{"id":"fa88c332-14a5-44a9-a62f-e8ea69745e36","arxiv_id":"2507.15808","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For n≥3, any short immersion of an n-dimensional domain into R^{2n} can be uniformly approximated by C^{1,θ} isometric immersions with θ close to 1/n in odd dimensions and 1/(n+1) in even dimensions.","lead":"A math paper improves the smoothness of length-preserving maps from n-dimensional shapes into 2n-dimensional space, reaching Hölder exponents close to 1/n in odd dimensions and 1/(n+1) in even dimensions. This narrows the gap between flexible and rigid isometric embeddings in high codimension.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Global normal frames are used at every substep (3.9), (3.20), (3.25), but they are supplied by Lemma 2.9, whose embedding hypothesis is never verified for the iterates u_m. This is the load-bearing gap: without a frame lemma for short immersions, the iteration is not defined.","rationale":"The reader's conditional verdict is appropriate. The central construction is a standard Nash-Kuiper iteration, and the only place where the text relies on an unproved structural input is the global normal frame. I checked whether this is merely cosmetic: for d=2n, parallelizability of Ω and the connectivity of V_{2n,n} suggest that a frame exists for any full-rank immersion, so the concern is probably repairable. However, the paper does not provide the needed lemma, and the derivative bounds on the frame are essential to the estimates. Thus I do not reject the paper; I keep the conditional verdict and request one missing lemma. My agreement is partial because the reader frames the issue as embedding versus immersion, whereas the essential missing piece is a frame lemma for the maps actually generated by the iteration.","tokens_in":27294,"tokens_out":27784,"duration_ms":317579,"concrete_test":"Prove and verify a variant of Lemma 2.9 for C^{N+1} immersions (not embeddings) u:Ω→R^{2n} with γ^{-1}Id≤∇u^t∇u≤γId on a bounded simply connected domain. Use the global frame of TΩ and the (n−1)-connectivity of V_{2n,n} to trivialize the normal bundle, then check whether the trivialization can be chosen so that [η_i]_k≤C(1+[u]_{k+1}). If the proof goes through, the uses at (3.9), (3.20), and (3.25) are justified and the theorem stands; if it does not, the iteration must be modified before Theorem 1.1 is established.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The proof hinges on Lemma 2.9 at every stage of the iteration. Equations (3.9), (3.20), and (3.25) require a global orthonormal normal frame for u_m or u_{m,i}, and the derivative bounds (3.12) are taken from Lemma 2.9. But Proposition 3.1 and Theorem 1.1 only provide short immersions; no proof is given that u_m is an embedding or that the normal bundle of u_m admits a frame with the stated bounds. This is not a cosmetic issue: the frame is the object used to define the perturbations, so without it the construction cannot even be written down. The gap is probably repairable: for d=2n, Ω is parallelizable and the Stiefel manifold V_{2n,n} is (n−1)-connected, so the normal bundle of any full-rank immersion is trivial and a frame should exist for immersions, not just embeddings. But the paper does not state or prove such a frame lemma, and the derivative estimates on the frame are not automatic from the cited references. Thus the central claim is conditional on a missing structural lemma.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper claims a local Nash-Kuiper theorem for isometric immersions of an n-dimensional simply connected domain Ω⊂R^n into R^{2n}: from any short immersion \\bar{u}∈C^{1,ε}, it constructs an isometric immersion u∈C^{1,θ} with θ=1/n−ε for odd n and θ=1/(n+1)−ε for even n, together with uniform approximation and explicit C^1 bounds (Theorem 1.1). The proof follows the standard Nash iteration: after mollification and a Källén-type decomposition of the metric error, the map is modified by n 'spiral' perturbations and then by (n−1)/2 or n/2 'corrugation' perturbations per stage, with carefully chosen frequencies λ_{m,i} and amplitudes δ_{m+1}^{1/2}, and an iterative integration-by-parts lemma. The constants b, ϑ, α, J are fixed explicitly in (3.2) and (3.4), and the parameter a is chosen large but independent of the stage index m. An auxiliary matrix decomposition (Proposition 4.1) is developed for possible future improvements in even dimensions.","tokens_in":27490,"tokens_out":9870,"duration_ms":104552,"significance":"If correct, Theorem 1.1 improves the Hölder exponent for high-codimension local isometric immersions from 1/(n+2)−ε (Cao–Székelyhidi [9]) to 1/n−ε in odd dimensions and 1/(n+1)−ε in even dimensions. The paper also provides explicit C^1 estimates, which is a useful quantitative feature, and the proof makes the dependence on ε and n transparent through the explicit parameter choices in (3.2)–(3.4). The construction appears to be forward-moving, with no circularity; the external lemmas cited are from the standard literature. The main obstacle is the use of a global normal frame lemma that is stated only for embeddings, while the iteration only produces immersions; this gap affects the definition of the perturbations and must be repaired before the result is fully established. The auxiliary Proposition 4.1 is not needed for Theorem 1.1.","major_comments":[{"comment":"The proof of Proposition 3.1 defines the perturbations using global normal frames (ζ_{i−1}, η_{i−1}) supplied by Lemma 2.9, but that lemma is stated only for an embedding u with a two-sided derivative bound γ^{−1}Id ≤ ∇u^t∇u ≤ γId. At each stage the text only knows that u_m is a short immersion; no argument is given that the iterates u_{m,i} are injective, nor that the normal bundle of these maps admits a global frame with the derivative estimates used in (3.12). Since (3.9), (3.20), and (3.25) are the definitions of the new maps, the iteration is not defined without an additional frame lemma for immersions. This is a load-bearing gap in the proof of Theorem 1.1.","section":"Section 2, Lemma 2.9; Section 3, Eqs. (3.9), (3.20), (3.25)"},{"comment":"The construction of u_0 uses the same Lemma 2.9 to choose normal vectors \\bar{ζ}_{i−1} and \\bar{η}_{i−1} for the Nash spirals defining \\bar{u}_i. The maps \\bar{u}_{i−1} are only known to be short immersions, not embeddings, so the same missing justification occurs already in the base step of the iteration.","section":"Section 3.4, display after Eq. (3.34)"}],"minor_comments":[{"comment":"The word 'folllows' in the sentence 'It then folllows from the linear independence...' is a typo for 'follows'.","section":"Section 2, proof of Lemma 2.5"},{"comment":"The phrase 'their slope become steeper' is grammatically incorrect; it should be 'their slopes become steeper'.","section":"Abstract and Remark 1.3"},{"comment":"The claim that the assumption on ε can be relaxed to 0<ε<2/n² appears inconsistent with the condition 0<ε<2/(3n+1) used in the proof of Proposition 3.1, since 2/(3n+1)<2/n² for n≥3; please clarify the intended range of ε.","section":"Remark 1.4 and proof of Proposition 3.1"},{"comment":"The induction is performed for a fixed p>2 and concludes 'by the arbitrariness of p>2', but this only yields a^j for j≤p−2 for each fixed p; an argument covering all j≥0 (for instance, a diagonal or inverse-limit argument) is missing. Since Proposition 4.1 is not used in the proof of Theorem 1.1, this does not affect the main result.","section":"Section 4, proof of Proposition 4.1"},{"comment":"The definition of β has different expressions for odd and even n, but the surrounding text writes 'for any n≥3' without specifying the branch; please make the case distinction explicit in the displayed equation.","section":"Section 3, Eq. (3.30)"}],"recommendation":"major_revision","confidential_remarks":"The only substantive gap I see is the missing frame lemma for immersions. If the author can prove a version of Lemma 2.9 under the standing hypotheses, or directly show that the normal bundle of the iterates admits a global frame with the required derivative bounds, the main theorem would be established. Section 4 should be repaired or its auxiliary status clarified, but it is not needed for the main claim."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper genuinely moves the known Hölder threshold in high-codimension isometric immersion: from 1/(n+2)-epsilon to 1/n-epsilon in odd dimensions and to 1/(n+1)-epsilon in even dimensions, with explicit C1 bounds whose monotonicity in the metric error matches the geometry. That is a real step, not a routine parameter push. The proof combines Nash spirals, Kuiper corrugations, and the iterated integration-by-parts machinery from [10], and the decomposition in Lemma 3.4 is honestly worked out. Section 4 is candid about why the even-dimensional case cannot be pushed further with the present method; that kind of limitation statement is rare and useful.\n\nWhere the paper is soft: the global normal-frame lemma. Lemma 2.9 is stated for embeddings with a two-sided derivative bound, while Proposition 3.1 and Theorem 1.1 only know u_m is a short immersion. Equations (3.9), (3.20), and (3.25) use normalized normal vector fields at every substep, and the derivative bounds (3.12) come straight from Lemma 2.9. As written, the iteration is not defined for maps that stop being embedded. I do not think this is fatal. For a simply connected domain Ω in R^n mapped into R^{2n}, the normal bundle of any full-rank immersion is trivial because the Stiefel manifold V_{2n,n} is (n-1)-connected, so a global frame exists. What is missing is the derivative estimate on such a frame, which does not follow automatically from the cited references. That is repairable, but the paper must state and prove an immersion-version frame lemma. Relatedly, the exposition relies heavily on \"direct computation\" flags; I did not verify every estimate, and a referee will need to check the parameter inequalities in (3.28)-(3.30) carefully. I saw no circularity: the improved exponent is not used as an input, and the external lemmas are from the right literature.\n\nBottom line: for people working in Nash-Kuiper convex integration, this is a useful, serious contribution. It is not ready as written because of the frame gap, but the central construction is credible and the repair route is clear. Send it to a competent referee; ask for an immersion-version frame lemma and a pass on the delegated estimates, then it should be publishable.","headline":"A real step up in Hölder regularity for high-codimension isometric immersions, but the construction currently hangs on a normal-frame lemma that is only stated for embeddings.","tokens_in":28091,"tokens_out":2752,"would_cite":true,"duration_ms":30891,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["53C42","53C40"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that for $n\\ge 3$, any short $C^{1,\\varepsilon}$ immersion of a smooth bounded simply connected domain in $\\mathbb{R}^n$ into $\\mathbb{R}^{2n}$ admits a uniformly close $C^{1,\\theta}$ isometric immersion, with…","keywords":["isometric immersion","Nash–Kuiper theorem","Hölder regularity","convex integration","high codimension","short immersion","strong embedding theorem","iterative integration by parts"],"falsifier":"Apply the construction to a short immersion of a disk into $\\mathbb{R}^4$ that is not injective, or whose derivative has no uniform lower bound, and compute the smallest singular value of $\\nabla u_m$ for the first few iterates; if it approaches zero, or if some $u_m$ self-intersects, Lemma 2.9 cannot supply the normal frames used in (3.20) and (3.25), and the iteration as written is not defined.","tokens_in":27041,"feed_emoji":"📐","tokens_out":14263,"duration_ms":126394,"temperature":0.7,"pith_summary":"The paper establishes a sharper local Nash–Kuiper theorem in high codimension. For $n\\ge 3$, any $C^{1,\\varepsilon}$ short immersion of a smooth bounded simply connected domain $\\Omega\\subset\\mathbb{R}^n$ into $\\mathbb{R}^{2n}$ is shown to admit a uniformly close $C^{1,\\theta}$ isometric immersion, with $\\theta=1/n-\\varepsilon$ when $n$ is odd and $\\theta=1/(n+1)-\\varepsilon$ when $n$ is even. This improves the previously known Hölder exponent $1/(n+2)-\\varepsilon$. The proof also yields explicit $C^1$ bounds that grow with the initial metric error, matching the intuition that a larger deficit requires steeper corrections. A corollary gives infinitely many $C^{1,\\theta}$ isometric embeddings of any $C^1$ metric on such domains into $\\mathbb{R}^{2n}$.","feed_headline":"Isometric immersions gain Hölder exponent 1/n in odd dimensions","feed_subtitle":"Sharpening the local Nash–Kuiper threshold from 1/(n+2) to 1/n for odd n and 1/(n+1) for even n.","key_machinery":"The construction is an iterative scheme in which each step reduces the metric error using two oscillatory building blocks: Nash-spirals and Kuiper-corrugations. A Källén-type decomposition (Lemma 2.3) writes the current metric error as a sum of rank-one primitive metrics with smooth positive coefficients, and an iterative integration-by-parts lemma (Lemma 2.7) rewrites a high-frequency oscillatory symmetric matrix as the symmetric gradient of a small perturbation plus a lower-order remainder in a complementary span. The perturbation formulas (3.20) and (3.25) rely on global normal vector frames along the intermediate immersions, supplied by Lemma 2.9. For even $n$, Proposition 4.1 provides a refined matrix decomposition that absorbs the cross terms generated by the $n/2$ Nash-spiral steps.","core_discovery":"The central claim is Theorem 1.1: the local Nash–Kuiper approximation scheme for isometric immersions into $\\mathbb{R}^{2n}$ can be run with a higher Hölder exponent than previously known, and the gain depends on the parity of the dimension. In odd dimensions the final map is $C^{1,1/n-\\varepsilon}$; in even dimensions the scheme reaches $C^{1,1/(n+1)-\\varepsilon}$. The even-dimensional obstruction is identified explicitly: the $n/2$ Nash-spiral steps used there force a larger growth $\\Lambda^{n/2}$ of the oscillation parameter, whereas odd dimensions need only $\\Lambda^{(n-1)/2}$. To handle this, Section 4 develops a modified matrix decomposition (Proposition 4.1) in which the $\\Lambda^{-1/2}$ error terms are absorbed into the symmetric-matrix decomposition through an implicit-function argument. The paper also records explicit $C^1$ estimates showing that the constructed immersion's slope increases when the initial short map is further from being isometric.","pith_inferences":["If the global normal-frame issue can be resolved, the same iteration would likely extend from local domains to compact manifolds without changing the Hölder exponent, since all estimates are local with constants depending only on the geometry.","The even-dimensional decomposition in Proposition 4.1 suggests a concrete route to removing the even-dimensional deficit: reduce the oscillation growth from $\\Lambda$ to $\\sqrt{\\Lambda}$ in the $n/2$ Nash-spiral steps, absorbing the larger errors through the implicit-function decomposition; this modification is not carried out in the paper.","The parity-dependent exponents are probably an artifact of the construction rather than a geometric threshold, and a numerical experiment on a simple $n=4$ short immersion could test whether the $\\Lambda^{n/2}$ growth actually appears or whether a smaller growth suffices.","The quantitative $C^1$ bounds may transfer to stability questions near the rigidity threshold: when the initial metric error is small, the constructed isometric immersion is close in $C^1$ to the short immersion, suggesting a quantitative flexibility statement."],"forward_implications":["If Theorem 1.1 is correct, the known Hölder exponent for local high-codimension isometric immersions improves from $1/(n+2)-\\varepsilon$ to $1/n-\\varepsilon$ in odd dimensions and to $1/(n+1)-\\varepsilon$ in even dimensions.","Corollary 1.6 yields infinitely many $C^{1,\\theta}$ isometric embeddings of any $C^1$ metric on a smooth bounded simply connected domain into $\\mathbb{R}^{2n}$, a local isometric analogue of the strong embedding theorem.","The explicit $C^1$ estimates make the construction quantitative: for a given short immersion, the $C^1$ norm of the isometric approximation is controlled by the size of the initial metric error, so the method can track how much slope is needed for a prescribed accuracy.","The even-dimensional analysis isolates the exact term that lowers the exponent when $n$ is even, identifying a specific obstruction whose removal would close the gap to $1/n$."],"supporting_citations":[{"why":"establishes the previous Hölder exponent $1/(n+2)-\\varepsilon$ in this high-codimension setting, the comparison point improved by Theorem 1.1.","marker":"[9]"},{"why":"develops the iterative integration-by-parts lemma and the Kuiper-type corrugation ansatz adapted here.","marker":"[10]"},{"why":"supplies the symmetric-matrix decomposition and the mollification estimates that underpin the iteration.","marker":"[12]"},{"why":"provides the multiple-perturbation strategy and the symmetric positive definite matrix decomposition behind Lemma 2.3.","marker":"[29]"},{"why":"introduces the original perturbation iteration and spiral construction that the scheme generalizes.","marker":"[31]"},{"why":"introduces the corrugation building blocks used in the later steps of the construction.","marker":"[30]"},{"why":"gives the explicit periodic function used to define the Kuiper corrugations in Lemma 3.6.","marker":"[15]"},{"why":"supplies the lemma on global normal vector frames for embeddings that the perturbation formulas require.","marker":"[8]"},{"why":"provides a matrix decomposition with square-root cross terms adapted in Proposition 4.1.","marker":"[16]"}],"fun_headline_variants":["Nash–Kuiper sharpened: Hölder up to 1/n for odd n, 1/(n+1) for even","Parity-dependent Hölder boost: C^{1,1/n} in odd, C^{1,1/(n+1)} in even","Explicit C^1 slope growth for high-codimension isometric immersions","Odd dimensions get Hölder exponent 1/n for isometric immersions"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The iteration needs global normal vector frames along every intermediate map, but the lemma that produces them requires the map to be an embedding with a uniform two-sided bound on its derivative; the theorem assumes only a short immersion, and the paper does not prove the iterates stay injective with such a bound.","fun_headline_variants_meta":{"raw":{"variants":["Nash–Kuiper sharpened: Hölder up to 1/n for odd n, 1/(n+1) for even","Parity-dependent Hölder boost: C^{1,1/n} in odd, C^{1,1/(n+1)} in even","Explicit C^1 slope growth for high-codimension isometric immersions","Odd dimensions get Hölder exponent 1/n for isometric immersions"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.002084,"raw_usage":{"total_tokens":8110,"prompt_tokens":958,"completion_tokens":7152,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":574,"completion_tokens_details":{"reasoning_tokens":7041}},"tokens_in":574,"tokens_out":7152,"duration_ms":49105,"temperature":1.0,"reasoning_tokens":7041,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T15:23:23.245543+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Apply the construction to a short immersion of a disk into $\\mathbb{R}^4$ that is not injective, or whose derivative has no uniform lower bound, and compute the smallest singular value of $\\nabla u_m$ for the first few iterates; if it approaches zero, or if some $u_m$ self-intersects, Lemma 2.9 cannot supply the normal frames used in (3.20) and (3.25), and the iteration as written is not defined.","supporting_citations":[{"cited_title":"Cao and L","cited_arxiv_id":null,"evidence_quote":"establishes the previous Hölder exponent $1/(n+2)-\\varepsilon$ in this high-codimension setting, the comparison point improved by Theorem 1.1."},{"cited_title":"Conti, C","cited_arxiv_id":null,"evidence_quote":"supplies the symmetric-matrix decomposition and the mollification estimates that underpin the iteration."},{"cited_title":"K¨ all´ en, Isometric embedding of a smooth compact manifold with a metric of low regularity","cited_arxiv_id":null,"evidence_quote":"provides the multiple-perturbation strategy and the symmetric positive definite matrix decomposition behind Lemma 2.3."},{"cited_title":"Nash,C 1 isometric imbeddings","cited_arxiv_id":null,"evidence_quote":"introduces the original perturbation iteration and spiral construction that the scheme generalizes."},{"cited_title":"Kuiper, OnC 1-isometric imbeddings","cited_arxiv_id":null,"evidence_quote":"introduces the corrugation building blocks used in the later steps of the construction."},{"cited_title":"De Lellis, D","cited_arxiv_id":null,"evidence_quote":"gives the explicit periodic function used to define the Kuiper corrugations in Lemma 3.6."},{"cited_title":"Cao and D","cited_arxiv_id":null,"evidence_quote":"supplies the lemma on global normal vector frames for embeddings that the perturbation formulas require."},{"cited_title":"De Lellis and D","cited_arxiv_id":null,"evidence_quote":"provides a matrix decomposition with square-root cross terms adapted in Proposition 4.1."}],"review_version":1}