{"id":"d76c0f9c-0d71-4723-91fd-589ff6de07c0","arxiv_id":"2608.08930","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Heat kernels on Abelian covers and their principal-bundle lifts share the same leading Gaussian large-time profile, with all polynomial coefficients determined by the base cover.","lead":"This mathematics paper derives sharp long-time asymptotics for heat kernels on Abelian covers of compact manifolds and of principal bundles. It shows that on principal bundles, after exponentially small fiber corrections, heat propagation follows the same Gaussian profile as the underlying Abelian cover.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Uniform subelliptic estimate (4.12) in Lemma 4.9 is asserted rather than proved; the C^ℓ-exponential remainder bounds in Theorems 1.5–1.6 collapse to L^2 bounds if the θ-uniform constants fail.","rationale":"The reader's weakest assumption identifies the same place where the main new conclusion loses its force. Theorems 1.5 and 1.6 are strictly stronger than Theorem 1.4: they assert pointwise and diffusive expansions with C^ℓ remainders. The reduction H^{P,H}_θ = H^M_θ + E_θ in equation (4.14) is exact and elementary; the only input needed beyond the base theorems is that E_θ decays exponentially in C^ℓ uniformly in θ. Lemma 4.9 is precisely that input. The proof of Theorem 4.3 gives the uniform L^2 gap in detail, and the Borel–Weil/Floquet framework is standard. However, the uniform subelliptic estimate (4.12) is asserted, not established. The compactness and smooth-dependence argument is not a proof of uniform constants: compactness can turn pointwise-in-θ subellipticity into local uniform bounds only if the constants are already continuous in θ, which requires the standard parameter-dependent subelliptic theorem or an explicit bracket computation; the paper supplies neither. If the constants in (4.12) are finite for each θ but unbounded as θ varies, then the L^2→H^m smoothing bounds for e^{-∆H_θ/2} are not uniform, and the C^ℓ exponential decay fails, leaving only the distributional Theorem 1.4. The identification of the coefficients k^P_j = k^M_j and P^P_j = P^M_j could still be true pointwise, but the asserted uniform differentiability of the remainders would be unsupported. This is a conditional-acceptance gap, not a fatal flaw: the ingredients are standard and likely fillable, so the verdict should remain CONDITIONAL, i.e. UNCHANGED relative to the reader. I found no stronger objection. Lemma 4.7's wording that density implies the intersection equals G×{0} is imprecise, since density of the intersection in the open subset would suffice for the argument, but this does not affect the conclusion.","tokens_in":38012,"tokens_out":16579,"duration_ms":165292,"concrete_test":"Use the standard subelliptic estimate with parameters (Hörmander, The Analysis of Linear Partial Differential Operators III, Thm 22.2.1) to derive (4.12) directly, without citing [CL24] mutatis mutandis: on a finite atlas of P0 trivializing Lθ with the smooth families ηθ from Proposition 2.1, verify that the horizontal vector fields satisfy the Hörmander bracket condition with step bounded uniformly in θ, and that the first-order terms involving ηθ have uniformly bounded coefficients. Then compute the resulting constants C_m and confirm they are independent of θ. Apply this check to the model G=U(1), P0=S^1×S^2 with the standard monopole connection (curvature B=ω_{S^2}) and d=1; if the constants stay finite uniformly in θ, Lemma 4.9 is sound, while if they diverge, the C^ℓ remainder claims in Theorems 1.5–1.6 must be weakened.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim needs the remainder E_P(t,u,v) of e^{-t∆H_P} orthogonal to G-invariant functions to decay exponentially in C^{ℓ_u,ℓ_v}(K_u×K_v) in Theorems 1.5 and 1.6. Lemma 4.9 is the only place where this is obtained. Its proof has two inputs: (i) the uniform L^2 spectral gap (4.11), which is proved via Theorem 4.3; and (ii) the uniform subelliptic inequality (4.12), ∥u∥_{H^m(P0)} ≤ C_m ∥(Id+∆H_θ)^{r_m}u∥_{L^2(P0)}. Inequality (4.12) is not proved. The text asserts that because ∆H_θ and ∆H_0 differ by first-order terms and all data are smooth on compact P0×U(1)^d, the constants are uniform, with a 'mutatis mutandis' citation to [CL24]. That is plausible but non-obvious: subellipticity of an individual operator does not automatically give quantitative estimates with constants continuous in a parameter unless one tracks the Hörmander bracket step and the lower-order perturbations, and the estimates must also be uniform across the family of flat bundles Lθ after local trivializations. If (4.12) fails, the short-time smoothing bound ∥e^{-∆H_θ/2}∥_{L^2→H^m}≤C_m can fail, and the exponential L^2 decay gives only exponential decay in L^2, not in C^ℓ; the pointwise expansions in Theorems 1.5 and 1.6, with their differentiated remainders, would not follow. This is exactly the kind of uniformity that the rest of the paper proves carefully for λ0(θ,k), so the gap is specific and fillable, but it should be supplied.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper studies large-time asymptotics of heat kernels on Abelian covers of compact manifolds and on Abelian covers of principal bundles equipped with a horizontal Laplacian. For Abelian covers, it proves three complementary expansions: a distributional expansion of the heat semigroup, a local pointwise expansion on compact sets, and a global diffusive expansion exhibiting a Gaussian profile. For principal bundles, under density of holonomy in G×Z^d and global nondegeneracy of the curvature, it proves analogous expansions for the horizontal heat kernel and shows that the leading coefficients coincide with those of the underlying Abelian cover, with exponentially decaying corrections from nontrivial fiber modes. The proofs combine Floquet theory with the Borel–Weil calculus.","tokens_in":38389,"tokens_out":7044,"duration_ms":66269,"significance":"If correct, the bundle-level results significantly extend the classical Lott and Kotani–Sunada theory by providing full asymptotic expansions that are uniform simultaneously in the Abelian and compact-group variables. The paper contains a genuinely new spectral gap theorem for the twisted Borel–Weil operators (Theorem 4.3), a detailed proof of the Abelian-cover expansions, an explicit computation of the first correction coefficient C1, and a nonvanishing result for all C_j. The main weakness is that a central uniformity estimate controlling the C^ℓ remainders is asserted rather than proved, leaving a gap that is structurally important but appears fillable.","major_comments":[{"comment":"The uniform subelliptic estimate (4.12) is asserted rather than proved. The proof states that because Δ^H_θ and Δ^H_0 differ by first-order terms and all data are smooth on compact P_0×U(1)^d, the constants are uniform, with a 'mutatis mutandis' reference to [CL24]. This is not automatic: subellipticity of an individual operator does not by itself give quantitative estimates with constants continuous in a parameter unless one tracks the Hörmander bracket step and the lower-order perturbations, and here the estimates must also be uniform across the family of flat bundles L_θ after local trivializations. This estimate is load-bearing: it is the only input that upgrades the exponential L^2 decay from Theorem 4.3 to the C^{ℓ_u,ℓ_v} remainder bounds in Theorems 1.5 and 1.6. If (4.12) fails, the short-time smoothing bound ∥e^{-Δ^H_θ/2}∥_{L^2→H^m} ≤ C_m can fail, and the pointwise expansions with differentiated remainders would not follow. Please supply a proof or a precise citation with the uniformity statement.","section":"§4.6, Lemma 4.9, Eq. (4.12)"}],"minor_comments":[{"comment":"The notation in the statement of Lemma 4.9 is confusing: 'P_G F(u) := ∫_G F(ug)dg' is followed by 'Note that P_G = p_0^* F_0', where F_0 was not defined in this section and clashes with the flag bundle F_0 introduced earlier. Please clarify the notation and define F_0 consistently.","section":"§4.6, Lemma 4.9"},{"comment":"There is a typo in the phrase 'the nondegeneracy. of the curvature' (stray period); also, the sentence 'The estimate is understood in any local trivialization of the family of flat line bundles L_θ → M_0' is vague and should specify the finite cover and the norm used.","section":"§4.6, proof of Lemma 4.9"},{"comment":"The sentence 'the curvature is not globally degenerate since B vanishes on S' is ambiguous; since B vanishes on S, the curvature is degenerate, so this should read 'is degenerate' or 'is not globally nondegenerate'.","section":"§4.8"},{"comment":"The paper is strongly dependent on the preprints [CLM26] and [CL24]; for example, the Floquet setup, the computation of C1, and the Borel–Weil calculus are taken from these sources. A short appendix summarizing the exact statements used, particularly the subellipticity assertion from [CL24], would improve self-containedness.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The manuscript's central new contribution is the spectral gap for the twisted Borel–Weil family (Theorem 4.3) and the resulting reduction of the horizontal heat kernel asymptotics to the Abelian-cover case. The uniform subelliptic estimate in Lemma 4.9 is the one point where the argument is not fully supplied; it is plausibly fillable by a quantitative version of Hörmander's theorem following [CL24], but it must be proved. Given the heavy reliance on two recent preprints by the same group, the editor may also wish to verify that those preprints are publicly available and that the cited statements indeed imply the asserted uniformity."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Punchline: the genuinely new result is the principal-bundle reduction—Theorems 1.5 and 1.6 identify the whole polynomial part of the horizontal heat kernel on P with the heat kernel on the underlying Abelian cover M, and the nontrivial fiber modes only contribute exponentially small remainders. The structure of the argument is sound, and the identification of all coefficients with the base cover is a real step beyond Lott and Kotani–Sunada.\n\nThe Abelian-cover half is not dramatically novel: the existence of the expansions is known, and the author says so. The contribution there is the functional-analytic upgrade—uniform differentiated remainder estimates and a distributional expansion—and those arguments are carried out in reasonable detail. The computation of C1 is concrete, and the nonvanishing lemma is credible.\n\nThe soft spot is exactly the one flagged in the stress test. Lemma 4.9 is the only place where the pointwise C^ell exponential decay of the non-invariant fiber modes is established, and its proof of the uniform subelliptic estimate (4.12) is a paragraph, not a proof. It is plausible that smooth dependence on theta in lower-order terms plus compactness of P0 and U(1)^d gives uniform constants, but that is not automatic: subellipticity of each operator does not by itself give constants continuous in a parameter unless you track the Hörmander bracket step and the local trivializations of L_theta. If the uniformity fails, the L^2 spectral gap from Theorem 4.3 still gives exponential decay in L^2, but the differentiated pointwise expansions in Theorems 1.5 and 1.6 would not follow. This is a specific, fillable gap, not a reason to reject. The rest of the spectral gap argument is substantial and mostly self-contained.\n\nThe magnetic case in Section 4.8 is sketched and leans on an unproved twisted version of Helffer–Kordyukov; that is a smaller issue, and the reduction to the known estimate is explicit. The reliance on the author's own preprints [CLM26] and [CL24] is not a flaw by itself, but because the uniform subelliptic estimate is asserted 'mutatis mutandis' from an unpublished source, a referee should ask for a self-contained proof.\n\nWho is this for? People working on long-time heat kernel asymptotics on covers, and on horizontal diffusions on principal bundles. It is a moderate, clearly written step in an established program. My recommendation: send it to peer review; the work deserves referee time. I would not desk reject. Ask for the missing uniformity proof in Lemma 4.9, and then accept.","headline":"The genuinely new principal-bundle reduction is plausible and worth refereeing; the load-bearing uniformity gap in Lemma 4.9 is asserted rather than proved, but it is specific and likely fillable.","tokens_in":38940,"tokens_out":3384,"would_cite":true,"duration_ms":33906,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["58J35","35K08","53C29","58J65"],"pacs":[],"model":"deepseek-v4-flash","headline":"The horizontal heat kernel on an Abelian cover of a principal bundle has the same full large-time asymptotic expansion as the heat kernel on the underlying Abelian cover, to every polynomial order.","keywords":["heat kernel","Abelian cover","principal bundle","horizontal Laplacian","Floquet theory","Borel–Weil calculus","large-time asymptotics","spectral gap"],"falsifier":"Compute the first eigenvalue $\\lambda_0(\\theta,k)$ of $\\Delta_{\\theta,k}$ for a fixed nontrivial $k$ along a sequence $\\theta_j$ of Floquet parameters approaching the boundary of a trivializing neighborhood on a manifold with dense holonomy and globally nondegenerate curvature; if $\\inf_j \\lambda_0(\\theta_j,k) = 0$, the uniform spectral gap of Theorem 4.3 fails and the local and diffusive pointwise expansions of Theorems 1.5 and 1.6 cannot hold with their stated $C^\\ell$ remainders.","tokens_in":37753,"feed_emoji":"🔥","tokens_out":5549,"duration_ms":60678,"temperature":0.7,"pith_summary":"The paper establishes three complementary large-time expansions for heat kernels on Abelian covers of compact manifolds—a distributional correlation expansion, a local pointwise expansion, and a global diffusive Gaussian expansion—and then proves the same structure for horizontal heat kernels on principal bundles pulled back to such covers. The central claim is that under density of the lifted holonomy in $G \\times \\mathbb{Z}^d$ and global nondegeneracy of the curvature, all leading coefficients are determined by the underlying Abelian cover, while every nontrivial $G$-representation mode enters only through exponentially small remainders. A reader should care because it reduces the long-time behavior of a diffusion in a bundle with a compact fiber to the geometry of a flat cover: the effective covariance is the Hessian of the lowest Floquet eigenvalue, and the fiber variables disappear from the asymptotic series.","feed_headline":"Heat on principal bundles matches the base cover to all orders","feed_subtitle":"Fiber modes decay exponentially; the Gaussian profile and every coefficient come from the Abelian cover's heat kernel.","key_machinery":"The argument is carried by two parallel decompositions: Floquet theory on the Abelian cover, which turns the covered Laplacian into a family of magnetic-type operators $\\Delta_\\theta$ with lowest eigenvalue $\\lambda_0(\\theta)$, and the Borel–Weil calculus on the principal bundle, which decomposes the horizontal Laplacian into operators $\\Delta_{\\theta,k}$ acting on fiberwise holomorphic sections of line bundles over the flag bundle $F_0 = P_0/T$. The load-bearing result is the uniform spectral gap for the twisted family: under global nondegeneracy of the curvature one gets $\\lambda_0(\\theta,k) \\gtrsim |k|$ for large $k$, and under dense holonomy one excludes zero eigenvalues for all $(\\theta,k)\\neq(0,0)$. This gap makes every nontrivial $G$-type decay exponentially, leaving the $k=0$ mode, which is isomorphic to the base Abelian-cover Laplacian, to generate the entire asymptotic series; the Hessian $D^2\\lambda_0(0)$ then supplies the Gaussian covariance through stationary phase.","core_discovery":"For the Abelian cover $M$, the heat kernel admits a full asymptotic expansion in three regimes: weak correlations, local pointwise values on compact sets, and a global expansion on the diffusive scale $|n-m| = O(\\sqrt{t})$ with a Gaussian profile whose covariance is $H = D^2\\lambda_0(0)$, the Hessian at $\\theta = 0$ of the lowest Floquet eigenvalue of the twisted Laplacian. On the pulled-back principal bundle $P \\to M$ with a horizontal Laplacian, the paper proves that, assuming dense holonomy and globally nondegenerate curvature, the same expansions hold with coefficients that are pullbacks of the base coefficients: locally $k^P_j(u,v) = k^M_j(p(u),p(v))$ and on the diffusive scale $P^P_j(z,u_0,v_0) = P^M_j(z,p(u_0),p(v_0))$. The mechanism is a uniform spectral gap $\\lambda_0(\\theta,k) \\geq C$ for all nontrivial representation weights $k$, so all polynomial-order contributions come from the fiberwise constant mode, and dependence on the compact fiber variables is exponentially small.","pith_inferences":["If the uniform subelliptic estimate (4.12) is the only obstruction, the same reduction should hold for other equivariant geometric heat equations on such covers, such as horizontal Hodge Laplacians acting on forms, once an analogous spectral gap for nontrivial representation modes is available.","The weaker holonomy hypothesis isolated in Remark 4.8—density of the $G$-projection of holonomy elements with trivial $\\mathbb{Z}^d$-component—suggests that fiberwise constancy of the leading terms may survive even when the full closed holonomy group is not dense; isolating exactly where that weaker assumption enters the spectral-gap proof would identify the minimal hypothesis.","A numerical check on a low-dimensional example, such as a $U(1)$-bundle over a torus cover with a nondegenerate magnetic field, could compare the horizontal heat kernel's local expansion with the base cover's expansion at fixed large time; matching through order $t^{-1}$ would support the full coefficient pullback beyond the theorem's proof.","The same two-parameter Floquet–Borel–Weil mechanism may apply to heat semigroups acting on sections of associated vector bundles, not just functions, yielding tensor-valued Gaussian profiles with the same base covariance $H$ and representation-dependent exponential corrections."],"forward_implications":["On any Abelian cover of a compact manifold, the heat kernel has a full asymptotic expansion to arbitrary order, with remainder estimates uniform after differentiation in both variables, in both the local and diffusive regimes.","For a principal bundle over such a cover with dense lifted holonomy and globally nondegenerate curvature, the horizontal heat kernel is asymptotically constant along the compact fibers: every coefficient in the expansion depends only on the projection to the base cover.","Nontrivial $G$-representation modes contribute only exponentially small remainders, so the same expansions hold for the Laplace–Beltrami heat kernel of the connection metric, since the vertical Laplacian is spectrally separated and commutes with the horizontal part.","The effective Gaussian covariance is computable from the base geometry as $D^2\\lambda_0(0) = 2\\|D_v\\eta_0\\|^2_{L^2(M_0)}$, and it is positive definite because the covering representation is surjective.","In the magnetic case $G = U(1)$ with a magnetic field degenerating to finite order along a hypersurface, a polynomial lower bound $\\lambda_0(\\theta,k) \\gtrsim |k|^{2/(r+2)}$ still yields a spectral gap and the same structural expansions."],"supporting_citations":[{"why":"Supplies the Borel–Weil semiclassical calculus on principal bundles, including subelliptic estimates and curvature-dependent eigenvalue bounds that the paper adapts uniformly in the Floquet parameter.","marker":"[CL24]"},{"why":"Supplies the Abelian-cover Floquet setup, the weight spaces $B^{s,r}$, and the companion dynamical motivation for combining Floquet and Borel–Weil decompositions.","marker":"[CLM26]"},{"why":"Established the leading Gaussian diffusive profile for heat kernels on Abelian covers, the baseline that Theorems 1.3 and 1.6 refine to arbitrary order.","marker":"[Lot99]"},{"why":"Provides the local central limit theorem and power-series expansions for heat kernels on manifolds with cocompact Abelian group actions, the precedent for the local and diffusive expansions.","marker":"[KS00]"},{"why":"Gives magnetic spectral lower bounds for $U(1)$ bundles with hypersurface degeneracies, used in Section 4.8 to extend the spectral-gap argument beyond global nondegeneracy.","marker":"[HK09]"},{"why":"Provides the stationary phase lemma used to extract the expansion coefficients from the Floquet integral over the torus of dual parameters.","marker":"[Hör03]"}],"fun_headline_variants":["Fiber heat dies out; base cover sets every heat coefficient","On principal bundles, heat asymptotics are pure base cover","Gaussian heat profile on fiber bundles: base cover decides all","Exponential fiber decay: Abelian base dictates heat expansion","Heat on principal bundles: only the base cover leaves a mark"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the subelliptic estimates for the twisted horizontal Laplacians hold uniformly in both the Floquet parameter $\\theta$ and the representation weight $k$ (estimate (4.12)); if that uniformity fails, the exponential decay of nontrivial fiber modes is only known in $L^2$ and the pointwise $C^\\ell$ expansions lose the differentiability they claim.","fun_headline_variants_meta":{"raw":{"variants":["Fiber heat dies out; base cover sets every heat coefficient","On principal bundles, heat asymptotics are pure base cover","Gaussian heat profile on fiber bundles: base cover decides all","Exponential fiber decay: Abelian base dictates heat expansion","Heat on principal bundles: only the base cover leaves a mark"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000594,"raw_usage":{"total_tokens":2774,"prompt_tokens":928,"completion_tokens":1846,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":544,"completion_tokens_details":{"reasoning_tokens":1762}},"tokens_in":544,"tokens_out":1846,"duration_ms":16177,"temperature":1.0,"reasoning_tokens":1762,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T04:19:45.119810+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the first eigenvalue $\\lambda_0(\\theta,k)$ of $\\Delta_{\\theta,k}$ for a fixed nontrivial $k$ along a sequence $\\theta_j$ of Floquet parameters approaching the boundary of a trivializing neighborhood on a manifold with dense holonomy and globally nondegenerate curvature; if $\\inf_j \\lambda_0(\\theta_j,k) = 0$, the uniform spectral gap of Theorem 4.3 fails and the local and diffusive pointwise expansions of Theorems 1.5 and 1.6 cannot hold with their stated $C^\\ell$ remainders.","supporting_citations":[],"review_version":1}