{"id":"c78a268b-2c75-49c3-8e55-61969bceec88","arxiv_id":"1908.06568","paper_version":4,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The authors characterize, up to technical hypotheses, which moduli of continuity allow exceptional (Denjoy-type) C^{1,α} actions on the circle.","lead":"The authors prove that an integrability condition on the modulus of continuity is sufficient for the existence of exceptional C^{1,α} circle diffeomorphisms, and nearly necessary for actions of free abelian groups. The result unifies all previously known Denjoy counterexamples and yields new ones, with a partial converse tied to McDuff's question.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Proof of Theorem 2.10 uses condition (2) in the wrong direction: for α(x)=x^{0.6}, d=2, all hypotheses hold but the key quantity A(s,k,y) is unbounded, so the central claim is not proved.","rationale":"The reader correctly identified condition (2) as the most technical and least explained part of Theorem 2.10. However, the reader's concern was that the condition might be an artifact with no examples showing failure. My stress-test found a stronger, internal problem: the condition has the wrong inequality direction for the proof's central claim. The displayed estimate in the proof of Theorem 2.10 reduces the uniform boundedness of A(s,k,y) to bounding u/α(u^{d+1}/α(u)^d), while condition (2) bounds the reciprocal expression α(u^{d+1}/α(u)^d)/u. This is not a matter of missing examples or an extra technical assumption; it is a false implication in the proof. The counterexample α(x)=x^{3/5}, d=2, satisfies every hypothesis of Theorem 2.10 yet makes the claimed uniform bound diverge, so the proof cannot be repaired by a minor clarification. The theorem may still be true, and the construction may be salvageable with a corrected condition, but the central claim is unsupported as written. Therefore the paper should not be accepted in its current form. This is an internal correctness risk, not a disagreement with the surrounding literature.","tokens_in":24095,"tokens_out":16653,"duration_ms":162691,"concrete_test":"Set d=2, G=Z^2, ρ the standard rotation action, and α(x)=x^{3/5}. Take the proof's length sequence ℓ_y^k=1/ν(|y|_0+k) with ν(x)=x^3α(1/x)^2=x^{9/5}. For a generator s with |sy|_0=|y|_0+1 and j=|y|_0+k, compute A(s,k,y)=α(ℓ_y^k)^{-1}|1−ν(j)/ν(j+1)|. Since α(ℓ_j)=j^{-27/25} and 1−ν(j)/ν(j+1) ∼ (9/5)j^{-1}, A ∼ (9/5)j^{2/25} → ∞. All hypotheses of Theorem 2.10 are satisfied, so this directly falsifies the claim in the proof and shows condition (2) does not imply the needed bound.","verdict_should_be":"REJECT","load_bearing_attack":"In the proof of Theorem 2.10, after defining ν(x)=x^{d+1}α(1/x)^d and ℓ_y^k=g'(0)/ν(|y|_0+k), the proof must show that A(s,k,y)=α(ℓ_y^k)^{-1}|1−ℓ_{sy}^k/(s'(y)ℓ_y^k)| is uniformly bounded. The displayed estimate gives A ≍ 1/(j α(1/ν(j))) with j≍|y|_0+k. Writing u=1/j, this is u/α(u^{d+1}/α(u)^d). Condition (2) is sup α(u^{d+1}/α(u)^d)/u < ∞, i.e. the denominator is bounded above; this gives a lower bound on A, not the required upper bound. The needed hypothesis is sup u/α(u^{d+1}/α(u)^d) < ∞. The d=1 case is not indicative, since there monotonicity of ν(x)/x gives the needed lower bound. For a concrete admissible case, d=2, α(x)=x^{0.6}: we have ∫_0^1 x^{-1.2}dx < ∞ and condition (2) holds because the quotient equals y^{0.08} ≤ 1, but A ≈ 1.8 j^{0.08} → ∞. Thus the key claim in the proof is false, and the main construction of exceptional C^{1,α} actions for d>1 is not established by the paper as written.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper studies the construction of exceptional (Denjoy-type) actions of finitely generated groups on the circle in regularity C^{1,α}. The main result, Theorem 1.1, asserts that if a finitely generated group G with spherical growth O(n^{d-1}) admits a C^{1,α} action ρ with a free orbit and uniformly bounded log-derivatives, and if α satisfies ∫_0^1 α^{-d} < ∞ plus a technical condition (2) when d>1, then ρ can be C^1-approximated by exceptional C^{1,α} actions with uniformly bounded C^{1,α} constants. Theorem 1.2 gives the single-diffeomorphism version under ∫_0^1 α^{-1} < ∞. The paper also proves partial converses: exceptional C^{1,α} actions of Z^d impose the integrability ∫_0^1 α^{-1}(t)/t^{d+1} dt < ∞ under endpoint-derivative hypotheses (Theorem 1.3), and a complementary sufficiency theorem under a regularity condition on α (Theorem 1.6). The proofs use a blow-up construction with carefully chosen equivariant interval maps; the d=1 case is proved separately in Appendix A.","tokens_in":24430,"tokens_out":8049,"duration_ms":75923,"significance":"If correct, Theorem 1.1 would unify and generalize Herman's construction of C^{1,α} exceptional diffeomorphisms, recover all previously known moduli for single diffeomorphisms, and produce new moduli such as x log(1/x)(log log 1/x)^{1+ε}. The converse results are a useful step toward McDuff's length-spectrum question. The d=1 construction in Appendix A is self-contained, detailed, and appears sound; no parameters are fitted, and the fundamental estimate is used honestly. However, the d>1 group-action part rests on a technical estimate whose proof uses condition (2) in the wrong direction, so the central claim for d>1 is not established as written.","major_comments":[{"comment":"The proof of the claim that A(s,k,y) is uniformly bounded uses condition (2) in the wrong direction. From the displayed estimate the proof obtains A(s,k,y) ≲ 1/(j α(1/ν(j))). With u=1/j this is u/α(u^{d+1}/α(u)^d). Condition (2) bounds α(u^{d+1}/α(u)^d)/u from above, which gives a lower bound on this ratio, not the required upper bound. The needed hypothesis would be sup_{01, are not established by the present proof.","section":"§2.3, Theorem 2.10 proof, equation (2)"},{"comment":"Because the proof of Theorem 2.10 fails at the estimate just discussed, the advertised applications to free abelian groups of rank d>1 and to moduli such as those of Deroin–Kleptsyn–Navas are unsupported. Remark 2.11 explicitly tells the single-diffeomorphism reader to ignore the d>1 case, but the main theorem is stated for all d. The authors should either repair the estimate with a corrected condition and re-check the examples in Corollary 2.16 against that condition, or restrict the main theorem to the cases actually proved.","section":"Corollary 2.16 and Remark 2.11"}],"minor_comments":[{"comment":"The sentence 'we see that ‖f'−1‖ indeed tends to 1' contradicts the preceding statement that ‖f'−1‖ vanishes outside the intervals and also contradicts the claimed C^1 convergence to a rotation; it should read 'tends to 0'.","section":"Appendix A, proof of Theorem 1.2, final paragraph"},{"comment":"The sentence 'the claim trivially implies lim_{x→∞} ν(x+1)/ν(x) = 1' is not a direct consequence of the uniform boundedness of A; the earlier bound involving ν'/ν is what gives this limit. A one-line derivation would clarify the exposition.","section":"§2.3, proof of Theorem 2.10"},{"comment":"The spherical growth condition is stated as 'at most c n^{d−1}' without specifying the finite generating set; since the spherical growth function depends on the generating set, the statement should clarify whether the bound is assumed for some or every generating set.","section":"Theorem 1.1 and §2.2"},{"comment":"The abstract writes ∫ 1/α^d while the theorem writes ∫ 1/α(x)^d dx; please unify the notation for the integrability condition.","section":"Abstract and Theorem 1.1"}],"recommendation":"major_revision","confidential_remarks":"The d=1 part of the paper appears sound and is a credible contribution. The d>1 group-action theorem, however, relies on a directional error in condition (2), and the examples claimed to satisfy that condition need to be re-examined against the corrected inequality. I would ask the authors to fix this before further consideration."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Let me get straight to it: the paper is a good-faith, mostly readable contribution to the Denjoy counterexample story, and the appendix proof for single diffeomorphisms is solid. But the proof of the main theorem for d≥2 has a direction error in the key estimate. The stress-test note is right: condition (2) is used the wrong way.\n\nWhat is actually new: Theorem 1.1 gives a group-level generalization of Herman's blow-up construction; Theorem 1.2 recovers all known moduli and adds new ones like x log(1/x)(log log 1/x)^{1+ε}. The converse results (Theorems 1.3 and 1.4) relating integrability of α^{-1}/t^{d+1} to interval length bounds are genuinely new and useful. The appendix is a careful self-contained proof for G=Z, and the fundamental estimate (Lemma 3.1) is a nice tool.\n\nThe soft spot is real. In the proof of Theorem 2.10, the claim that A(s,k,y) is uniformly bounded reduces to bounding 1/(j α(1/ν(j))) from above. With u=1/j, that is u/α(u^{d+1}/α(u)^d). Condition (2) asserts sup α(...)/u < ∞, which bounds this quantity from below, not above. For d=2 and α(x)=x^{0.6}, all hypotheses hold, yet A ~ j^{0.08} → ∞. So the C^{1,α} regularity of the constructed blow-ups for d>1 is not proved. This isn't a cosmetic gap; it is the load-bearing estimate for the group-action theorem. The d=1 case is handled separately and passes, and the converses don't rely on this step. There is also a minor typo in the appendix where ||f'-1|| is said to tend to 1 instead of 0.\n\nWho is this for: people interested in exceptional circle actions, Denjoy theory, and regularity thresholds. The d=1 results and the converse section are worth reading and citing. The d>1 generalization is attractive but currently unproved as written. A serious referee should send it back for a fix or a weakened statement. The paper deserves peer review because it is substantial and the flaw may be repairable, but the burden is on the authors to sort out the inequality.","headline":"A substantial paper with a genuinely useful single-diffeomorphism proof and new converses, but the main d≥2 group-action theorem rests on an estimate that uses condition (2) in the wrong direction and is not proved as written.","tokens_in":24941,"tokens_out":4430,"would_cite":true,"duration_ms":41176,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["37E10","37C05","37C15"],"pacs":[],"model":"deepseek-v4-flash","headline":"The integrability of the modulus of continuity decides when exceptional circle actions exist.","keywords":["exceptional circle diffeomorphisms","moduli of continuity","Denjoy counterexamples","C^{1,α} actions on the circle","spherical growth","wandering intervals","length spectrum","free abelian group actions"],"falsifier":"Compute $A(s,k,y)=|1-\\ell_{sy}/(s'(y)\\ell_y)|/\\alpha(\\ell_y)$ for the paper's length choice $\\ell_y=g'(0)/\\nu(\\|y\\|_0+k)$ with $\\nu(x)=x^{d+1}\\alpha(1/x)^d$, using a concave $\\alpha$ that satisfies $\\int_0^1 \\alpha(x)^{-d}\\,dx<\\infty$ but violates condition (2). If for some generator $s$ and some sequence $y_k$ the values $A(s,k,y_k)$ are unbounded, the key estimate of Theorem 2.10 fails; a computation showing boundedness for all such $\\alpha$ would indicate that condition (2) is removable rather than a genuine boundary.","tokens_in":23917,"feed_emoji":"🌀","tokens_out":14908,"duration_ms":140304,"temperature":0.7,"pith_summary":"This paper asks which moduli of continuity can appear in the derivatives of exceptional circle diffeomorphisms, the dynamical systems Denjoy constructed as counterexamples to smooth rigidity. The main theorem states that if a finitely generated group has spherical growth at most $c n^{d-1}$ and admits a $C^{1,\\alpha}$ action with a free orbit and uniformly bounded logarithms of derivatives, then the integrability condition $\\int_0^1 \\alpha(x)^{-d}\\,dx < \\infty$ (together with a technical inequality when $d>1$) guarantees a sequence of exceptional $C^{1,\\alpha}$ actions converging to it in the $C^1$ topology. For a single diffeomorphism this yields an exceptional $C^{1,\\alpha}$ diffeomorphism at every irrational rotation number whenever $\\int_0^1 dx/\\alpha(x) < \\infty$, recovering all previously known attainable moduli and adding new ones such as $\\alpha(x)=x\\log(1/x)(\\log\\log(1/x))^{1+\\varepsilon}$. In the opposite direction, the paper shows that exceptional $C^{1,\\alpha}$ actions of $\\mathbb{Z}^d$ with natural derivative hypotheses force $\\int_0^1 \\alpha^{-1}(t)/t^{d+1}\\,dt < \\infty$. Together the two directions frame the long-standing question of exactly which moduli admit Denjoy counterexamples.","feed_headline":"When 1/α is integrable, Denjoy counterexamples exist","feed_subtitle":"New proof covers all known moduli, adds x log(1/x)(log log 1/x)^(1+ε), and gives partial converses.","key_machinery":"The machinery is the orbit blow-up construction, refined to $C^{1,\\alpha}$ regularity. One assigns to each point $y$ of a free orbit a positive length $\\ell_y$ with total mass tending to zero, replaces $y$ by an interval of that length, and interpolates the derivative of the lifted map by a bump function so that the derivative distortion on the interval is governed by the quotient $\\ell_{sy}/(s'(y)\\ell_y)$. The paper chooses $\\ell_y = g'(0)/\\nu(\\|y\\|_0+k)$ with $\\nu(x)=x^{d+1}\\alpha(1/x)^d$, where $\\|y\\|_0$ is the orbit distance; the growth hypothesis makes $\\sum_y \\ell_y$ finite, and the monotonicity of $\\nu(x)/x$ makes the ratios $\\ell_{sy}/\\ell_y$ tend to $1$. The load-bearing estimate is the uniform bound on $A(s,k,y) = |1-\\ell_{sy}/(s'(y)\\ell_y)|/\\alpha(\\ell_y)$; concavity of $\\alpha$ turns that bound into a uniform $C^{1,\\alpha}$ seminorm for the lifted derivatives.","core_discovery":"The central claim is that the regularity border for exceptional actions is encoded by integrability of the reciprocal modulus. More precisely, Theorem 1.1 asserts that under the stated hypotheses, the given action can be blown up at a free orbit to produce nontrivial exceptional actions $\\rho_k$ with $\\rho_k\\to\\rho$ in $C^1$ and with $\\sup_k [\\rho_k(g)']_\\alpha < \\infty$ for each $g$. The case $G=\\mathbb{Z}$ gives the one-dimensional statement: for every irrational rotation number $\\theta$, the condition $\\int_0^1 dx/\\alpha(x)<\\infty$ is sufficient for the existence of a $C^{1,\\alpha}$ exceptional diffeomorphism with rotation number $\\theta$, arbitrarily $C^1$-close to a rotation. The paper also proves a partial converse: for exceptional $C^{1,\\alpha}$ actions of $\\mathbb{Z}^d$ that are semi-conjugate to rotations with matched derivatives on the minimal set, the integral $\\int_0^1 \\alpha^{-1}(t)/t^{d+1}\\,dt$ must converge, and under the extra regularity condition $\\sup_t \\alpha(t)/(t\\alpha'(t))<\\infty$ this integrability is again sufficient.","pith_inferences":["The two integrability conditions bracket the classical open case $\\alpha(x)=x\\log(1/x)$: the sufficient integral diverges and the necessary integral diverges too, so the theorem neither constructs nor rules out such diffeomorphisms; settling the open case would require understanding endpoint derivatives of wandering intervals rather than moduli alone.","Condition (2) is the most fragile premise for $d>1$; if it is an artifact, the construction should work for moduli that violate it, and the natural check is to run the same length formula with a slowly varying modulus where the supremum diverges yet the $\\alpha$-seminorm bound still appears to hold.","Using the paper's Proposition 2.12 relating spherical and word growth for nilpotent groups, one could restate the main hypothesis as polynomial word growth of degree $d$ and carry the construction over to virtually nilpotent groups of arbitrary nilpotency class, up to the same technical condition on $\\alpha$."],"forward_implications":["Every irrational rotation number admits a $C^{1,\\alpha}$ exceptional diffeomorphism for every concave modulus with $\\int_0^1 dx/\\alpha(x)<\\infty$, recoverable arbitrarily $C^1$-close to a rotation.","All previously known attainable moduli, including H\\\"older moduli and Herman's $\\alpha(x)=x(\\log 1/x)^{1+\\varepsilon}$, are recovered, and new moduli such as $x\\log(1/x)(\\log\\log(1/x))^{1+\\varepsilon}$ are shown attainable.","If an exceptional diffeomorphism has wandering interval lengths with $\\ell_{i+1}/\\ell_i\\to 1$ and is $C^{1,\\alpha}$, then $\\alpha(\\ell_i)\\ge A/i$ for some $A>0$; in particular moduli like $x\\log(1/x)$ are impossible in this class.","For $\\mathbb{Z}^d$, exceptional $C^{1,\\alpha}$ actions with derivative matching force $\\int_0^1\\alpha^{-1}(t)/t^{d+1}\\,dt<\\infty$, while the same condition plus $\\sup_t\\alpha(t)/(t\\alpha'(t))<\\infty$ suffices, so the two statements are near-converses.","The main approximation theorem applies to every finitely generated group with spherical growth $O(n^{d-1})$, giving exceptional $C^{1,\\alpha}$ approximations for non-abelian groups as well as for $\\mathbb{Z}$."],"supporting_citations":[{"why":"Supplies Herman's construction of $C^1$ and $C^{1,\\alpha}$ exceptional diffeomorphisms close to rotations, which the paper's blow-up method generalizes to group actions.","marker":"[14]"},{"why":"Gives Denjoy's original wandering-interval counterexamples and the $C^0$ blow-up procedure that all later constructions refine.","marker":"[9]"},{"why":"Provides the standard treatment of exceptional actions, the fundamental estimate, and the known lower-bound examples such as Exercise 4.1.26 that the converse results sharpen.","marker":"[24]"},{"why":"Shows a concave modulus can be replaced by a smooth concave majorant, justifying the assumption that $\\alpha$ is differentiable in the proof.","marker":"[22]"},{"why":"Gives intermediate-regularity $C^{1,\\alpha}$ exceptional actions of free abelian groups that the $\\mathbb{Z}^d$ corollaries extend and recover.","marker":"[10]"},{"why":"Supplies the standard deformation and circle-diffeomorphism background used in the appendix for the single-diffeomorphism construction.","marker":"[17]"}],"fun_headline_variants":["Integrability of modulus yields exceptional actions","Exceptional circle actions if 1/α integrable","Partial converse: abelian groups force integrability","All known moduli covered by integrability criterion"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"For $d>1$, the proof relies on the technical inequality $\\sup_{0