claude-fable-5-2 high: bw-holonomy-note
Reviewer record, not the publication verdict. The response includes unaccepted proposals and source limits. Missing antecedent proofs are not automatically defects in the paper. The initial request for Lemmas 8.4–8.5 was our inventory error; the journal ends at Lemma 8.3. The curated report records final dispositions.
Completed visible assessment, preserved as a readable check record. It is a model assessment, not a formal certificate. This document contains no private reasoning stream.
No obstruction to note Theorem 0.2 or to journal Theorem 0.3 was found, but the 2005 preliminary note does not prove its theorem as written. Its Lemma 1.1 is false as stated (explicit counterexample A x id). The version actually used in Lemma 1.2 and Step 2 is a different, unproved and also false Lipschitz form, and several cocycle base points are wrong. The Step 3 identity is incorrect. Step 4 fails when the center expands and gammagammahat is near 1. All of these are false auxiliary claims or repairable proof steps, not a false principal theorem. An affine one-step estimate that tracks base and fiber distances separately restores Lemma 1.2, the Cauchy estimates and uniform C^1 convergence under exactly the hypotheses nu^theta < gammagammahat, theta-Hoelder E^c and theta < delta; these repairs are mine and are labelled as new. The C^{1+beta} claim survives only with an unspecified beta. The journal's C^{1+delta} branch additionally depends on unwritten transfers: from genuine coherent foliations on compact M to the fake foliations of F_r, from center-to-center holonomy to C^1 subfoliation, and on the Hoelder exponent of the fake center bundle. It cites a 2006 preprint that postdates the recorded cutoff and whose relation to the supplied 2005 note is unresolved. The C^2 branch does not depend on the note. No external tools, web search or other sources were used.
H1
Checked statement: Supporting note (June 11 2005, preliminary version) Theorem 0.2: f is C^{1+delta}, partially hyperbolic, dynamically coherent on compact M, nu^theta < gamma*gammahat, with theta in (0,delta) a Hoelder exponent for E^c; then the stable holonomy between center transversals is C^{1+beta} for some beta>0. Checked together with the standing definitions (note p.1), the construction of pi^s and h^s_n (note p.4) and Step 1 (note p.5).
Returned assessment: repairable_gap
CHECKED. (a) Rate consistency. T pi at x_n deviates projectively by O(nu_n(x)^theta), because tangent planes of center leaves at stable distance nu_n differ by that much when E^c is theta-Hoelder. Pull-back by f_*^{-n} on P(E^c) expands by at most ((gamma gammahat)n(x))^{-1}. So nu^theta < gamma gammahat is the right pointwise rate, and theta < delta lets the delta-Hoelder error of Tf be absorbed. I found no counterexample to Theorem 0.2. Under the stated hypotheses the C^1 conclusion follows from the repaired arguments recorded under H2 and H3 (my own repairs, not the note's proof). (b) Step 1 is correct in substance. Properties 1 and 3 give d(pi(x_n), h^s(x_n)) <= C nu_n(x), both points lie on W^c(q_n), and f^{-n} expands center distances by at most gamma_n(.)^{-1} along nearby orbits. Hence d(h_n x, h x) <= C (nu/gamma)n(x) <= C kappa^n. (c) Coherence is used exactly where stated: W^c(q) lies in W^cs(p), and h^s maps center leaves to center leaves. NOT VERIFIED OR DEFECTIVE. (i) Uniformity in p, q, x is not in the statement of Theorem 0.2 and appears only inside the proof; the journal consumer needs it. (ii) Existence of pi^s with properties 1-3 is asserted by a one-paragraph sketch. Property 2 is stated only projectively and only at the base point p; Step 3 silently needs a linear or norm version. (iii) Step 1 prints gamma{-n}(x) and nu{-n}(x), which are products along the backward orbit of x. The correct quantities are gamma_n(x)^{-1} and nu_n(x). With non-constant rate functions this matters, since bunching is pointwise along the same orbit segment. (iv) Minor: (1) on p.1 omits gamma from the list of functions; mu and muhat in the note's (6) are undefined in the note; the letter delta is reused for the Lemma 1.1 threshold. (v) The C^{1+beta} part rests on Step 4, which fails as written (see H3). Limitation: I cannot remove later knowledge from model weights; no later result was used, and the repairs are derived directly from the supplied hypotheses.
Premises used
- Note definitions (1)-(4), pointwise non-strict rates, compact M
- Dynamical coherence as defined on note p.1
- theta-Hoelder continuity of E^c (hypothesis of Theorem 0.2)
- Existence and uniform local contraction of W^s (imported: [BP] = Brin-Pesin 1974, cited on note p.11; available_from 1974; date_basis: bibliography entry; version: Math. USSR Izv. 8; exact statement not supplied, so conditional)
- Note Theorem 0.1 = [PSW] Theorem B is quoted for context only and is not used in the proof of Theorem 0.2 (available_from 1997 per bibliography, Duke Math. J. 86; correction 2000 per journal p.489; exact hypotheses not supplied, so conditional)
Missing inputs
- A proof that pi^s_{p,q} exists with properties 1-3, including a linear version of property 2 (norm and sign control), with constants uniform in p and q
- An explicit statement of uniformity in Theorem 0.2
- Dated record identifying which version of the note is the journal's [BW06b] (see H3)
Scope changes
- The 'consequently' clause of note Theorem 0.1 and Corollaries 0.4 and 0.5 of the note were not audited.
Located concern
NOTE-2005 printed p.5, Step 1 display and following sentence
Cocycle base points and signs are misprinted: gamma_{-n}(x), nu_{-n}(x), then gamma_{-n} nu_n.
With the note's convention, alpha_{-n}(x) is a product over x_{-n},...,x_{-1}. The estimate actually involves gamma and nu along x_0,...,x_{n-1}. For non-constant functions these differ, and pointwise domination nu<gamma only controls (nu/gamma)_n(x).
Replace the bound by C*(nu/gamma)_n(x) <= C*kappa^n, using slack in the strict inequalities to evaluate gamma at nearby points.
None after correction; C^0 convergence of h_n to h holds uniformly.
Located concern
NOTE-2005 printed p.2 (Theorem 0.2) and p.4 (properties 1-3)
Uniformity is not stated in the theorem, and the pi^s properties are asserted rather than proved, with property 2 only projective.
Step 3 on p.9 estimates the norms of T pi v_n - v_n and of T(f pi f^{-1}) v_n as vectors, which property 2 as stated does not give.
State the theorem with constants uniform in p, q in W^s(p,1) and x in W^c_loc(p). Add to property 2 the linear bound |T pi_{p,q} u - P u| <= C d(p,q)^theta |u| for a fixed local identification P.
Repairable; needed for the interface with journal Proposition 3.1(vii).
H2
Checked statement: Note Lemma 1.1 (forward and backward iterated estimates for f_* on P(E^c), with the 'in particular' clause), note Lemma 1.2 (d(v_k, v'k) <= C nu_k(x)^theta for k <= n), and the Step 2 Cauchy estimate for h^s{n*}. This is the note's own Lemma 1.1, not the journal's Lemma 1.1.
Returned assessment: claim_false
FALSE AUXILIARY CLAIM (not a false principal theorem). Lemma 1.1 is false as stated and its induction is invalid. (1) The case n=1 as proved gives only d(v_1,w_1) <= (gamma gammahat)^{-1} d(v,w)^theta. Because theta<1 this is trivially true for small d(v,w) and uses no fiber-Lipschitz information. The last printed line of the n=1 proof drops the exponent theta and does not follow from the preceding line. (2) 'The other cases are proved inductively' fails: iterating a theta-power bound gives exponent theta^n. (3) Counterexample to part 1 (part 2 is symmetric). Take f = A x id on T^2 x S^1, A = [[2,1],[1,1]], lambda its expanding eigenvalue, nu = nuhat = 1/lambda, gamma = gammahat = 1 (allowed by the note's non-strict (3)). All note hypotheses hold: smooth, dynamically coherent, E^c smooth, nu^theta < 1. P(E^c(x)) is a single point v_x. Take y on the unstable leaf of x at distance eps. Then d((v_x)k, (v_y)k) >= d(x_k, y_k) = lambda^k eps. Given any threshold delta0, set d0 = min(delta0, 1/8) and eps = d0lambda^{-(n-1)}. The hypothesis holds for k <= n-1, but the conclusion would require lambdad0 <= d0^theta * lambda^{-(n-1)theta}, false for large n. With strict rates gamma = gammahat = 1-eta the same example works for small eta. (4) The 'in particular' clause does not follow from the stated theta-power conclusion either. (5) Lemma 1.2 and Step 2 actually use a different, unstated version: exponent 1 and no additive base-point term. That version is also false: in the same example with y on the stable leaf, one backward step multiplies the distance by lambda > (gamma gammahat)^{-1} = 1. (6) Base-point and time-direction errors: Lemma 1.2 prints (gamma gammahat){k-n}(x) where the product must be based at x_n; Step 2 prints (gamma gammahat){-n}(x) where ((gamma gammahat)n(x))^{-1} is needed. With non-constant functions the printed products run over the wrong orbit segment. NEW SELF-CONTAINED REPAIR (mine, not in the source). Separate base and fiber distances. One-step estimate: for a, b close, zeta in P(E^c(a)), xi in P(E^c(b)), the fiber distance satisfies D(f^{-1} zeta, f_^{-1} xi) <= (gamma gammahat)(f^{-1}a)^{-1} D(zeta, xi) + K d(a,b)^theta. Proof: project xi into E^c(a) at cost O(d(a,b)^theta), using theta-Hoelder E^c. Apply the fiber Lipschitz constant norm/conorm <= (gamma gammahat)^{-1} of the projectivised T f^{-1} on E^c. Then use that f_^{-1} is Lipschitz in fibers of P(TM) and delta-Hoelder in the base, with delta > theta. For Lemma 1.2: the base points x_k and z_k = f^{-(n-k)} pi(x_n) satisfy d(x_k, z_k) <= C' nu_k(x), by stable contraction plus center pull-back with (nu/gamma){n-k} <= 1. Let rho = sup nu^theta/(gamma gammahat) < 1 and A >= max(C, K'/(1-rho)). Backward induction gives D{k-1} <= nu_{k-1}(x)^theta * (A rho + K') <= A nu_{k-1}(x)^theta. So Lemma 1.2 holds for all n, with a larger uniform constant. For Step 2: apply the same recursion to pi_ v_n and f_^{-1} pi_ f_* v_n, whose base points lie on W^c(q_n) at distance O(nu_n). At level k the base distance is <= C (nu/gamma){n-k}(x_k) nu_k(x). This gives D(h{n*} v, h_{n+1*} v) <= C n sigma^{n-1} with sigma = max(rho, kappa^theta) < 1. That is summable, so h_{n*} is uniformly Cauchy. CHECKED AND FINE: the fiber Lipschitz constant of projectivised Tf and Tf^{-1} on E^c is at most (gamma gammahat)^{-1} at the appropriate point; the final inductive line of Lemma 1.2 is pointwise consistent at x_{k-1} once a correct one-step bound is available.
Premises used
- nu^theta < gamma gammahat pointwise, hence rho < 1 by compactness
- nu < gamma (partial hyperbolicity), hence kappa < 1
- E^c theta-Hoelder and T W^c = E^c
- Tf delta-Hoelder with delta > theta
- Properties 1-3 of pi^s (asserted in the note, not proved)
Missing inputs
- A correct statement and proof of the one-step fiber estimate with additive base term (supplied above as a new repair)
- A metric convention on P(TM) and local identification of fibers making '(x,xi)' with xi in P(E^c(y)) meaningful
Scope changes
- Verdict claim_false refers to the auxiliary note Lemma 1.1 only. Lemma 1.2 and the Step 2 Cauchy conclusion are assessed as repairable, not false.
Located concern
NOTE-2005 printed p.6, Lemma 1.1 statement and proof (final display and 'proved inductively')
Lemma 1.1 is false as stated; the n=1 bound cannot be iterated, the final n=1 line drops the exponent theta, and the 'in particular' clause is unsupported.
f = A x id on T^2 x S^1 with y on the unstable leaf of x at distance eps = d0lambda^{-(n-1)}: the hypothesis holds for k<n, but d(v_n,w_n) >= lambdad0 exceeds the claimed bound (gamma gammahat)_n^{-1} eps^theta for large n. Induction from the n=1 bound yields exponent theta^n.
Replace by the affine one-step estimate D' <= (gamma gammahat)^{-1} D + K d(base)^theta for directions in E^c over nearby base points, and track base distances separately (new repair).
Lemma 1.1 cannot be cited as stated. Lemma 1.2 and the Step 2 Cauchy property survive with larger uniform constants via the repair.
Located concern
NOTE-2005 printed p.7 (Lemma 1.2 proof, display using Lemma 1.1) and p.8 (Step 2 display)
These apply an exponent-1, no-additive-error version of Lemma 1.1 that was neither stated nor proved, and with wrong cocycle base points.
The exponent-1 version fails in A x id for x, y on a common stable leaf: one backward step expands d by lambda > (gamma gammahat)^{-1}. The v_k and v'_k of Lemma 1.2 always have distinct base points, at stable distance about nu_k.
Use the affine recursion. The base distance is O(nu_k) in Lemma 1.2 and O((nu/gamma)_{n-k} nu_k) in Step 2. Write the cocycles as ((gamma gammahat)n(x))^{-1} and (gamma gammahat){-(n-k)}(x_n).
Conclusions of Lemma 1.2 and Step 2 hold after repair; the Step 2 rate becomes n*sigma^n instead of the printed (nu^theta/(gamma gammahat))_n.
H3
Checked statement: Completion of the note proof (Lemma 1.3, Step 3 derivative convergence, Step 4 Hoelder continuity of Th^s) and its interfaces with journal Theorem 0.3 (printed pp.455-456) and the C^{1+delta} clause of Proposition 3.1(vii) as justified on printed p.467.
Returned assessment: repairable_gap
LEMMA 1.3 (p.8): correct once Lemma 1.2 holds in the repaired form: phi = log|Tf w| is delta-Hoelder on P(TM), and sum_k nu_k^{theta delta} converges. STEP 3 (p.9): the displayed identity is wrong. |Th_n v| = R_n |T pi v_n| with R_n = prod|Tf v_k| / prod|Tf v'k|. But |Th{n+1} v| = R''_n |T(f^{-1} pi f) v_n|, where R''n uses v''k = f*^{-(n-k)}(f^{-1} pi f)* v_n, not v'_k. The note prints 'f pi f^{-1}', an inverse typo, and uses the same prefactor R_n for both. New repair: write the difference as R_n(|T pi v_n| - |T(f^{-1} pi f) v_n|) + (R_n - R''_n)|T(f^{-1} pi f) v_n|. The first term is O(nu_n^theta) given a norm version of property 2. For the second, the H2 recursion gives D(v'_k, v''_k) <= C nu_k^theta (n-k) sigma^{n-k-1}, so |log(R_n/R''_n)| <= C n^{1+delta} max(nubar^theta, sigma)^{delta(n-1)}, which is summable. So |Th_n v| is uniformly Cauchy. Together with Step 2 and a sign check, Th_n converges uniformly to a bundle isomorphism (bounds K^{-1}, K), and h^s is C^1 with uniform constants and a uniform modulus. STEP 4 (pp.9-11): not established as written. (i) It invokes the false Lemma 1.1 forward over n steps without checking the closeness hypothesis. (ii) rho = (gamma gammahat)^{(2+eps)/theta} ignores expansion of base distances inside the center leaf. Example with the note's hypotheses satisfied: f = A^3 x A on T^2 x T^2, E^s = both stable lines, E^c = unstable line of A, E^u = unstable line of A^3, gamma and 1/gammahat near lambda. Then gamma gammahat is near 1, rho is near 1, and for p' in W^c(p, rho_n) chosen with lambda^n d(p,p') = 1/4 the base distance of the images is 1/4, while the claimed bound (gamma gammahat)n^{1+eps} tends to 0. (iii) Further slips: 'q in W^c(p,1)' should be W^s; the choice of n is inconsistent ([rho_n, rho{n-1}) versus W^c(p, rho_n)); the series formula for log|Th^s| is said to be shown in Step 3 but is not written there, though it is derivable by dominated convergence. New repair for existence of some beta>0: |Th - Th_n| <= C sigma1^n uniformly, and Th_n is theta-Hoelder along the leaf with constant <= C Lambda^n. Choosing n proportional to log(1/d(x,x')) gives a Hoelder exponent beta > 0. The explicit beta' on p.11 is not verified. Step 4 is not needed by the journal consumers, which use only uniform C^1 or Lipschitz. INTERFACES. (a) Hypotheses match: journal (7) gives nu^theta < gamma gammahat and nuhat^theta < gamma gammahat (the latter for f^{-1}); journal (8) gives theta-Hoelder E^c via the criterion of [PSW97, Th. A]. (b) Genuine versus fake: note Theorem 0.2 is stated for a dynamically coherent diffeomorphism of a compact manifold with genuine W^c and W^cs. Journal p.467 applies 'the main result there' to the fiberwise extension F_r on TM (M discrete, non-compact fibers) and its fake foliations. This needs a uniform-estimates version of the note for bundle maps, plus theta-Hoelder continuity of the fake center distribution with the same theta. Proposition 3.1(vi) asserts only unspecified Hoelder continuity. Both look obtainable because (7),(8) are strict on compact M and F_r is C^1-close to Tf, but neither is written. (c) Holonomy versus subfoliation: the note concludes C^1 holonomy between center leaves; p.467 asserts that W^u C^1-subfoliates W^cu. New transfer argument: for any C^1 transversal tau in a cs (or cu) leaf through a point y' of a center leaf, projection along strong leaves from that center leaf to tau is differentiable at y', with derivative the linear projection along E^s. Composing with the uniformly C^1 center-to-center holonomy shows holonomy to arbitrary C^1 transversals is C^1 with continuous derivative. With C^1 leaves this yields C^1 foliation charts. This needs joint continuity of Th in (p,q,x), which follows from uniform convergence. (d) Consumers in the journal (author-text sections 6-7; printed pages not supplied): Lemma 6.1 needs a uniform Lipschitz stable holonomy between fake center leaves; section 7 needs absolute continuity with bounded Jacobians of the fake unstable foliation in fake cu leaves. Both follow from uniform C^1 subfoliation. CHRONOLOGY. Journal p.488 cites [BW06b] as 'A note on unstable holonomy between centers, Preprint, 2006', which is after the recorded cutoff 2005-10-11. The supplied note is titled 'stable holonomy', dated June 11 2005 and marked 'Preliminary version'. Identity of versions is unresolved. The 2005 note is admissible by date but has the defects above. The C^2 branch of Proposition 3.1(vii) does not use the note. Limitation: later knowledge in model weights cannot be removed; none was used to close any step.
Premises used
- Repaired Lemma 1.2 and Step 2 from H2
- delta-Hoelder continuity of Tf and of log|Tf w| on P(TM)
- Journal (6),(7),(8) on printed pp.455-456
- Bundle regularity criterion sup beta1/(alpha1^a alpha2) < 1 as stated in the author text (journal printed p.466 not supplied as image)
- Imported, conditional: [PSW97] Theorem A (Hoelder exponent of E^c); available_from 1997; date_evidence: bibliography on journal printed p.489 and note p.11; date_basis: publication year; version: Duke Math. J. 86 (1997) with correction Duke Math. J. 105 (2000); exact hypotheses not supplied
- Imported, conditional: [PSW97],[PSW00] for the C^2 branch of Proposition 3.1(vii); same dating; not audited
Missing inputs
- Dated copy of the 2006 [BW06b] version actually cited by the journal, to see whether Lemma 1.1, Step 3 and Step 4 were corrected and whether it treats the non-coherent, fiberwise setting
- A written version of the note's theorem for F_r with uniform estimates and fake foliations
- A statement that the fake center distribution is theta-Hoelder with the theta of (7),(8)
- Linear or norm version of property 2 of pi^s
- Verification of the explicit exponent beta'
- Printed journal pages for sections 5-7 to confirm consumer wording
Scope changes
- Consumers Lemma 6.1 and section 7 were checked only against the author-version text; journal printed pages for them were not supplied.
- Journal printed pp.487-489 were inspected only for the reference list and dates.
Located concern
NOTE-2005 printed p.9, first display of Step 3
The identity for | |Th_n v| - |Th_{n+1} v| | uses the same distortion prefactor for both terms, and prints f pi f^{-1} for f^{-1} pi f.
Th_{n+1} = Tf^{-n} T(f^{-1} pi f) Tf^n, so its backward norm product runs along f_^{-(n-k)}(f^{-1} pi f)_ v_n. These directions differ from v'k = f^{-(n-k)} pi_ v_n and even have different base points.
Add the term (R_n - R''_n)|T(f^{-1} pi f) v_n| and bound it by the affine recursion plus Hoelder distortion; this decays geometrically in n (new repair). Add a norm version of property 2.
Uniform C^1 convergence of h_n, and hence uniformly C^1 stable holonomy between center leaves, holds after repair.
Located concern
NOTE-2005 printed pp.9-11, Step 4
The Hoelder estimate relies on the false Lemma 1.1 and on a scale rho_n that does not control expansion of base distances along the center; the explicit beta' is unsupported.
f = A^3 x A with E^c the unstable line of A and gamma gammahat near 1: for p' in W^c(p, rho_n) with lambda^n d(p,p') = 1/4, d(f_^n(p,xi), f_^n(p',xi)) >= 1/4, contradicting the claimed bound (gamma gammahat)_n^{1+eps} for large n.
Standard interpolation between the uniform C^0 rate of Th_n to Th and the exponentially growing Hoelder constant of Th_n (new repair); this gives some unspecified beta > 0.
The C^{1+beta} claim holds only with an unspecified beta after repair. There is no effect on journal Theorem 0.3 or Proposition 3.1(vii), which need only uniform C^1.
Located concern
BW journal printed p.467, paragraph beginning 'When 1 < k < 2'; printed p.455 last paragraph before Theorem 0.3; printed p.488 reference [BW06b]
Transfer steps are asserted without proof: (1) from a theorem about genuine coherent foliations on compact M to fake foliations of F_r; (2) from C^1 holonomy between center leaves to C^1 subfoliation; (3) theta-Hoelder fake center bundle with the theta of (7),(8); (4) the cited 2006 preprint postdates the 2005-10-11 cutoff and differs in title from the supplied 2005 preliminary note.
Note Theorem 0.2 (printed p.2) assumes 'dynamically coherent diffeomorphism of a compact manifold', and its conclusion concerns only holonomy between center transversals. Proposition 3.1(vi) gives unspecified Hoelder regularity. Reference list: '[BW06b] A note on unstable holonomy between centers, Preprint, 2006'.
State and prove the note's theorem for fiberwise maps with uniform estimates and an invariant center foliation subfoliating cs leaves; record the theta-Hoelder exponent of the fake center bundle from the p.466 criterion with a = theta; add the short transversal-projection argument for C^1 charts (new repair). Supply a dated version record for [BW06b].
The C^{1+delta} branch (Theorem 0.3, Corollary 0.4, C^{1+delta} clause of Theorem 5.1) is conditional on these transfers. I found no obstruction and each appears repairable. The C^2 branch (Theorems 0.1, 5.1 for C^2) is unaffected.
Outside the audit
- Exact statements and hypotheses of [PSW97] Theorems A and B and of [PSW00] (not supplied)
- C^2 branch of Proposition 3.1(vii) and the remaining parts of Proposition 3.1
- Journal sections 2 and 4-8 beyond identifying consumers of Proposition 3.1(vii) from the author-version text
- Journal printed p.466 (regularity criterion), not supplied as an image
- Note Corollaries 0.4 and 0.5 and the introductory conjugacy example
- The explicit exponent beta' and the intermediate sums on note pp.10-11
- Full construction of pi^s and its extension to f^n(W^c_loc(p))
- The 2006 version of [BW06b]
Additional supporting archive items are identified in the original report. View the manuscript hashes and repository snapshot.