{"id":"4871273b-6c49-4a9b-b5ca-62b84c6b7b79","arxiv_id":"2505.19113","paper_version":1,"verdict":"REJECT","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"high","formal_verification":"none","parameter_count":0,"one_line_summary":"A heat kernel Gaussian estimate, L1 Liouville theorem, uniqueness, eigenvalue bounds, and a Li-Yau gradient estimate are derived on weighted manifolds with lower N-Ricci curvature in the epsilon-range.","lead":"This paper proves Gaussian upper and lower bounds for the heat kernel on weighted Riemannian manifolds under a variable, epsilon-parametrized lower N-Ricci curvature bound, and derives Liouville, uniqueness, eigenvalue, and gradient estimates from them. The heat kernel bounds are the load-bearing result; the final Li-Yau-type gradient estimate contains a visible gap in the proof.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The on-diagonal lower bound in Theorem 1.1 is unproven: the step at (4.11) applies the Harnack inequality (4.8) with s=0, where the (t-s)/s term diverges, so the diagonal estimate (4.12) and hence the full lower bound do not follow.","rationale":"The reader's verdict REJECT is justified, but I identify a more direct obstacle to the central claim than the reader's primary criticisms. The reader's weakest_assumption, the two-sided weight bound, is an explicit hypothesis rather than an unproven step; the application-level flaws in Theorems 1.2 and 1.3 are serious, yet the heat-kernel lower bound itself contains an invalid Harnack application that threatens Theorem 1.1 directly. In the proof of the lower bound, the comparison 1 = u(x,0) <= C u(x,t/2) invokes Proposition 4.3 with s=0, while that proposition is only proved for 0 < s < t and its constant contains (t-s)/s, which diverges at s=0. Thus the on-diagonal estimate (4.12) is unsupported and the lower estimate in Theorem 1.1 does not follow. This concern is independent of the other issues and is not a typographical artifact: the Harnack machinery in the paper genuinely requires positive time separation. I therefore maintain the reader's REJECT verdict, but with a different primary justification; the concrete test above settles the matter by direct substitution.","tokens_in":24734,"tokens_out":21748,"duration_ms":186967,"concrete_test":"Substitute s=0 into inequality (4.8) of Proposition 4.3 and observe that the term (t-s)/s is infinite, so the Harnack constant is not finite and the step 1 = u(x,0) <= exp(C'_11(...)) u(x,t/2) used at (4.11) is unjustified; this directly invalidates the on-diagonal lower bound (4.12) as written.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The lower-bound half of the central heat-kernel estimate rests on an invalid Harnack application. Proposition 4.3 (and Theorem 3.7 behind it) yields a finite Harnack constant only for 0 < s < t: inequality (4.8) contains the term (t-s)/s, which blows up as s tends to 0. In the proof of Theorem 1.1, after constructing u(y,t) = P_t g(y) with g supported in B_x(2 sqrt t), the authors assert at (4.11): 1 = u(x,0) <= exp(C'_11(...)) u(x,t/2). This compares a solution at time 0 to a positive time, exactly the regime excluded by Proposition 4.3. Consequently the chain from (4.11) to (4.12) cannot be justified by the Harnack inequality proved earlier; the on-diagonal lower bound H(x,x,t/2) >= exp(-C t) V_x(sqrt t)^{-1} is not established, and the lower estimate (4.15) in Theorem 1.1 collapses. This is not a minor typo: the parabolic Harnack inequality genuinely requires positive times, and no limiting argument with s->0 can produce a finite constant because of the divergent (t-s)/s term. A repair would need a different argument, for example comparing times t/4 and t/2 and using the semigroup identity, but that is absent from the manuscript. Since Theorem 1.1 is the paper's central claim, this unresolved gap alone prevents acceptance of the advertised result.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper studies weighted Riemannian manifolds with lower N-Ricci curvature in the ε-range and a two-sided bound on e^{2(1-ε)φ/(n-1)}. The main result (Theorem 1.1) is a pair of Gaussian upper and lower bounds for the φ-heat kernel. The authors then apply the heat kernel bounds to prove an L1 Liouville theorem for φ-subharmonic functions, L1 uniqueness for the φ-heat equation, eigenvalue lower bounds for Δφ, and a Li-Yau-type gradient estimate under an Lp bound on |∇φ|. The proofs follow the standard Davies/Saloff-Coste framework, using comparison theorems of Lu-Minguzzi-Ohta and local Sobolev inequalities of Fujitani.","tokens_in":24976,"tokens_out":9519,"duration_ms":83421,"significance":"If Theorem 1.1 were proved, it would be a meaningful extension of heat kernel estimates to the ε-range curvature condition, with several geometric consequences. The paper also attempts a partial answer to a question of Ohta on negative N in Section 7. The structure is standard and the reliance on prior comparison and Sobolev results is explicit. However, as detailed below, the lower-bound proof and the Liouville theorem proof contain invalid steps, so the advertised results are not presently established.","major_comments":[{"comment":"The application of Proposition 4.3 with s=0 is invalid because the Harnack inequality (4.8) is stated only for 0<s<t and contains the divergent term (t-s)/s. Consequently the estimate 1 = u(x,0) ≤ e^{C} u(x,t/2) cannot be justified, and this breaks the derivation of the diagonal lower bound (4.12) and the lower bound (4.15) in Theorem 1.1.","section":"Section 4, Eq. (4.11)"},{"comment":"Even if the Harnack application were valid, the Cauchy-Schwarz step in (4.11) yields a bound on H(x,x,t) with volume factor V_x(2√t), not a bound on H(x,x,t/2) with volume factor V_x(√2t) as claimed in (4.12). The time and volume arguments need to be reconciled before the stated diagonal lower bound follows.","section":"Section 4, Eq. (4.12)"},{"comment":"The parabolic mean value inequality (3.1) is applied to the time-independent function h(x,t)≡h(x), which is only assumed to be a nonnegative L1 subharmonic function and is not a solution of the φ-heat equation. Therefore (5.4) is not justified. This estimate is used to control the boundary integrals in Proposition 5.3, so the L1-Liouville theorem (Theorem 1.2) and Theorem 5.5 are not established by the given proof.","section":"Section 5, Eq. (5.4)"},{"comment":"Uniqueness for the auxiliary equation (7.8) is attributed to the Liouville theorem of Section 5, but that theorem concerns L1 subharmonic functions and does not directly imply uniqueness for bounded solutions of this linear parabolic equation. A standard maximum principle or semigroup argument is needed for the uniqueness claim.","section":"Section 7, Lemma 7.1"},{"comment":"The assumption in Theorem 2.6 is stated as 0<a≤ e^{2(ε-1)φ/(N-1)} ≤ b, while every other theorem in the paper uses e^{2(1-ε)φ/(n-1)}. Since Theorem 2.6 underpins the local Sobolev inequality used throughout, this inconsistency must be resolved.","section":"Section 2, Theorem 2.6"}],"minor_comments":[{"comment":"The title contains \"it's application\"; it should read \"its application\".","section":"Title"},{"comment":"The word \"kernal\" is used instead of \"kernel\", for example in Section 3 and in the statement of Theorem 1.1.","section":"Throughout"},{"comment":"The exponential term in (4.14) appears to have an unbalanced parenthesis and an unclear factor involving √2t/a; please check the formula.","section":"Section 4, Eq. (4.14)"},{"comment":"The formula for λ_k in (6.7) has an extra parenthesis after the exponential term; the expression should be cleaned up.","section":"Section 6, Theorem 6.2"},{"comment":"The proof of Lemma 3.4 is only sketched; since this weighted Poincaré inequality is a key ingredient, the authors should provide full details or a precise reference to the argument in [30].","section":"Section 3, Lemma 3.4"}],"recommendation":"major_revision","confidential_remarks":"The paper is a serious attempt and the results are plausible, but the current proof has several load-bearing gaps. I recommend major revision rather than outright rejection because the main approach is standard and the errors are localized. However, the authors must supply a valid proof of the heat kernel lower bound and repair the boundary-term estimates in Proposition 5.3 before the paper can be considered further."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should know: this is a standard-template heat-kernel paper with a real gap in the lower-bound half, and the advertised Liouville and gradient applications inherit trouble.\n\nWhat is new is the class of curvature bounds: the Lu-Minguzzi-Ohta epsilon-range N-Ricci condition. Pushing the Saloff-Coste / Wu-Wu / Fujitani Moser-Davies machinery into that class is a legitimate extension, and the upper-bound argument mostly follows the known route with the right ingredients cited: comparison theorems, local Sobolev inequality, Davies' double integral estimate. If the gaps I list below get fixed, this would be a useful subfield contribution. The citation pattern looks normal; the external theorems are taken as black boxes, which is fair enough for this kind of paper.\n\nThe bigger problem is the diagonal lower bound. Proposition 4.3's Harnack inequality only gives a finite constant for 0 < s < t, and the term (t-s)/s diverges at s=0. At (4.11) the proof compares u(x,0) with u(x,t/2) using that Harnack inequality, with no limiting argument. The stress-test note is right: (4.11)-(4.12) do not follow, and the lower-bound half of Theorem 1.1 collapses. That is load-bearing, not a typo.\n\nProposition 5.3 applies the parabolic mean value inequality to a time-independent subharmonic function h. That function is not a heat solution, and Remark 3.3's extension to upper solutions does not cover a subsolution unless the sign convention is carefully checked; as written it is not. So the L1-Liouville theorem and the uniqueness theorem depending on it are not established.\n\nSection 7 has a concrete algebraic error. With delta = 2/(2n+1), one computes Z = delta, and the coefficient Z((alpha-J)/alpha)^2 - delta is never nonnegative for finite alpha and J<=1. The maximum-principle bound at (7.27) does not follow. There is also a minor exponent mismatch in Theorem 2.6's assumption.\n\nNone of this is a takedown. The upper bound may be salvageable, and the overall structure is honest and follows established templates. But as it stands, the central theorem and the two headline applications are not proven. I would send this to a serious referee, not desk-reject it, because there is enough substance in the framework and the fixable upper-bound part. I would not cite it until the gaps are closed. If the authors resubmit after repair, the same places need rereading: (4.11), Proposition 5.3, and the delta algebra in Section 7.","headline":"A promising but incomplete extension: the heat-kernel upper bound is plausible, but the lower bound and two applications have real gaps.","tokens_in":25582,"tokens_out":3746,"would_cite":false,"duration_ms":35953,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["53C21","58J35"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves two-sided Gaussian heat kernel bounds on weighted Riemannian manifolds with lower N-Ricci curvature in the ε-range, under a two-sided bound on the weight, and derives Liouville, uniqueness, spectral, and gradient-estimate…","keywords":["weighted Riemannian manifold","N-Ricci curvature","ε-range curvature bound","heat kernel estimate","Gaussian upper and lower bounds","parabolic Harnack inequality","L1-Liouville theorem","Li-Yau gradient estimate"],"falsifier":"Compute the $\\phi$-heat kernel explicitly on a model warped-product space where $\\mathrm{Ric}_N^\\phi$ equals the $\\varepsilon$-range bound with equality, and check whether the Gaussian exponent and the volume prefactor in Theorem 1.1 are reproduced; a discrepancy in the power of $V_x(\\sqrt t)$ or in the rate $d^2/(4(1+\\varepsilon)t)$ would falsify the estimate.","tokens_in":24416,"feed_emoji":"🔥","tokens_out":9150,"duration_ms":66628,"temperature":0.7,"pith_summary":"This paper proves that on a complete weighted Riemannian manifold whose N-Ricci curvature satisfies the lower bound $\\mathrm{Ric}_N^\\phi \\ge K e^{4(\\varepsilon-1)\\phi/(n-1)}$ with $\\varepsilon$ in the admissible range, and whose weight satisfies $0<a\\le e^{2(1-\\varepsilon)\\phi/(n-1)}\\le b$, the $\\phi$-heat kernel obeys explicit two-sided Gaussian estimates. The upper bound decays like $\\exp(-d^2/4(1+\\varepsilon)t)$ with volume-square-root prefactors, and the lower bound is of the form $\\exp(-C t - C d^2/t)V_x(\\sqrt t)^{-1}$. From this single kernel estimate the paper derives an $L^1_\\phi$-Liouville theorem, $L^1_\\phi$-uniqueness for the weighted heat equation, eigenvalue lower bounds for $\\Delta_\\phi$, and a Li–Yau-type gradient estimate under a weighted $L^p$ bound on $|\\nabla\\phi|^2$.","feed_headline":"Gaussian heat bounds proved for weighted manifolds under ε-range curvature","feed_subtitle":"Two-sided control of the weight yields volume doubling, L¹-Liouville, and eigenvalue bounds from a single heat kernel estimate.","key_machinery":"The load-bearing object is the $\\phi$-heat kernel $H^\\phi(x,y,t)$, the minimal positive fundamental solution of $(\\partial_t-\\Delta_\\phi)u=0$, with $\\Delta_\\phi=\\Delta-\\langle\\nabla\\phi,\\nabla\\cdot\\rangle$ self-adjoint on $L^2(\\mu)$. The machinery consists of four linked inequalities: the $\\phi$-Laplacian comparison and Bishop–Gromov volume comparison for the $\\varepsilon$-range bound; the resulting local Sobolev, Neumann–Poincaré, and volume-doubling estimates; the parabolic mean value inequality and Moser's Harnack inequality; and Davies' double-integral estimate, combined with a Li–Yau-type Harnack inequality. Each step needs the two-sided weight bound $0<a\\le e^{2(1-\\varepsilon)\\phi/(n-1)}\\le b$ to keep the curvature parameter $c=(1-\\varepsilon^2(N-n)/(N-1))/(n-1)$ effective.","core_discovery":"The central assertion is Theorem 1.1: for all $x,y\\in M$ and $t>0$, the minimal positive heat kernel of $\\Delta_\\phi$ satisfies a Gaussian upper bound of the form $C(\\varepsilon)E'_2\\exp(2D_2\\sqrt{K_\\varepsilon(q,10\\sqrt t)}\\,/\\sqrt t)\\,(V_x(\\sqrt t)V_y(\\sqrt t))^{-1/2}\\exp(-d^2(x,y)/4(1+\\varepsilon)t)$ and a lower bound of the form $C'_{12}\\exp(-C'_{13}t - C'_{14}d^2(x,y)/t)V_x(\\sqrt t)^{-1}$. Here the constants $E'_2,D_2$ depend on $a,b,n,\\nu$; the $C'_i$ depend on $n,\\nu,a,b,c,K$; and $C(\\varepsilon)\\to\\infty$ as $\\varepsilon\\to0$, while $V_x(\\sqrt t)$ is the weighted volume of the geodesic ball of radius $\\sqrt t$ centered at $x$. The proof is a comparison-geometry chain: the $\\varepsilon$-range curvature condition gives a $\\phi$-Laplacian comparison and a Bishop–Gromov volume comparison; these feed a local Sobolev inequality and Moser iteration, yielding the parabolic mean value and Harnack inequalities; Davies' double-integral estimate gives the Gaussian upper bound, and a Li–Yau-type Harnack inequality converts the upper bound into the lower bound.","pith_inferences":["The operative quantity in the estimates is the ratio $b/a$; this suggests that the same proof strategy could tolerate a slowly growing weight, with the ratio entering only through explicit constants and yielding polynomial volume-growth corrections instead of the exponential factors here.","Because the $\\varepsilon$-range formulation was designed for weighted Finsler and Lorentzian settings, the same Harnack-to-kernel route may transfer to those geometries once a parabolic Harnack inequality and a Davies-type double-integral estimate are available.","Testing the gradient estimate on explicit model weights with known heat kernels would show whether $p>n$ is a genuine threshold or an artifact of the proof."],"forward_implications":["Any nonnegative $L^1_\\phi(\\mu)$-integrable $\\phi$-subharmonic function is constant; in particular, every $L^1(\\mu)$ harmonic function is constant.","Every $L^1_\\phi$ solution of the weighted heat equation is uniquely determined by its initial data.","The eigenvalues of $\\Delta_\\phi$ admit explicit lower bounds; when $K=0$, $\\lambda_k \\ge C(k+1)^{2c/(c+1)}/d^2$ with $C$ depending only on $n,\\nu,a,b$ and $d$ the diameter.","A Li–Yau-type gradient estimate holds for positive solutions of the weighted heat equation under the constraint $\\|\\nabla\\phi\\|_{L^p(\\mu)}\\le V$ with $p>n$, partially answering the negative-dimensional $N$-Ricci question."],"supporting_citations":[{"why":"Supplies the $\\phi$-Laplacian comparison and Bishop–Gromov volume comparison for the $\\varepsilon$-range curvature bound that the whole proof builds on.","marker":"[23]"},{"why":"Provides the local Sobolev inequality and Neumann–Poincaré inequality on weighted manifolds under the same $\\varepsilon$-range assumptions.","marker":"[6]"},{"why":"Is the source of the parabolic mean value, Moser Harnack, and weighted Poincaré machinery used in Sections 3 and 4.","marker":"[30]"},{"why":"Supplies the heat-kernel framework on smooth metric measure spaces, including stochastic completeness and integration-by-parts arguments used for the $L^1$-Liouville theorem.","marker":"[39]"},{"why":"Provides the template for deriving Gaussian upper bounds of heat kernels from mean value inequalities and Davies' estimate.","marker":"[29]"},{"why":"Supplies Davies' double-integral estimate for the heat kernel on metric measure spaces used in the upper bound proof.","marker":"[38]"},{"why":"Provides the $L^1$-Liouville and uniqueness arguments that convert the heat kernel bound into Theorems 1.2 and 5.5.","marker":"[17]"},{"why":"Supplies the Li–Yau eigenvalue estimate argument adapted in Section 6.","marker":"[19]"},{"why":"Supplies the technique for Li–Yau gradient bounds under integral curvature assumptions adapted in Section 7.","marker":"[42]"}],"fun_headline_variants":["Gaussian heat kernel bounds on weighted manifolds with ε-range Ricci","Heat kernel Gaussian control under ε-range weighted curvature","Gaussian heat kernel bounds for weighted manifolds with ε-range curvature","Under ε-range Ricci: Gaussian heat kernel bounds on weighted manifolds"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof collapses if the stretching factor $e^{2(1-\\varepsilon)\\phi(x)/(n-1)}$ is not trapped between two fixed positive constants $a$ and $b$ on the whole manifold, since that trap is what makes the Laplacian comparison, volume comparison, and Sobolev inequality quantitative.","fun_headline_variants_meta":{"raw":{"variants":["Gaussian heat kernel bounds on weighted manifolds with ε-range Ricci","Heat kernel Gaussian control under ε-range weighted curvature","Gaussian heat kernel bounds for weighted manifolds with ε-range curvature","Under ε-range Ricci: Gaussian heat kernel bounds on weighted manifolds"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001128,"raw_usage":{"total_tokens":4728,"prompt_tokens":1021,"completion_tokens":3707,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":637,"completion_tokens_details":{"reasoning_tokens":3636}},"tokens_in":637,"tokens_out":3707,"duration_ms":19236,"temperature":1.0,"reasoning_tokens":3636,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T14:21:31.714028+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the $\\phi$-heat kernel explicitly on a model warped-product space where $\\mathrm{Ric}_N^\\phi$ equals the $\\varepsilon$-range bound with equality, and check whether the Gaussian exponent and the volume prefactor in Theorem 1.1 are reproduced; a discrepancy in the power of $V_x(\\sqrt t)$ or in the rate $d^2/(4(1+\\varepsilon)t)$ would falsify the estimate.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the $\\phi$-Laplacian comparison and Bishop–Gromov volume comparison for the $\\varepsilon$-range curvature bound that the whole proof builds on."},{"cited_title":"Fujitani, Analysis of harmonic functions under lower bounds ofN−weighted Ricci curvature with ε−range,Journal of Mathematical Analysis and Applications, vol","cited_arxiv_id":null,"evidence_quote":"Provides the local Sobolev inequality and Neumann–Poincaré inequality on weighted manifolds under the same $\\varepsilon$-range assumptions."},{"cited_title":"Saloff-Coste, Aspects of Sobolev-Type Inequalities","cited_arxiv_id":null,"evidence_quote":"Is the source of the parabolic mean value, Moser Harnack, and weighted Poincaré machinery used in Sections 3 and 4."},{"cited_title":"Wu and P","cited_arxiv_id":null,"evidence_quote":"Supplies the heat-kernel framework on smooth metric measure spaces, including stochastic completeness and integration-by-parts arguments used for the $L^1$-Liouville theorem."},{"cited_title":"Saloff-Coste, Uniformly elliptic operators on Riemannian manifolds.J","cited_arxiv_id":null,"evidence_quote":"Provides the template for deriving Gaussian upper bounds of heat kernels from mean value inequalities and Davies' estimate."},{"cited_title":"Wu and P","cited_arxiv_id":null,"evidence_quote":"Supplies Davies' double-integral estimate for the heat kernel on metric measure spaces used in the upper bound proof."},{"cited_title":"Li, Uniqueness ofL 1 solutions for the Laplace equation and the heat equation on Riemannian manifolds,Journal of Differential Geometry, vol","cited_arxiv_id":null,"evidence_quote":"Provides the $L^1$-Liouville and uniqueness arguments that convert the heat kernel bound into Theorems 1.2 and 5.5."},{"cited_title":"Li, S.-T","cited_arxiv_id":null,"evidence_quote":"Supplies the Li–Yau eigenvalue estimate argument adapted in Section 6."},{"cited_title":"Zhang, M","cited_arxiv_id":null,"evidence_quote":"Supplies the technique for Li–Yau gradient bounds under integral curvature assumptions adapted in Section 7."}],"review_version":1}