{"id":"e54a9114-432c-44e0-8b73-533273fffef5","arxiv_id":"2412.21048","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The paper proves solvability of div u = f in Hardy, BMO, and Lipschitz spaces for p greater than n/(n+1), and a Korn inequality in Hardy-Sobolev spaces.","lead":"Given a function with zero average on a bounded domain, this paper constructs vector fields whose divergence equals it while staying controlled in Hardy, BMO, and Lipschitz spaces, covering endpoint cases where the classical L^p result fails. It also proves a new Korn inequality in Hardy-Sobolev spaces, answering an open question raised in 2023.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Koskela–Saksman embedding is invoked for local h^p without proof; Theorem 3.4 and the Korn inequality depend on it.","rationale":"The reader and I identify the same load-bearing assumption: the unproved extension of the Koskela–Saksman Sobolev embedding from global H^p(R^n) to local h^p(Ω). This step is necessary for the stated solution space in Theorem 3.4 and again for the Korn inequality in Theorem 6.1. I do not see a more serious internal inconsistency. The paper's other main steps—boundedness of the Bogovskii operator on h^p and on Lipschitz/bmo spaces—are supported by kernel estimates and are plausible. The proof of Theorem 3.3 that T^*_{ij}1 ∈ Λ^α is somewhat terse but follows standard Calderón–Zygmund theory. The weighted case in Theorem 4.1 is also only sketched, but it is secondary because it explicitly defers to a weighted version of Theorem 3.1 and to [15, Theorem 8.7]; the primary unproved local embedding is the more serious gap. Since the reader already returned CONDITIONAL and my own reading agrees, the verdict should remain unchanged: the paper is promising but needs the local embedding written out before the main theorems can be considered fully verified.","tokens_in":600,"tokens_out":6346,"duration_ms":599331,"concrete_test":"Write out the local analogue of Theorem 1.1 in Koskela–Saksman: for n/(n+1)<p≤1 and every u in h^{1,p}_z(Ω), prove or disprove ||u||_{L^{p*}(Ω)} ≤ C ||∇u||_{h^p_z(Ω)}. A positive proof should use a local grand maximal function and cover the atomic decomposition of h^p; a counterexample would be an h^{1,p}_z function whose gradient is small in h^p_z but whose L^{p*} norm is not controlled—e.g., a sequence of atoms with increasingly high frequency and zero mean but nonzero low-frequency part. If the local embedding fails, Theorems 3.4 and 6.1 collapse; if it holds, the gap is merely expository.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The proof of Theorem 3.4 uses (3.14): ||u||_{h^p_z(Ω)} ≲ ||u||_{L^{p*}(Ω)} ≲ ||∇u||_{h^p_z(Ω)}. The second inequality is the Koskela–Saksman Sobolev embedding, proved in [19] for global H^p(R^n) only. The paper states in the proof of Theorem 3.4 that 'it is not difficult to see' that the same arguments apply to h^p(Ω), but supplies no proof. The same extension is used in the weighted theorem 4.1 (via [15]) and the Korn inequality (6.1) depends on the unweighted h^p version through (6.3)–(6.4). This is genuinely load-bearing: without ||u||_{h^p_z} ≤ C ||∇u||_{h^p_z}, estimate (3.13) controls only derivatives, not u, and the Hardy-Sobolev solution space h^{1,p}_{z,0} is not attained. Local h^p differs from H^p in exactly the relevant way: local Hardy functions may have nonvanishing averages and do not automatically satisfy the global vanishing-moment/maximal-function conditions used in [19]. Adapting the proof requires a local pointwise characterization or a local Riesz-potential bound; the paper does not provide it. The reader's conditional verdict is therefore appropriate, and I do not see a separate fatal flaw.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the divergence equation div u = f with zero boundary data on bounded Lipschitz domains. For n/(n+1)<p<=1 the authors prove that Bogovskii's explicit integral operator is bounded on local Hardy spaces h^p, and on weighted local Hardy spaces with A_1 weights; for the opposite endpoint they prove boundedness on Lipschitz spaces Lambda_alpha and bmo. As a byproduct they derive a Korn inequality for Hardy-Sobolev spaces. The arguments are based on kernel estimates for the associated singular integral operators, following and simplifying results of Komori, Ding-Han-Zhu, Goldberg, and Koskela-Saksman.","tokens_in":13177,"tokens_out":6378,"duration_ms":60377,"significance":"If the main theorems are fully established, they give a clean and useful endpoint extension of the classical Bogovskii theory, and the Korn inequality would answer a question left open in [22]. The paper is explicit about what it imports from the literature, and the kernel estimates in Lemmas 3.2 and 5.2 are mostly plausible and standard in structure. The proof of the unweighted Hardy case is elegant, and the extension to Lipschitz/BMO spaces by duality is natural. The central obstacle is the unproved local h^p version of the Koskela-Saksman Sobolev embedding, which is load-bearing for Theorem 3.4 and, through (6.3)-(6.4), for the Korn inequality. If that gap is closed, the paper would be a solid contribution.","major_comments":[{"comment":"The estimate ||u||_{h^p_z(Omega)} <= C||u||_{L^{p*}(Omega)} <= C||nabla u||_{h^p_z(Omega)} is load-bearing: without it, inequality (3.13) controls only the derivatives of the Bogovskii solution and does not show that u belongs to h^{1,p}_{z,0}(Omega). The second inequality is the Koskela-Saksman Sobolev embedding, proved in [19] for the global space H^p(R^n). The assertion that 'it is not difficult to see' that the same arguments apply to the local h^p(Omega) is not a proof, and the local spaces differ from H^p precisely in the vanishing-moment and maximal-function conditions used in [19]. This gap also propagates to Corollary 3.5 and to the Korn inequality in Theorem 6.1, which uses (6.3)-(6.4). Please supply a full proof of the local embedding, or a reference that contains it, before the main solvability theorem can be considered established.","section":"Section 3, proof of Theorem 3.4, Eq. (3.14)"},{"comment":"The weighted theorem is stated as a direct extension, but the two ingredients it invokes are not specified in enough detail to verify. A 'weighted version of Theorem 3.1' is said to follow easily from [18], and the weighted Sobolev embedding is attributed to [15, Theorem 8.7]. It is not clear which weighted Hardy-Sobolev space on the metric measure space (R^n, w dx) corresponds to h^{1,p}_{w,z}(Omega), nor that the hypotheses of [15, Theorem 8.7] hold for A_1 weights in the needed range. These points should be stated explicitly and proved where the cited results do not literally apply.","section":"Section 4, Theorem 4.1"},{"comment":"The verification of the kernel estimates (5.1)-(5.2) for N(x,y)=K(x,x-y) is delegated to 'the same arguments used in Lemma 3.2 and Theorem 3.3,' and the proof that S_0 1 belongs to dotLambda_alpha(R^n) is omitted with 'we omit details because they are exactly as those used in Theorem 3.3.' Since Lemma 5.2 is the key technical step behind Theorem 5.3, these arguments should be written out or replaced by precise references rather than left as an exercise.","section":"Section 5, proof of Lemma 5.2"},{"comment":"In the proof of (6.1), the transition from the integral identity to the operator expression (6.3) is summarized as 'integrating by parts in the usual way for singular integrals' with a reference to [1, Lemma 2.3]. This is the point at which the non-zero averages (partial u_i/partial x_j)_omega are separated and the boundary term is controlled. Since Theorem 3.1 is then applied to the composed operators T^*_{kj} without verifying their kernel estimates in the h^p_r(Omega) setting, more detail is needed here.","section":"Section 6, Theorem 6.1"}],"minor_comments":[{"comment":"The abstract contains a spelling typo: 'Lipschit z' should read 'Lipschitz.'","section":"Abstract"},{"comment":"The displayed estimate gives |K(y,x-y)| <= C|x-y|^{-n}; to obtain the second bound in (3.8) one should explicitly use |x-y| <= d and |x-y|^{-n} <= d|x-y|^{-(n+1)}.","section":"Section 3, Lemma 3.2"},{"comment":"The final sentence says 'applying the continuity of T^* in h^{n/(n+alpha)}'; it would help to state explicitly that this uses the dual pair (h^{n/(n+alpha)})^* = Lambda_alpha under the stated duality.","section":"Section 5, Theorem 5.1"},{"comment":"The norm notation in inequality (6.1) is garbled: the expression starting '\\(\\left\\{\\epsilon(u)\\right\\|\\)' should read '\\(\\|\\epsilon(u)\\|_{h^p_r(\\Omega)}\\)'.","section":"Section 6, Eq. (6.1)"},{"comment":"The inclusion L^{p*}(Omega) subset h^p(Omega) is asserted without a one-line justification; adding the simple Holder/maximal-function argument would make the constant dependence transparent.","section":"Section 3, proof of Theorem 3.4"}],"recommendation":"major_revision","confidential_remarks":"The stress-test concern raised by the reader lands: Eq. (3.14) is genuinely load-bearing, and the local h^p extension of [19] is asserted without proof. I did not find a separate fatal flaw beyond this embedding gap and the related sketchiness around the weighted case. The paper's self-citations to [1] are standard for the kernel identities and do not raise a novelty concern. The topic is well within the scope of math.AP."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a real paper. The Hardy-space solvability of div was already in Chen–Jiang–Yang, and the authors are upfront about that; what's new is the Lipschitz/bmo solvability and the Korn inequality for p<1, which answers an open question from Spector–Spector. The proofs use standard Calderón–Zygmund machinery and duality, and the kernel estimates in Lemmas 3.2 and 5.2 look plausible and checkable. I think the main theorems are correct, but there is one genuinely load-bearing gap.\n\nThe gap is in equation (3.14): the authors invoke the Koskela–Saksman Sobolev embedding for local Hardy spaces h^p(Ω), while [19] proves it for global H^p(R^n). The sentence 'it is not difficult to see' is doing real work here. Without that embedding, estimate (3.13) controls only ∇u, not u, so the solution space h^{1,p}_{z,0} is not attained. The weighted theorem 4.1 has a similar sketch, and the Korn inequality depends on the same step. This doesn't sink the paper—the method is standard and the gap is likely fillable—but it should be written out before the results are taken as verified.\n\nMinor points: Corollaries 3.5 and 5.4 are stated as 'by known arguments' and rely on a decomposition lemma from Galdi; that's fine. The paper is honest about prior work and doesn't oversell. No circularity concerns.\n\nFor whom? Harmonic analysts working on endpoint estimates for the divergence operator, and people interested in Korn-type inequalities in Hardy–Sobolev spaces. I'd send it to a serious referee—the new results deserve scrutiny—but the referee should be asked to check the local embedding step carefully.","headline":"Solid harmonic-analysis paper with a genuinely new Korn inequality; the main gap is a sketched Sobolev embedding for local Hardy spaces.","tokens_in":13741,"tokens_out":2129,"would_cite":true,"duration_ms":21242,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["42B30","26D10"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that the Bogovskii integral operator is bounded on local Hardy spaces h^p for n/(n+1) < p ≤ 1 and on Lipschitz Λ_α and bmo spaces, giving divergence-free solutions u supported in Ω whose derivatives are controlled by the…","keywords":["divergence equation","local Hardy spaces","Lipschitz spaces","BMO","Bogovskii operator","Korn inequality","Hardy-Sobolev spaces","singular integral operators"],"falsifier":"A concrete falsifier would be a function u with u and ∇u both belonging to h^p_z(Ω) but with u failing to lie in $L^{{p*}}$(Ω) for p* = np/(n−p); if such a function exists, equation (3.14) and the proof of the full $h^{{1,p}}$ estimate collapse.","tokens_in":12676,"feed_emoji":"📐","tokens_out":10251,"duration_ms":87706,"temperature":0.7,"pith_summary":"The paper proves that the divergence equation div u = f, with u vanishing on the boundary, is solvable in the borderline regularity scales where the classical L^p result fails: local Hardy spaces h^p for n/(n+1) < p ≤ 1, and Lipschitz Λ_α spaces as well as bmo at the p = ∞ endpoint. The proof shows that the Bogovskii integral operator, the standard explicit right inverse of the divergence for 1 < p < ∞, is bounded on these spaces whenever f has zero integral and is supported in Ω. A byproduct is a Korn inequality in Hardy-Sobolev spaces for the same range, which gives a positive answer to the previously open second Korn case at p = 1. The reason to care is that h^p and bmo/Lipschitz spaces are the natural replacements for $L^{1}$ and L^∞ in applications such as the Stokes problem, where the usual L^p theory stops working.","feed_headline":"Bogovskii operator is bounded on Hardy and Lipschitz spaces","feed_subtitle":"Extends div u = f solvability to p ≤ 1 and to the p = ∞ borderline, yielding a Korn inequality.","key_machinery":"The central object is the Bogovskii integral operator with kernel (3.2), which is the explicit right inverse of the divergence operator and the engine of every estimate in the paper. The derivative formula ∂u_i/∂x_j = T_ij f + ω_ij f splits the solution into a smooth term and a singular integral T_ij; the kernel of T_ij obeys the decay bound min(|x-y|^{-n}, |x-y|^{-n-1}) and a Hölder regularity condition in the y-variable (3.9). A singular-integral theorem adapted from Komori and from Ding-Han-Zhu then gives h^p boundedness once T^*1 belongs to every Lipschitz space Λ_α, which the authors verify by splitting the kernel into a Calderón-Zygmund piece and a smooth remainder. For the Korn inequality, the auxiliary identity φ(y) − φ_ω = −∫_Ω G(x,y)·∇φ(x) dx transfers the symmetric strain tensor into control of the full gradient, after integrating by parts and applying the same singular-integral bounds.","core_discovery":"On its own terms, the paper establishes that for a bounded Lipschitz domain Ω the Bogovskii operator, defined through the kernel G(x,y) = ∫$_0^{1}$ (x-y)/s ω(y+(x-y)/s) ds/s^n with ω a smooth bump of integral one supported inside Ω, is bounded from h^p_z(Ω) to $h^{{1,p}}$_{z,0}(Ω)^n for n/(n+1) < p ≤ 1, and from Λ_α(ℝⁿ) (respectively bmo(ℝⁿ)) to the corresponding first-derivative spaces for 0 < α < 1, provided f has zero integral and is supported in Ω. Consequently every such f admits a vector field u supported in Ω with div u = f and ‖∇u‖ controlled by ‖f‖ in the same scale. The derivative of the solution reduces to the identity ∂u_i/∂x_j = T_ij f + ω_ij f, where T_ij is a singular integral whose kernel satisfies the size and smoothness conditions that make it bounded on h^p and, by duality, on Lipschitz and bmo spaces. A final theorem transfers the same singular-integral machinery into the Korn inequality ‖∇u‖_{h^p_r} ≤ C(‖ε(u)‖_{h^p_r} + ‖u‖_{h^p_r}) for Hardy-Sobolev vector fields.","pith_inferences":["If the local Hardy-Sobolev embedding asserted from [19] fails, the estimate (3.13) would lose the ‖u‖_{h^p_z} term, though the control of ‖∇u‖ might survive; checking this transfer is the first step to strengthening or weakening the result.","Because the star-shaped case is the only domain ingredient in the proof, the same arguments could plausibly extend the Hardy and Lipschitz results to John domains, where the L^p theory already works.","One can test the sharpness of the range n/(n+1) < p by trying to push the singular-integral theorem below that threshold; the failure of the p = 1 L^1 case suggests the Hardy range is the natural limit."],"forward_implications":["For any f in h^p_z(Ω) with zero integral and n/(n+1) < p ≤ 1, the divergence equation has a solution u in h^{1,p}_{z,0}(Ω)^n with full norm control (3.13), extending the Bogovskii theorem into the range where the L^p result is false.","With an A1 weight, the same construction solves the equation in weighted Hardy spaces h^p_{w,z}(Ω), with constants depending only on the domain, p, n, and the weight.","For f in Λ_α or bmo with compact support and zero integral, the solution u has first derivatives in the same space, and the compact-support condition on f is necessary for the particular solution constructed in the paper.","The Korn inequality (6.1) holds in Hardy-Sobolev spaces for n/(n+1) < p ≤ 1, answering the open question about the second Korn case at p = 1."],"supporting_citations":[{"why":"Supplies the explicit integral operator with kernel (3.2) that generates the divergence solution.","marker":"[3]"},{"why":"Provides the detailed derivative identity (3.3) and the template for deriving Korn inequalities from the operator.","marker":"[1]"},{"why":"Defines the local Hardy spaces h^p and records the dualities (h^p)* = Λ_α and (h^1)* = bmo used throughout.","marker":"[13]"},{"why":"Provides the Hardy-Sobolev embedding used in (3.14) to control ‖u‖_{h^p_z} by ‖∇u‖_{h^p_z}.","marker":"[19]"},{"why":"Adapts Komori's singular-integral result to local Hardy spaces, giving the boundedness theorem used for T_ij.","marker":"[9]"},{"why":"Base result on Calderón-Zygmund operators on H^p that [9] modifies.","marker":"[16]"},{"why":"Gives the domain decomposition lemma used to pass from star-shaped to Lipschitz domains.","marker":"[12]"}],"fun_headline_variants":["Bogovskii bounded on Hardy and Lipschitz spaces","Div equation solved on Hardy and Lipschitz spaces","Solving div u = f on Hardy and Lipschitz spaces","Korn inequality from Hardy-Sobolev solutions","Endpoint Hardy and Lipschitz div solvability"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the Hardy-Sobolev embedding proved for the global spaces H^p(ℝⁿ) in the work cited as [19] carries over to the local spaces h^p(ℝⁿ) and h^p(Ω); the paper states this without supplying the proof, and both the full estimate (3.13) and the Korn inequality depend on it.","fun_headline_variants_meta":{"raw":{"variants":["Bogovskii bounded on Hardy and Lipschitz spaces","Div equation solved on Hardy and Lipschitz spaces","Solving div u = f on Hardy and Lipschitz spaces","Korn inequality from Hardy-Sobolev solutions","Endpoint Hardy and Lipschitz div solvability"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000825,"raw_usage":{"total_tokens":3632,"prompt_tokens":993,"completion_tokens":2639,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":609,"completion_tokens_details":{"reasoning_tokens":2573}},"tokens_in":609,"tokens_out":2639,"duration_ms":16703,"temperature":1.0,"reasoning_tokens":2573,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T23:04:45.566705+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A concrete falsifier would be a function u with u and ∇u both belonging to h^p_z(Ω) but with u failing to lie in $L^{{p*}}$(Ω) for p* = np/(n−p); if such a function exists, equation (3.14) and the proof of the full $h^{{1,p}}$ estimate collapse.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the explicit integral operator with kernel (3.2) that generates the divergence solution."},{"cited_title":"G., Divergence Operator and Related Ine qualities, SpringerBriefs in Mathematics, Springer, 2017","cited_arxiv_id":null,"evidence_quote":"Provides the detailed derivative identity (3.3) and the template for deriving Korn inequalities from the operator."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines the local Hardy spaces h^p and records the dualities (h^p)* = Λ_α and (h^1)* = bmo used throughout."},{"cited_title":", Saksman, E., Pointwise characterizations of Hard y-Sobolev functions, Mathemat- ical Research Letters 15, 727–744 (2006)","cited_arxiv_id":null,"evidence_quote":"Provides the Hardy-Sobolev embedding used in (3.14) to control ‖u‖_{h^p_z} by ‖∇u‖_{h^p_z}."},{"cited_title":"S , Zhu Y","cited_arxiv_id":null,"evidence_quote":"Adapts Komori's singular-integral result to local Hardy spaces, giving the boundedness theorem used for T_ij."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Base result on Calderón-Zygmund operators on H^p that [9] modifies."},{"cited_title":"P., An introduction to the mathematical theory of the N avier-Stokes equations","cited_arxiv_id":null,"evidence_quote":"Gives the domain decomposition lemma used to pass from star-shaped to Lipschitz domains."}],"review_version":1}