on $\mathbb R^d$, $d>2$, where $\delta$ is the Dirac delta function and $J,g$ are given functions. We provide conditions on $J, g$ that ensure the deconvolution $G(x)$ to decay as $( x \cdot \Sigma^{-1} x)^{-(d-2)/2}$ for large $|x|$, where $\Sigma$ is a positive-definite diagonal matrix. This extends a recent deconvolution theorem on $\mathbb Z^d$ proved by the author and Slade to the possibly anisotropic, continuum setting while maintaining its simplicity. Our motivation comes from studies of statistical mechanical models on $\mathbb R^d$ based on the lace expansion. As an example, we apply our theorem to a self-repellent Brownian motion in dimensions $d>4$, proving its critical two-point function to decay as $|x|^{-(d-2)}$, like the Green function of the Laplace operator $\Delta$.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

27 extracted references · 23 canonical work pages

[2]

Bolthausen, W

E. Bolthausen, W. Koenig, and C. Mukherjee. Self-repell ent Brownian bridges in an interacting Bose gas. Preprint, https://arxiv.org/pdf/2405.08753, (2024). 11

arXiv 2024
[1]

Bolthausen, R

E. Bolthausen, R. van der Hofstad, and G. Kozma. Lace expa nsion for dummies. Ann. I. Henri Poincar´ e Probab. Statist., 54:141–153, (2018)

work page 2018
[3]

Brydges, T

D.C. Brydges, T. Helmuth, and M. Holmes. The continuous- time lace expansion. Commun. Pure Appl. Math. , 74:2251–2309, (2021)

work page 2021
[4]

Brydges and T

D.C. Brydges and T. Spencer. Self-avoiding walk in 5 or mo re dimensions. Commun. Math. Phys., 97:125–148, (1985)

work page 1985
[5]

Chatterjee, J

S. Chatterjee, J. Hanson, and P. Sosoe. Subcritical conn ectivity and some exact tail exponents in high dimensional percolation. Commun. Math. Phys. , 403:83–153, (2023)

work page 2023
[6]

L.C. Evans. Partial Diﬀerential Equations . American Mathematical Society, Providence, 2nd edition, (2010)

work page 2010
[7]

Fitzner and R

R. Fitzner and R. van der Hofstad. Mean-ﬁeld behavior for nearest-neighbor percolation in d > 10. Electron. J. Probab. , 22:1–65, (2017)

work page 2017
[8]

Fitzner and R

R. Fitzner and R. van der Hofstad. NoBLE for lattice trees and lattice animals. J. Stat. Phys. , 185:paper 13, (2021)

work page 2021

Show all 27 references

[9]

T. Hara. Decay of correlations in nearest-neighbor self -avoiding walk, percolation, lattice trees and animals. Ann. Probab., 36:530–593, (2008)

2008
[10]

T. Hara, R. van der Hofstad, and G. Slade. Critical two-p oint functions and the lace expansion for spread-out high-dimensional percolation and related mode ls. Ann. Probab., 31:349–408, (2003)

2003
[11]

Hara and G

T. Hara and G. Slade. Mean-ﬁeld critical behaviour for p ercolation in high dimensions. Commun. Math. Phys. , 128:333–391, (1990)

1990
[12]

Hara and G

T. Hara and G. Slade. On the upper critical dimension of l attice trees and lattice animals. J. Stat. Phys. , 59:1469–1510, (1990)

1990
[13]

Hara and G

T. Hara and G. Slade. Self-avoiding walk in ﬁve or more di mensions. I. The critical behaviour. Commun. Math. Phys. , 147:101–136, (1992)

1992
[14]

Heydenreich and R

M. Heydenreich and R. van der Hofstad. Progress in High-Dimensional Percolation and Random Graphs. Springer International Publishing Switzerland, (2017)

2017
[15]

Heydenreich, R

M. Heydenreich, R. van der Hofstad, G. Last, and K. Matzk e. Lace expansion and mean-ﬁeld behavior for the random connection model. Preprint, https://arxiv.org/pdf/1908.11356, (2023)

2023 arXiv
[16]

van der Hofstad and A.A

R. van der Hofstad and A.A. J´ arai. The incipient inﬁnit e cluster for high-dimensional unoriented percolation. J. Stat. Phys. , 114:625–663, (2004)

2004
[17]

Kozma and A

G. Kozma and A. Nachmias. The Alexander–Orbach conject ure holds in high dimensions. Invent. Math., 178:635–654, (2009)

2009
[18]

Kozma and A

G. Kozma and A. Nachmias. Arm exponents in high dimensio nal percolation. J. Am. Math. Soc., 24:375–409, (2011)

2011
[19]

Lawler and V

G.F. Lawler and V. Limic. Random Walk: A Modern Introduction . Cambridge University Press, Cambridge, (2010)

2010
[20]

Y. Liu. A general approach to massive upper bound for two -point function with application to self-avoiding walk torus plateau. Preprint, https://arxiv.org/pdf/2310.17321, (2023). 12

2023 arXiv
[21]

Y. Liu, J. Park, and G. Slade. Universal ﬁnite-size scal ing in high-dimensional critical phenomena. In preparation, (2024)

2024
[22]

Liu and G

Y. Liu and G. Slade. Gaussian deconvolution and the lace expansion. To appear in Probab. Theory Related Fields. Preprint, https://arxiv.org/pdf/2310.07635, (2023)

2023 arXiv
[23]

Liu and G

Y. Liu and G. Slade. Gaussian deconvolution and the lace expansion for spread- out models. To appear in Ann. Inst. H. Poincar´ e Probab. Statist. Preprint, https://arxiv.org/pdf/2310.07640, (2023)

2023
[24]

A. Sakai. Lace expansion for the Ising model. Commun. Math. Phys. , 272:283–344, (2007). Correction: A. Sakai. Correct bounds on the Ising lace-expa nsion coeﬃcients. Commun. Math. Phys., 392:783–823, (2022)

2007
[25]

A. Sakai. Application of the lace expansion to the ϕ 4 model. Commun. Math. Phys. , 336:619– 648, (2015)

2015
[26]

G. Slade. A simple convergence proof for the lace expans ion. Ann. I. Henri Poincar´ e Probab. Statist., 58:26–33, (2022)

2022
[27]

G. Slade. The near-critical two-point function and the torus plateau for weakly self-avoiding walk in high dimensions. Math. Phys. Anal. Geom. , 26:article 6, (2023). 13

2023

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.

[1] [2]

Bolthausen, W

E. Bolthausen, W. Koenig, and C. Mukherjee. Self-repell ent Brownian bridges in an interacting Bose gas. Preprint, https://arxiv.org/pdf/2405.08753, (2024). 11

arXiv 2024

[2] [1]

Bolthausen, R

E. Bolthausen, R. van der Hofstad, and G. Kozma. Lace expa nsion for dummies. Ann. I. Henri Poincar´ e Probab. Statist., 54:141–153, (2018)

work page 2018

[3] [3]

Brydges, T

D.C. Brydges, T. Helmuth, and M. Holmes. The continuous- time lace expansion. Commun. Pure Appl. Math. , 74:2251–2309, (2021)

work page 2021

[4] [4]

Brydges and T

D.C. Brydges and T. Spencer. Self-avoiding walk in 5 or mo re dimensions. Commun. Math. Phys., 97:125–148, (1985)

work page 1985

[5] [5]

Chatterjee, J

S. Chatterjee, J. Hanson, and P. Sosoe. Subcritical conn ectivity and some exact tail exponents in high dimensional percolation. Commun. Math. Phys. , 403:83–153, (2023)

work page 2023

[6] [6]

L.C. Evans. Partial Diﬀerential Equations . American Mathematical Society, Providence, 2nd edition, (2010)

work page 2010

[7] [7]

Fitzner and R

R. Fitzner and R. van der Hofstad. Mean-ﬁeld behavior for nearest-neighbor percolation in d > 10. Electron. J. Probab. , 22:1–65, (2017)

work page 2017

[8] [8]

Fitzner and R

R. Fitzner and R. van der Hofstad. NoBLE for lattice trees and lattice animals. J. Stat. Phys. , 185:paper 13, (2021)

work page 2021