{"id":"07b66c4d-ff18-4cdf-897e-25d724187fcd","arxiv_id":"2501.18149","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A Sobolev map between manifolds with kp<m can be approximated by smooth maps exactly when its generic restrictions to floor(kp)-dimensional spheres are VMO-homotopically trivial, and relaxing the target dimension gives approximation by maps smooth off structured singular sets.","lead":"This paper proves a complete if-and-only-if condition for when a Sobolev map between manifolds can be approximated by smooth maps, expressed through a new topological property called extendability. The result unifies and extends two decades of partial answers, and also describes the best possible approximation when smooth approximation fails.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Manifold reduction in §10 appears to assume an unproved inheritance of extendability under the tubular-neighborhood retraction; this step is load-bearing for Theorems 1.14 and 1.17.","rationale":"The reader's verdict accepted the paper with moderate confidence, citing the un-audited final approximation chapter as the main limitation. This stress-test focuses on a specific unproved step inside that chapter: the reduction from compact manifolds to open sets via the smooth retraction ~Π. The reader's identified weakest assumption was the generic composition lemma (Proposition 2.8/2.14), which the available text supports in some detail. The more fragile point appears to be the inheritance of extendability under the change of ambient dimension from m to κ, since the extendability property is dimension-sensitive and the higher homotopy groups of N are not controlled by the (ℓ,m) hypothesis. This is not a demonstrated contradiction, but it is a concrete, checkable gap in the logical chain leading to the central claims. Therefore the appropriate verdict is conditional: accept the Euclidean case and the forward direction, and require verification of the manifold-reduction step before fully accepting Theorems 1.14 and 1.17.","tokens_in":866,"tokens_out":1605,"duration_ms":368713,"concrete_test":"Inspect §10.8–10.9 for a lemma proving that (ℓ,e)-extendability of u:M→N implies (ℓ,e+κ-m)-extendability of v = u∘~Π on the tubular neighborhood U⊂R^κ. If no such lemma is stated, perform an independent obstruction-theory check in the model case M = S^3⊂R^4, N = S^2, k=1, p=5/2 (so ℓ=2): take a 2-extendable u and a generic Lipschitz γ:K^4→U for which β = ~Π∘γ sends the boundary of the single 4-cell to S^3 with Hopf number 1, then determine whether the obstruction in H^4(K^4,K^2;π_3(S^2)) can be made to vanish by suitable choices in the definition of extendability. Non-vanishing would disprove the inheritance claim; vanishing would confirm the omitted step.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"Theorems 1.14 and 1.17 for a compact manifold M are reduced to the open-set case by extending u to v = u∘~Π on a tubular neighborhood U⊂R^κ of M, and the introduction asserts that v 'inherits the same extendability properties as u'. This assertion is load-bearing because the main approximation theorem is first proved for bounded Lipschitz open sets in R^m, and the tubular neighborhood has dimension κ>m. To apply the open-set theorem to v, one needs v to be (ℓ,κ)-extendable, while the hypothesis on u only gives (ℓ,m)-extendability. Extending from the ℓ-skeleton to a κ-dimensional complex involves additional obstruction groups H^{j+1}(K^κ,K^ℓ;π_j(N)) for j≥m. Since β = ~Π∘γ maps K^κ into the m-dimensional manifold M, these higher obstructions are not automatically controlled by (ℓ,m)-extendability. The visible text does not contain a proof of the inheritance lemma, nor a statement of it, so the manifold case of Theorem 10.4 is not independently verifiable from the material provided.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript develops a framework for the strong approximation of Sobolev maps between manifolds, based on a new notion of genericity by composition: for each Sobolev map one constructs a summable 'detector' function w such that u∘γ is Sobolev or VMO for every Lipschitz map γ with w∘γ integrable (Fuglede maps). Using this screening tool and a VMO homotopy theory, the authors define ℓ-extendability and its refinement (ℓ,e)-extendability for maps u∈W^{k,p}(M;N) with kp<m. The central claims are that u∈H^{k,p}(M;N) if and only if u is (⌊kp⌋,m)-extendable (Theorem 1.14), and that (⌊kp⌋,e)-extendability exactly characterizes approximability by maps that are smooth off structured singular sets of rank m−e−1 (Theorem 1.17). The monograph also contains cohomological criteria (Hurewicz degree and currents), local-to-global criteria, and corollaries on weak density. The proof strategy is to establish the approximation theorem first for bounded Lipschitz open sets in R^m and then pass to compact manifolds by extending u to a tubular neighborhood of M in an ambient Euclidean space.","tokens_in":75406,"tokens_out":7652,"duration_ms":83568,"significance":"If correct, the results give a precise and largely self-contained answer to a long-standing question: the exact topological and analytical obstruction to strong approximation of Sobolev maps by smooth maps, unifying and extending work of Bethuel, Hang–Lin, White, and the authors' earlier paper. The detector/Fuglede-map formalism is an original and potentially reusable tool, and the visible chapters contain detailed, largely self-contained proofs with explicit constructions and examples. The claims are falsifiable and involve no fitted parameters. The VMO machinery, the cohomological criteria, and the approximation constructions in Chapters 2–7 are internally coherent and appropriately reference prior work. The main unresolved issue is the reduction from manifolds to open sets in Chapter 10, which is load-bearing for the central theorems stated for compact manifolds.","major_comments":[{"comment":"The reduction of Theorems 1.14 and 1.17 from a compact manifold M to the open-set case proceeds by extending u to v:=u∘Π on a tubular neighborhood U⊂R^κ of M, and the introduction asserts that v 'inherits the same extendability properties as u'. No statement or proof of this inheritance lemma is visible in the provided text. This step is load-bearing: the open-set theorem requires v to be (⌊kp⌋,κ)-extendable, while the hypothesis on u gives only (⌊kp⌋,m)-extendability with κ>m. Since β:=Π∘γ maps a κ-dimensional complex into the m-dimensional manifold M, the higher obstructions H^{j+1}(K^κ,K^ℓ;π_j(N)) for j≥m are not controlled by (ℓ,m)-extendability. The authors should either prove the inheritance statement or replace this reduction by an argument that does not require extendability in dimension κ.","section":"§10.9 and Introduction, p. 10–11"},{"comment":"The higher-order generic composition result is proved only for smooth maps Φ with bounded derivatives, and Proposition 2.16, the W^{k,p} analogue of Proposition 2.10, is stated with its proof omitted ('The details of the argument are omitted'). The later approximation theorems for k≥2 rely on higher-order composition and chain-rule estimates in the opening, thickening, and adaptive-smoothing constructions. The text should state precisely which regularity of Φ is needed in each application, and either supply the missing proof of Proposition 2.16 or give a complete reference. As written, the higher-order part of the framework is not independently verifiable from the material provided.","section":"§2.3, Propositions 2.14 and 2.16"},{"comment":"The transversal-perturbation axiom (c) in Definition 5.10 is used to prove the averaging estimate of Proposition 5.15, which in turn underpins the approximation of Fuglede maps and the stability arguments in Chapters 5, 6, and 10. For the Euclidean translation example the property is immediate, and for the tubular-neighborhood retraction it is plausible, but the verification for general compact manifolds is not given in the visible text. Since this is the only mechanism by which the constructions pass from R^m to manifolds, a short verification of (c) for Example 5.13 should be included.","section":"§5.2, Definition 5.10 and Proposition 5.11"}],"minor_comments":[{"comment":"The word 'Lispchitz' appears and should be corrected to 'Lipschitz'.","section":"§2.3, proof of Proposition 2.14"},{"comment":"The phrase 'symetric difference' should read 'symmetric difference'.","section":"§2.1, p. 17"},{"comment":"The statement refers to 'every summable function w : R^2→R', but detectors are [0,+∞]-valued; the codomain should be [0,+∞].","section":"Example 2.25"},{"comment":"The class R_i(M;N) is used in the introduction before the formal definition of structured singular sets appears; it would help the reader to state the definition of R_i(M;N) explicitly at first use rather than only in the introduction.","section":"Introduction, Definition 3.17"},{"comment":"In the proof, the path t↦Π∘v_{i,δt} uses the convention v_{i,0}=v_i; this convention is introduced only in the middle of the proof and should be stated before it is used.","section":"Chapter 4, Proposition 4.11"}],"recommendation":"major_revision","confidential_remarks":"This is a serious and substantial monograph, and I found no evidence of circularity: extendability is not defined as approximability, and the main theorems require real work in both directions. The central concern is the manifold reduction in §10.9: the inheritance of extendability under the tubular-neighborhood retraction is asserted but not proved in the visible text, and this is exactly the step that lets Theorems 1.14 and 1.17 go beyond the open-set case. If the authors can supply that lemma (or restructure the reduction), the paper is likely acceptable. I would also ask them to close the small gaps around Proposition 2.16 and the verification of Definition 5.10(c) for manifolds. The citation practice appears appropriate."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should know this paper is the real thing, not a routine extension. The authors build a new framework around Fuglede maps and generic composition: for each Sobolev map u they produce a summable detector w such that composing u with any Lipschitz map that satisfies w∘γ∈L1 lands in the right Sobolev or VMO class. That detector idea is genuinely novel and much more flexible than the old generic-restriction approach. They use it to define ℓ-extendability and (ℓ,e)-extendability, and then prove the expected characterization: a map in W^{k,p}(M;N) is strongly approximable by smooth maps iff it is (⌊kp⌋,m)-extendable, and approximable by maps with singular sets of rank m-e-1 iff it is (⌊kp⌋,e)-extendable. These are new results for compact manifolds and for higher k, and they subsume Bethuel's k=1 work on the ball.\n\nThe visible chapters are carefully written. Proposition 2.8 and its higher-order variant are solid; the VMO-detector machinery in Chapters 3 and 4 is coherent; the simplex-restriction characterization (Theorem 1.12) and the cohomological criterion (Theorem 1.13) are clean and genuinely useful. The paper credits prior work properly — Bethuel, Hang–Lin, White, and their own earlier paper — and I found no circularity or parameter fitting. The authors even include a limitation statement: Proposition 2.16's proof is omitted (\"details are omitted\"), which is honest but worth noting.\n\nThe soft spot is exactly where the stress-test points. For the manifold case, the proof reduces to a bounded Lipschitz open set by extending u to v = u∘~Π on a tubular neighborhood of dimension κ>m. The introduction asserts that v \"inherits the same extendability properties as u\". That inheritance is load-bearing: the open-set theorem is proved for (ℓ,m)-extendability, while v would need (ℓ,κ)-extendability, and the extension from the ℓ-skeleton to a κ-dimensional complex involves higher obstruction groups not controlled by the hypothesis on u. No proof of this inheritance lemma appears in the visible text, and the final approximation chapter (10.4–10.9) is not fully visible for audit. I cannot say the claim is false — the higher-order adaptation may handle it — but as it stands the main theorems are not independently verifiable from the material provided. That is a presentation gap, not a demonstrated error.\n\nThis is a monograph for specialists: anyone working on Sobolev maps between manifolds, calculus of variations, or the topology of VMO maps will want the framework. It deserves a serious referee, and I would accept it for peer review despite my own skepticism about the unverified step; the referee needs to check Chapter 10 closely, especially the manifold reduction. I would cite the generic-composition lemma for my own work even before the full proof is sorted out.","headline":"A serious, substantial monograph that introduces a genuinely new generic-composition framework and sharp if-and-only-if criteria for strong approximation on manifolds, but the manifold reduction in the final chapter rests on an unproved inheritance claim that needs close checking.","tokens_in":75925,"tokens_out":2296,"would_cite":true,"duration_ms":26752,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["58D15","46E35","58C25","47H11","26A99","55S35"],"pacs":[],"model":"deepseek-v4-flash","headline":"A Sobolev map can be approximated by smooth maps exactly when it passes a generic extendability test, and only the integer part of kp matters.","keywords":["Sobolev maps between manifolds","strong approximation of smooth maps","Fuglede maps","extension property","VMO maps","Hurewicz currents","distributional Jacobian","higher-order Sobolev spaces"],"falsifier":"A decisive calculation is to take the degree-one radial map $u(x)=x/|x|$ on $B^m$ into $S^{m-1}$ with $m-1<kp<m$: the theory predicts non-approximability because it is not $(\\lfloor kp\\rfloor,m)$-extendable, and indeed its distributional Jacobian is a nonzero delta at the origin. One can implement the detector selection numerically, sample generic spheres, and check whether any smooth sequence converges strongly in $W^{k,p}$; if such a sequence exists, the dichotomy collapses.","tokens_in":74942,"feed_emoji":"🧭","tokens_out":8187,"duration_ms":79542,"temperature":0.7,"pith_summary":"This monograph answers, for maps between compact manifolds, the question of when a Sobolev map can be strongly approximated by smooth maps. The answer is an extendability condition: when the domain dimension $m$ exceeds $kp$, a map $u\\in W^{k,p}(M;N)$ lies in the closure of smooth maps if and only if $u$ is $(\\lfloor kp\\rfloor,m)$-extendable, meaning that for every generic Lipschitz map from an $m$-dimensional simplicial complex into $M$, the composition $u\\circ\\gamma$ is homotopic in the VMO sense to a continuous map defined on the whole complex. More generally, $u$ is $(\\lfloor kp\\rfloor,e)$-extendable for $e$ between $\\lfloor kp\\rfloor+1$ and $m$ if and only if $u$ can be approximated by maps that are smooth except on structured singular sets of rank $m-e-1$. The framework makes the obstruction quantitative: only the integer part of $kp$ controls approximability, and all approximability criteria are tested through compositions with a single summable \"detector\" function. If correct, this reduces the approximation problem for any given map to a topological test on generic spheres or simplices.","feed_headline":"One extendability test decides smooth approximability","feed_subtitle":"For kp<m, a Sobolev map can be smoothed exactly when it is (floor(kp),m)-extendable; only the integer part matters.","key_machinery":"The load-bearing object is the Fuglede map/detector pair. For every Sobolev function $u$ there is a summable function $w$—the detector—such that any Lipschitz map $\\gamma$ with $w\\circ\\gamma$ integrable is a Fuglede map, and the composition $u\\circ\\gamma$ is Sobolev with the chain rule holding almost everywhere. This converts the classical notion of generic restriction (almost every sphere, almost every translation) into a single composition condition that is stable under limits and works uniformly on spheres, simplices, and simplicial complexes. On top of this, the paper builds VMO-homotopy, where maps are compared by continuous paths in VMO, and the $(\\ell,e)$-extendability conditions that package local and global topological obstructions. The final chapter's opening, thickening, adaptive smoothing, and shrinking constructions are what convert extendability into actual approximating sequences.","core_discovery":"On its own terms, the paper establishes Theorems 1.14 and 1.17: for $kp<m$, a map $u\\in W^{k,p}(M;N)$ belongs to $H^{k,p}(M;N)$ if and only if $u$ is $(\\lfloor kp\\rfloor,m)$-extendable, and for each $e\\in\\{\\lfloor kp\\rfloor+1,\\dots,m\\}$ it is $(\\lfloor kp\\rfloor,e)$-extendable if and only if it is the $W^{k,p}$ limit of maps in $R_{m-e-1}(M;N)$. The $(\\ell,e)$-extendability condition is defined through Fuglede maps: a summable detector $w$ on $M$ selects Lipschitz maps $\\gamma$ from an $e$-dimensional simplicial complex such that $w\\circ\\gamma$ is integrable, and the condition asks that $u\\circ\\gamma$ restricted to the $\\ell$-skeleton be VMO-homotopic to the restriction of a continuous map on the whole complex. The proof runs through opening, thickening, adaptive smoothing, and shrinking, which turn extendability into an explicit approximating sequence, first on Lipschitz open sets and then on compact manifolds via a smooth retraction of a tubular neighborhood. For maps into a sphere $S^n$ with $n\\le kp<n+1$, the paper also shows that $\\lfloor kp\\rfloor$-extendability is equivalent to the vanishing of the distributional Jacobian $d(u^\\#\\omega_{S^n})=0$.","pith_inferences":["The detector-based genericity suggests an algorithmic route: sample Lipschitz maps $\\gamma$ from a suitable family, check numerical integrability of $w\\circ\\gamma$ and homotopy triviality of $u\\circ\\gamma$, and thereby certify approximability of a given map without resolving the full homotopy groups of $N$.","The equality $H^{k,p}(M;N)=H^{1,\\lfloor kp\\rfloor}(M;N)$ suggests that in higher-order variational problems the topological selection of minimizers is governed by first-order data, which could simplify regularity arguments for polyharmonic maps.","The same framework may extend to fractional Sobolev maps $W^{s,p}$ with $sp$ playing the role of $kp$; if the chain of arguments adapts, only $\\lfloor sp\\rfloor$ should matter for smooth approximability.","A natural testable extension is to apply the extendability criterion to maps with singularities along lower-dimensional strata and ask whether the hierarchy $R_{m-e-1}$ gives a complete stratification of the space $W^{k,p}(M;N)$.","The paper's own Chapter 10 adaptation of the higher-order detector lemma to manifolds is the place to look for a potential gap; if that adaptation fails for some Lipschitz polytope, the equivalence between extendability and approximability would need a new foundation."],"forward_implications":["If correct, deciding whether a given Sobolev map is approximable by smooth maps reduces to checking extendability on generic simplices or spheres, without assuming anything global about the homotopy groups of the target manifold.","Approximability depends only on $\\lfloor kp\\rfloor$: $u\\in H^{k,p}(M;N)$ if and only if $u\\in H^{1,\\lfloor kp\\rfloor}(M;N)$, so higher-order regularity beyond the integer part of $kp$ does not change the topological obstruction.","Every map in $W^{k,p}(M;N)$ with $kp<m$ can be approximated by maps that are smooth outside a structured singular set of dimension $m-\\lfloor kp\\rfloor-1$, and approximation by maps with fewer singularities is possible exactly when the corresponding extendability test is passed.","For sphere-valued maps, the extendability criterion becomes cohomological: $d(u^\\#\\omega)=0$ in the sense of currents is equivalent to being in the smooth-approximation class, giving a computable obstruction in the range $n\\le kp<n+1$.","The same machinery gives a concrete handle on the weak density problem, including cases where weak density fails and the obstruction depends on analytical energy rather than only on topology."],"supporting_citations":[{"why":"Supplies the VMO homotopy theory and the density of bounded uniformly continuous maps used to define extendability and to compare VMO homotopy with classical homotopy.","marker":"[26]"},{"why":"Introduces the modulus/genericity-by-measures viewpoint from which the detector concept is drawn.","marker":"[44]"},{"why":"Identifies the global obstruction on manifolds that motivates the stronger $(\\ell,e)$-extendability notion and provides the transversal perturbation framework.","marker":"[60]"},{"why":"Provides the higher-order opening technique and the earlier $k\\ge2$ approximation scheme that this work extends.","marker":"[16]"},{"why":"Settles the $k=1$ density criterion that the new framework recovers as a special case.","marker":"[7]"},{"why":"Supplies the $kp\\ge m$ density argument and the trace-type obstructions that motivate the critical case.","marker":"[100]"},{"why":"Gives the generic composition lemma for translations that Proposition 2.8 generalizes to detectors on manifolds.","marker":"[108]"},{"why":"Introduces the opening construction that the present work adapts into the detector-based approximation scheme.","marker":"[24]"}],"fun_headline_variants":["Extendability test decides if Sobolev maps smooth out","Sobolev maps smooth iff (floor kp, m)-extendable","One extendability condition captures smooth approximability","Smooth approximation requires just one extendability test"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The entire machinery assumes the generic composition lemma and its higher-order, manifold version: for every Sobolev map there really is a single summable detector $w$ that makes all admissible compositions Sobolev and obeys the chain rule; if this lemma fails for the Lipschitz domains or polytopes used later, the equivalence between extendability and approximability loses its foundation.","fun_headline_variants_meta":{"raw":{"variants":["Extendability test decides if Sobolev maps smooth out","Sobolev maps smooth iff (floor kp, m)-extendable","One extendability condition captures smooth approximability","Smooth approximation requires just one extendability test"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000332,"raw_usage":{"total_tokens":1927,"prompt_tokens":1104,"completion_tokens":823,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":720,"completion_tokens_details":{"reasoning_tokens":754}},"tokens_in":720,"tokens_out":823,"duration_ms":7667,"temperature":1.0,"reasoning_tokens":754,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T00:27:27.746200+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A decisive calculation is to take the degree-one radial map $u(x)=x/|x|$ on $B^m$ into $S^{m-1}$ with $m-1<kp<m$: the theory predicts non-approximability because it is not $(\\lfloor kp\\rfloor,m)$-extendable, and indeed its distributional Jacobian is a nonzero delta at the origin. One can implement the detector selection numerically, sample generic spheres, and check whether any smooth sequence converges strongly in $W^{k,p}$; if such a sequence exists, the dichotomy collapses.","supporting_citations":[{"cited_title":"Fuglede,Extremal length and functional completion, Acta Math.98 (1957), 171–219.↑6, 19, 37 Preprint – January 2025 BIBLIOGRAPHY 205","cited_arxiv_id":null,"evidence_quote":"Introduces the modulus/genericity-by-measures viewpoint from which the detector concept is drawn."},{"cited_title":"Hang and F.-H","cited_arxiv_id":null,"evidence_quote":"Identifies the global obstruction on manifolds that motivates the stronger $(\\ell,e)$-extendability notion and provides the transversal perturbation framework."},{"cited_title":"Schoen and K","cited_arxiv_id":null,"evidence_quote":"Supplies the $kp\\ge m$ density argument and the trace-type obstructions that motivate the critical case."},{"cited_title":"160 (1988), 1–17.↑12, 19, 49, 70, 142","cited_arxiv_id":null,"evidence_quote":"Gives the generic composition lemma for translations that Proposition 2.8 generalizes to detectors on manifolds."}],"review_version":1}