{"id":"0608094d-00c9-405b-b024-2980865c5cc2","arxiv_id":"2412.02848","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"Fractional Hardy inequalities in doubling metric measure spaces self-improve in both the power p and the regularity theta, via an equivalence with Hardy inequalities in hyperbolic fillings.","lead":"This paper proves that fractional Hardy inequalities in doubling metric spaces self-improve in two parameters at once. It works by translating the fractional inequality into a classical Hardy inequality on a hyperbolic filling and then improving that local inequality.","discovery_kind":"unification","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The equivalence Theorem 1.4 depends on an energy-only trace/extension comparison from [5] that the paper does not explicitly establish; if the extension gradient bound requires the full Besov norm, Proposition 5.15's reduction breaks.","rationale":"The reader's weakest assumption is the same: the Björn–Björn–Shanmugalingam trace theorem quoted as Theorem 4.2, with the precise parameter relation (1.5), is the load-bearing external premise. I agree that this is the most security-critical point. The rest of the architecture—localization to compact Z_R, Koskela–Zhong p-self-improvement on the filling, and the new δ-regularizable weight theorem—is coherent, and the flagged issues (the factor in (6.5) and the absorption step in Proposition 6.3) appear to be repairable typos or missing details rather than structural flaws. My concrete concern is slightly sharper than the reader's: Proposition 5.15 uses the extension operator as if its gradient energy were controlled by the Besov seminorm alone, whereas the paper only quotes a full-norm control from Theorem 4.2. If the [5] extension does not have the energy-only bound, the reverse implication in Theorem 1.4 is not established as written. This does not lower my confidence in the theorem itself—[5] very likely contains the needed estimate—but it is the single step where a hidden assumption could invalidate the central claim, and it should be verified explicitly before the proof is accepted as complete.","tokens_in":38162,"tokens_out":44319,"duration_ms":472477,"concrete_test":"Re-derive the extension estimate from [5, Theorem 1.1] (or inspect the extension construction in [5, Section 10]) for the exact parameter relation β/ε=p(1−θ). Check whether ∫_{X_ε} g_{E f}^p dμβ ≲ Energy_{B^θ_{p,p}}(f) := ∫_Z∫_Z |f(z)−f(w)|^p / (d(z,w)^{θp}ν(B(z,d(z,w)))) dν(w)dν(z) holds for all f in a dense subclass (e.g. Lipschitz functions compactly supported in Z\\E). If the bound is only valid with ‖f‖_{L^p}^p added to the right-hand side, test whether functions u in Proposition 5.15 satisfy ‖u‖_{L^p(Z)}^p ≲ Energy(u) uniformly; any proof of this must not use the fractional Hardy inequality on Z\\E being proved, since that would be circular.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central bridge is Theorem 1.4, which says a fractional (θ,p)-Hardy inequality on Z\\E is equivalent to a p-Hardy inequality on the hyperbolic filling X_ε\\E whenever β/ε=p(1−θ). This rests entirely on Theorem 4.2, quoted from [5, Theorem 1.1]. For the forward direction Proposition 5.6, the trace bound ∫_{X_ε} g_u^p dμβ ≳ ‖Tu‖_{B^θ_{p,p}}^p is needed. For the reverse direction Proposition 5.15, the proof invokes boundedness of the extension operator as ∫_{X_ε} g_{Eu}^p dμβ ≲ ∫_Z∫_Z |u(z)−u(w)|^p / (d(z,w)^{θp}ν(B(z,d(z,w)))) dν(w)dν(z), i.e. it uses only the Besov seminorm (energy), not the full Besov norm ‖u‖_{L^p}^p + energy. However, Theorem 4.2 as stated in the paper controls the Newton–Sobolev norm, including the L^p term of Ef, by the full Besov norm. The paper does not justify dropping the L^p term of u. If the extension operator of [5] only satisfies a full-norm bound, then Proposition 5.15 would not follow for functions whose L^p mass is not controlled by the fractional energy, and the self-improvement transfer from X_ε back to Z would fail. This is a load-bearing gap because every subsequent step in Theorem 1.3 passes through Theorem 1.4. It is likely repairable—e.g. by using a fractional Poincaré inequality for functions vanishing near E or by checking that the [5] extension satisfies an energy-only bound—but the paper does not supply that verification.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves a simultaneous self-improvement result for fractional Hardy inequalities in complete doubling metric measure spaces: if a bounded open set Z\\E satisfies a (θ0,p0)-Hardy inequality, then it satisfies (θ,p)-Hardy inequalities for all (θ,p) in a neighborhood of (θ0,p0). The proof is built on a Caffarelli-Silvestre-type equivalence (Theorem 1.4) between the fractional Hardy inequality on Z\\E and a classical p-Hardy inequality on a uniformized hyperbolic filling X_ε\\E with measure μ_β, where β/ε = p(1−θ). This equivalence is obtained from trace/extension theory of Björn-Björn-Shanmugalingam. To move θ independently of p, the authors develop a new weighted self-improvement theory for p-Hardy inequalities with δ-regularizable weights (Theorem 1.7) and combine it with the Koskela-Zhong self-improvement in p. A localization argument reduces the general complete case to the compact case. Applications include a sufficient condition in terms of Assouad codimensions and the example of the punctured unit ball.","tokens_in":38501,"tokens_out":29632,"duration_ms":282808,"significance":"The result is significant: it removes geodesicity and reverse-doubling assumptions that were present in the recent work [36], and it gives self-improvement in both the differentiability parameter θ and the integrability p simultaneously, with a radius depending only on structural constants. The hyperbolic-filling equivalence is a conceptual contribution with likely further applications, and the weighted self-improvement theorem for regularizable weights is a new tool of independent interest. The paper is written in a careful, theorem-proof style; all arguments are derivations from stated assumptions, with explicit constant dependencies and no fitted or empirical input. The main weaknesses are a small number of local proof gaps and notational ambiguities that I believe are repairable without altering the central claims.","major_comments":[{"comment":"Proposition 5.15, Eq. (5.17): The step 'by the boundedness of the extension operator' relies on the estimate ∫ g_{Eu}^p dμ_β ≲ ∫∫ |u(z)-u(w)|^p / (d(z,w)^{θp} ν(B(z,d(z,w)))) dν(w)dν(z), i.e. an energy-only extension bound. With the notation fixed in (2.2), Theorem 4.2 states exactly this bound, but the accompanying phrase 'bounded extension operators' suggests a full Besov-norm bound. Since the equivalence Theorem 1.4 and hence Theorem 1.3 pass through this estimate, please clarify the norm convention and confirm that [5, Theorem 1.1] indeed provides the energy-only bound, or add a direct proof for this case.","section":"Section 5, Proposition 5.15, Eq. (5.17)"},{"comment":"In Proposition 6.3(i), the absorption of the exterior term into (1/2)∫ |u|^p/d(z,E)^{θp} dν requires a bound on sup_{z∈supp u} d(z,E)^{θp} relative to R; the condition 'R > (2 C_{θ,p} C_1 (1+C_ν))^{1/(θp)}' alone does not ensure this. The argument should either choose R large relative to the diameter of Z\\E and the distance from the reference point to E, or first rescale so that d(z,E) is uniformly controlled. Without this, the localization to a compact doubling subset is not fully justified.","section":"Section 6.1, Proposition 6.3(i)"}],"minor_comments":[{"comment":"Equation (6.5): the displayed δ_p is p/C0 (2C1)^{-1/p}, but the absorption in the proof of Theorem 1.7 requires p/(2C0)(2C1)^{-1/p}. Please correct the factor of 2.","section":"Section 6.2, Eq. (6.5)"},{"comment":"The concluding constant in Theorem 1.7 should depend on Cp (the Hardy constant in the assumption), not only on p, Cμ, and the p-admissibility constants of w; the proof in Section 3 exhibits this dependence, so the statement should be amended.","section":"Theorem 1.7 statement"},{"comment":"Using the notation ||u||^p_{B^θ_{p,p}(X,μ)} for the energy seminorm conflicts with the standard norm meaning of the symbol; consider denoting the energy by [u]^p_{B^θ} or E_p(u) to avoid ambiguity, especially in Theorem 4.2.","section":"Section 2.2, Eq. (2.2)"},{"comment":"Proposition 5.15 states 1 ≤ p < ∞, whereas the Hardy inequality in Definition 2.7 is stated for 1 < p < ∞; please align the ranges.","section":"Proposition 5.15 statement"}],"recommendation":"major_revision","confidential_remarks":"The paper is well suited to the journal. The main risk is the reliance on the energy-only form of the Björn-Björn-Shanmugalingam trace theorem; if the authors can confirm this from [5] or prove it, the paper should be accepted after the local corrections listed. I do not see evidence of novelty concealment or unacknowledged dependencies beyond the cited external theorems."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The main theorem is the real thing: if a bounded complement in a complete doubling space satisfies a (θ0,p0)-fractional Hardy inequality, it satisfies (θ,p)-Hardy inequalities for an open neighborhood of (θ0,p0) in both parameters. That was not known before, and the approach is genuinely new. The Caffarelli–Silvestre-style bridge between fractional Hardy inequalities and p-Hardy inequalities in the hyperbolic filling is a clean idea, and the weighted self-improvement result for δ-regularizable weights is a substantive extension beyond the distance-weight results of Lehrbäck and Koskela. The examples, including the punctured ball, are a nice payoff.\n\nThe stress-test concern about Theorem 4.2 does not land. The paper defines the Besov ‘norm’ in (2.2) as the double integral energy only, and Theorem 4.2 is stated with that notation. So the extension bound in Proposition 5.15 is energy-only by construction; there is no missing L^p term to drop. The notation is potentially confusing but the mathematics is consistent.\n\nThere are two real but minor issues. First, equation (6.5) gives δp = p/C0(2C1)^{-1/p}, while Theorem 1.7 requires δ ≤ p/(2C0)(2C1)^{-1/p}. As written, the application of Theorem 1.7 to the constructed weights is not justified because the weights are only known to be δ-regularizable for the larger δ. This is clearly a typo, but it needs fixing before the proof is fully clean. Second, the absorption step in Proposition 6.3(i) is telegraphic: the constant C1 is asserted to be independent of z, which works if one uses R > 2 diam(Z\\E) so that d(z0,z) < R/2, plus doubling to compare balls centered at z and z0 with radii differing by a factor depending only on k through a geometric sum. The written estimate skips those details but the argument is sound.\n\nThe paper deserves a serious referee. The main theorem is important, the architecture is coherent, and the flaws are repairable presentation issues, not load-bearing gaps. I would send it out.","headline":"Strong paper: proves the natural two-parameter self-improvement for fractional Hardy inequalities in doubling metric measure spaces via a genuinely new hyperbolic-filling equivalence; the trace-theory worry in the stress test is a misreading, and the remaining issues are typo-level.","tokens_in":39074,"tokens_out":6512,"would_cite":true,"duration_ms":61277,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35R11","26D10","28A75","30L15","31C15","31E05","35A23","46E35"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves that fractional Hardy inequalities self-improve in both parameters at once, in every complete doubling metric measure space, with a proof that routes through hyperbolic fillings and weighted local Hardy inequalities.","keywords":["fractional Hardy inequality","self-improvement","metric measure space","hyperbolic filling","Besov space","Newton-Sobolev space","weighted Hardy inequality","regularizable weights"],"falsifier":"A direct falsifier is a counterexample: a complete doubling space $Z$, a closed $E$ with $Z\\setminus E$ bounded, and a parameter point $(\\theta_0,p_0)$ at which the fractional Hardy inequality holds, together with a sequence $(\\theta_k,p_k)$ converging to $(\\theta_0,p_0)$ at which it fails. The theorem asserts no such sequence exists. A concrete place to look is a Euclidean domain with a tangential cusp or a porous boundary, where the validity set of fractional Hardy inequalities is known or suspected to have a non-open boundary; if any such domain shows a failure sequence, Theorem 1.3 is false.","tokens_in":37916,"feed_emoji":"📐","tokens_out":15613,"duration_ms":142681,"temperature":0.7,"pith_summary":"This paper proves that fractional Hardy inequalities self-improve in both parameters at once. Specifically, if $Z$ is a complete doubling metric measure space and $E$ is closed with bounded complement $Z\\setminus E$, and $Z\\setminus E$ satisfies a $(\\theta_0,p_0)$-Hardy inequality, then it satisfies a $(\\theta,p)$-Hardy inequality for every $(\\theta,p)$ in a neighborhood of $(\\theta_0,p_0)$. The improvement radius depends only on the structural constants, and no geodesicity or reverse-doubling assumption is needed. The proof works by an extension-problem argument: the fractional inequality on $Z$ is equivalent, through the hyperbolic filling of $Z$, to a classical $p$-Hardy inequality in a geodesic filling space with a weighted measure. This bridge, together with a new weighted self-improvement theorem for $p$-Hardy inequalities, yields the two-parameter result and new examples such as the punctured ball. The authors note the results are for the integral case $p=q$, leaving the pointwise $(\\theta,p,q)$-Hardy setting outside the scope.","feed_headline":"Fractional Hardy inequalities self-improve in two parameters at once","feed_subtitle":"If a domain satisfies one fractional Hardy inequality, it satisfies nearby ones, with no extra geometry.","key_machinery":"The load-bearing object is the uniformized hyperbolic filling $(X_\\varepsilon, d_\\varepsilon, \\mu_\\beta)$ of a compact doubling space $Z$, a geodesic metric graph whose boundary is bi-Lipschitz equivalent to $Z$. A trace theorem identifies the Besov space $B^{\\theta}_{p,p}(Z,\\nu)$ (a fractional Sobolev-type space) with the traces of the Newton-Sobolev space $N^{1,p}(X_\\varepsilon,\\mu_\\beta)$ (a Sobolev space defined via upper gradients), with comparable energies, precisely when $\\beta/\\varepsilon = p(1-\\theta)$; this identity is the bridge that turns the nonlocal Hardy inequality into a local one. The second mechanism is the weighted self-improvement theorem: a domain satisfying a $p$-Hardy inequality with respect to $\\mu$ also satisfies it with respect to any $p$-admissible weight that is $\\delta$-regularizable at Whitney scales for sufficiently small $\\delta$. The weights needed to change $\\beta$ are of the form $d_\\varepsilon(\\cdot,Z)^{\\sigma}$, and they are shown to be $\\delta$-regularizable because $Z$ is porous in the filling. The parameter relation then transfers the improved $p$ and $\\beta$ back to improved $\\theta$ and $p$ on $Z$.","core_discovery":"The central discovery is the equivalence stated in Theorem 1.4: for a compact doubling space $Z$ with $\\mathrm{diam}(Z)<1$, a closed set $E$, and parameters satisfying $\\beta/\\varepsilon = p(1-\\theta)$, the complement $Z\\setminus E$ satisfies the $(\\theta,p)$-fractional Hardy inequality if and only if the complement $X_\\varepsilon\\setminus E$ satisfies the classical $p$-Hardy inequality in the uniformized hyperbolic filling $X_\\varepsilon$ with respect to the measure $\\mu_\\beta$. The trace and extension operators between the Newton-Sobolev space $N^{1,p}(X_\\varepsilon,\\mu_\\beta)$ and the Besov space $B^{\\theta}_{p,p}(Z,\\nu)$ provide the norm comparisons that make the equivalence two-sided. From this, the authors obtain Theorem 1.3 by combining two self-improvement mechanisms in the filling: the classical open-endedness of $p$-Hardy inequalities, and a new weighted self-improvement result (Theorem 1.7) for $p$-admissible, $\\delta$-regularizable weights. Since changing $\\theta$ while keeping $p$ fixed corresponds to perturbing the parameter $\\beta$, the weighted result is exactly what is needed to turn the one-parameter curve into a full two-parameter neighborhood.","pith_inferences":["The same filling equivalence could transfer other local results to nonlocal ones, for instance fractional Poincar\\'e or Sobolev inequalities, whenever the trace relation $\\beta/\\varepsilon = p(1-\\theta)$ is available.","The paper explicitly limits itself to the integral case $p=q$; extending the filling machinery to Triebel-Lizorkin energies with $p \\neq q$ is a natural next step that would likely require a multi-parameter weighted self-improvement.","The $\\delta$-regularizability condition is local and checkable; because distance-to-porous-set weights satisfy it, the proof gives a concrete recipe for building new weighted Hardy inequalities in the filling and hence new fractional Hardy inequalities on the boundary, beyond the examples in Section 7.","On the punctured ball, the theorem predicts the best constant $C_{\\theta,p}$ stays finite on a full neighborhood of any point with $\\theta p < n$; numerically tracking this constant as $\\theta p$ approaches $n$ could probe the sharp quantitative dependence of the improvement radius $\\varepsilon_0$ on the structural constants."],"forward_implications":["In every complete doubling metric measure space, a bounded domain satisfying a fractional Hardy inequality at one parameter point automatically satisfies it on a whole open neighborhood of that point in the $(\\theta,p)$-plane.","The equivalence with local Hardy inequalities in the hyperbolic filling is a transfer principle: results for classical $p$-Hardy inequalities immediately yield fractional Hardy inequalities; the paper demonstrates this by deriving a sufficient condition in terms of Assouad codimensions and showing the punctured unit ball in $\\mathbb{R}^n$ satisfies $(\\theta,p)$-Hardy exactly when $\\theta p < n$.","The new weighted self-improvement theorem stands on its own: $p$-Hardy inequalities are stable under replacing the measure by a $p$-admissible, $\\delta$-regularizable weight, which generalizes the previously known stability for distance weights to a broader class including powers of distance to porous sets.","The results are new even in Euclidean, sub-Riemannian, and manifold settings, where no geodesicity or reverse-doubling restriction was previously required for two-parameter self-improvement."],"supporting_citations":[{"why":"Supplies the trace and extension operators between N^{1,p}(X_epsilon, mu_beta) and B^theta_{p,p}(Z,nu) with two-sided norm bounds; this is the bridge for Theorem 1.4.","marker":"[5]"},{"why":"Establishes that p-Hardy inequalities self-improve in p (open-endedness), used in the filling to get self-improvement along the curve.","marker":"[44]"},{"why":"Prior weighted self-improvement for distance weights d_Omega^beta; the paper's Theorem 1.7 generalizes it, and it is the starting point for the weighted theory.","marker":"[47]"},{"why":"Gives the Assouad-codimension sufficient condition for p-Hardy inequalities, which via Theorem 1.4 yields the fractional examples (Proposition 7.2 and Example 7.8).","marker":"[46]"},{"why":"The extension-problem analogy that motivates relating nonlocal energies to local ones via traces.","marker":"[13]"},{"why":"Proves the Poincar\\'e inequality is open-ended; used inside the proof of Theorem 1.7 to obtain (q,p)-Poincar\\'e from (1,q)-Poincar\\'e.","marker":"[39]"},{"why":"Supplies the measure-decay estimate for porous sets used to show distance-to-porous weights are delta-regularizable.","marker":"[8]"}],"fun_headline_variants":["Two-parameter self-improvement for fractional Hardy","Fractional Hardy self-improves in both p and theta","Hyperbolic fillings prove Hardy inequality self-improvement","Open-endedness for fractional Hardy in two parameters","Self-improve p and theta via hyperbolic fillings"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is the trace/extension theorem for hyperbolic fillings: a two-sided energy comparability between the Besov space on $Z$ and the Newton-Sobolev space on the filling, at the exact parameter relation $\\beta/\\varepsilon = p(1-\\theta)$. If that comparability fails for some parameter range, the equivalence between fractional and local Hardy inequalities breaks and the main theorem has no bridge.","fun_headline_variants_meta":{"raw":{"variants":["Two-parameter self-improvement for fractional Hardy","Fractional Hardy self-improves in both p and theta","Hyperbolic fillings prove Hardy inequality self-improvement","Open-endedness for fractional Hardy in two parameters","Self-improve p and theta via hyperbolic fillings"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000204,"raw_usage":{"total_tokens":1450,"prompt_tokens":1067,"completion_tokens":383,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":683,"completion_tokens_details":{"reasoning_tokens":308}},"tokens_in":683,"tokens_out":383,"duration_ms":4050,"temperature":1.0,"reasoning_tokens":308,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T23:03:43.948828+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A direct falsifier is a counterexample: a complete doubling space $Z$, a closed $E$ with $Z\\setminus E$ bounded, and a parameter point $(\\theta_0,p_0)$ at which the fractional Hardy inequality holds, together with a sequence $(\\theta_k,p_k)$ converging to $(\\theta_0,p_0)$ at which it fails. The theorem asserts no such sequence exists. A concrete place to look is a Euclidean domain with a tangential cusp or a porous boundary, where the validity set of fractional Hardy inequalities is known or suspected to have a non-open boundary; if any such domain shows a failure sequence, Theorem 1.3 is false.","supporting_citations":[{"cited_title":"Björn, J","cited_arxiv_id":null,"evidence_quote":"Supplies the trace and extension operators between N^{1,p}(X_epsilon, mu_beta) and B^theta_{p,p}(Z,nu) with two-sided norm bounds; this is the bridge for Theorem 1.4."},{"cited_title":"Koskela, X","cited_arxiv_id":null,"evidence_quote":"Establishes that p-Hardy inequalities self-improve in p (open-endedness), used in the filling to get self-improvement along the curve."},{"cited_title":"Lehrbäck.Self-improving properties of weighted Hardy inequalities.Advances in Calculus of Variations, vol","cited_arxiv_id":null,"evidence_quote":"Prior weighted self-improvement for distance weights d_Omega^beta; the paper's Theorem 1.7 generalizes it, and it is the starting point for the weighted theory."},{"cited_title":"Lehrbäck.Hardy inequalities and Assouad dimensions.J","cited_arxiv_id":null,"evidence_quote":"Gives the Assouad-codimension sufficient condition for p-Hardy inequalities, which via Theorem 1.4 yields the fractional examples (Proposition 7.2 and Example 7.8)."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"The extension-problem analogy that motivates relating nonlocal energies to local ones via traces."},{"cited_title":"Keith, X","cited_arxiv_id":null,"evidence_quote":"Proves the Poincar\\'e inequality is open-ended; used inside the proof of Theorem 1.7 to obtain (q,p)-Poincar\\'e from (1,q)-Poincar\\'e."},{"cited_title":"Björn, N","cited_arxiv_id":null,"evidence_quote":"Supplies the measure-decay estimate for porous sets used to show distance-to-porous weights are delta-regularizable."}],"review_version":1}