{"id":"9a07dacf-18bf-4e85-80e3-aa49c696e00d","arxiv_id":"1908.00595","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"For self-adjoint operators comparable to a positive-homogeneous semi-elliptic operator, the heat kernel satisfies off-diagonal bounds with the Legendre-Fenchel transform of the symbol, under three hypotheses.","lead":"This paper develops a variant of Davies' method to prove off-diagonal heat kernel estimates for anisotropic (semi-elliptic) differential operators. The main theorem gives exponential decay bounds involving the Legendre-Fenchel transform of the operator's symbol, under three abstract hypotheses.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection: Theorem 8.2 is sound under its stated hypotheses; Hypothesis 6.3 is restrictive but honestly qualified as a scope limitation.","rationale":"The reader's weakest-assumption identification (Hypothesis 6.3) is accurate: it is the restrictive condition that lifts the μ_Λ < 1 restriction, and the paper verifies it only under extra smoothness and constant-principal-coefficient assumptions. However, this is a limitation of scope, not a flaw in the conditional theorem. My independent review of the proof of Theorem 8.2 found the chain of estimates sound, including the use of the Gagliardo-Nirenberg inequality for Λ^κ, the application of Hypothesis 6.3, and the absorption of lower-order terms. The reader's stated reasons for CONDITIONAL were presentation gaps: the 'careful study reveals' statement in Theorem 9.2 and omitted proofs of Lemmas 4.1 and 6.3. These are addressable and do not bear on the correctness of the central estimate. In particular, the uniformity-in-s constants in Theorem 9.2 follow from Definition 9.1, which requires the scaled forms to satisfy Hypotheses 6.1-6.2 with the same constants for all s; the subsequent lemmas depend only on those constants and on μ_Λ, so the uniformity is automatic. The omitted proofs are routine. Therefore my read does not change the reader's verdict: the paper merits conditional acceptance, but the conditions are presentational, not mathematical. The concrete test proposed targets the one genuinely delicate omitted step, the W-term bound in Lemma 11.6, to ensure that the verification of Hypothesis 6.3 under (C.4)-(C.5) has no hidden gap.","tokens_in":35015,"tokens_out":30414,"duration_ms":260442,"concrete_test":"As a verification step, independently re-derive the omitted W-term estimate in Lemma 11.6: explicitly bound |W(λ,φ,f)| by ε Q_{Λ^κ}(f) + M_ε(1+R(λ))^κ ||f||^2 using Lemma A.5 item 2b for all admissible multi-indices with |α+β : m| < 2κ. If the 'similar argument' fails for any admissible combination of multi-indices, then Hypothesis 6.3's verification under (C.4)-(C.5) would be incomplete, which would affect Proposition 11.7 and hence the concrete applicability of Theorem 8.2 to super-semi-elliptic operators.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is a conditional theorem: under Hypotheses 6.1, 6.2 and 6.3, Theorem 8.2 establishes off-diagonal heat-kernel estimates with the Legendre-Fenchel transform of the reference symbol. I traced the proof of Theorem 8.2 and found no internal inconsistency. The proof correctly uses Lemma 5.3 for the positive-homogeneous operator Λ^κ (whose homogeneous order is μ_Λ/κ < 1 by the choice of κ), applies Hypothesis 6.3 to f_t = e^{-tH_{λ,φ}}f, and absorbs the (1+R(λ))^{μ/2} factor into t^{-μ/2} and the exponential by enlarging M; the final minimization over λ in Lemma 8.1 is valid. The most restrictive condition, Hypothesis 6.3, is exactly where the paper's own remarks (Section 11.2 and Remarks 8-10) acknowledge limited verification: it is established only under (C.4)-(C.5), i.e., smooth coefficients and constant principal part, and the optimal smoothness is left open. This narrows the advertised extension to bounded measurable coefficients when μ_Λ ≥ 1, but this is a scope limitation, not a defect in the central proof. The alleged presentation gaps (Theorem 9.2's 'careful study reveals' and the omitted proofs of Lemmas 4.1 and 6.3) are fillable: Lemma 6.3 follows immediately from Lemma 6.2 and Hypothesis 6.2, Lemma 4.1 is a standard Fourier characterization of anisotropic Sobolev spaces, and the uniformity-in-s constants in Theorem 9.2 depend only on the constants in Hypotheses 6.1-6.2, which are identical for the scaled forms by Definition 9.1. No load-bearing flaw was identified.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops an anisotropic analogue of Davies' perturbation method for heat-kernel estimates. It introduces positive-homogeneous constant-coefficient operators as reference operators and considers self-adjoint variable-coefficient operators whose forms are comparable to such a reference operator. Under three hypotheses (form comparability, a twisted-form comparison inequality with a separating family of functions, and a high-power perturbation estimate), Theorem 8.2 proves an off-diagonal estimate for the heat kernel with the Legendre-Fenchel transform R# of the reference symbol. Section 9 removes the exponential factor Mt for homogeneous operators when μΛ<1; Section 10 proves Hölder regularity and analytic continuation under the same restriction; Section 11 applies the theory to super-semi-elliptic operators, giving a bounded measurable coefficient result when μΛ<1 and a result for μΛ≥1 under additional smoothness and constant-principal-part assumptions.","tokens_in":35403,"tokens_out":8066,"duration_ms":83505,"significance":"If the results stand, the paper substantially extends Davies' elliptic theory to a natural anisotropic setting, providing a unified mechanism for off-diagonal estimates governed by the Legendre-Fenchel transform. The proof of Theorem 8.2 is detailed and checkable: the twisted semigroup bounds, the ultracontractive step via Lemma 5.3, and the final minimization over λ are coherent, and the constants are tracked in terms of the structural hypotheses. The paper is honest about the restrictive nature of Hypothesis 6.3 and explicitly leaves the optimal smoothness open (Remark 9); this is a scope limitation rather than a defect. There are no fitted parameters, and the central estimate is a genuine conditional theorem not already contained in the authors' prior work [23].","major_comments":[],"minor_comments":[{"comment":"The phrase 'a careful study reveals' is not a proof that every constant in the preceding lemmas is uniform in s. Since Definition 9.1 supplies exactly the uniformity needed, this is fillable, but the argument should be written out: one should state explicitly that the constants in Lemmas 7.1-7.3, Lemma 8.1 and Theorem 8.2 depend only on the constants appearing in Hypotheses 6.1-6.2, and that those constants are identical for Q_s by Definition 9.1.","section":"§9, proof of Theorem 9.2"},{"comment":"The introduction's statement that the article extends the theory to operators with bounded measurable coefficients is too broad: for μΛ ≥ 1, Hypothesis 6.3 is verified only under the additional smoothness condition (C.4) and constant principal coefficients (C.5). The abstract and introduction should qualify the bounded measurable coefficient claim so that the conditional scope of the μΛ ≥ 1 results is visible from the outset.","section":"§11.2 and Remarks 8-10"},{"comment":"Lemma 4.1 is stated without proof. A short Fourier argument or an explicit reference for the anisotropic Sobolev space characterization would make the paper more self-contained, especially because the equivalence of norms in Lemma 4.2 relies on it.","section":"§4, Lemma 4.1"},{"comment":"The proof of Lemma 6.3 is omitted. One sentence using Lemma 6.2 and Hypothesis 6.2 would suffice, since (13) follows immediately from (12) and the representation Q_{λ,φ}(f) = ⟨H_{λ,φ}f,f⟩.","section":"§6, Lemma 6.3"},{"comment":"In the chain of inequalities leading to the ultracontractive bound, the absorption of the factor (1+R(λ))^{μΛ/2} into t^{-μΛ/2} and the exponential exp(M(1+R(λ))t/2) is not immediate and requires an explicit justification; as written, it seems to involve an enlargement of M and a constant depending on κ and μΛ.","section":"§8, proof of Theorem 8.2"},{"comment":"The final sentence of the proof states the bound for all x,y ∈ V, but the kernel is defined on Ω; it should read x,y ∈ Ω.","section":"§8, Lemma 8.1"},{"comment":"The heading 'When μΛ = |1, 2m| ≥ 1' contains a typo; the notation used elsewhere is |1 : 2m|.","section":"§11.2, heading"}],"recommendation":"minor_revision","confidential_remarks":"The paper is sound in its central claim: Theorem 8.2 is a valid conditional theorem, and the proof has been traced without finding a load-bearing error. The reliance on the authors' prior paper [23] for Proposition 3.3 and Legendre-Fenchel properties is legitimate and does not make the main argument circular. The main caveat for the editor is scope: the advertised extension to bounded measurable coefficients really only holds for μΛ < 1; for larger homogeneous orders the verification of Hypothesis 6.3 requires (C.4)-(C.5), and this should be made more prominent in the abstract and introduction."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: this paper does what it claims. It extends Davies' functional-analytic method for off-diagonal heat kernel estimates to positive-homogeneous (semi-elliptic) operators, and the main theorem (8.2) is correct as stated. The proof is careful, and the abstract framework with the twisted semigroup and Legendre-Fenchel optimization generalizes cleanly. This is a real new result, not a repackaging of the earlier Levi-parametrix work [23].\n\nWhat I found strongest: the paper is honest about what it does and doesn't prove. Hypothesis 6.3 is explicitly called restrictive, and the verification in Section 11.2 requires smooth coefficients and constant principal part. Remarks 8-10 spell out the open questions. That level of candor is rare and welcome. The reliance on [23] for the positive-homogeneous operator machinery is appropriate; the self-citation is there because the framework genuinely comes from that paper, while Theorem 8.2 builds something new on top.\n\nThe soft spots are minor and presentation-level. Lemma 4.1's proof is omitted but the statement is a standard Fourier characterization. Lemma 6.3 follows almost immediately from Lemma 6.2 and Hypothesis 6.2. In Theorem 9.2, the uniformity in s is asserted via 'a careful study reveals' rather than shown; this is the one place I'd like to see the details spelled out. None of these affect the central argument, and the stress-test note is right that they are fillable.\n\nWho gets value: anyone working on higher-order heat kernel estimates, anisotropic Sobolev spaces, or semigroup methods for non-elliptic operators. It deserves serious refereeing. I would send it to a strong analyst familiar with Davies' method and expect it to be accepted after minor revision. The paper is a useful, careful extension, and the limitations are stated honestly.","headline":"A genuine and careful extension of Davies' method to semi-elliptic operators, with a sound central theorem and honest limitation statements; only minor presentation gaps.","tokens_in":35911,"tokens_out":2956,"would_cite":true,"duration_ms":29764,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35K08","35K25","35H30"],"pacs":[],"model":"deepseek-v4-flash","headline":"Heat kernels of semi-elliptic operators obey off-diagonal estimates whose exponent is the Legendre-Fenchel transform of the operator's symbol, under three hypotheses on the operator and its powers.","keywords":["heat kernel estimates","semi-elliptic operators","quasi-elliptic operators","Legendre-Fenchel transform","positive-homogeneous operators","Davies method","off-diagonal estimates","anisotropic operators"],"falsifier":"A concrete test: on $R^{3}$ take the reference operator Λ = −∂_{x1}^2 − ∂_{x2}^2 + ∂_{x3}^6, whose homogeneous order is µΛ = 7/6 > 1, and add a bounded measurable, non-smooth coefficient perturbation satisfying Conditions (C.1)–(C.3). Check whether Hypothesis 6.3 holds for κ = 2; if it fails while Hypotheses 6.1 and 6.2 hold, the theorem does not apply, and a numerical check of whether the heat kernel obeys the bound (15) would show whether the µΛ ≥ 1 regime is genuinely inaccessible without extra smoothness.","tokens_in":34802,"feed_emoji":"🔥","tokens_out":10087,"duration_ms":99231,"temperature":0.7,"pith_summary":"This paper extends Davies' twisted-semigroup method for heat-kernel estimates from uniformly elliptic higher-order operators to semi-elliptic (anisotropic) operators built from positive-homogeneous constant-coefficient models. The target is an off-diagonal Gaussian-type bound in which the anisotropic distance is encoded by the Legendre-Fenchel transform R# of the reference symbol R, with prefactor $t^{{-µΛ}}$ where µΛ is the homogeneous order. The paper shows that three hypotheses — form comparability with a reference positive-homogeneous operator, a twisted form-comparison inequality, and a perturbation estimate for a suitable power of H — are enough to produce the kernel and the bound. The third hypothesis is the load-bearing one: it lifts the old restriction µΛ < 1 (the analogue of d/2m < 1) and is verified here only under extra smoothness and constant principal coefficients, leaving optimal smoothness open.","feed_headline":"Semi-elliptic heat kernels obey Legendre-Fenchel estimates","feed_subtitle":"Extension of Davies' technique proves off-diagonal kernel bounds for anisotropic operators, even when the homogeneous order exceeds 1.","key_machinery":"The machinery is Davies' twisted-semigroup method recast in a coordinate-free, multi-parameter form. One conjugates $e^{{-tH}}$ by multiplication operators $e^{{±λ(φ)}}$ for λ in the dual space and φ chosen so that φ(x) − φ(y) = x − y; the twisted form is Q_{λ,φ}(f) = Q($e^{{-λ(φ)}}$f, $e^{{λ(φ)}}$f). Hypotheses 6.1 and 6.2 control these twisted forms uniformly, yielding exponential $L^{2}$ bounds in (1 + R(λ)) t. Hypothesis 6.3, a perturbation estimate for the κ-th power of H with κ = min{n : µΛ/n < 1}, supplies the ultracontractivity needed to convert those $L^{2}$ bounds into $L^{1}$ → L^∞ kernel bounds; optimizing over λ then produces the Legendre-Fenchel transform R#. The homogeneous-order parameter µΛ = tr E plays the role of d/2m and controls the on-diagonal decay.","core_discovery":"The central result is Theorem 8.2: if a self-adjoint variable-coefficient operator H is comparable to a positive-homogeneous operator Λ with symbol R and homogeneous order µΛ, and if the twisted semigroups $e^{{λ(φ)}}$ $e^{{-tH}}$ $e^{{-λ(φ)}}$ satisfy the form bounds of Hypotheses 6.1, 6.2 and 6.3, then $e^{{-tH}}$ has an integral kernel K_H satisfying |K_H(t,x,y)| ≤ C $t^{{-µΛ}}$ exp(−t M R#((x−y)/t) + M t) for all x,y in the domain and all t>0. The exponent R# is the support function sup_λ {λ((x−y)/t) − M R(λ)}, so the shape of the off-diagonal decay is forced by the symbol's own dilation geometry. The paper also proves Hölder regularity and analytic continuation of the kernel when µΛ < 1, removes the M t term when H is homogeneous in the dilation sense, and applies the abstract theorem to super-semi-elliptic operators, recovering known elliptic results as special cases.","pith_inferences":["If Hypothesis 6.3 can be verified for coefficients of finite smoothness, the same framework would resolve the open question identified in the paper and put genuinely variable-coefficient anisotropic operators with µΛ ≥ 1 on the same footing as the elliptic theory.","The Legendre-Fenchel exponent suggests that the correct off-diagonal 'distance' for semi-elliptic heat kernels is not Euclidean but the support function of the symbol's level sets; one could test this numerically for Λ = −∂_{x1}^2 + ∂_{x2}^4 on R^2 by comparing the kernel's spatial decay with R#.","The homogeneous-case scaling argument that removes the M t term may extend to yield two-sided estimates or sharp constants by optimizing over the dilation parameter s, a step the paper does not take.","Because the bound implies L^p holomorphy and p-independent spectra, the hypotheses provide a template for studying functional calculi and Riesz transforms of anisotropic divergence-form operators with measurable coefficients."],"forward_implications":["Under Hypotheses 6.1–6.3, the kernel bound gives L^1 → L^∞ control with the anisotropic distance t R#((x−y)/t), so the semigroup e^{-tH} extends to a strongly continuous semigroup on L^p for all 1 ≤ p < ∞ with spectrum independent of p.","When µΛ < 1, Hypothesis 6.3 is automatic, so the estimates hold for every super-semi-elliptic operator with bounded measurable coefficients; in addition the kernel is jointly Hölder continuous of order (1 − µΛ)/2 and extends analytically in time to the right half-plane.","If H is homogeneous under the anisotropy of Λ and µΛ < 1, the spurious M t term disappears, giving the sharper bound |K_H(t,x,y)| ≤ C t^{-µΛ} exp(−t M R#((x−y)/t)).","For super-semi-elliptic operators with smooth coefficients and constant principal coefficients, the estimate holds for arbitrary µΛ, covering anisotropic operators of high homogeneous order."],"supporting_citations":[{"why":"Supplies the abstract twisted-semigroup method for uniformly elliptic operators with measurable coefficients that the paper adapts to the semi-elliptic setting.","marker":"[9]"},{"why":"Introduces positive-homogeneous operators, the homogeneous order µΛ, and the Legendre-Fenchel transform R#, and gives the parametrix-based kernel estimates that this paper extends to measurable coefficients.","marker":"[23]"},{"why":"Provides the scaling argument used in Section 9 to remove the M t term for homogeneous H.","marker":"[2]"},{"why":"Establishes dimension-order counterexamples showing that without extra hypotheses the estimate can fail when d/2m > 1, motivating Hypothesis 6.3.","marker":"[8]"},{"why":"Gives the approach for the limiting case µΛ = 1 that the paper's theory partially extends and cites as an alternative route.","marker":"[26]"}],"fun_headline_variants":["Davies' heat-kernel bounds now cover anisotropic operators","Heat-kernel estimates for anisotropic operators via Legendre-Fenchel","Off-diagonal heat-kernel bounds in semi-elliptic settings","Extending Davies' heat-kernel method to anisotropic PDEs"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole scheme rests on Hypothesis 6.3, the assumption that the reference form's κ-th power is controlled by the twisted κ-th power of H; the paper itself calls it 'much more subtle, difficult to verify and restrictive,' and verifies it only when the coefficients are smooth and the principal coefficients are constant.","fun_headline_variants_meta":{"raw":{"variants":["Davies' heat-kernel bounds now cover anisotropic operators","Heat-kernel estimates for anisotropic operators via Legendre-Fenchel","Off-diagonal heat-kernel bounds in semi-elliptic settings","Extending Davies' heat-kernel method to anisotropic PDEs"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000583,"raw_usage":{"total_tokens":2761,"prompt_tokens":983,"completion_tokens":1778,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":599,"completion_tokens_details":{"reasoning_tokens":1706}},"tokens_in":599,"tokens_out":1778,"duration_ms":10985,"temperature":1.0,"reasoning_tokens":1706,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T15:45:41.231617+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A concrete test: on $R^{3}$ take the reference operator Λ = −∂_{x1}^2 − ∂_{x2}^2 + ∂_{x3}^6, whose homogeneous order is µΛ = 7/6 > 1, and add a bounded measurable, non-smooth coefficient perturbation satisfying Conditions (C.1)–(C.3). Check whether Hypothesis 6.3 holds for κ = 2; if it fails while Hypotheses 6.1 and 6.2 hold, the theorem does not apply, and a numerical check of whether the heat kernel obeys the bound (15) would show whether the µΛ ≥ 1 regime is genuinely inaccessible without extra smoothness.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the abstract twisted-semigroup method for uniformly elliptic operators with measurable coefficients that the paper adapts to the semi-elliptic setting."},{"cited_title":"Positive-homogeneous operators, heat kernel estimates and the Legendre-Fenchel transform","cited_arxiv_id":null,"evidence_quote":"Introduces positive-homogeneous operators, the homogeneous order µΛ, and the Legendre-Fenchel transform R#, and gives the parametrix-based kernel estimates that this paper extends to measurable coefficients."},{"cited_title":"Barbatis and E.B","cited_arxiv_id":null,"evidence_quote":"Provides the scaling argument used in Section 9 to remove the M t term for homogeneous H."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes dimension-order counterexamples showing that without extra hypotheses the estimate can fail when d/2m > 1, motivating Hypothesis 6.3."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the approach for the limiting case µΛ = 1 that the paper's theory partially extends and cites as an alternative route."}],"review_version":1}