{"id":"14b274fc-855e-4410-aae8-7510cbff666a","arxiv_id":"2502.01153","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For an Edwards-Wilkinson interface with D(t)=B^2(t), the single-point height distribution scales as t^{3/4} and its exact scaling function has exponential tails, a robust feature for general z>1.","lead":"This paper computes exactly how the height of a one-dimensional Edwards-Wilkinson interface spreads when its diffusivity itself diffuses as the square of a Brownian motion. It finds the height grows like time to the 3/4 and a symmetric distribution with exponential tails that persists across a family of linear interface models.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek-v4-flash","headline":"General-z exponential tail in Section 5 is asserted without proof; z=2 branch-point claim is numerically checked but unproven, leaving the advertised robustness across z>1 unsupported.","rationale":"The central z=2 calculation (Sections 2–4) is internally consistent and I found no error in the Feynman–Kac reduction, the Riccati solution, or the matching of prefactors; the p→0 limit of (44) gives ⟨V⟩=4/3, as expected. The z=2 exponential tail is the least secure step in the derivation because the branch-point structure at k=±ia is verified numerically rather than proven, but this can be supplied from standard Bessel-function properties, so I do not consider it a real correctness risk. The genuine load-bearing concern is the general-z claim in Section 5: it is the paper's advertised 'robust' result, yet no derivation or numerical evidence is given for the exponential tail at any z≠2. A reader cannot check the claim from the manuscript alone. The small-H summary in Eq. (51) is an obvious sign error relative to Eq. (48) and Fig. 4; it is cosmetic and does not affect (46). For these reasons I agree with the reader's CONDITIONAL verdict but only partially with the stated weakest assumption: the z=2 branch-point gap is minor, while the general-z assertion is the main soft spot. The concrete check I propose — high-precision evaluation of G_4(H) from the exact representation and comparison with the first-zero prediction — would settle whether the general-z tail holds.","tokens_in":13401,"tokens_out":25449,"duration_ms":208421,"concrete_test":"For z=4 (Mullins-Herring), compute G_4(H) from Eq. (65) (with the correct substitution p=k^2/2) by high-precision numerical quadrature for |H| up to, say, 40, and fit log G_4(H) versus H. Compare the fitted decay rate with a_4 = y_1^{(z)} / (4z/((2z-1)√2)) where y_1^{(z)} is the first positive zero of J_{-z/(2z-1)}; a non-exponential tail or a rate different from a_4 would falsify Section 5. Also independently derive the z=2 tail by substituting the 0F1 representation and locating the first zero of J_{-2/3}, which should reproduce a=0.6592248 and b=0.7592287.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The exact z=2 result (46) is derived cleanly, and the exponential tail for z=2 rests on the claim that F(k)=k^{2/3}I_{-2/3}(8k/(3√2)) has its nearest zero at k=ia, a=0.659..., with F(k)≈b(k^2+a^2) near k=ia. This is checked numerically (Fig. 2) but not proven. The missing proof is, however, standard: F(k) is proportional to the entire hypergeometric function 0F1(;1/3;c^2k^2/4), whose zeros lie on the imaginary axis at k=i y_n/c with y_n the real simple zeros of J_{-2/3}; the first zero gives a≈0.659, so the z=2 tail is on solid ground. The load-bearing gap is Section 5: after deriving the exact representation (65), the paper states it is not difficult to show that for large |H|, the scaling function G_z(H) has an exponential tail, with no singularity analysis, no decay exponent, and no check for any z≠2. Since the abstract advertises this robustness across z>1 as a main result, the general-z exponential-tail claim is unsupported. If for some z the nearest singularity of the z-analogue of F were not a simple imaginary-axis zero, the tail could be different. Additionally, Eq. (51)'s small-H line is internally wrong: (48) gives the H^2 coefficient as -d1/2! ≈ -0.4095, not +d2/24 ≈ +0.4368, contradicting Fig. 4; this is a sign/typographical error, not a failure of the integral representation.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the one-dimensional Edwards-Wilkinson (EW) interface in the presence of a stochastic diffusivity D(t)=B^2(t), where B(t) is a one-dimensional Brownian motion. By adapting the backward Feynman-Kac formalism to treat the explicit time-dependent kernel in the variance functional, the authors obtain the exact Laplace transform of the distribution of the scaled variance, Eq. (44), and hence an exact integral representation for the scaled height distribution G(H), Eq. (46). They derive the large-|H| exponential tail G(H) ~ const * e^{-a|H|}/sqrt(|H|) with a=0.6592248, and a quadratic approach to G(0) at small H. The paper also sketches a generalization to a family of linear interface models with dynamical exponent z>1, claiming an exponential tail for all z>1. The z=2 result is exact and internally consistent; the main weakness is the unsupported general-z claim and the numerically verified rather than rigorously proven branch-point structure used for the tail.","tokens_in":13723,"tokens_out":4652,"duration_ms":46455,"significance":"If the results hold, the z=2 exact solution is a significant contribution to the diffusing-diffusivity literature, providing a rare exact tagged-monomer/height distribution for an interacting system driven by a noise with time-dependent stochastic diffusivity. The explicit integral representation (46), the exact Laplace transform (44), and the scaling form (50) are analytically checkable and likely to be useful. The paper's derivation of the Laplace transform is self-contained and the connection to Brownian functionals is clean. However, the advertised robustness of exponential tails for all z>1 is not substantiated, and the asymptotic analysis at z=2 rests on a numerically checked but unproved singularity structure. These gaps do not undermine the exact z=2 result but do limit the paper's broader claims.","major_comments":[{"comment":"The claim that G_z(H) has an exponential tail for every z>1 is asserted without derivation. After Eq. (65), the paper states that 'it is not difficult to show' the tail, but no singularity analysis, decay exponent, or numerical check for any z≠2 is provided. Since the abstract advertises the exponential tail as robust across the whole family z>1, this is a load-bearing claim. Please provide the actual singularity analysis for the integrand in Eq. (65) for general z, or explicitly restrict the claim to z=2. If the general proof is not available, the abstract and Section 5 should be revised so that the unproved part is not presented as a main result.","section":"Section 5, Eq. (65)"},{"comment":"The large-H tail is derived from the assertion that F(k)=k^{2/3}I_{-2/3}(8k/(3\\sqrt{2})) has its nearest singularities on the imaginary axis at k=±ia, with F(k)≈b(k^2+a^2) near k=ia, and is verified numerically in Fig. 2 but not proved. Because the exponential decay rate a and the prefactor depend on this singular structure, the tail result is not fully rigorous. Please provide a proof (e.g., using the fact that the zeros of I_{-ν} are purely imaginary, so the integrand has square-root branch points) or cite a reference where this property is established. Without such a justification, the tail in (47) should be described as a strongly supported conjecture or verified numerically.","section":"Section 4, Eq. (47)"}],"minor_comments":[{"comment":"The small-H line in Eq. (51) contains a sign and coefficient error: from Eq. (48), the H^2 coefficient is -d1/2! ≈ -0.4095, not +d2/24 ≈ +0.4368. This contradicts Fig. 4 and should be corrected.","section":"Eq. (51)"},{"comment":"The vertical axis in Figs. 3 and 4 is labeled P(H), but the plotted function is G(H) as defined in Eq. (46). Please relabel to avoid notation inconsistency.","section":"Fig. 3 and Fig. 4"},{"comment":"The symbol Γ is used both for the coupling constant in Eq. (14) and for the Gamma function in the definition of A_z. The sentence after Eq. (55) clarifies this, but the formula would be less prone to misreading if the coupling constant were denoted by a different symbol (e.g., \\gamma).","section":"Eq. (55)"}],"recommendation":"major_revision","confidential_remarks":"The exact z=2 result is a solid piece of work and the derivation is transparent. The main concern is the unsupported general-z exponential-tail claim in Section 5, which is presented as a main result in the abstract. I recommend major revision, primarily to either prove the general-z statement or to restrict the main claims to z=2 with the generalization presented as a conjecture. The small-H typo in Eq. (51) and the unproved branch-point structure in Section 4 should also be addressed. The paper should be publishable after these revisions."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The exact z=2 result is the real thing. For the EW interface with D(t)=B^2(t), the paper computes the full single-point height distribution, obtaining the t^{3/4} scaling and an exponential tail with exponent a≈0.659. That is a clean, new exact solution in the diffusing-diffusivity literature, and the first for an interacting chain. The central technical step, a modified Feynman-Kac treatment of the weighted Brownian functional V=∫_0^1 B^2(u)/√(1-u) du, is handled carefully. I checked the coefficient matching in Eq. (35), the c1=0 boundary condition, and the reduction of the general-z formula back to z=2; all are internally consistent. The exact Fourier integral (46) is a solid representation, and the numerical checks in Figs. 2–4 are convincing. The self-citations to refs. [31,32,34] are appropriate; those prior results are independent and parameter-free, so there is no circularity.\n\nThe soft spots are real but fixable. First, Eq. (51) has a sign error: from (48) the H^2 coefficient is -d1/2! ≈ -0.4095, not +d2/24 ≈ +0.4368. The text and Fig. 4 are consistent with the negative sign, so this is a typographical slip in the summary line, but it should be corrected. Second, the general-z exponential tail in Section 5 is asserted without proof. The paper says \"it is not difficult to show\" but gives no singularity analysis, no decay exponent, and no check for any z≠2. Since the abstract presents this robustness across z>1 as a main result, that is a load-bearing gap for the generalization. For z=2, the branch-point claim is numerically checked but unproven. That said, the missing proof is standard: F(k) is proportional to a hypergeometric function whose zeros lie on the imaginary axis at the zeros of J_{-2/3}, so the first zero gives a≈0.659. Thus the z=2 tail is on solid ground, and the paper would be stronger if it sketched this argument or cited a proof.\n\nWho is this for? Researchers working on Brownian yet non-Gaussian diffusion, Rouse polymer dynamics, and exact distributions of Brownian functionals. They will find the z=2 result useful and will want to cite it. The general-z claim needs either proof or more cautious wording. A serious referee should engage: the paper deserves peer review, and after fixing the typo and addressing the z>1 gap, it would be a good contribution.","headline":"The exact z=2 height distribution is a genuine new result and the derivation is clean; the advertised z>1 generalization is asserted rather than proven.","tokens_in":810,"tokens_out":959,"would_cite":true,"duration_ms":23650,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["60J65","60H10","82C31","82C41"],"pacs":[],"model":"deepseek-v4-flash","headline":"For a one-dimensional Edwards–Wilkinson interface whose diffusivity is the square of a Brownian motion, the paper obtains the exact single-point height distribution, whose scaled form is symmetric and decays as a non-Gaussian exponential.","keywords":["Edwards-Wilkinson model","diffusing diffusivity","Brownian functional","Feynman-Kac formula","height distribution","exponential tail","Rouse chain","dynamical exponent"],"falsifier":"Numerically integrate the exact integral (46) at large $H$ and check whether $\\log G(H)+a|H|+\\tfrac12\\log|H|$ approaches a constant with $a=0.6592248$; equivalently, compute $F(iq)=(iq)^{2/3}I_{-2/3}(8iq/(3\\sqrt{2}))$ near $q=a$ and verify the branch-point approximation $F(iq)\\approx b(q^2+a^2)$ with $b=0.7592287$. If a different singularity is closer to the real axis, or the branch structure differs, the exponential tail fails.","tokens_in":13193,"feed_emoji":"📈","tokens_out":11933,"duration_ms":113945,"temperature":0.7,"pith_summary":"An interface described by the Edwards–Wilkinson equation normally has Gaussian height fluctuations. When the diffusivity is itself stochastic—taken here as $D(t)=B^2(t)$ with $B(t)$ a Brownian motion—the paper shows the height at a fixed point still obeys an exact scaling law, but the scaled distribution $G(H)$ is no longer Gaussian. The typical height grows like $t^{3/4}$, and $G(H)$ is symmetric: it tends to a nonzero constant quadratically near $H=0$ and decays like $e^{-a|H|}/\\sqrt{|H|}$ for large $H$, with $a=0.6592248$. The same exponential-tail form is argued to hold for a family of linear interface models with dynamical exponent $z>1$. The result is presented as the first exact tagged-monomer height distribution in an interacting chain driven by time-dependent stochastic diffusivity, connecting diffusing-diffusivity physics to interface growth.","feed_headline":"Exact height law: interface with random diffusivity","feed_subtitle":"For D(t)=B^2(t), heights scale as t^{3/4} with a non-Gaussian exponential tail.","key_machinery":"The load-bearing object is the Brownian functional $\\mathcal{V}=\\int_0^1 B^2(u)(1-u)^{-1/2}\\,du$, whose distribution $Q(\\mathcal{V})$ controls the height distribution through Gaussian conditioning. The paper computes the Laplace transform of $Q(\\mathcal{V})$ by a backward Feynman–Kac equation adapted to the time-dependent weight $(1-u)^{-1/2}$. The solution uses the ansatz $\\phi_p(x,w)=f(w)e^{-g(w)x^2/2}$, which reduces the PDE to a Riccati equation for $g$; a Hopf–Cole transformation $s''(w)-4p\\,s(w)/\\sqrt{w}=0$ turns it into a linear equation solved by modified Bessel functions $I_{\\pm 2/3}$. The explicit Laplace transform (44) then feeds the Fourier representation of $G(H)$ in Eq. (46), from which the small- and large-$H$ behaviors are read off.","core_discovery":"The central result is an exact expression for the single-point height distribution of the one-dimensional Edwards–Wilkinson equation with $D(t)=B^2(t)$ and flat initial condition. Writing the variance for a fixed diffusivity history as $V(t)=t^{3/2}\\,\\mathcal{V}/\\sqrt{2\\pi\\Gamma}$, where $\\mathcal{V}=\\int_0^1 B^2(u)(1-u)^{-1/2}\\,du$, the paper computes the Laplace transform of the distribution of $\\mathcal{V}$ by an adapted backward Feynman–Kac method. This yields the scaling form $p(h,t)=(2\\pi\\Gamma)^{1/4}\\,t^{-3/4}\\,G((2\\pi\\Gamma)^{1/4}h/t^{3/4})$, with $G(H)$ given exactly by the Fourier integral $G(H)=\\frac{3^{1/3}}{\\sqrt{2}\\,\\Gamma(1/3)}\\int_{-\\infty}^{\\infty}\\frac{dk}{2\\pi}\\frac{e^{ikH}}{\\sqrt{k^{2/3}I_{-2/3}(8k/(3\\sqrt{2}))}}$. From that integral the paper extracts $G(H)\\simeq d_0+(d_2/24)H^2$ near $H=0$ with $d_0=0.387$, and $G(H)\\simeq \\text{const}\\cdot e^{-a|H|}/\\sqrt{|H|}$ as $|H|\\to\\infty$, with $a=0.6592248$. It further asserts the same exponential-tail structure for general linear interface models $\\partial_t h=-\\Gamma(-\\partial_x^2)^{z/2}h+\\sqrt{2D(t)}\\,\\eta$ for any $z>1$, with a $z$-dependent decay exponent.","pith_inferences":["An extension the paper leaves implicit: in a finite Rouse chain of length $L$, the variance functional is dominated by the slowest relaxation mode, so the exponential tail should survive only for $t\\ll L^z$ and cross over to Gaussian behavior at longer times; this can be tested by Brownian dynamics simulation.","The same adapted Feynman–Kac machinery applies to other positive self-similar diffusivity processes such as $|B(t)|^\\alpha$; one would expect a family of exact scaling functions whose tail exponents depend on $\\alpha$, interpolating between the results here and the single-particle case.","Because the exact $G(H)$ decays exponentially rather than quadratically in the exponent, the large-deviation rate for atypical heights is linear in $|H|$; this is a concrete prediction that could be checked in experiments or simulations of stochastic-diffusivity interface growth."],"forward_implications":["At late times, the tagged-monomer position distribution in a Rouse chain with $D(t)=B^2(t)$ is the same non-Gaussian scaling law, so diffusing-diffusivity anomalies that are known for a single particle persist in an interacting polymer.","The typical height scales as $t^{3/4}$, intermediate between the $t^{1/2}$ of ordinary Edwards–Wilkinson growth and the linear-in-$t$ scaling of a free particle with $D(t)=B^2(t)$.","The scaled distribution has a nonzero density at $H=0$ with a quadratic correction, and an exponential rather than Gaussian tail, so large-height events are far more likely than in the constant-diffusivity Edwards–Wilkinson model.","Every linear interface model with dynamical exponent $z>1$, including $z=4$ Mullins–Herring, inherits the exponential tail, with a $z$-dependent decay exponent.","The exact Fourier representation (46) gives a closed-form route to all moments of the scaled height through the coefficients $d_n$ in Eq. (49)."],"supporting_citations":[{"why":"Defines the discrete Rouse chain whose late-time tagged-monomer distribution is the physical target of the calculation.","marker":"[1]"},{"why":"Defines the Edwards–Wilkinson interface whose single-point height distribution is computed exactly.","marker":"[3]"},{"why":"Introduces the diffusing-diffusivity mechanism that motivates the choice D(t)=B^2(t).","marker":"[16]"},{"why":"Lays out the backward Feynman–Kac method for Brownian functionals that the computation adapts.","marker":"[33]"},{"why":"Supplies the specific adaptation of the Feynman–Kac method for time-dependent weights such as (1-u)^{-1/2} in the variance functional.","marker":"[34]"}],"fun_headline_variants":["Exact height law: Brownian diffusivity in 1D Edwards-Wilkinson","Exact t^3/4 distribution for interface with random D","Brownian diffusivity leads to exact non-Gaussian height tail","Random diffusivity gives exact 1D EW height distribution"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The exponential-tail claim for large $|H|$ rests on the nearest singularities of the integrand in Eq. (46) being square-root branch points at $k=\\pm ia$ with $F(k)\\approx b(k^2+a^2)$; the paper verifies this numerically but does not prove it, and the analogous tail for general $z$ is asserted without derivation.","fun_headline_variants_meta":{"raw":{"variants":["Exact height law: Brownian diffusivity in 1D Edwards-Wilkinson","Exact t^3/4 distribution for interface with random D","Brownian diffusivity leads to exact non-Gaussian height tail","Random diffusivity gives exact 1D EW height distribution"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001019,"raw_usage":{"total_tokens":4369,"prompt_tokens":1083,"completion_tokens":3286,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":699,"completion_tokens_details":{"reasoning_tokens":3210}},"tokens_in":699,"tokens_out":3286,"duration_ms":24079,"temperature":1.0,"reasoning_tokens":3210,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-09T16:23:23.294801+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Numerically integrate the exact integral (46) at large $H$ and check whether $\\log G(H)+a|H|+\\tfrac12\\log|H|$ approaches a constant with $a=0.6592248$; equivalently, compute $F(iq)=(iq)^{2/3}I_{-2/3}(8iq/(3\\sqrt{2}))$ near $q=a$ and verify the branch-point approximation $F(iq)\\approx b(q^2+a^2)$ with $b=0.7592287$. If a different singularity is closer to the real axis, or the branch structure differs, the exponential tail fails.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines the discrete Rouse chain whose late-time tagged-monomer distribution is the physical target of the calculation."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines the Edwards–Wilkinson interface whose single-point height distribution is computed exactly."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduces the diffusing-diffusivity mechanism that motivates the choice D(t)=B^2(t)."},{"cited_title":"Sci.892076","cited_arxiv_id":null,"evidence_quote":"Lays out the backward Feynman–Kac method for Brownian functionals that the computation adapts."}],"review_version":1}