Pith. sign in

REVIEW 6 minor 1 cited by

Spatial decorrelation of KPZ from narrow wedge

T0 review · 0 major / 6 minor · reviewed 2026-08-06 · 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 KPZ equation from a narrow wedge has fixed-time spatial covariance asymptotic to t/x, so the spatial average converges to Brownian motion only under a sqrt(N log N) normalization.

desk verdict A solid, carefully proved new asymptotic (Cov ~ t/x) for KPZ from narrow wedge; relies on imported estimates but no load-bearing error seen. read the letter →

arxiv 2506.23065 v1 pith:5OJBB5Y7 submitted 2025-06-29 math.PR

Yu Gu , Fei Pu This is my paper · ORCID
classification math.PR MSC 60H1560H0760F05
keywords KPZequationnarrowwedgeinitialdataspatialcovariancedecaystochasticheataverageCLTBrownianmotionlimitWienerchaoscontinuumdirectedrandompolymer
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 establishes the exact fixed-time decorrelation rate for the one-dimensional KPZ equation started from a narrow wedge, the droplet-like initial data concentrated at one point. The central result is that for any fixed $t>0$, the spatial covariance $\operatorname{Cov}[h(t,x),h(t,0)]$ is asymptotic to $t/x$ as $x\to\infty$, a slower, nonintegrable decay than the $x^{-2}$ tail of the Airy2 process and the exponential tail of the flat case. Because the covariance tail is not integrable, the usual central-limit scaling fails: the spatial average over $[0,N]$ must be normalized by $\sqrt{N\log N}$, and with that normalization it converges to a Brownian motion. The proof's insight is that the dominant correlations come from the noise very close to the origin, where the two polymer paths share the environment, and that the conditioned polymer-overlap quantity admits a sharp small-time approximation.

What carries the argument

The central object is the scaled Green's function of the stochastic heat equation, $\bar{G}(t,x;s,y)=G(t,x;s,y)/p_{t-s}(x-y)$, where $p$ is the Gaussian heat kernel and $G$ is the SHE Green's function. The Clark–Ocone formula writes $h(t,x)-\mathbb{E}[h(t,x)]$ as an Itô integral whose integrand is $A(t,x;s,y)=\mathbb{E}[\bar{G}(t,x;s,y)\bar{G}(s,y;0,0)/\bar{G}(t,x;0,0)\mid\mathcal{F}_s]$, a conditional expectation describing how the two polymer paths from $x$ and $0$ share noise near the origin. The covariance becomes the integral of $A(t,x;s,y)A(t,0;s,y)$ against explicit heat kernels. Proposition 3.1 is the load-bearing estimate: for small $s$, $A(t,x;s,y)$ is close to $g_t(x,y)=\mathbb{E}[\bar{G}(t,x;0,y)/\bar{G}(t,x;0,0)]$, with error controlled by $s^{1/4}$ and distance terms. Lemma 2.1 then reduces the remaining expectation to a ratio that converges to $1$, which justifies passing the limit under the integral and yields the constant $t$.

What would settle it

Fix $t>0$ and estimate $\operatorname{Cov}[h(t,x),h(t,0)]$ numerically from a fine-grid simulation of the stochastic heat equation with delta initial data for large $x$; if $x$ times the estimated covariance does not approach $t$, Theorem 1.1 is wrong. A more targeted check is the small-time estimate (3.7): if numerical moments of $\bar{G}(s,0;0,0)-1$ grow faster than $s^{1/4}$ as $s\downarrow 0$, the controlling estimate in the proof fails.

Watch

Extended reading notes

Core claim

On its own terms, the paper's discovery is Theorems 1.1 and 1.2. Theorem 1.1 states that for each fixed $t>0$, $\operatorname{Cov}[h(t,x),h(t,0)]\sim t/x$ as $x\to\infty$, where $h=\log Z$ and $Z$ is the solution to the stochastic heat equation with delta initial data at $0$. The proof expresses the covariance through Itô isometry as an integral over $s\in(0,t)$, $y\in\mathbb{R}$ of the product $A(t,x;s,y)A(t,0;s,y)$ against explicit Gaussian kernels, where $A$ is a conditional expectation of ratios of the scaled SHE Green's function. The core calculation shows that after rescaling $s\mapsto s/x^2$, $y\mapsto y/x$, the integrand converges to a deterministic limit, yielding the constant $t$. Theorem 1.2 then proves that the centered spatial average, divided by $\sqrt{N\log N}$, converges in finite-dimensional distributions to $\sqrt{2}$ times a standard Brownian motion, with the first Wiener chaos component carrying the entire Gaussian limit.

Load-bearing premise

The argument depends on uniform bounds on how close the normalized Green's function of the stochastic heat equation stays to 1 at tiny times and over the shifted positions used in the rescaling; if those imported bounds fail in this regime, the limit under the integral is unjustified.

Editorial extensions

If this is right

  • For fixed $t$, the covariance has a $1/x$ tail, so spatial correlations do not decay fast enough for a standard central limit theorem; the variance of the spatial average grows like $2t\,N\log N$.
  • The centered spatial average divided by $\sqrt{N\log N}$ converges in finite-dimensional distribution to $\sqrt{2}$ times a standard Brownian motion as $N\to\infty$.
  • All higher Wiener chaos components of the spatial average vanish in $L^2$; the Gaussian limit comes entirely from the first chaos.
  • The fixed-time decay $t/x$ and the long-time KPZ scaling limit to the Airy2 process do not commute; taking $t\to\infty$ first would give a covariance decaying like $x^{-2}$ instead.

Reading between the lines

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

  • Not claimed in the paper: retaining the $s^{1/4}$ error from Proposition 3.1 could yield the rate of convergence to $t/x$, as well as a rate for the variance in Theorem 1.2.
  • Not claimed in the paper: the conditioned-overlap picture suggests the $1/x$ tail with coefficient $t$ is tied to the delta initial profile; a natural test is to rerun the argument for narrow Gaussian initial data and let the width tend to zero.
  • Not claimed in the paper: the first-chaos dominance for spatial averages hints that a full space-time Gaussian limit may hold after averaging over long spatial windows, although the paper only proves convergence of finite-dimensional distributions in time.
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.

Referee Report

0 major / 6 minor

Summary. The paper studies the KPZ equation with narrow wedge initial data and determines the fixed-time spatial decorrelation rate of the height function. Theorem 1.1 states that for every fixed t>0, Cov[h(t,x),h(t,0)] ~ t/x as x→∞, so the covariance is non-integrable in the spatial variable. Theorem 1.2 then proves that the spatial average of the centered height over [0,N], normalized by sqrt(N log N), converges in finite-dimensional distributions to sqrt(2) times standard Brownian motion. The proof uses the Clark-Ocone formula to write the centered height as a stochastic integral whose integrand involves a conditioned continuum directed polymer quantity A(t,x;s,y). After rescaling s↦s/x² and y↦y/x, the authors show that the small-s contribution dominates, approximate A by its deterministic limit g_t(x,y) via Proposition 3.1, and obtain the constant 2 by dominated convergence using the fixed-s limit in Lemma 3.4. The CLT for the spatial average is reduced to showing that the first Wiener chaos component dominates; Proposition 4.1 computes the limiting covariance of that component as 2(t1∧t2).

Significance. If correct, Theorem 1.1 gives the first precise fixed-time spatial decorrelation rate for narrow-wedge KPZ, showing a non-integrable t/x decay that is qualitatively different from the 1/x² decay of the Airy_2 process; the paper also correctly notes that the x→∞ and t→∞ limits do not commute. Theorem 1.2 identifies the anomalous sqrt(N log N) normalization and the Brownian limiting process, matching the earlier SHE result of [7]. The proof strategy is clear and largely self-contained: the Clark-Ocone representation reduces the covariance to a polymer-overlap quantity, and the error terms are tracked with explicit exponents such as (3.18) and (3.19), which vanish in the regime β<1/2. The paper honestly imports two key estimates, the uniform moment bound (2.2) from [1] and the small-s estimate (3.7) from [7]; on reading, these references do cover the parameter regime used here. The main covariance claim is further corroborated by the variance computation in Corollary 3.5.

minor comments (6)
  1. [Section 4, proof of Proposition 4.1, Step 3] The long displayed formula in Step 3 contains unreadable placeholder glyphs (for example, the string beginning with "⌟⟨rro..."), so the dominated-convergence step cannot be checked from the manuscript as printed. Please restore the full formula.
  2. [Section 3, Proposition 3.1 proof, estimate for I4] In the bound for ∥Gbar(t,x;s,√(s/t)y)-Gbar(t,x;0,0)∥_k, the terms Gbar(1,x;s,0) and Gbar(1,x;0,0) should presumably be Gbar(t,x;s,0) and Gbar(t,x;0,0), respectively.
  3. [Equation (3.11)] The displayed estimate for ∥Gbar(t,x;s,y)-Gbar(t,x;0,y)∥_k appears to omit a small-s term: when x=y the right-hand side is 0, while the left-hand side is generally of order s^{1/4}. The intended statement of [1, Lemma 3.6] should be reproduced correctly, and the proof should be checked against the corrected statement; the rest of the argument seems to include the extra s^{1/4} term separately.
  4. [Lemma 3.4, after (3.20)] In the Jensen-inequality display, the factor Gbar(s/x^2, y; 0,0) should be Gbar(s/x^2, z; 0,0), since the integration variable in the denominator is z rather than y.
  5. [Theorem 1.2 statement and Section 1.1] The convergence symbols around "fdd" in Theorem 1.2 are corrupted in the manuscript text, and Section 1.1 contains a duplicated "Indeed, Indeed". These should be typeset and proofread.
  6. [Remark 4.2] Remark 4.2 states the total-variation bound (4.10) without proof. Since this bound is not used in the main argument, the statement should either be proved or explicitly labeled as a heuristic claim to be developed elsewhere.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: Cov ∼ t/x follows from an exact Clark–Ocone representation plus independent SHE Green's-function estimates; no equation in the chain reduces to the claimed covariance.

full rationale

The derivation is self-contained modulo standard, externally checkable SHE input, and no step reduces to its own output. The covariance identity (3.2) is exact: Itô isometry applied to the Clark–Ocone representation of h (published tool [10, Prop. 6.3]), with A(t,x;s,y) in (3.1) defined so the integrand equals E[D_{s,y}h(t,x)|F_s] via the heat-kernel identity (2.3); the rescaling (3.15) is a pure change of variables (s↦s/x², y↦y/x). The approximation A≈g_t in Proposition 3.1 is proved, not assumed: g_t in (3.4) is the heuristic s→0 limit and (3.5) gives explicit L^k errors (s^{1/4}+d(xs^{1/4},ys^{1/4})+d(xs^{1/4},0)) that vanish in the regime used. Lemmas 3.2–3.4 are proved from the exact identity of Lemma 2.1 and from estimates (2.2), (3.7), (3.17), all statements about the SHE Green's function alone—none contains Cov ∼ t/x. The final dominated-convergence integral evaluates to ∫_0^∞ (1/√(πt²s))e^{-s/(4t²)}ds = 2, so Cov ∼ t/x; the conclusion never appears as an input. Theorem 1.2 combines Corollary 3.5 (from Theorem 1.1) with Proposition 4.1, whose first-chaos computation is independent of Theorem 1.1, so there is no circular dependence between the two theorems. Self-citations [7], [10], [23] are tool/comparison citations: (3.7) from [7, Lemma A.4] is parameter-free with assumptions excluding the target result, and [23] only motivates. Flagged non-circularity items: Remark 4.2 omits details of an alternative Malliavin–Stein proof of (4.10); Remark 4.3 leaves tightness for a functional CLT open; the proof relies on imported uniform positive/negative moment bounds (2.2) and small-s estimate (3.7) in the regime s=x^{-2}, y=x^{-1}. These are verification risks or omissions, not circular steps.

Assumptions & free parameters 0 free parameters · 8 assumptions · 0 invented entities

The central claim is derived from the SHE framework; no free parameters are fitted. The load-bearing inputs are external uniform moment/continuity estimates on the SHE Green's function and the Clark-Ocone representation. These are hypotheses inherited from [1], [7], and [14]; the paper does not prove them.

assumptions (8)
  • domain assumption Cole-Hopf solution of KPZ and narrow wedge initial data are well-defined
    Throughout Sections 1 and 3, h=log Z with Z solving SHE with delta initial condition; well-posedness is taken as known.
  • domain assumption Uniform positive and negative moment bounds for \bar G (equation (2.2))
    Taken from [1, Lemma 3.2]; used repeatedly in Propositions 3.1, Lemmas 3.3 and 3.4.
  • domain assumption Small-time estimate \bar G(s,0;0,0)-1 has L^k norm O(s^{1/4})
    Equation (3.7), cited to [7, Lemma A.4]; drives all error terms.
  • domain assumption Spatial and temporal continuity estimates for \bar G from [1, Lemmas 3.4, 3.6]
    Used to bound differences like \bar G(t,x;s,y)-\bar G(t,x;0,y).
  • domain assumption Negative moment bound for I1 from [14, Cor 4.9]
    Used in Proposition 3.1 for the inverse of I1.
  • domain assumption Clark-Ocone formula representation (3.1) from [10, Proposition 6.3]
    Starting point of the covariance computation.
  • standard math Ito isometry and L2 orthogonality of Wiener chaos components
    Used in Sections 3 and 4 to convert covariance to integrals and to isolate first chaos.
  • standard math Heat kernel identities pt(x)pt(y)=2p2t(x+y)p2t(x-y) and scaling
    Used to rewrite covariance integrals; no independent empirical input.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Spatial decorrelation of KPZ from narrow wedge." pith.science (2026). https://pith.science/paper/5OJBB5Y7

@misc{pith2026250623065,
  author       = {Pith},
  title        = {Pith review of: Spatial decorrelation of KPZ from narrow wedge},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5OJBB5Y7}},
  note         = {Machine review of arXiv:2506.23065}
}
abstract

We study the spatial decorrelation of the solution to the KPZ equation with narrow wedge initial data. For fixed $t>0$, we determine the decay rate of the spatial covariance function, showing that ${\rm Cov}[h(t,x),h(t,0)]\sim \frac{t}{x}$ as $x\to\infty$. In addition, we prove that the finite-dimensional distributions of the properly rescaled spatial average of the height function converge to those of a Brownian motion.

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

  1. Two-time spatial decorrelation for the flat KPZ fixed point

    math.PR 2026-07 conditional novelty 6.0 of 10

    The two-time spatial covariance of the flat KPZ fixed point decays as exp(-c|x|^3), and normalized spatial averages converge to a Gaussian process with covariance equal to the space-integrated two-time correlation.

Reference graph

Works this paper leans on

28 extracted references · 26 canonical work pages · cited by 1 Pith paper

  1. [7]

    Chen, L., Khoshnevisan, D., Nualart, D., and Pu, F. (2022). Spatial ergodicity and central limit theorems for parabolic Anderson model with delta initial condition. J. Funct. Anal., 282(2), 109290. 2, 7, 17, 18

    work page 2022

  2. [1]

    The Green's function of the parabolic Anderson model and the continuum directed polymer

    Alberts, T., Janjigian, C., Rassoul-Agha, F. and Seppäläinen, T.: The Green’s function of the parabolic Anderson model and the continuum directed polymer. arXiv preprint arXiv:2208.11255 (2022). 4, 5, 8, 10, 15

    work page Pith review arXiv 2022

  3. [2]

    and Quastel, J

    Alberts, T., Khanin, K. and Quastel, J. (2014). The continuum directed random polymer. Journal of Statisti- cal Physics, 154(1), 305-326. 1

    work page 2014

  4. [3]

    and Quastel, J

    Amir, G., Corwin, I. and Quastel, J. (2011). Probability distribution of the free energy of the continuum directed random polymer in 1+ 1 dimensions. Comm. Pure Appl. Math. 64, 466–537

    work page 2011

  5. [4]

    and Ferrari, P.L

    Basu, R., Busani, O. and Ferrari, P.L. (2023). On the exponent governing the correlation decay of the Airy1 process. Comm. Math. Phys. 398 1171–1211. 1 20 YU GU AND FEI PU

    work page 2023

  6. [5]

    and Zheng, G

    Bolaños Guerrero, R., Nualart, D. and Zheng, G. (2021). Averaging 2d stochastic wave equation. Electron. J. Probab. 26, pp.1-32. 2

    work page 2021

  7. [6]

    and Nualart, D

    Chen, L., Hu, Y . and Nualart, D. (2021). Regularity and strict positivity of densities for the nonlinear sto- chastic heat equation. Mem. Amer. Math. Soc. 273 no. 1340, v+102 pp. 4

    work page 2021

  8. [8]

    Chen, L., Khoshnevisan, D., Nualart, D., Pu, F. (2023). Central limit theorems for spatial averages of the stochastic heat equation via Malliavin-Stein’s method. Stoch. Partial Differ. Equ. Anal. Comput. 11 122–

    work page 2023

Show all 28 references
  1. [9]

    Chen, L., Khoshnevisan, D., Nualart, D., Pu, F. (2022). Central limit theorems for parabolic stochastic partial differential equations. Ann. Inst. Henri Poincaré Probab. Stat.58, no. 2, 1052–1077 2

    2022

  2. [10]

    Chen, L., Khoshnevisan, D., Nualart, D., Pu, F. (2021). Spatial ergodicity for SPDEs via Poincaré-type inequalities.Electron. J. Probab.26, 1-37. 2, 5

    2021

  3. [11]

    Corwin, I. (2012). The Kardar-Parisi-Zhang equation and universality class. Random Matrices Theory Appl. 1 no. 1, 1130001, 76 pp. 2

    2012

  4. [12]

    and Virág, B

    Dauvergne, D., Ortmann, J. and Virág, B. (2022) The directed landscape. Acta Math. 229 no. 2, 201–285. 2

    2022

  5. [13]

    Delgado-Vences, F., Nualart, D., and Zheng, G. (2020). A central limit theorem for the stochastic wave equation with fractional noise. Ann. Inst. Henri Poincaré Probab. Stat.56 no. 4, 3020–3042.3020-3042. 2

    2020

  6. [14]

    and Lê, K

    Hu, Y . and Lê, K. (2022). Asymptotics of the density of parabolic Anderson random fields.Ann. Inst. Henri Poincaré Probab. Stat.58 no. 1, 105–133. 7

    2022

  7. [15]

    and Viitasaari, L

    Huang, J., Nualart,D. and Viitasaari, L. (2020). A central limit theorem for the stochastic heat equation. Stochastic Process. Appl. 130, no. 12, 7170–7184.2, 17

    2020

  8. [16]

    (2020).Gaussian fluctuations for the stochastic heat equation with colored noise

    Huang, J., Nualart, D., Viitasaari, L., Zheng, G. (2020).Gaussian fluctuations for the stochastic heat equation with colored noise. Stoch. Partial Differ. Equ. Anal. Comput.8 no. 2, 402–421. 2

    2020

  9. [17]

    and Zhang, Y .-C

    Kardar, M., Parisi, G. and Zhang, Y .-C. (1986). Dynamic scaling of growing interfaces. Phys. Rev. Lett. 56

    1986

  10. [18]

    and Nualart, D.(2023)

    Kuzgun, S. and Nualart, D.(2023). Feynman-Kac formula for iterated derivatives of the parabolic Anderson model. Potential Anal. 59 , no. 2, 651–673. 17

    2023

  11. [19]

    and Remenik, D

    Matetski, K., Quastel, J. and Remenik, D. (2021). The KPZ fixed point. Acta Math. 227 115–203. 2

    2021

  12. [20]

    Nualart, D. (2006). The Malliavin Calculus and Related Topics. Springer, New York.3

    2006

  13. [21]

    Nualart, D., Xia, P., Zheng, G. (2022). Quantitative central limit theorems for the parabolic Anderson model driven by colored noises. Electron. J. Probab.27 Paper No. 120, 43 pp. 2

    2022

  14. [22]

    and Spohn, H

    Prähofer, M. and Spohn, H. (2002). Scale invariance of the PNG droplet and the Airy process. J. Stat. Phys. 108 (2002): 1071-1106. 1

    2002

  15. [23]

    arXiv e-prints (2023): arXiv-2311

    Pu, F.: Ergodicity, CLT and asymptotic maximum of the Airy 1 process. arXiv e-prints (2023): arXiv-2311. 2

    2023

  16. [24]

    Quastel, J. (2012). Introduction to KPZ. Current developments in mathematics, 2011, 125–194. International Press, Somerville, MA 2

    2012

  17. [25]

    and Spohn, H

    Quastel, J. and Spohn, H. (2015). The one-dimensional KPZ equation and its universality class. J. Stat. Phys. 160 no. 4, 965–984. 2

    2015

  18. [26]

    Stroock, D. W. (1987). Homogeneous chaos revisited. Séminaire de Probabilités, XXI, 1–7. Lecture Notes in Math., 1247 13

    1987

  19. [27]

    Virág, B. (2020). The heat and the landscape I. arXiv:2008.07241 2

    2020 arXiv

  20. [28]

    The KPZ equation and the directed landscape

    Wu, Xuan. The KPZ equation and the directed landscape. arXiv preprint arXiv:2301.00547 (2023). 2, 3 (Yu Gu) DEPARTMENT OF MATHEMATICS , U NIVERSITY OF MARYLAND , C OLLEGE PARK , MD 20742, USA (Fei Pu) LABORATORY OF MATHEMATICS AND COMPLEX SYSTEMS , S CHOOL OF MATHEMATICAL SCI-...

    2023 arXiv

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.