{"id":"3f94b309-9f28-4640-a70e-bc9d4eb622b7","arxiv_id":"1908.07895","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Fractional heat semigroup kernels on metric measure spaces satisfy Euclidean-style decay, yielding Strichartz regularity and capacity-based trace criteria for fractional dissipative Cauchy problems.","lead":"This paper proves that fractional heat semigroup kernels on metric measure spaces obey the same off-diagonal decay as on Euclidean space, then uses those bounds to prove regularity and measure-extension theorems for fractional dissipative equations. It is worth reading because the subordination method removes the need for Fourier transforms, so the results cover Lie groups, manifolds, and weighted Euclidean spaces.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Unproved good-lambda inequality in Lemma 5.7 is the most load-bearing gap: Theorem 5.10's Wolff-potential characterization depends on it, and the asserted modification of Adams–Hedberg is not justified.","rationale":"I read the paper in good faith and found the core fractional-kernel program largely coherent: Proposition 2.8 follows from (A1) plus lower density, and the space-time estimates of Section 3 are plausible given those kernels. The reader's verdict of CONDITIONAL is appropriate. The most load-bearing unresolved point I can identify is not the two-sided Gaussian assumption (A4) itself, which is an explicit hypothesis for Sections 4–5, nor the spherical capacity bound in Theorem 4.3, which is likely salvageable by using a test function supported on a ball of radius t0^{1/(2α)}+r0. The sharpest gap is Lemma 5.7: the paper asserts a good-λ inequality (5.3) without proof and then builds Lemma 5.8 and Theorem 5.10 on it. That inequality is highly nontrivial, and the scaling mismatch between the r^{-Q} maximal operator and the parabolic volume r^{Q+2α} makes the 'slight modification' of Adams–Hedberg's result suspect. Since Theorem 5.10 is presented as a main result and provides the explicit Wolff-potential characterization advertised in the abstract, this unproved inequality is the single most load-bearing concern. The concrete test I propose would settle whether (5.3) holds in the simplest Euclidean model; a failure there would require removing or substantially revising Theorem 5.10. The reader had flagged this lemma as one of several technical problems; I agree with that identification, though I weight it more heavily than the A4/Theorem 4.3 issues.","tokens_in":42961,"tokens_out":32313,"duration_ms":283154,"concrete_test":"Verify (5.3) directly for the Euclidean fractional heat kernel with L=-Δ, α=1/2, Q=n, taking ν to be Lebesgue measure restricted to a parabolic ball B_r^{(α)}(r^{2α},0). Compute both sides of the good-λ inequality for p=2 and several choices of a>1, ε, λ, r; if the left-hand side exceeds the right-hand side for any admissible choice, the stated inequality is false. Alternatively, attempt a rigorous proof of (5.3) from the two-sided Gaussian bound using the standard covering/Calderón–Zygmund arguments of Adams–Hedberg; if the proof requires an additional doubling condition on ν or a restriction on p, then Lemma 5.7 and Theorem 5.10 must be amended accordingly.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"Lemma 5.7 claims the Lp equivalence ||Mα ν||_{Lp(G)} ≈ ||(e^{-tL^α})^*ν||_{Lp(G)} for all p∈(1,∞), where Mα is the parabolic fractional maximal operator. The proof contains no derivation of the asserted good-λ inequality (5.3); it merely says 'we slightly modify [1, (3.6.1)] to get' the inequality. This is an omitted proof of a nontrivial estimate, and the inequality is the engine for Lemma 5.8 and Theorem 5.10, the paper's 'second main result' characterizing trace spaces via the Hedberg–Wolff potential Pαp. The concern is not merely cosmetic: the claimed good-λ exponent (Q+2α)/Q and the maximal operator Mαν(x)=sup_r r^{-Q}ν(B_r^{(α)}(r^{2α},x)) have incompatible scaling with the parabolic volume r^{Q+2α}, so the reduction to Adams–Hedberg's Euclidean inequality is not routine. If (5.3) requires ν to be doubling, or holds only for a restricted range of p, or is simply false, then Lemma 5.7 fails and Theorem 5.10 collapses. Since the abstract advertises a broad 'characterize a nonnegative Radon measure ν' result, this missing step is load-bearing for a substantive part of the central claim. Other issues (A4 assumptions, Theorem 4.3's spherical bound, Lemma 3.2's exponent) are either explicit hypotheses, fixable, or peripheral to the main capacity characterization, whereas Lemma 5.7's gap is a missing proof of a key equivalence.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops a Fourier-free, subordination-based approach to the fractional heat semigroup e^{-tL^alpha} on a metric measure space (M,d,mu) with upper and lower density bounds. Under heat-kernel assumptions (A1)-(A4), it derives pointwise estimates for the fractional heat kernel and its fractional derivatives, then uses them to prove space-time and Strichartz-type estimates for the fractional dissipative equation, to define an L^p-capacity on the parabolic space M x (0,infty), and to characterize nonnegative Radon measures nu for which the solution operator maps L^p(M) into L^q(M x (0,infty), nu). The central advertised results are the decay estimates in Propositions 2.8 and 2.11 and the capacity trace criteria in Theorems 5.3, 5.4, and 5.10.","tokens_in":43299,"tokens_out":6292,"duration_ms":130870,"significance":"If fully correct, the paper would provide a genuinely non-Fourier route to Euclidean-type off-diagonal estimates for fractional heat semigroups on general metric measure spaces and would extend the Chang-Xiao and Jiang-Xiao-Yang-Zhai capacity theory to this setting. The subordination derivation in Sections 2 and 3 is clean and self-contained, and the examples in Section 2.1 give useful context. The capacity duality in Proposition 4.1 is also a well-structured part of the paper. However, the current manuscript contains several load-bearing gaps: an unproved good-lambda inequality, an unjustified ball-measure lower bound in Theorem 4.3, and an exponent mismatch in Lemma 3.2(ii). These issues prevent the advertised conclusions from being fully supported as written.","major_comments":[{"comment":"The asserted good-lambda inequality (5.3) is the engine of Lemma 5.7, and Lemma 5.7 is essential for Lemma 5.8 and therefore for Theorem 5.10, the paper's second main result. The proof only says that the inequality follows by 'slightly modify[ing] [1, (3.6.1)]', but no derivation is given. The reduction to Adams-Hedberg is not routine: the parabolic ball B_r^{(alpha)}(r^{2alpha},x) has volume of order r^{Q+2alpha}, whereas the maximal operator M_alpha in Lemma 5.7 uses the normalization r^{-Q}, so the scaling of the claimed constant epsilon^{(Q+2alpha)/Q} needs independent verification. The paper should either prove (5.3) from the kernel estimates and the density assumptions, or state precisely the extra hypotheses under which it holds (e.g., doubling of nu or a restricted range of p). As written, the proof of Theorem 5.10 has a missing nontrivial step.","section":"Section 5.3, Lemma 5.7, Eq. (5.3)"},{"comment":"The upper estimate Cap_p^{(alpha)}(B_{r0}^{(alpha)}(t0,x0)) lesssim (t0^{1/2alpha}+r0)^beta is proved using e^{-tL^alpha} 1_{B_{r0}(x0)}(x) greatersim r0^beta/(t0^{1/2alpha}+r0)^beta. This inequality is obtained by integrating the lower kernel bound over B(x0,r0), which requires mu(B(x0,r0)) greatersim r0^beta. Assumption (1.3) only gives mu(B(x0,r0)) greatersim r0^{beta*}, and (1.2) gives the opposite inequality. Unless beta=beta* or an additional lower density condition with exponent beta is assumed, the displayed lower bound is unjustified. This gap matters for Theorem 5.3, where the ball condition nu(B_r^{(alpha)}) lesssim r^{qbeta/p} is claimed to be equivalent to the capacitary condition.","section":"Section 4.2, Theorem 4.3 (upper bound)"},{"comment":"Lemma 3.2(ii) states ||L^{theta/2} e^{-tL^alpha} phi||_{L^p} lesssim t^{-theta - beta*(1/r-1/p)/2alpha} ||phi||_{L^r}, but the proof, using Proposition 2.11, yields the exponent -theta/(2alpha) - beta*(1/r-1/p)/(2alpha) in place of -theta. The mismatch is not cosmetic: Lemma 3.2(ii) is used in Theorems 3.5, 3.6, and 3.8, and the admissible-triplet computations in Section 3 depend on the correct exponent. The authors should correct the statement and then re-check the exponents in the subsequent estimates, including the displayed bounds in Theorem 3.8(i), which appear to use yet another exponent for L^alpha e^{-tL^alpha}.","section":"Lemma 3.2(ii) and its consequences"},{"comment":"The proof of Lemma 5.1 invokes Proposition 4.1(ii) for the level sets E_j = {(t,x): e^{-tL^alpha} f(x) >= 2^j}. Proposition 4.1 is stated and proved only for compact subsets K of M+. The paper does not explain why the capacitary extremal measure exists for these level sets, which are not shown to be compact. Since Lemma 5.1 is used in the proofs of Theorems 5.3 and 5.4, this approximation/extension step should be supplied or replaced by a limiting argument.","section":"Lemma 5.1"}],"minor_comments":[{"comment":"The abstract announces results under Gaussian upper estimates, but Sections 4 and 5 require the two-sided bound (A4). The scope should be stated accurately in the abstract and introduction.","section":"Abstract and Introduction"},{"comment":"The definition of admissible triplet appears to be printed twice with the same displayed condition; if one copy is intended to define the 'generalized admissible triplet', the two conditions should be distinguished.","section":"Definition 3.3"},{"comment":"In the displayed formula inside the proof, the kernel is written with |x| instead of d(x,y), and there is a 'beta*+2alpha nu' typo; these should be fixed.","section":"Proof of Lemma 3.2(ii)"},{"comment":"There are typographical issues such as 'Randon' for 'Radon', 'Naiver-Stokes' for 'Navier-Stokes', and 'tempreratue' in a reference title; these should be corrected in a final version.","section":"Throughout"},{"comment":"Remark 5.9 claims Lemmas 5.7 and 5.8 hold for general metric measure spaces under (1.2), but Lemma 5.8 uses the Lie-group structure and the dyadic cube construction in its proof; the remark should indicate what modifications are needed or be removed.","section":"Section 5.3.2, Remark 5.9"}],"recommendation":"major_revision","confidential_remarks":"The paper's main ideas are promising and the subordination-based kernel estimates in Section 2 are sound, but the current version has several load-bearing gaps, especially the unproved good-lambda inequality in Lemma 5.7 and the unjustified use of mu(B) greatersim r^beta in Theorem 4.3. These are fixable in principle, but they require substantial additional work rather than local editing. The self-citation [25] is used only as an example domain and does not create circularity."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"This paper tries to do a lot: extend the fractional heat semigroup kernel bounds of Miao–Yuan–Zhang and Jiang–Xiao–Yang–Zhai from Euclidean space and stratified groups to general metric measure spaces with upper and lower densities. The subordination route is standard, but the systematic transfer of the bounds and the Fourier-free off-diagonal estimates are genuinely new, and the PDE applications in Sections 2–3 look like they work. The kernel estimate in Proposition 2.8, the derivative bound in Proposition 2.10, and the space-time estimates that follow are clean, and I didn't find any circularity. Credit where it's due: the opening sections are a solid piece of work.\n\nThe trouble starts in Section 4. Theorem 4.3's upper spherical capacity bound uses the lower bound μ(B(x0,r0)) ≳ r0^β, which is not what (1.3) gives unless β=β*. That's a concrete error, not a style quibble. If the intended statement is only for β=β*, fine, but say so.\n\nThe bigger problem is Lemma 5.7. The asserted good-λ inequality (5.3) is the engine for Lemma 5.8 and hence for the Wolff-potential characterization in Theorem 5.10, and the proof simply says 'we slightly modify [1, (3.6.1)]'. That is not a proof of a nontrivial estimate, and the scaling looks off: with the parabolic ball volume r^{Q+2α}, an exponent (Q+2α)/Q in the good-λ term is not the routine reduction to Adams–Hedberg. If this inequality fails, the equivalence in Lemma 5.7 collapses, and with it the main capacity result. The rest of Section 5 is built on that.\n\nThere are also smaller issues: Lemma 3.2(ii) states a time exponent -θ where the proof gives -θ/2α, and Theorem 3.8(i) seems to use yet another exponent. These look like fixable typos, but they indicate the manuscript wasn't checked carefully. The reliance on (A4) for all of Sections 4–5 is explicit, so that's not a flaw per se, but it narrows the advertised scope.\n\nWho is this for? Harmonic analysts working on heat kernels and trace inequalities will want to know the kernel bounds and the general framework. The capacity results are the most advertised, and they're the least supported. I'd send it to a serious referee — the core ideas are worth engaging — but it needs major revision, not copy-editing. The referee should be asked to verify Lemma 5.7 from scratch and to fix the ball-mass exponent in Theorem 4.3. If those get repaired, this becomes a useful paper.","headline":"Solid kernel-estimate core, but the capacity half rests on an unproved good-lambda inequality and a ball-mass misstep; deserves referee attention but needs real revision.","tokens_in":43845,"tokens_out":8455,"would_cite":false,"duration_ms":73469,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["31E05","47D03","35K05","31C15"],"pacs":[],"model":"deepseek-v4-flash","headline":"Fractional heat semigroups on metric measure spaces inherit Euclidean-type kernel decay, and that decay controls regularity and capacity trace theorems for fractional dissipation.","keywords":["fractional heat semigroup","metric measure spaces","Gaussian heat kernel estimates","fractional dissipative equations","space-time estimates","Strichartz estimates","Lp-capacity","trace embeddings"],"falsifier":"On a metric measure space satisfying (A1), (1.2) and (1.3) but not (A4), evaluate the subordination integral at $t=d(x,y)^{2\\alpha}$: if $K^L_{\\alpha,t}(x,y)$ decays faster than $t/(t^{1/(2\\alpha)}+d(x,y))^{\\beta+2\\alpha}$ for some sequence of points, then Proposition 2.9 and every capacity upper bound depending on it fail; such a space would separate the upper-bound theory of Sections 2–3 from the capacity theory of Sections 4–5.","tokens_in":42740,"feed_emoji":"🔥","tokens_out":9812,"duration_ms":508131,"temperature":0.7,"pith_summary":"The paper aims to prove that the fractional heat semigroup $e^{-tL^\\alpha}$ on a metric measure space behaves, in its kernel, like the Euclidean fractional heat semigroup: whenever the ambient measure obeys upper and lower power-law ball estimates $\\mu(B(x,r))\\lesssim r^\\beta$ and $\\mu(B(x,r))\\gtrsim r^{\\beta_*}$, and the original heat kernel has Gaussian upper bounds, the fractional kernel satisfies $K^L_{\\alpha,t}(x,y)\\lesssim t/(t^{1/(2\\alpha)}+d(x,y))^{\\beta_*+2\\alpha}$. It then uses this single estimate to derive space-time and Strichartz estimates for the Cauchy problem $\\partial_t u+L^\\alpha u=f$, giving regularity for fractional dissipative equations on spaces where Fourier analysis is unavailable. In a second thread, under a matching two-sided Gaussian bound, it develops an $L^p$-capacity on space-time and characterizes exactly which Radon measures $\\nu$ on $M\\times(0,\\infty)$ admit the trace embedding $e^{-tL^\\alpha}: L^p(M)\\to L^q(M\\times(0,\\infty),\\nu)$. A sympathetic reader would care because the package of Euclidean results for fractional dissipation—kernel decay, Strichartz estimates, capacity criteria—is transplanted to manifolds, Lie groups, and other metric spaces by a Fourier-free argument.","feed_headline":"Fractional heat kernels get Euclidean decay on metric spaces","feed_subtitle":"A Fourier-free proof yields space-time, Strichartz, and capacity trace theorems for fractional diffusion.","key_machinery":"The machine that carries the argument is the subordination representation of the fractional heat semigroup, $K^L_{\\alpha,t}(x,y)=\\int_0^\\infty \\eta^\\alpha_t(s)p_s(x,y)\\,ds$, where $\\eta^\\alpha_t$ is the scaled stable density satisfying $\\eta^\\alpha_t(s)\\simeq t/s^{1+\\alpha}$ for large $s$. It does the work of transferring decay: the Gaussian factor in $p_s$ contributes $e^{-d(x,y)^2/s}$, the density assumption contributes the volume factor $\\mu(B(x,\\sqrt{s}))^{-1}\\lesssim s^{-\\beta_*/2}$, and the remaining $s$-integral is evaluated by the change of variable $r=d(x,y)/\\sqrt{s}$ to produce the combined denominator $t^{1/(2\\alpha)}+d(x,y)$. The same mechanism, fed with the derivative estimate (A2) and the fractional-power identity (1.5), produces the time-derivative and $L^{\\theta/2}$ kernel bounds, and with the two-sided bound (A4) it produces the lower kernel bound used for capacities.","core_discovery":"The central discovery is that the subordination formula $K^L_{\\alpha,t}(x,y)=\\int_0^\\infty \\eta^\\alpha_t(s)p_s(x,y)\\,ds$, with the stable-type kernel $\\eta^\\alpha_t(s)=t^{-1/\\alpha}\\eta^\\alpha_1(s/t^{1/\\alpha})$ obeying $\\eta^\\alpha_t(s)\\lesssim t/s^{1+\\alpha}$, turns a one-sided Gaussian upper bound on $p_s$ into sharp two-sided-shape bounds on the fractional kernel: $K^L_{\\alpha,t}(x,y)\\lesssim t/(t^{1/(2\\alpha)}+d(x,y))^{\\beta_*+2\\alpha}$ (Proposition 2.8), and with an additional derivative bound on $p_s$, $|L^{\\theta/2}K^L_{\\alpha,t}(x,y)|\\lesssim (t^{1/(2\\alpha)}+d(x,y))^{-(\\beta_*+\\theta)}$ (Proposition 2.11). These pointwise estimates are the load-bearing input for the space-time estimates of Section 3 and for the capacity and trace theorems of Sections 4 and 5. Under the two-sided Gaussian assumption (A4) the same subordination argument gives the matching lower bound $K^L_{\\alpha,t}(x,y)\\gtrsim t/(t^{1/(2\\alpha)}+d(x,y))^{\\beta+2\\alpha}$, which is what makes the $L^p$-capacity of parabolic balls comparable to $r^{\\beta_*}$ from below and $(t_0^{1/(2\\alpha)}+r_0)^\\beta$ from above. In the paper's own formulation, the fractional solution operator $e^{-tL^\\alpha}$ has Euclidean-type off-diagonal decay, and the $L^p$-capacity characterization of trace embeddings is valid on metric measure spaces with finite densities.","pith_inferences":["A direct extension the paper does not state is that the same proof should give kernel bounds for other subordinated semigroups, such as $e^{-tL^{1/2}}$ or relativistic-type fractional powers, whenever the subordinator has the same scaling and two-sided decay.","In the non-Ahlfors case $\\beta\\neq\\beta_*$, the spherical capacity upper bound involves $(t_0^{1/(2\\alpha)}+r_0)^\\beta$ while the lower bound involves $r_0^{\\beta_*}$; a natural testable refinement would be to locate the transition radius where the two exponents cross, which the paper leaves implicit.","Because the kernel estimates are Fourier-free and use only (A1)–(A3), one could try to verify the pointwise bounds for jump-type nonlocal Dirichlet forms once their heat kernels satisfy comparable Gaussian-type bounds, connecting this capacity theory to known two-sided jump-kernel estimates.","Sharpness of the exponent $\\beta_*+2\\alpha$ could be probed by taking $M=\\mathbb{R}^n$ and $L=-\\Delta$, where the bounds collapse to the classical Euclidean estimates; that suggests the metric-space result is dimensionally optimal in the Ahlfors-regular case."],"forward_implications":["The admissible-triplet space-time estimates $\\|e^{-tL^\\alpha}\\phi\\|_{L^q(I;L^p)}\\lesssim\\|\\phi\\|_{L^r}$ and the companion estimates for the Duhamel term give well-posedness and decay for the Cauchy problem (1.6) on any metric measure space satisfying the density and Gaussian upper assumptions.","The Strichartz-type bound $\\|G(F)\\|_{L^{\\tilde q}((0,\\infty);L^{\\tilde p})}\\lesssim\\|F\\|_{L^q((0,\\infty);L^p)}$, together with the exponential-integrability and Hölder estimates of Theorem 3.8, transfers the Euclidean regularity package for fractional dissipation to the metric setting.","The $L^p$-capacity on $M\\times(0,\\infty)$ is dual to the adjoint fractional heat semigroup, is subadditive, and has parabolic-ball capacity comparable to $r^{\\beta_*}$ from below and $(t_0^{1/(2\\alpha)}+r_0)^\\beta$ from above.","The trace theorem gives a quantitative characterization: for $1