{"id":"c25a7026-ada6-4d5b-b529-0aecbfcc6dd6","arxiv_id":"2507.06237","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A Li-Yau gradient estimate, a Harnack inequality, and an a priori bound are established for positive solutions of the Finslerian logarithmic Schrödinger equation under lower curvature bounds.","lead":"This paper proves a new control inequality, called a Li-Yau estimate, for a nonlinear heat-like equation on curved spaces where distances may be asymmetric. The inequality leads to a Harnack inequality and a boundedness result for positive solutions, tools useful in PDE and geometric analysis.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The cut-off argument relies on an unproved Laplacian comparison lower bound for Δ∇uφ, cited from an unreviewed preprint; the central estimate is conditional on it.","rationale":"The reader's weakest assumption matches the point I find most load-bearing. Tracing the proof: (3.22) gives the lower bound on Δ∇uL, and the passage to (3.23) replaces φΔ∇uL by B·L using the displayed lower bound on Δ∇uφ. This is the only step where the non-Riemannian tensor bound and the Laplacian comparison enter globally, and it is needed in both cases of the maximum-principle argument. The cited theorem is not stated or proved, so the central estimate is conditional on an external result whose hypotheses may not match Theorem 1.1 exactly. Secondary issues do not change the verdict: Theorem 4.1 and 4.2 apply the estimate with A=a, violating the strict hypothesis A>a_+, though the case-2 bound appears continuous and a limiting argument could repair the applications. There are also typographical concerns about exponents in the Cauchy step, but these are hard to adjudicate from the typeset. The overall strategy is coherent and the result is a plausible extension of Wang's Riemannian theorem, so the appropriate disposition remains conditional on verification of the comparison input.","tokens_in":18092,"tokens_out":21585,"duration_ms":197199,"concrete_test":"Compare the hypotheses and conclusion of Shen's Theorem 1.1 (arXiv:2312.06617) with the cut-off estimate used here: verify that the stated comparison holds for the linearized Laplacian Δ∇uφ (reference direction ∇u) on forward geodesic balls in a forward complete Finsler manifold with finite misalignment and mRicN≥−K, with uniform constants depending only on N, α, K, R, K0. In particular, confirm that the reference vector may be taken to be the solution-dependent direction ∇u and that C0 is controlled by F(U)+F*(T)+F(div_C(V))≤K0. If the comparison is not stated for Δ∇u or requires extra regularity/support hypotheses on φ, the proof of (3.23) has a gap.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The maximum-principle proof of Theorem 1.1 controls the cut-off term through the inequality Δ∇uφ ≥ −(C1/R)[C(N,α)√(K(2R)/C(N,α)) coth(R√(K(2R)/C(N,α))) + C0(K0,α)] − αC2/R², attributed to 'the Laplacian comparison theorem ... established by B. Shen. See Theorem 1.1 in [16]'. The theorem is not stated, and [16] is an arXiv preprint. The reference direction is ∇u, the solution-dependent gradient direction, so the comparison must hold uniformly for this direction, and C0 must depend on the non-Riemannian tensor bound F(U)+F*(T)+F(div_C(V))≤K0. If Shen's result applies only to the nonlinear Laplacian Δf=div(∇f), to reference directions coming from geodesic fields, or to cut-offs supported away from the cut locus, the B·L term in (3.23) is unjustified and the proof of (1.2) breaks. The text also invokes Calabi's trick to make φ smooth, but on a Finsler manifold the distance function is only Lipschitz and no approximation argument is supplied for the maximum principle at points where r is not smooth.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript claims a local Li-Yau type gradient estimate (Theorem 1.1) for positive bounded solutions of the Finslerian logarithmic Schrödinger equation (1.1) on a forward complete non-compact Finsler metric measure space, assuming mixed weighted Ricci curvature bounded below and a uniform bound on certain non-Riemannian tensors. The proof combines Bochner-type formulas with a maximum principle and a cut-off function, and it relies on a Laplacian comparison theorem due to B. Shen (reference [16]). Two applications are given: a Harnack inequality (Theorem 4.1) and an a priori boundedness result for a stationary equation (Theorem 4.2).","tokens_in":18341,"tokens_out":10065,"duration_ms":98808,"significance":"If the proof is correct, this would be a meaningful generalization of Wang's Riemannian gradient estimates to Finsler geometry with general time-dependent coefficients, and the explicit constants in the estimate would provide a useful tool. The manuscript makes a serious attempt at handling the nonlinear Finsler Laplacian, and the algebraic core of Theorem 3.1 appears plausible. However, the central estimate is conditional on an unstated external comparison theorem from an unreviewed preprint, and the applications as written violate the hypotheses of the main theorem. The paper also leaves several technical gaps in the maximum-principle step and in the definition of constants. For these reasons I cannot recommend acceptance in the present form.","major_comments":[{"comment":"The lower bound for Δ∇uφ is obtained by quoting 'the Laplacian comparison theorem ... established by B. Shen. See Theorem 1.1 in [16]', but the theorem is not stated in the manuscript, and [16] is an arXiv preprint rather than a peer-reviewed reference. The bound is used for the solution-dependent direction ∇u, and the constant C0 is asserted to depend on K0, the bound on F(U)+F*(T)+F(div_C(V)). It is not apparent from the text whether Shen's theorem applies to this nonlinear setting, to this class of reference directions, and to cut-offs on forward geodesic balls with the stated constants. Because the B·L term in (3.23) is what makes the maximum-principle argument close, the validity of (1.2) is directly conditional on this unverified external result. The author should state Shen's theorem precisely and either prove the needed version or show explicitly that the hypotheses of [16, Theorem 1.1] are satisfied in the present situation.","section":"Section 3, cut-off argument before (3.23)"},{"comment":"The proof says: 'By Calabi’s trick [5], we can assume further that φ is smooth on the forward metric ball Bp(2R).' On a Finsler manifold the distance function r is only Lipschitz, and no approximation argument is supplied to justify the pointwise maximum-principle identities ∇∇uH(x0,t0)=0, Δ∇uH(x0,t0)≤0 at a maximum point that may lie at a non-smooth point of r. Since these identities are used to derive (3.23) and hence the final estimate, this is a load-bearing regularity gap. The discussion in Remark 3.4 asserts that one can approximate by smooth functions, but it does not explain how the maximum-principle step with the cut-off is recovered in the limit, particularly for the solution-dependent reference direction ∇u.","section":"Section 3, 'By Calabi's trick' before (3.23)"},{"comment":"Theorem 1.1 requires A > a_+ on Bp(2R)×(0,∞) and A to be a positive constant. In the proof of Theorem 4.1 the author applies Theorem 1.1 'in case 2 with A = a = 0', and in Theorem 4.2 sets A = a = 2. Both choices violate the hypothesis A > a_+ (since a_+=0 in the former and a_+=2 in the latter), and the expression (1.2) contains 4(A-[a]_+) in a denominator. Even if the case-2 bound (3.29) could be used formally without that hypothesis, the theorems as stated do not apply, and the transition from (1.2) to the Harnack and boundedness claims is not justified by the stated hypotheses.","section":"Section 4, Theorems 4.1 and 4.2"},{"comment":"The Harnack constant T in Theorem 4.1 is displayed as T = 4N[ B + NC1^2/R^2 + 2E/N + 1/2 (-[-2K-2-|logD|]_+) ], but the constant E is not defined in the theorem statement; it is introduced only inside the proof of Theorem 1.1 as a constant chosen large enough to satisfy (3.20) and (3.21). The theorem therefore does not give an explicit Harnack constant in terms of the data, and it is unclear how E depends on a, b, K0, α, D, and the other bounds. This needs to be clarified or the statement revised so that T is expressed entirely through the hypotheses.","section":"Section 4, Theorem 4.1 statement"}],"minor_comments":[{"comment":"After the inequality (1.2), the line 'ut/u = F(∇f)^2 + (A+a)f − 2ft' is incorrect as a definition of ut/u; it appears to be a garbled version of L/t = F^2(∇f)+(A+a)f+2(E+b)−2ft. Please correct the formula and make the roles of L, E, and M in the theorem statement coherent.","section":"Theorem 1.1 statement"},{"comment":"The notation [h]_+ is defined as a global supremum over Bp(2R)×(0,∞), but in (1.2) it is used pointwise, e.g., [a(x,t)]_+. These two uses are inconsistent; please introduce separate notation for the pointwise positive part and the global sup.","section":"Notation, beginning of Section 1 and use in (1.2)"},{"comment":"In the definition of global misalignment, the expression 'αM(x,M)' should presumably be 'αM(x,U)'; as written it is not meaningful.","section":"Definition 2.2"},{"comment":"The sentence 'Choose the same cut-off function φ as below' is confusing because no cut-off function was introduced before; it should say 'as follows' or the passage should be reworded.","section":"Section 3, after (3.22)"},{"comment":"The theorem writes '(L,F,µ)' for the manifold, whereas the rest of the paper uses M; please use consistent notation.","section":"Theorem 4.2 statement"},{"comment":"The constants C(N,α) and C0(K0,α) appear in the lower bound for Δ∇uφ but are never defined or characterized. Even aside from the external theorem, give at least the defining inequalities or the exact expressions used.","section":"Section 3, cut-off Laplacian bound"},{"comment":"The proof of Lemma 2.10(3) is very abbreviated: the displayed integration by parts in s is not carried out, and the reader is told to use the weak formulation of Definition 2.2 without seeing the details. This is acceptable as a sketch but should be expanded for a journal submission.","section":"Lemma 2.10(3) proof"}],"recommendation":"major_revision","confidential_remarks":"The central dependence on Shen's unpublished arXiv preprint [16] is a significant concern for a journal submission: the paper under review does not state the comparison theorem on which the main estimate rests, and the applications do not satisfy the stated hypotheses of the main theorem. I would encourage the editor to check the status of [16] and to ask the author to either prove the needed comparison result or state it in full with hypotheses, and to repair the constant choices in Theorems 4.1 and 4.2. The manuscript also needs careful proofreading of formulas and notation."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this is a genuine, incremental extension of Wang's Riemannian Li–Yau estimate to the Finslerian logarithmic Schrödinger equation. The proof is long and mostly coherent, and the two applications are sensible. But the main theorem is conditional on a Laplacian comparison theorem quoted from an unreviewed preprint, and the applications gloss over the strict A > a_+ hypothesis. Both issues are fixable, but they need referee attention.\n\nWhat's new: the specific estimate for a Finslerian logarithmic Schrödinger equation with time-dependent coefficients, plus a Harnack inequality and a boundedness result. The structure follows Wang and uses Shen's comparison results, so the novelty is incremental but real. I don't see any circularity in the derivation.\n\nWhere I have doubts. The cut-off argument in (3.23) needs a lower bound on Δ_{∇u}φ for a cut-off of the distance function, with reference direction ∇u. The paper cites “Theorem 1.1 in [16]” for this and does not state the theorem. Since [16] is an arXiv preprint, the proof is not self-contained, and if Shen's theorem does not apply to arbitrary reference directions like ∇u, the B·L term is unjustified. Calabi's trick is also invoked to make φ smooth without an approximation argument for the nonsmooth distance function, which is the same kind of gap. These are load-bearing, not cosmetic. A referee should verify Shen's comparison theorem and ask the author to state it precisely.\n\nSecond, the applications: Theorem 4.1 and Theorem 4.2 both apply the case bounds with A = a, which violates the strict A > a_+ hypothesis of Theorem 1.1. The paper's Remark 3.5 anticipates using the case bounds directly, but the text does not clearly separate the hypotheses needed for case 1 and case 2. This is a presentation issue, but it should be corrected.\n\nMinor: the constant M in Theorem 1.1 is never tied to the E used in the proof, and there are typographical inconsistencies in the key displays. These reduce reliability but are easy to fix.\n\nWho it's for: specialists in Finsler PDE and geometric analysis. It deserves a serious referee; with the comparison theorem stated or verified and the A = a issue clarified, it would be a solid contribution. I recommend sending it to peer review rather than desk-rejecting.","headline":"A serious but conditional Finslerian extension of Wang's Li–Yau estimate; the main theorem leans on an unstated comparison theorem from an unreviewed preprint, and the applications violate the strict A > a_+ hypothesis.","tokens_in":18867,"tokens_out":6027,"would_cite":false,"duration_ms":58206,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35K55","53C60","58J35"],"pacs":[],"model":"deepseek-v4-flash","headline":"A Li-Yau gradient estimate holds for the Finslerian logarithmic Schrödinger equation under mixed weighted Ricci curvature bounded below, yielding Harnack and boundedness results.","keywords":["gradient estimate","Li-Yau estimate","logarithmic Schrödinger equation","Finsler geometry","mixed weighted Ricci curvature","Harnack inequality","Laplacian comparison theorem","boundedness"],"falsifier":"Take a forward complete Finsler manifold with $m\\mathrm{Ric}_N\\ge -K$, for instance a Finsler metric on $\\mathbb{R}^n$ with a prescribed S-curvature, and an explicit positive bounded solution of (1.1). Compute both sides of (1.2) on a grid in $B_p(R)\\times(0,\\infty)$; if the left side ever exceeds the right side while the tensor bound $F(U)+F^*(T)+F(\\mathrm{div}_C(V))\\le K_0$ holds, the theorem is false. A more targeted check is to verify the cut-off Laplacian bound $\\Delta_{\\nabla u}\\phi \\ge -B$ used in the proof on that metric, since that is the step where the comparison theorem enters.","tokens_in":17854,"feed_emoji":"📐","tokens_out":11164,"duration_ms":103130,"temperature":0.7,"pith_summary":"This paper establishes a Li-Yau type gradient estimate for positive bounded solutions of the logarithmic Schrödinger equation on forward complete non-compact Finsler manifolds, a setting where the metric is a norm rather than a quadratic form and the Laplacian is nonlinear. The main theorem bounds the combination $F^2(\\nabla u)/u^2 + (A+a)\\log(u/D) - 2u_t/u$ by an explicit expression built from the time $t$, the Ricci lower bound $K$, the radius $R$, and the bounds on the coefficients and non-Riemannian tensors. It matters because such pointwise gradient control is the usual bridge from curvature assumptions to quantitative analytic consequences: here it yields a Harnack inequality comparing values of a solution at different points and times, and an explicit a priori bound for stationary solutions. The estimate also covers coefficients $a(x,t)$ and $b(x,t)$ that depend on both position and time, rather than being constant.","feed_headline":"Curvature yields gradient bound for Finsler log-Schrödinger flow","feed_subtitle":"Pointwise Li-Yau inequality gives Harnack and boundedness for positive solutions.","key_machinery":"The central mechanism is the maximum-principle cut-off argument applied to $H=\\phi L$ with $L=t[F^2(\\nabla f)+(A+a)f+2(E+b)-2f_t]$ and a cut-off $\\phi$ equal to 1 on $B_p(R)$ and 0 outside $B_p(2R)$. At the maximum point, the proof controls $\\phi\\Delta_{\\nabla u}L$ from below by $BL$; the constant $B$ comes from a new Laplacian comparison theorem for Finsler distance functions, together with the reverse triangle inequality for the misalignment $\\alpha$ and the uniform tensor bound $F(U)+F^*(T)+F(\\mathrm{div}_C(V))\\le K_0$. The quadratic-in-$L$ term in the Bochner inequality, involving $h=F^2(\\nabla f)/L$, is what forces the $4N[\\cdots]$ structure of the final estimate. Large constants $A,E$ are chosen so that lower-order coefficient terms become nonnegative, allowing the two-case split that completes the bound.","core_discovery":"The paper proves Theorem 1.1: if $u$ is a positive bounded solution of $(\\Delta-\\partial_t)u + a(x,t)u\\log u + b(x,t)u = 0$ on a forward complete non-compact Finsler manifold with finite misalignment, $m\\mathrm{Ric}_N \\ge -K$ on $B_p(2R)$, and $F(U)+F^*(T)+F(\\mathrm{div}_C(V))\\le K_0$, then with $f=\\log(u/D)\\le 0$ the quantity $F^2(\\nabla u)/u^2 + (A+a)f - 2u_t/u$ is bounded on $B_p(R)\\times(0,\\infty)$ by the explicit right-hand side of (1.2): a combination of $4N[1/t + [a]_+ + B + N C_1^2/R^2 + 2[M-a\\log D]_+/N]$ and a term built from $-[A-2K-2-|\\log D|]_+$ and $[\\Delta_{\\nabla u}a + a_t]_+/(4(A-[a]_+))$. The proof runs the maximum principle on $\\phi L$, where $L=t[F^2(\\nabla f)+(A+a)f+2(E+b)-2f_t]$, using the Bochner formula and a new Laplacian comparison theorem to control $\\Delta_{\\nabla u}\\phi$. As applications the paper derives a local Harnack inequality (Theorem 4.1) and, for the stationary equation $\\Delta u+2u\\log u+V(x)u=0$ with $m\\mathrm{Ric}_N\\ge 0$, an explicit boundedness estimate for positive solutions (Theorem 4.2).","pith_inferences":["Not pursued in the paper: the same cut-off argument should adapt to other nonlinear Finslerian parabolic equations whose nonlinearity is a Lipschitz perturbation of the Laplacian, as long as the same comparison and tensor bounds hold.","The two-case split in the proof suggests the constant $4N$ and the term involving $-(A-2K-2-|\\log D|)_+$ are not optimized; a reader could try to sharpen them by choosing $A$ and the cut-off differently.","A direct consequence the author leaves implicit is that letting $R\\to\\infty$ under global curvature and tensor bounds converts the local estimate into a global one, in the spirit of the stationary boundedness application; this is a testable route to Liouville-type results under mild growth assumptions on solutions.","Restricting the new Laplacian comparison theorem to a Riemannian metric should make the non-Riemannian tensor terms vanish and recover the Riemannian predecessor estimate, which would isolate exactly what the Finsler generalization adds."],"forward_implications":["For the heat-type equation $(\\Delta-\\partial_t)u + b(x,t)u=0$, the gradient estimate directly implies the local Harnack inequality $u(x_1,t_1)\\le u(x_2,t_2)(t_2/t_1)^{2N}\\exp\\{(t_2-t_1)T+S(x_1,x_2,t_2-t_1)\\}$ for $x_1,x_2\\in B_p(R)$ and $0\\le t_1\\le t_2$.","For the stationary equation $\\Delta u+2u\\log u+V(x)u=0$ with $m\\mathrm{Ric}_N\\ge 0$ and bounded $V$, $\\Delta_{\\nabla u}V$, and $F_{\\nabla u}(\\nabla_{\\nabla u}V)$, every positive solution is bounded above by an explicit exponential of the bounds on $V$ and its Hessian and gradient.","The estimate is local and quantitative: it holds on $B_p(R)\\times(0,\\infty)$ and degenerates in the expected way as $t\\to 0$ through $1/t$ and as $R\\to\\infty$ through the constant $B$.","Because the coefficients $a$ and $b$ are allowed to be time-dependent, the theorem covers parabolic equations not accessible to earlier constant-coefficient treatments."],"supporting_citations":[{"why":"Supplies the new Laplacian comparison theorem for Finsler distance functions and the regularity theory for the Finslerian logarithmic Schrödinger equation.","marker":"[16]"},{"why":"Provides the Bochner formula, definitions of the Finsler Laplacian and Hessian, and the weighted Ricci curvature framework used throughout the proof.","marker":"[13]"},{"why":"The Riemannian Li-Yau gradient estimate whose method and theorem this paper generalizes to the Finsler setting.","marker":"[19]"},{"why":"Li-Yau estimates on Finsler manifolds and the mollification and regularization argument behind Lemma 2.10.","marker":"[20]"},{"why":"Prior Li-Yau estimates for a nonlinear parabolic equation on Finsler metric measure spaces, giving the comparison point for the new result.","marker":"[17]"},{"why":"The standard smoothing device that lets the cut-off function be treated as smooth on the forward geodesic ball.","marker":"[5]"}],"fun_headline_variants":["Weighted Ricci yields Li-Yau bound for Finsler log-Schrödinger","Li-Yau bound gives Harnack and boundedness for Finsler flow","Curvature yields Li-Yau and Harnack for Finsler log-Schrödinger","Finsler log-Schrödinger flow: Li-Yau bound from curvature","Gradient bound and Harnack for Finsler log-Schrödinger"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The estimate depends on a comparison theorem for the Laplacian of distance on Finsler manifolds being available, and on the non-Riemannian correction tensors being uniformly bounded; if either condition fails, the lower bound on the cut-off Laplacian and therefore the final inequality do not follow.","fun_headline_variants_meta":{"raw":{"variants":["Weighted Ricci yields Li-Yau bound for Finsler log-Schrödinger","Li-Yau bound gives Harnack and boundedness for Finsler flow","Curvature yields Li-Yau and Harnack for Finsler log-Schrödinger","Finsler log-Schrödinger flow: Li-Yau bound from curvature","Gradient bound and Harnack for Finsler log-Schrödinger"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001651,"raw_usage":{"total_tokens":6578,"prompt_tokens":990,"completion_tokens":5588,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":606,"completion_tokens_details":{"reasoning_tokens":5487}},"tokens_in":606,"tokens_out":5588,"duration_ms":38142,"temperature":1.0,"reasoning_tokens":5487,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T19:42:59.121837+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a forward complete Finsler manifold with $m\\mathrm{Ric}_N\\ge -K$, for instance a Finsler metric on $\\mathbb{R}^n$ with a prescribed S-curvature, and an explicit positive bounded solution of (1.1). Compute both sides of (1.2) on a grid in $B_p(R)\\times(0,\\infty)$; if the left side ever exceeds the right side while the tensor bound $F(U)+F^*(T)+F(\\mathrm{div}_C(V))\\le K_0$ holds, the theorem is false. A more targeted check is to verify the cut-off Laplacian bound $\\Delta_{\\nabla u}\\phi \\ge -B$ used in the proof on that metric, since that is the step where the comparison theorem enters.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the Bochner formula, definitions of the Finsler Laplacian and Hessian, and the weighted Ricci curvature framework used throughout the proof."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"The Riemannian Li-Yau gradient estimate whose method and theorem this paper generalizes to the Finsler setting."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Li-Yau estimates on Finsler manifolds and the mollification and regularization argument behind Lemma 2.10."},{"cited_title":"and Zhu, Y., Li–Yau estimates for a nonlinear parabolic equation on Finsler metric measure spaces, Manuscripta Math","cited_arxiv_id":null,"evidence_quote":"Prior Li-Yau estimates for a nonlinear parabolic equation on Finsler metric measure spaces, giving the comparison point for the new result."},{"cited_title":"Hopf’s maximum principle with an application to Riemannian geometry, Duke Math","cited_arxiv_id":null,"evidence_quote":"The standard smoothing device that lets the cut-off function be treated as smooth on the forward geodesic ball."}],"review_version":2}