{"id":"c9f96bb7-8672-4717-9ea8-0396520ef66f","arxiv_id":"2506.06875","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"New pointwise bounds on the fractional gradient of the fractional heat kernel yield global Bessel-potential regularity, a compactness result, and existence for a fractional KPZ equation with nonlocal gradient term.","lead":"This paper proves new regularity estimates for solutions of the fractional heat equation with zero boundary data, and applies them to show existence of solutions for a fractional Kardar-Parisi-Zhang growth model with nonlocal gradients. The key technical step is a pointwise bound on the fractional gradient of the associated heat kernel, which controls how much smoothness the solution gains near the boundary and at small times.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 3.2 proves the kernel estimate only for ρ < min{1,2s}, while Theorems 1.1, 1.4 and 4.2 claim it up to max{1,2s}; the downstream regularity results inherit this unsupported range.","rationale":"The single most load-bearing concern is not the imported character of Lemma 3.1—the cited estimates (3.2)–(3.3) are standard for sufficiently regular domains and the paper explicitly assumes 'regular boundary'—but the internal mismatch between the proved and announced ranges for ρ. Theorem 3.2 is the engine of the paper; its proof fixes ρ ∈ [s, min{1,2s}) and nowhere extends beyond. Theorems 1.1, 1.4, and 4.2 nevertheless state the conclusions for ρ ∈ [s, max{1,2s}), which contains strictly larger values in both parameter regimes (s < 1/2 and s > 1/2). Since Section 4 invokes (3.7)/(3.8) directly and no supplementary estimate covers the missing strip, the main regularity theorems are currently unproved in their stated generality. The KPZ existence results (Theorems 1.6 and 1.8) use ρ = s, which lies in the proved range, so the paper's principal application is not jeopardized; this is why the verdict remains conditional rather than reject. A corrected version that either narrows the statements to ρ < min{1,2s} or extends the kernel estimate would resolve the issue. The concrete test isolates the exact excluded case and determines which repair is needed. The reader's rationale already noted this range mismatch, though the reader's formal 'weakest_assumption' focused on the imported kernel estimates rather than on this internal gap; hence partial agreement.","tokens_in":65361,"tokens_out":8572,"duration_ms":86290,"concrete_test":"Analytically re-derive Theorem 3.2 in the excluded regime: take s = 1/3, ρ = 0.8, so ρ ∈ [s, max{1,2s}) = [1/3, 1) but ρ > min{1,2s} = 2/3. Run the proof through the derivation of (3.27) with σ = 2s + ρ − 1 and check whether the final bound (3.7) is obtained. If the proof fails at the step imposing ρ < min{1,2s}, the statements of Theorems 1.1, 1.4, and 4.2 must be restricted to ρ < min{1,2s}; if it succeeds, the restriction in Theorem 3.2 is an artifact and the announced range is justified.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The proof of Theorem 3.2 begins with the explicit restriction s ≤ ρ < min{1,2s}, and every subsequent estimate (3.28)–(3.37) is derived under that restriction. Yet Theorem 1.1 and Theorem 4.2 (and Theorem 1.4) state the same pointwise bound for every ρ ∈ [s, max{1,2s}). For s < 1/2 these intervals differ (2s vs 1), and for s > 1/2 they differ (1 vs 2s). Since Theorem 3.2 is the only source of the kernel estimate, every later statement allowing ρ outside [s, min{1,2s}) lacks proof. For example, with s = 0.3, ρ = 0.8 is in [s, max{1,2s}) = [0.3, 1) but not in [s, min{1,2s}) = [0.3, 0.6); Theorem 4.2(1) nevertheless applies (3.7) at this ρ. This is not a purely cosmetic typo because the proof of Theorem 4.2 uses (3.7)/(3.8) at the stated ρ and no alternative estimate is supplied for ρ ≥ min{1,2s}. The KPZ application uses ρ = s and therefore survives a restriction, but the advertised global regularity results for the full range are unsupported unless Theorem 3.2 is extended or the statements are narrowed.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops global regularity theory for the fractional heat equation with homogeneous exterior Dirichlet conditions on a bounded domain. The main tool is a new pointwise estimate for the fractional gradient of the Dirichlet heat kernel, stated for ρ ∈ [s, max{1,2s}). From this estimate the authors derive global regularity of (−Δ)^{ρ/2}w in parabolic Bessel potential and fractional Sobolev spaces, obtain L^p estimates with explicit constants, and prove compactness of the data-to-solution map (w_0,h) ↦ w. In the second part, these results are applied to a fractional KPZ equation with nonlocal gradient term |(−Δ)^{s/2}u|^q, where existence of small-time solutions is proved by Schauder's fixed-point theorem. The proof of the kernel estimate is detailed and self-contained up to imported heat-kernel bounds; the regularity theorems are then obtained by combining this estimate with hyper-singular integral estimates proved in the appendix.","tokens_in":65680,"tokens_out":6087,"duration_ms":63599,"significance":"If the claimed estimates hold for the stated range, the paper is a substantial contribution: it extends the elliptic Calderón–Zygmund framework of [4] to the parabolic setting, gives global rather than local regularity results, and provides a compactness result that is used to obtain existence for a class of fractional KPZ problems. The proof of the kernel estimate in the range actually treated (ρ < min{1,2s}) is careful and technically involved, and the auxiliary integral estimates in Section 8 are of independent interest. The KPZ application is meaningful because the nonlinearity only uses ρ = s, so the main application is not harmed by the restriction discussed below. However, the advertised range ρ < max{1,2s} is not proved, and this affects the central regularity claims.","major_comments":[{"comment":"The proof of Theorem 3.2 begins with 'Fixed ρ > 0 such that s ≤ ρ < min{1,2s}' and every estimate (3.10)–(3.37) is derived under this restriction. Yet Theorem 1.1, Theorem 1.4 and Theorem 4.2 (and similarly Theorem 4.10 and Corollary 3.6) state the same pointwise bound for every ρ ∈ [s, max{1,2s}), which is strictly larger when s ≠ 1/2. For example, with s = 0.3, ρ = 0.8 lies in [s, max{1,2s}) = [0.3,1) but not in [s, min{1,2s}) = [0.3,0.6), and Theorem 4.2(1) applies estimate (3.7) at this ρ. Since Theorem 3.2 is the only source of the kernel estimate, all statements allowing ρ ≥ min{1,2s} are currently unsupported. The authors should either extend the proof to the wider range or restrict all statements to ρ ∈ [s, min{1,2s}); the KPZ application, which only uses ρ = s, does not by itself justify the stronger advertised range.","section":"Theorem 3.2, Theorem 1.1, Theorem 4.2"},{"comment":"In the case 2s + ρ < 1, the first bullet sets the threshold m ≤ (N+2s)/(2s−ρ), while the proof and the parallel statements in Theorem 1.5 and Corollary 4.13 use the threshold (N+2s)/(4s−1). For ρ = s and s < 1/3 these thresholds differ, and as stated the two bullets do not cover the parameter plane consistently; in particular the claimed range for r can be empty. This is a load-bearing inconsistency in a central theorem statement and should be corrected.","section":"Theorem 4.10(2)"},{"comment":"The two-sided heat-kernel estimate (3.2) and the gradient bound (3.3) are imported from [8,16,21,43] and are known in general for domains with C^{1,1} or otherwise sufficiently regular boundaries. The standing assumption in Section 1 is only 'regular boundary'. Since Lemma 3.1 is the sole input for Theorem 3.2 and hence for all downstream regularity, compactness and KPZ results, the precise boundary regularity needed for (3.2)–(3.3) must be stated and, if necessary, verified for the domains covered by Theorems 1.1, 1.4 and 4.10.","section":"Lemma 3.1"}],"minor_comments":[{"comment":"The title and abstract contain typographical artifacts ('RESUL TS', 'HEA T', 'A TION', 'majeur'); these should be cleaned in a revised version.","section":"Abstract and title"},{"comment":"The proof of Proposition 4.5 refers to 'Proposition 2.15', but the cited statement is Theorem 2.15.","section":"Proposition 4.5"},{"comment":"The preamble to Section 6 states 'q ≤ 1', whereas Theorems 6.2 and 6.4 require q > 1 in all branches; the text should read q ≥ 1 (with the status of q = 1 made explicit).","section":"Section 6"},{"comment":"In the first extension, 'Dirichlet heart kernel' should be 'Dirichlet heat kernel'.","section":"Section 7"}],"recommendation":"major_revision","confidential_remarks":"The main obstacle is a statement/proof range mismatch rather than an internal contradiction in the range actually proved. The paper is technically substantial and the KPZ application appears safe, but the current version cannot be accepted while Theorems 1.1, 1.4, 4.2 and 4.10 claim a range that is not established by the supplied proof. A revision that either proves the missing range or explicitly narrows the theorems would resolve the issue."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper's real contribution is a pointwise estimate for the fractional gradient of the Dirichlet heat kernel of the fractional Laplacian, and a careful descent from that estimate to global parabolic regularity, a compactness theorem, and existence for a fractional KPZ problem. The proof of the kernel estimate is genuinely new as far as I can tell, and the structure is coherent: representation formula, kernel bounds, hyper-singular integral estimates, then the nonlinear application. The authors also flag where they rely on imported kernel bounds and on their own elliptic framework, which is honest.\n\nThe main problem is a mismatch between what is proved and what is stated. Theorem 3.2 proves the kernel estimate only for rho < min{1,2s}. The introduction and Theorems 1.1, 1.4, 4.2 claim the same bound for rho < max{1,2s}. The proof of Theorem 4.2 applies (3.7)/(3.8) at values of rho outside the range established in Theorem 3.2. That is a load-bearing gap, not a typo: the regularity theorems as stated are unsupported for rho >= min{1,2s}. The fix is straightforward in principle — either extend the kernel estimate or narrow the statements. Since the KPZ application only needs rho = s, which lies in the proved range, a corrected version would almost certainly go through.\n\nTwo smaller issues. First, the paper says \"regular boundary\" but the imported heat-kernel estimates require C^{1,1} (or similar) boundary smoothness; this should be stated precisely, because the pointwise estimate and everything downstream collapses if those bounds fail. Second, the manuscript needs proofreading; there are typos and some notation slips. These are minor relative to the range gap.\n\nThe compactness theorem is a useful byproduct, and the application to KPZ is a legitimate payoff. I would not desk-reject this; it deserves a serious referee. My recommendation: send it out, with instructions to the authors to fix the range mismatch before acceptance, and to state the boundary hypothesis exactly. The central idea is solid and the paper is worth the community's time.","headline":"A genuinely new kernel estimate and a solid parabolic regularity program, but the advertised range rho < max{1,2s} is not proved; restrict statements or fill the gap.","tokens_in":66233,"tokens_out":2109,"would_cite":true,"duration_ms":22771,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35B05","35K15","35B40","35K55","35K65"],"pacs":[],"model":"deepseek-v4-flash","headline":"New pointwise estimates on the fractional gradient of the Dirichlet heat kernel yield global regularity for the fractional heat equation and, via a compactness argument, short-time existence for a fractional KPZ problem with nonlocal…","keywords":["fractional heat equation","heat kernel estimates","Bessel potential spaces","fractional Sobolev spaces","Kardar-Parisi-Zhang equation","nonlocal gradient","compactness","pointwise kernel bounds"],"falsifier":"Find a bounded domain with the smoothness assumed here, or with only slightly less smoothness, where the bound $|\\nabla_x P_\\Omega(x,y,t)|\\le C(\\delta(x)^{-1}\\wedge t^{-1/(2s)})P_\\Omega(x,y,t)$ fails uniformly in $t$ and $x$; concretely, one could check this estimate near a boundary corner or in a $C^1$ domain with an explicit or numerically computed heat kernel for small $s$. A failure there would remove the foundation of Theorem 3.2 and of the regularity and existence theorems built on it.","tokens_in":65120,"feed_emoji":"🔥","tokens_out":9590,"duration_ms":87884,"temperature":0.7,"pith_summary":"The paper proves global regularity, up to the boundary, for solutions of the fractional heat equation $w_t+(-\\Delta)^s w=h$ on a bounded domain with zero exterior Dirichlet data, when the data are merely integrable ($h\\in L^m$, $w_0\\in L^\\sigma$). The engine is a new pointwise bound on the fractional gradient $(-\\Delta)^{\\rho/2}P_\\Omega(x,y,t)$ of the Dirichlet heat kernel for $\\rho\\in[s,\\min\\{1,2s\\})$, with explicit powers of the distance to the boundary and of small time. Feeding this bound into the representation formula for $w$ places $w$ in parabolic Bessel potential spaces $L^r(0,T;L_0^{\\rho,r}(\\Omega))$ and in fractional Sobolev spaces, with constants controlled by the data norms, and it makes the solution map $(w_0,h)\\mapsto w$ compact. As an application, the authors obtain short-time existence for a Kardar–Parisi–Zhang equation with fractional diffusion and the nonlocal gradient term $|(-\\Delta)^{s/2}u|^q+f$. If correct, this gives a parameter regime in which $L^1$-type data still yield solutions with genuine fractional regularity.","feed_headline":"Fractional heat kernel bound yields global regularity","feed_subtitle":"The pointwise estimate for the fractional gradient turns L^1 data into Bessel-space regularity and short-time KPZ existence.","key_machinery":"The load-bearing object is the Dirichlet heat kernel $P_\\Omega(x,y,t)$ of the fractional Laplacian on a bounded domain, together with its standard two-sided estimate $P_\\Omega(x,y,t)\\asymp (1\\wedge \\delta^s(x)/\\sqrt t)(1\\wedge \\delta^s(y)/\\sqrt t)\\,t/(t^{1/(2s)}+|x-y|)^{N+2s}$ and gradient estimate $|\\nabla_xP_\\Omega|\\le C(\\delta(x)^{-1}\\wedge t^{-1/(2s)})P_\\Omega$. The new mechanism is Theorem 3.2's pointwise control of $(-\\Delta)^{\\rho/2}_xP_\\Omega$: the proof splits the fractional Laplacian into integrations over $\\mathbb{R}^N\\setminus\\Omega$ and $\\Omega$, uses the gradient bound on the auxiliary function $\\Theta_{y,t}(x)=(t^{1/(2s)}+|x-y|)^{N+\\sigma}P_\\Omega$, and chooses $\\sigma=2s+\\rho-1$ for $s\\le 1/2$ or $\\sigma=\\rho$ for $s>1/2$ to balance the singular terms. The kernel bound is then transferred to solutions by the representation formula and by $L^p$ estimates for hyper-singular convolution-type integrals; compactness later uses a Marcinkiewicz-space characterization of Bessel potential spaces in terms of the difference quotient $(u(x)-u(y))|x-y|^{-(N/q+s)}$.","core_discovery":"On the paper's own terms, the central discovery is Theorem 3.2: for $s\\in(0,1)$ and $\\rho\\in[s,\\min\\{1,2s\\})$, the fractional Laplacian of the Dirichlet heat kernel satisfies the pointwise estimates (3.7) for $s\\le 1/2$ and (3.8) for $s>1/2$, with the factor $(\\delta^s(y)/\\sqrt t\\wedge 1)$ multiplying terms such as $\\delta^{s-\\rho}(x)(t^{1/(2s)}+|x-y|)^{-(N+s)}$, $t^{(2s-1)/(2s)}\\log(D/|x-y|)$, and $|x-y|^{2s-1}$ over the natural kernel denominator. The paper then derives from these bounds, via the representation $w(x,t)=\\int_\\Omega w_0(y)P_\\Omega(x,y,t)\\,dy+\\int_0^t\\int_\\Omega h(y,\\tau)P_\\Omega(x,y,t-\\tau)\\,dy\\,d\\tau$ and estimates for hyper-singular integrals, that $(-\\Delta)^{\\rho/2}w\\in L^r(\\Omega_T)$ with quantitative bounds depending only on the data norms; in particular $w\\in L^r(0,T;L_0^{\\rho,r}(\\Omega))$ for the admissible $r$, both inside and outside $\\Omega$. With further analysis, the same estimates give compactness of the data-to-solution map and existence for the fractional KPZ problem.","pith_inferences":["Because the pointwise kernel bound isolates the boundary factors $\\delta^{s-\\rho}(x)$ and logarithmic terms, it suggests analogous weighted or boundary-Harnack estimates for parabolic nonlocal problems; one could test whether the factor $\\delta^{s-\\rho}$ is optimal by comparing with explicit kernels in domains where the Green function is known.","The authors note that the restriction $s>1/4$ appears naturally in the proof and leave open whether it is technical. A natural test is to run the same time-integration argument for $s\\le 1/4$: if the $t^{(2s-1)/(2s)}$ singularities cannot be compensated by the kernel denominator, the threshold is genuine, whereas an improved small-time kernel bound would remove it.","The compactness proof, built on a Marcinkiewicz bound for $|u(x)-u(y)||x-y|^{-(N/q+s)}$, likely transfers to other nonlocal parabolic equations with $L^1$ data, such as fractional $p$-Laplacian or measure-data problems where classical Rellich compactness is unavailable.","The short-time existence result for the KPZ problem is local in time; extending it to global time would require a priori bounds preventing finite-time blowup, which are not derived here."],"forward_implications":["For $s>1/4$ and $h\\in L^m(\\Omega_T)$, the fractional gradient $(-\\Delta)^{\\rho/2}w$ belongs to $L^r(\\Omega_T)$ for a range of exponents $r$ above $m$, with norm bounded by $C(\\Omega,T)\\|h\\|_{L^m}$; in particular $w\\in L^r(0,T;L_0^{\\rho,r}(\\Omega))$.","For $h=0$, an $L^\\sigma$ initial datum gives $(-\\Delta)^{\\rho/2}w(\\cdot,t)\\in L^p(\\Omega)$ with explicit time singularities, and $w\\in L^p(0,T;W_0^{s,p}(\\Omega))$ when the stated integrability condition on $\\sigma$ holds.","The data-to-solution operator $\\Phi:(h,w_0)\\mapsto w$ is compact as a map into $L^q(0,T;L_0^{\\rho,q}(\\Omega))$ for $q<\\hat\\kappa_{s,\\rho}$, which is what allows Schauder fixed-point arguments.","For the fractional KPZ problem with nonlocal gradient, if $f\\in L^m$ or $u_0\\in L^\\sigma$ in the stated parameter ranges, a weak solution exists on a short time interval $[0,T^*)$.","For $s>1/2$, the same regularity transfers to a fractional heat equation perturbed by a drift term satisfying the stated integrability condition."],"supporting_citations":[{"why":"Supplies the elliptic global-regularity framework and Green-function estimates that this paper extends to the parabolic setting.","marker":"[4]"},{"why":"Provides prior global regularity results for the fractional heat and KPZ equations, including weighted estimates used in the proof.","marker":"[8]"},{"why":"Gives sharp heat-kernel estimates for the fractional Laplacian perturbed by gradient operators, cited as a source for Lemma 3.1.","marker":"[16]"},{"why":"Provides Dirichlet heat-kernel estimates for the fractional Laplacian with gradient perturbation, cited for the kernel bounds in Lemma 3.1.","marker":"[21]"},{"why":"Provides Dirichlet heat-kernel estimates for stable processes with singular drift, cited for the kernel bounds in Lemma 3.1.","marker":"[43]"},{"why":"Supplies the theory of the fractional heat equation used to control the auxiliary potential $W$ by its initial data in the proof of Theorem 3.2.","marker":"[12]"},{"why":"Provides the Marcinkiewicz-space characterization of Bessel potential spaces used in the compactness proof.","marker":"[35]"},{"why":"Establishes existence and uniqueness of weak solutions to the fractional heat equation, the foundational existence result on which the paper builds.","marker":"[47]"}],"fun_headline_variants":["Fractional kernel bound drives global regularity","Pointwise heat kernel estimate yields KPZ existence","Kernel bound gives Bessel-space regularity for fractional heat","New fractional gradient bound proves global regularity"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument imports the sharp two-sided heat-kernel and gradient estimates for the Dirichlet fractional heat kernel in a regular bounded domain (Lemma 3.1) instead of proving them; if those estimates fail, or require boundary smoothness beyond what 'regular boundary' guarantees, the pointwise kernel bound and all downstream regularity, compactness, and KPZ existence results collapse.","fun_headline_variants_meta":{"raw":{"variants":["Fractional kernel bound drives global regularity","Pointwise heat kernel estimate yields KPZ existence","Kernel bound gives Bessel-space regularity for fractional heat","New fractional gradient bound proves global regularity"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000274,"raw_usage":{"total_tokens":1692,"prompt_tokens":1050,"completion_tokens":642,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":666,"completion_tokens_details":{"reasoning_tokens":585}},"tokens_in":666,"tokens_out":642,"duration_ms":6859,"temperature":1.0,"reasoning_tokens":585,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T05:47:58.126922+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Find a bounded domain with the smoothness assumed here, or with only slightly less smoothness, where the bound $|\\nabla_x P_\\Omega(x,y,t)|\\le C(\\delta(x)^{-1}\\wedge t^{-1/(2s)})P_\\Omega(x,y,t)$ fails uniformly in $t$ and $x$; concretely, one could check this estimate near a boundary corner or in a $C^1$ domain with an explicit or numerically computed heat kernel for small $s$. A failure there would remove the foundation of Theorem 3.2 and of the regularity and existence theorems built on it.","supporting_citations":[{"cited_title":"Abdellaoui, A.J","cited_arxiv_id":null,"evidence_quote":"Supplies the elliptic global-regularity framework and Green-function estimates that this paper extends to the parabolic setting."},{"cited_title":"Abdellaoui, I","cited_arxiv_id":null,"evidence_quote":"Provides prior global regularity results for the fractional heat and KPZ equations, including weighted estimates used in the proof."},{"cited_title":"Bogdan, T","cited_arxiv_id":null,"evidence_quote":"Gives sharp heat-kernel estimates for the fractional Laplacian perturbed by gradient operators, cited as a source for Lemma 3.1."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides Dirichlet heat-kernel estimates for the fractional Laplacian with gradient perturbation, cited for the kernel bounds in Lemma 3.1."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides Dirichlet heat-kernel estimates for stable processes with singular drift, cited for the kernel bounds in Lemma 3.1."},{"cited_title":"Bonforte, Y","cited_arxiv_id":null,"evidence_quote":"Supplies the theory of the fractional heat equation used to control the auxiliary potential $W$ by its initial data in the proof of Theorem 3.2."},{"cited_title":"Gu, P.-L","cited_arxiv_id":null,"evidence_quote":"Provides the Marcinkiewicz-space characterization of Bessel potential spaces used in the compactness proof."},{"cited_title":"Leonori, I","cited_arxiv_id":null,"evidence_quote":"Establishes existence and uniqueness of weak solutions to the fractional heat equation, the foundational existence result on which the paper builds."}],"review_version":1}