{"id":"412d1649-e7cc-41c3-9031-a7e6f2f6c3c6","arxiv_id":"1908.06786","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For every Bernstein function f, the f-subordinated heat semigroup satisfies Besov/Triebel-Lizorkin smoothing with rate [f^{-1}(1/t)]^{-d/2}.","lead":"Bochner subordination turns the heat semigroup into a flexible family of semigroups, and this paper proves that all of them smooth functions in Besov and Triebel-Lizorkin spaces at a rate determined by the subordinator. The estimates extend known heat-flow bounds to fractional Laplacians and other Levy generators, and give a route to well-posedness for fractional nonlinear heat equations.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 7.2's change of variables is wrong (for f(x)=x it gives t^{-r-1} instead of t^{-r}); Corollary 5.2's rate comparison lacks a valid proof as written.","rationale":"The paper's central construction, Theorem 5.1, is proved from (4.1), Theorem 2.3 and subordinator moments; that part appears sound. The strongest claim, however, also asserts the quantitative rate via Corollary 5.2, whose proof depends on Lemma 7.2. That lemma's displayed substitution is algebraically wrong and fails already for f(x)=x. Because the final rate is part of the central claimed contribution, this is a load-bearing gap in the written proof, though not evidence that the theorem is false. The reader's identified gap in Corollary 5.4 (missing uniform bound for W^{(m)}_r, r>1) is real but concerns an extension and is also repairable. Both issues justify a conditional acceptance: the core idea is correct, but the extended rate and higher-order corollary need corrected proofs. Thus I do not move the reader's verdict.","tokens_in":14422,"tokens_out":22200,"duration_ms":224130,"concrete_test":"Recompute Lemma 7.2 for f(x)=x, where μ_t=δ_t and S_t=t. The true value is E[S_t^{-r}]=t^{-r}. Substitute f(x)=x into the displayed formula in Lemma 7.2: the integral equals t^{-r-1}, so the change-of-variables step is incorrect. Then attempt to re-derive the claimed bounds with the correct Jacobian; if the bounds follow with adjusted constants, Corollary 5.2 can be repaired, otherwise the rate comparison stays unproved.","verdict_should_be":"UNCHANGED","load_bearing_attack":"In Lemma 7.2 (Appendix), the proof transforms E[(S_t)^{-r}] = (1/Γ(r))∫_0^∞ e^{-t f(x)} x^r dx/x by 'changing variables y=f(x)' and writes the integrand as e^{-ty} [f^{-1}(y)]^r dy with a prefactor. The correct Jacobian yields e^{-ty} [f^{-1}(y)]^{r-1} (f^{-1})'(y) dy, not [f^{-1}(y)]^r dy. For the heat semigroup f(x)=x, S_t≡t, so E[S_t^{-r}]=t^{-r}, while the displayed formula gives t^{-r-1}. The subsequent bounds in Lemma 7.2, and hence Corollary 5.2's assertion that the smoothing constant in Theorem 5.1 is comparable to [f^{-1}(1/t)]^{-d/2}, are therefore not justified by the written proof. The underlying moment estimate is very likely true and repairable by splitting the integral at 1/t and using monotonicity of f^{-1}, but the paper should supply the corrected calculation. Separately, the proof of Corollary 5.4 uses (5.4) for r>1 without the needed uniform bound on W^{(m)}_r; this is a second gap in an extension claim.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper uses Bochner subordination to prove caloric smoothing estimates for semigroups with Fourier multiplier e^{-t f(|ξ|^2)} acting on Besov and Triebel--Lizorkin spaces. The main result, Theorem 5.1, bounds the A^{s+d}_{p,q}-norm of the subordinate semigroup by c(E[(S^f_t)^{-d/2}] + P(S^f_t>1)) times the A^s_{p,q}-norm, and Corollary 5.2 compares the resulting constant with [f^{-1}(1/t)]^{-d/2} under a doubling condition on the Bernstein function f. Section 4 treats fractional powers, Section 5 extends the method to general Bernstein functions and to higher-order fractional powers β>0, and Section 6 sketches an application to a nonlinear heat equation. An appendix collects moment estimates for subordinators.","tokens_in":14722,"tokens_out":12055,"duration_ms":107801,"significance":"If the proofs are completed, the paper gives a unified and quantitative extension of known caloric smoothing estimates, with explicit dependence on the Bernstein function through f^{-1}. The subordination approach is elegant and the paper correctly relies on external results (Triebel's (4.1), Baaske--Schmeisser's (5.4)) rather than assuming the target estimates. The explicit rates and the connection to subordinator moments are valuable, and the application to a nonlinear heat equation is potentially useful. However, the current manuscript contains two nontrivial gaps in the proof of the rate comparison and in the higher-order extension, so the results are not yet fully established as written.","major_comments":[{"comment":"The change of variables in the proof of Lemma 7.2 is incorrect. Starting from E[(S_t)^{-r}] = Γ(r)^{-1} ∫_0^∞ e^{-t f(x)} x^r dx/x and setting y=f(x) gives dx/x = (f^{-1}(y))^{-1}(f^{-1})'(y) dy, so the integrand should be e^{-ty}(f^{-1}(y))^{r-1}(f^{-1})'(y) dy, not e^{-ty}(f^{-1}(y))^r dy with prefactor 1/(rΓ(r)). For f(x)=x the displayed formula yields E[S_t^{-r}]=t^{-r-1} instead of t^{-r}. Consequently the upper and lower bounds stated in the lemma are not established by the written proof. Since Corollary 5.2 relies directly on this lemma, the proof of the rate comparability in (5.3) is currently incomplete. The underlying statement is plausible and repairable by a correct Jacobian and a different splitting, but the calculation must be redone; in particular, the lower bound also needs an argument on a subinterval, since monotonicity alone gives the wrong inequality direction on (0,1/t).","section":"§7, Lemma 7.2"},{"comment":"The passage 'If we use (5.4) instead of (4.1)... we get immediately' is not immediate and needs an additional uniform bound. In the proof scheme of Theorem 4.1, the integral over r>1 is controlled by applying the d=0 estimate to W_{r-1}, which requires uniform boundedness of W^{(m)}_{r-1} on A^s_{p,q} for all r>1. Estimate (5.4) is quoted only for t∈(0,1], and the manuscript neither states nor proves the required uniform bound for t>1. Without such a bound, the subordination integral for r>1 is not controlled, so (5.5) is not proved as written. The later sentence saying that the cases p=∞ 'should be clear' for the lifting extension is also not a proof.","section":"§5, Corollary 5.4"},{"comment":"The deduction of (5.2) from (5.1) is not justified by the two observations given. The proof states that P(S_t>1)≤1 and that E[S_t^{-d/2}]→∞ as t→0, but a uniform constant for all t∈(0,1] requires a lower bound on E[S_t^{-d/2}] on the whole interval. Such a bound follows from monotonicity of the subordinator (S_t≤S_1, so E[S_t^{-d/2}]≥E[S_1^{-d/2}]>0) whenever E[S_1^{-d/2}] is finite, but this argument is absent; if that expectation is infinite, the assertion should be stated as vacuous or handled differently. As written, the 'in particular' claim is not derived.","section":"§5, Theorem 5.1, proof of (5.2)"}],"minor_comments":[{"comment":"Theorem 5.1 refers to 'Lemma 4.1' but the relevant statement is Theorem 4.1; Corollary 5.2 refers to 'Corollary 5.1', which does not exist and should read 'Theorem 5.1'.","section":"Cross-references"},{"comment":"The duality argument defines q'=q/(1-q), which is undefined for q=1; the case q=1 should be treated separately or the convention q'=∞ should be stated explicitly.","section":"§2, Lemma 2.2(b)"},{"comment":"The footnote and the displayed condition use the letter t both for the argument of f^{-1} and for the time parameter of the subordinator, which is confusing in the proof of the moment estimate.","section":"§7, Lemma 7.2"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is within scope, and the subordination technique is a genuine and useful contribution. The main estimate (5.1) is likely correct, but the current text contains two nontrivial gaps in derived claims: the moment lemma's change of variables and the uniform boundedness needed for Corollary 5.4. Both are repairable without changing the scope, so I recommend major revision rather than rejection. The self-citation [6] as the source of the moment estimate is appropriate; the flaw is in the reproduced proof, not in the attribution."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should know this paper before citing it: the main theorem, Theorem 5.1, is a clean and apparently correct caloric smoothing estimate for every subordinated heat semigroup W_t^f, with constant controlled by E[(S_t^f)^{-d/2}] + P(S_t^f > 1). That is new and unifies the known Gaussian, Cauchy–Poisson, and generalized Gaussian results. The proof is short and transparent: Bochner subordination plus Triebel's estimate (4.1) plus the moment bounds. The paper is honestly written and credits prior work properly.\n\nThe soft spots are real but localized. First, Lemma 7.2 contains a wrong change of variables. The paper writes E[(S_t)^{-r}] = 1/Γ(r) ∫ e^{-t f(x)} x^r dx/x and then 'changing variables y=f(x)' gives e^{-ty}[f^{-1}(y)]^r dy, but the correct Jacobian yields e^{-ty}[f^{-1}(y)]^{r-1}(f^{-1})'(y) dy. For f(x)=x, the printed formula gives r t^{-r-1}, while the true moment is t^{-r}. So the rate comparison in Corollary 5.2, and the lower bound in Lemma 7.2, are not justified by the proof as written. The underlying estimate is very likely true and repairable by splitting the integral at 1/t and using the doubling-type condition, but the paper should supply the corrected calculation.\n\nSecond, Corollary 5.4 needs a uniform bound for the higher-order semigroup W_t^{(m)} on A^s_{p,q} for all t>1, because (5.4) is only stated for 0<t≤1. The paper says 'immediately', but the integral over r>1 in the subordination formula needs that uniform bound. Again, this is probably fixable because W_t^{(m)} is convolution with a Schwartz kernel and should be bounded on the full scale for t>1, but it is not stated or proved.\n\nThe statement of Theorem 6.1 is given without proof, with only a reference to a literally transferred argument. That is acceptable in a note but should be flagged more prominently.\n\nFor whom: anyone working on Besov/Triebel–Lizorkin smoothing estimates, nonlinear heat equations, or subordinated semigroups will get value from the main theorem. The paper deserves a serious referee, but the referee should require the Lemma 7.2 correction and the missing uniform bound before publication. My own verdict: the central result holds up; the extensions need repair.","headline":"A genuinely useful subordination-based smoothing estimate for arbitrary Bernstein functions, with a sound core theorem but two fixable gaps in the rate comparison and the higher-order corollary.","tokens_in":15276,"tokens_out":2922,"would_cite":true,"duration_ms":28890,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["46E36","60J35","35K25","35K55","60G51"],"pacs":[],"model":"deepseek-v4-flash","headline":"The authors prove that Bochner subordination turns the heat semigroup's caloric smoothing into a general estimate for every Bernstein function, with a constant comparable to $[f^{-1}(1/t)]^{-d/2}$ for small $t$.","keywords":["caloric smoothing","Besov spaces","Triebel–Lizorkin spaces","Bochner subordination","Bernstein functions","subordinators","fractional Laplace operator","nonlinear heat equation"],"falsifier":"Take $f(\\lambda)=\\lambda^\\alpha$ with $0<\\alpha<1$, choose a dyadic block $u$ with $\\widehat u$ supported on an annulus away from zero, and compute both sides of the claimed estimate (5.1) or (5.3) using the exact moment formula of Lemma 7.1; if the ratio of the right-hand constant to the actual norm is unbounded as $t\\to0$, the theorem fails. For the higher-order extension, test whether $\\sup_{r>1}\\|W_r^{(m)}\\|_{A^s_{p,q}\\to A^s_{p,q}}$ is finite; if it is infinite, the uncontrolled $r>1$ part of the subordination integral invalidates Corollary 5.4.","tokens_in":14210,"feed_emoji":"♨️","tokens_out":16212,"duration_ms":141771,"temperature":0.7,"pith_summary":"This paper sets out to prove a general smoothing theorem: take any Bernstein function $f$ — an increasing function with alternating derivatives, equivalently the Laplace exponent of a subordinator — and subordinate the heat semigroup to get $W_t^f$, the semigroup whose Fourier multiplier is $e^{-t f(|\\xi|^2)}$. The authors show that in Besov and Triebel–Lizorkin spaces $A^s_{p,q}$, this semigroup gains $d$ derivatives with a time-dependent constant, and that for small time the constant is comparable to $[f^{-1}(1/t)]^{-d/2}$ under a mild doubling condition on $f$. This unifies the classical smoothing estimates for the Gauss–Weierstraß, Cauchy–Poisson, and stable semigroups, and provides the missing estimate for the nonlinear Cauchy problem with fractional powers of the Laplacian. A sympathetic reader should care because the mechanism is simple and general: subordination transports a known smoothing rate from one semigroup to a whole class.","feed_headline":"Bernstein functions dictate Besov smoothing for Levy semigroups","feed_subtitle":"Bochner subordination transfers the heat semigroup's smoothing rate to every fractional heat semigroup.","key_machinery":"The workhorse is Bochner's subordination formula $$W_t^f u = \\int_0^\\infty W_r u \\,\\mu_t^f(dr),$$ where $\\mu_t^f$ is the unique convolution semigroup of probability measures on $[0,\\infty)$ with Laplace transform $\\int_0^\\infty e^{-\\lambda r}\\mu_t^f(dr)=e^{-t f(\\lambda)}$, and $W_r$ is the heat semigroup. The proof feeds the known heat-semigroup smoothing estimate $\\|W_r u|A^{s+d}\\|\\le c r^{-d/2}\\|u|A^s\\|$ into this integral, uses the contraction property of $W_r$ on the function spaces for $r\\le 1$, and controls the moments $\\mathbb{E}[(S_t^f)^{-d/2}]$ by a comparison with $[f^{-1}(1/t)]^{-d/2}$ (Lemma 7.2). The inverse Bernstein function $f^{-1}$ is the object that carries the smoothing rate: it converts the time scale of the subordinator into the regularity scale of the semigroup.","core_discovery":"On the paper's own terms, the central discovery is Theorem 5.1 and Corollary 5.2. For every Bernstein function $f$, the $f$-subordinated heat semigroup (extended by lifting when the multiplier is not smooth at the origin) satisfies, for all $d\\ge 0$, $s\\in\\mathbb{R}$, and admissible $1\\le p,q\\le\\infty$, $$\\|W_t^f u\\,|\\,$A^{{s+d}}$_{p,q}\\| \\le c\\Big(\\mathbb{E}\\big[(S_t^f)^{-d/2}\\big] + \\mathbb{P}(S_t^f>1)\\Big)\\|u\\,|\\,A^s_{p,q}\\|,$$ where $S_t^f$ is the subordinator with Laplace exponent $f$. When $f$ obeys $\\liminf_{\\lambda\\to0} f(2\\lambda)/f(\\lambda)>1$, the right-hand side is comparable to $[f^{-1}(1/t)]^{-d/2}\\,\\|u\\,|\\,A^s_{p,q}\\|$ for $0<t\\le1$. Applying the same argument to the higher-order generalized Gauss–Weierstraß semigroup yields the corresponding estimate for $e^{-t(-\\Delta)^\\beta}$ with $\\beta>0$, and this in turn extends the existence and uniqueness theorem for the nonlinear equation $\\partial_t u + (-\\Delta)^\\beta u = \\operatorname{div}[u^2]$ from integer $\\beta$ to all real $\\beta\\ge 1$.","pith_inferences":["A direct testable extension not pursued in the paper: the comparability of the constant with $[f^{-1}(1/t)]^{-d/2}$ suggests a general dictionary in which the small-time behaviour of the inverse Bernstein function dictates the smoothing rate, so slowly varying $f$ should produce logarithmic corrections to the pure power rate.","Because the proof only uses the triangle inequality for the norm and a known smoothing estimate for the base semigroup, the same subordination argument should work for any uniformly bounded, translation-invariant semigroup on a Banach scale with its own caloric estimate; the missing ingredient in the higher-order case is the long-time boundedness for $r>1$.","One could numerically verify the stated rate on dyadic test functions for the relativistic semigroup $f(\\lambda)=\\sqrt{\\lambda+1}-1$; the exact Fourier multiplier makes the norms computable, and a mismatch between the observed blow-up and the predicted $[f^{-1}(1/t)]^{-d/2}$ rate would localise any hidden defect in the lifting construction."],"forward_implications":["Every Lévy semigroup whose characteristic exponent is $f(|\\xi|^2)$ with $f$ a Bernstein function inherits a caloric smoothing estimate in $B^s_{p,q}$ and $F^s_{p,q}$ spaces.","Under the doubling condition on $f$, the smoothing rate for $0<t\\le1$ is $[f^{-1}(1/t)]^{-d/2}$; for $f(\\lambda)=\\lambda^\\alpha$ this recovers the familiar $t^{-d/(2\\alpha)}$ rate for stable semigroups.","The same subordination argument gives the smoothing rate for the higher-order fractional heat semigroup $e^{-t(-\\Delta)^\\beta}$ for all real $\\beta>0$, not only integer $m$.","The nonlinear fractional heat equation $\\partial_t u + (-\\Delta)^\\beta u = \\operatorname{div}[u^2]$ has a unique mild solution for all real $\\beta\\ge 1$, and a strong solution under the parameter conditions stated in Theorem 6.1.","The estimates transfer to homogeneous and hybrid Besov and Triebel–Lizorkin spaces, as sketched in Remark 5.3."],"supporting_citations":[{"why":"Provides the definitions of Besov and Triebel–Lizorkin spaces and the base caloric smoothing estimate (4.1) for the heat semigroup that the proof starts from.","marker":"[15]"},{"why":"Supplies the higher-order smoothing estimate (5.4) and the mild/strong solution theorem for the integer-order nonlinear heat equation that the paper extends to real $\\beta$.","marker":"[1]"},{"why":"Contains the Lévy–Khintchine representation of Bernstein functions and the abstract subordination machinery used in Section 3.","marker":"[11]"},{"why":"Gives the stable-subordinator moment formula that Lemma 7.1 re-derives and that fixes the constant in the fractional case.","marker":"[8]"},{"why":"Provides the moment upper bound for general subordinators, transformed into Lemma 7.2, which drives the comparability with $f^{-1}(1/t)$.","marker":"[6]"},{"why":"Introduced the subordination principle that underpins the whole argument.","marker":"[4]"}],"fun_headline_variants":["Bernstein functions dictate smoothing rates for subordinated heat semigroups","Fractional heat semigroups inherit Besov smoothing from Bernstein subordination","Subordination bridges heat to fractional semigroups: Besov smoothing rates","Bernstein functions set the caloric smoothing scale in Besov spaces","Every Bernstein subordination yields Besov smoothing for heat semigroups"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument for powers $\\beta>0$ assumes that the generalized heat semigroup stays bounded on the function spaces for all times $r>1$; the paper cites an estimate that is stated only for $0<r\\le 1$, so without a separate proof of that long-time bound the integral over $r>1$ in the subordination formula may not be controlled.","fun_headline_variants_meta":{"raw":{"variants":["Bernstein functions dictate smoothing rates for subordinated heat semigroups","Fractional heat semigroups inherit Besov smoothing from Bernstein subordination","Subordination bridges heat to fractional semigroups: Besov smoothing rates","Bernstein functions set the caloric smoothing scale in Besov spaces","Every Bernstein subordination yields Besov smoothing for heat semigroups"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000546,"raw_usage":{"total_tokens":2610,"prompt_tokens":942,"completion_tokens":1668,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":558,"completion_tokens_details":{"reasoning_tokens":1575}},"tokens_in":558,"tokens_out":1668,"duration_ms":13608,"temperature":1.0,"reasoning_tokens":1575,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T12:36:55.932290+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take $f(\\lambda)=\\lambda^\\alpha$ with $0<\\alpha<1$, choose a dyadic block $u$ with $\\widehat u$ supported on an annulus away from zero, and compute both sides of the claimed estimate (5.1) or (5.3) using the exact moment formula of Lemma 7.1; if the ratio of the right-hand constant to the actual norm is unbounded as $t\\to0$, the theorem fails. For the higher-order extension, test whether $\\sup_{r>1}\\|W_r^{(m)}\\|_{A^s_{p,q}\\to A^s_{p,q}}$ is finite; if it is infinite, the uncontrolled $r>1$ part of the subordination integral invalidates Corollary 5.4.","supporting_citations":[{"cited_title":"Triebel, Theory of Function Spaces IV (Birkh¨ auser, Monog raphs in Mathematics vol","cited_arxiv_id":null,"evidence_quote":"Provides the definitions of Besov and Triebel–Lizorkin spaces and the base caloric smoothing estimate (4.1) for the heat semigroup that the proof starts from."},{"cited_title":"Baaske and H.-J","cited_arxiv_id":null,"evidence_quote":"Supplies the higher-order smoothing estimate (5.4) and the mild/strong solution theorem for the integer-order nonlinear heat equation that the paper extends to real $\\beta$."},{"cited_title":"Schilling, R","cited_arxiv_id":null,"evidence_quote":"Contains the Lévy–Khintchine representation of Bernstein functions and the abstract subordination machinery used in Section 3."},{"cited_title":"Sato, L´ evy Processes and Infinitely Divisible Distributions, 2n d","cited_arxiv_id":null,"evidence_quote":"Gives the stable-subordinator moment formula that Lemma 7.1 re-derives and that fixes the constant in the fractional case."},{"cited_title":"Deng, R.L","cited_arxiv_id":null,"evidence_quote":"Provides the moment upper bound for general subordinators, transformed into Lemma 7.2, which drives the comparability with $f^{-1}(1/t)$."},{"cited_title":"Bochner, Diffusion equation and stochastic processes, Proc","cited_arxiv_id":null,"evidence_quote":"Introduced the subordination principle that underpins the whole argument."}],"review_version":1}