claude-fable-5-2 high: bw-geometry
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 false principal theorem was found, and no error was found in the statements of Theorems 0.1 and 0.3 or Corollaries 0.2 and 0.4; the corollaries follow from the theorems as claimed.
One auxiliary statement is false as literally printed. The cs/cu growth clauses of Proposition 3.1(iii) (journal pp. 464-465, 468) can fail for a metric that is adapted in the sense of p. 457, because (1)-(4) do not bound Tf on E^cs or E^cu when the bundles are far from orthogonal. An explicit linear example on T^4 is given under G2. The defect is repairable in two independent ways: normalising the metric, or restating the clause for center leaves, which is all the checked consumers (Lemmas 4.2 and 6.2) use. Both repairs are new and are not part of the original proof.
The remaining findings are missing justifications, not errors.
- The transfer of the Hoelder and C1-subfoliation theorems to F_r on the disjoint union of tangent spaces, with uniformity in p, is asserted rather than proved. The supplied PSW 2000 correction supports the C2 branch; its Frobenius step was verified, and it contains two harmless wording slips.
- Lemmas 4.1 and 4.4 are stated for functions on M but are applied to a Jacobian defined only along the fake foliations. A restatement fixes this.
- Item (vii) in the C^(1+delta) branch rests entirely on the unsupplied note [BW06b], whose version at the 2005-10-11 cutoff is unknown. Theorem 0.3 and Corollary 0.4 are therefore conditional.
- The paper's cited direct proof of absolute continuity in the pointwise setting, [AV08], is dated 2008 on journal p. 488 and cannot be used under the cutoff. The alternative 'Pesin theory' route is cited without an exact statement.
- The K property relies on [BP74, Prop. 5.1], which by the paper's own footnote 1 (p. 452) was proved for constant rates.
Section 4 was verified line by line, including uniformity of constants in n and p, and found correct apart from citation and interface defects.
Limits. No external tools, web search or other sources were used. Later knowledge cannot be removed from model weights, but no later theorem or later proof was used to close any obligation. Imported results are recorded with dates read from the supplied bibliography pages only. This is a scoped audit, not a complete line-by-line verification of the paper.
G1
Checked statement: Journal statements of Theorem 0.1 (p. 454), Corollary 0.2 (p. 455), Theorem 0.3 with (6)-(8) (pp. 455-456) and Corollary 0.4 (p. 456), together with the definitions of partial hyperbolicity (1)-(4) (pp. 452-453), volume-preserving, accessibility and essential accessibility (p. 453), center bunching (5) (p. 454), and the adapted-metric conventions (p. 457). Audited at the level of statements, internal consistency, deduction of the corollaries, and identification of the proof obligations for ergodicity and the Kolmogorov property. This is not a verification that Theorems 0.1 and 0.3 are true.
Returned assessment: no_issue_found
No defect was found in the four statements or in the deduction of the two corollaries; the theorems themselves remain conditional on the obligations listed below.
CHECKED. (a) Journal wording agrees with the author text for Theorem 0.1, Corollary 0.2, Theorem 0.3 and Corollary 0.4. Displays (1)-(8) were read from the images: (5) is nu < gammagammahat and nuhat < gammagammahat; (6) is mu(p) < |Tf v| on E^s and |Tf v| < muhat(p)^(-1) on E^u; (7) is nu^theta < gammagammahat and nuhat^theta < gammagammahat; (8) is nugamma^(-1) < mu^theta and nuhatgammahat^(-1) < muhat^theta, with theta in (0,delta). (b) Conformal case (p. 455). If Tf|E^c is conformal, set gamma = (1-eta)|Tf|E^c| and gammahat^(-1) = (1+eta)|Tf|E^c|. These are continuous, (1)-(4) persist for small eta by compactness and strictness, and gammagammahat = (1-eta)/(1+eta) exceeds sup nu and sup nuhat (both < 1 by compactness). So (5) holds; in particular it holds when dim E^c = 1. Corollary 0.2 follows from Theorem 0.1. (c) Corollary 0.4 from Theorem 0.3. The quantifier order is consistent. First fix any admissible gamma_0, gammahat_0 and choose theta in (0,delta) small enough for (8); this is possible since nu/gamma_0 and nuhat/gammahat_0 are uniformly < 1 while mu^theta, muhat^theta tend to 1 uniformly. Then tighten gamma >= gamma_0 and gammahat >= gammahat_0 towards |Tf|E^c|. This only decreases the left sides of (8), so (8) persists, and it makes gammagammahat exceed sup nu^theta and sup nuhat^theta (both < 1). So (7) holds for the same theta. (d) Theorem 0.3 is non-vacuous and consistent. (8) forces theta small and (7) forces gamma*gammahat close to 1 relative to nu^theta; these are compatible, for instance for near-isometric center. A fiber-contraction heuristic confirms the remark on p. 456 that (8) is the Hoelder condition for E^cu and E^cs: fiber contraction nu/gamma against inverse base Lipschitz constant mu^(-1) to the power theta. The cited [PSW97, Th. A] itself is not supplied. (e) 'Volume-preserving' means an invariant probability mu equivalent to a Riemannian volume m (p. 453). Essential accessibility is formulated for m. The two notions of null set coincide, so the Hopf step and the measure hypotheses are compatible. (f) The letter mu denotes the invariant measure on p. 453 and a rate function in (6) and (8) on p. 455-456. This is a cosmetic clash only.
PROOF OBLIGATIONS IDENTIFIED (from author text PDF pp. 27-28). O1. Theorem 5.1: density points of a bi-essentially saturated set are bisaturated. No invariant measure is needed. C2 with (5), or C^(1+delta) with (7)-(8). O2. Corollary 5.2. Needs Borel measurability of the density-point set and the Lebesgue density theorem. Both are elementary; checked. O3. Hopf step. Limsup Birkhoff averages of a continuous function are constant on W^s leaves (forward) and W^u leaves (backward); the Birkhoff theorem for mu gives equality mu-a.e., hence m-a.e.; a.e.-constancy for continuous functions implies ergodicity. Elementary steps checked. O4. K property. Triviality of the Pinsker algebra, using [BP74, Prop. 5.1] (Pinsker sets are bi-essentially saturated). That result must hold for pointwise rates (footnote 1, p. 452, records that BP74 assumes constant rates), for a measure merely equivalent to volume, and in the C^(1+delta) branch. It is not supplied and the extension is asserted, not proved. O5. Transverse absolute continuity of the genuine W^s and W^u with bounded Jacobians in the pointwise setting. The author text attributes it to Pesin theory and to Abdenur-Viana. O6. C^(1+delta) branch only: [BW06b], feeding Proposition 3.1(vii).
UNCHECKED. Sections 2 and 5-8 line by line; all imported theorems.
Premises used
- Compactness of M and continuity/strictness of (1)-(4), (5), (7), (8) (journal pp. 452-457)
- Lebesgue density theorem and Birkhoff ergodic theorem (classical, available long before 2005-10-11)
- Rokhlin-Sinai characterisation of K-automorphisms by triviality of the Pinsker algebra (classical, 1960s; exact statement not supplied)
- IMPORT [BP74] Brin-Pesin, Izv. 1974, Prop. 5.1: available_from 1974; date_evidence journal p. 488 bibliography; date_basis publication year; version: published article, constant-rate definition; exact hypotheses not supplied, so conditional
- IMPORT absolute continuity of W^s, W^u: [BP74], [PS72] (1972-1974, constant rates) are admissible; [AV08] is listed on journal p. 488 as 'Preprint, 2008', i.e. after the 2005-10-11 cutoff, version at cutoff unknown, so chronology unresolved and it cannot close the obligation; the 'Pesin theory' route (pre-cutoff literature) is stated without an exact theorem
- IMPORT [BW06b] Burns-Wilkinson note on unstable holonomy between centers: listed 'Preprint, 2006' on journal p. 488; not supplied; date and version relative to the cutoff unresolved; authors' own companion note
Missing inputs
- Journal pages 458-463 and 471-486 (Sections 2 and 5-8) as images
- [BP74] Prop. 5.1 with exact hypotheses, and a justification that it extends to pointwise rates, C^(1+delta), and a measure equivalent to volume
- A cutoff-admissible source with exact statement for transverse absolute continuity of W^s, W^u with bounded Jacobians in the pointwise partially hyperbolic C^(1+delta) setting
- [BW06b] (text and dated version) for Theorem 0.3 and Corollary 0.4
- [PSW97] Theorem A (Hoelder exponent of E^c under (8))
Scope changes
- Verdict is statement-level only; truth of Theorems 0.1 and 0.3 is conditional on O1, O4, O5, O6
- Theorem 0.3 and Corollary 0.4 additionally depend on an unsupplied note whose availability at the cutoff is unresolved
- Journal p. 487 mentions a recent Hammerlindl example and p. 488-489 list 2008 items; these post-cutoff items are not used in the audited statements and are disregarded
G2
Checked statement: Proposition 3.1 (journal pp. 464-465), items (i)-(vii) with the uniformity clause, and its proof (pp. 465-468): the bundle map F_r on TM with M discrete, graph-transform foliations of each T_pM, the regularity discussion for the C2 and C^(1+delta) branches, projection by exp_p, and the choice of r_1. The supplied PSW 2000 correction (pp. 105-106) was read as the cited support for C1 subfoliation in the C2 branch.
Returned assessment: repairable_gap
Items (i), (ii), (iv), (v) and the s/u clauses of (iii) check out at sketch level. The cs/cu clauses of (iii) are false as literally stated for the paper's notion of adapted metric, but repairable and not needed downstream in that form. Item (vi) and the C2 branch of (vii) rest on imports that are plausible but partly unsupplied. The C^(1+delta) branch of (vii) could not be verified.
CHECKED. (1) F_r. With F_r(p,v) = T_pf v + rho(|v|/r) g_p(v) and g_p = exp^(-1) f exp - T_pf, one has |Dg_p(v)| <= omega(|v|) and |g_p(v)| <= |v| omega(|v|), so property (c) holds uniformly in p. Uniform C^(1+delta) and C2 bounds on fibers follow the same way. (2) Existence and uniqueness. Each F_r(p,.) is a global C1-small perturbation of T_pf, with rates constant on the fiber T_pM. The pointwise domination therefore becomes sequence domination along the base orbit, and the global Hadamard-Perron scheme gives s, u, cs, cu foliations. Uniqueness among invariant families with bounded-slope graph leaves was verified directly: if v, w lie on a leaf of a competing cs family and z = W^u(v) meet W^cs(w), then F^n v - F^n z dominates F^n z - F^n w and lies in the u-cone, contradicting bounded slope unless v = z. The s case is symmetric under backward iteration. The journal wording 'functions with bounded derivatives' (p. 466) is the correct requirement; stable leaves away from the zero orbit need not be graphs of bounded functions. (3) (iv) Coherence. Rate characterisation gives W^s inside W^cs and W^u inside W^cu. Center leaves are transverse intersections of C1 graphs. (4) (v) Uniqueness. Verified independently. The fake stable leaf through 0 is a graph g over E^s(p) with g(0) = 0 and small slope, so its trace in the r-ball is star-shaped and connected. Points on it contract monotonically, never leave the r-ball, so F_r agrees with f there and the points lie in the genuine W^s(p). The trace is open and relatively closed in W^s(p) meet B(p,r), hence equals W^s(p,r). This does not depend on the extension F_r. (5) (i) and (ii). Follow from the epsilon/2 cones for F_r, D exp_p close to parallel transport, uniform continuity of the bundles, and f(B(p,2r_1)) inside B(f(p),r). (6) (iii), s and u clauses. Leaves are small-slope graphs over a fixed subspace in exponential coordinates, so leaf paths have length at most (1+O(eps)+O(r)) times ambient distance. Strictness of (2) and (4) and continuity of nu, nuhat absorb this and the change of base point from q to p. The induction is as in Lemma 1.1. Cocycle bookkeeping checked: nuhat_{-n}(p)^(-1) = nuhat(p_{-n})...nuhat(p_{-1}) and gamma_{-n}(p) = gamma(p_{-n})^(-1)...gamma(p_{-1})^(-1). (7) (vii), C2 branch. For F_r^(-1): alpha_1 = gammahat, beta_1 = gamma^(-1), alpha_2 = nu^(-1), so the criterion is nu < gammagammahat. For F_r it is nuhat < gammagammahat. Both match (5). The PSW correction was checked step by step. Its displayed bunching inequality |T^c f| / (m(T^u f) m(T^c f)) < 1 is the same condition. Its Frobenius step is sound: pushing E^u forward by the linear projection of the graph chart gives a C1 plane field on the plane E^cu_p that admits integral manifolds, hence is involutive, hence has C1 foliation charts, and uniqueness of integral manifolds identifies them with the projected W^u. In the BW setting the cu leaves are global graphs in T_pM and the approximating smooth bundles can be taken constant, so the transfer is plausible provided the C1 section theorem is applied with uniform estimates on the non-compact disjoint union.
NOT CHECKED OR CONDITIONAL.
- Full graph-transform details ([HPS77], [KH95] not supplied).
- Hoelder regularity of foliations ([PSW97] Theorem A and the foliation statement, not supplied).
- [PSW97] Theorem B body (only the correction is supplied).
- The C^(1+delta) branch of (vii) ([BW06b] not supplied).
- The asserted uniformity in p of (vi) and (vii). It is plausible from uniform estimates but not written out.
BRANCHES AND FAKE VERSUS GENUINE.
- (i)-(v): need only C1, no bunching. They concern fake foliations, except that (v) identifies the fake s and u leaves through p with genuine ones.
- (vi): needs C^(1+delta).
- (vii): needs either C2 with (5), or C^(1+delta) with (7)-(8). It concerns only fake foliations; they are genuine invariant foliations of F_r.
Premises used
- Journal Section 1.3 (p. 457): adapted metric defined only by (1)-(4); rescaling conventions
- IMPORT [HPS77] Hirsch-Pugh-Shub, LNM 583 (1977): plaque families, Theorem 6.1(e), C^r section theorem; available_from 1977; date_evidence journal p. 488; date_basis publication year; version: published monograph; exact statements not supplied, so conditional
- IMPORT [KH95] Katok-Hasselblatt (1995), Hadamard-Perron theorem; available_from 1995; date_evidence journal p. 488; date_basis publication year; version: first edition; not supplied, so conditional
- IMPORT [PSW97] Pugh-Shub-Wilkinson, Duke 86 (1997), Theorems A and B; available_from 1997; date_evidence journal p. 489; date_basis publication year; version: published article; not supplied, so conditional
- IMPORT [PSW00] correction, Duke 105 (2000) pp. 105-106: supplied; 'Received 16 August 1999' and copyright 2000 visible on p. 105; admissible; checked as described
- IMPORT [SSS92], [Has94], [Wil98] for pointwise section-theorem variants: pre-cutoff by bibliography dates (journal pp. 488-489); not supplied, so conditional
- IMPORT [BW06b]: 'Preprint, 2006' per journal p. 488; not supplied; version at the 2005-10-11 cutoff unknown, so unresolved
Missing inputs
- [BW06b] full text and dated version: exact theorem (C1 or only Lipschitz; between which transversals), its hypotheses (compactness, dynamical coherence, global versus pointwise bunching), and whether it applies to F_r on the disjoint union of tangent spaces
- [PSW97] Theorems A and B with exact hypotheses, to confirm the bunching condition is pointwise and that compactness enters only through uniform estimates
- [HPS77] Theorem 3.5 and Theorem 6.1(e) statements, to confirm the non-compact uniform version of the C1 section theorem with base a non-separable C1 manifold
- A written proof that the regularity constants in (vi) and (vii) are uniform in p
Scope changes
- Prop. 3.1(iii) cs/cu clauses hold only after the metric normalisation (repair A) or in the center-leaf form (repair B)
- Prop. 3.1(vii) in the C^(1+delta) branch is conditional on the unsupplied note [BW06b]
- Prop. 3.1(vi) and the C2 branch of (vii) are conditional on [PSW97] and [HPS77] in a uniform non-compact form
Located concern
BW journal p. 464, Prop. 3.1(iii)(a) cs clause; p. 465, (iii)(b) cu clause; p. 468, third and fourth displayed implications; p. 465 first paragraph of proof; p. 457 Section 1.3
False auxiliary clause as literally stated (metric-dependent); not a false principal theorem. An adapted metric is defined on p. 457 only by (1)-(4), which bound Tf on E^s, E^c, E^u separately. The proof (p. 465) arranges (1)-(4) only on eps-cones about 'the spaces in the splitting'. The cs and cu clauses of (iii), and the one-step inequalities for the cs and cu fake leaves on p. 468, need |Tf v| < gammahat^(-1) on a cone about E^cs and |Tf v| > gamma on a cone about E^cu. This does not follow from (1)-(4) when E^s and E^c (or E^c and E^u) are not nearly orthogonal.
Let B = [[2,1],[1,1]] with eigenvalues lambda > 1 > lambda^(-1), and f = B^2 x B on T^4. Take E^s and E^u to be the lambda^(-2) and lambda^2 eigenlines and E^c the span of the lambda^(-1) and lambda eigenlines. Choose a flat metric in which all four eigenvectors are unit, e_u and the lambda-eigenvector are orthogonal to everything, and e_s makes a small angle phi with the lambda^(-1)-eigenvector e_c. Then (1)-(4) hold with nu slightly above lambda^(-2), gamma < lambda^(-1), lambda < gammahat^(-1) < nuhat^(-1) < lambda^2. So f is C1 partially hyperbolic and the metric is adapted in the sense of p. 457. Here F_r is the linear map, so the fake foliations are the affine ones. For v = e_c - e_s: |v| = 2 sin(phi/2) and |Tf v| >= lambda^(-1) - lambda^(-2). Hence |Tf v|/|v| exceeds lambda^2, and therefore every admissible gammahat^(-1), once phi is small. So d(f q, f q') <= gammahat(p)^(-1) d(q,q') fails for q' in the fake cs leaf through q, at every scale r_1 and already for n = 1.
NEW REPAIR (not in the original proof), either of two. (A) Strengthen the normalisation of the adapted metric. Replace g by g'(v,w) = g(v_s,w_s) + g(v_c,w_c) + g(v_u,w_u), which keeps (1)-(4) with the same functions and makes the bundles orthogonal, so |Tf|E^cs| < gammahat^(-1) and m(Tf|E^cu) > gamma. Then smooth g'; strictness and compactness preserve all inequalities, including (5) or (7)-(8). Ergodicity and the K property do not depend on the metric. (B) Restate the second clauses of (iii)(a),(b) for points on a common fake center leaf. There only the cone about E^c and inequality (3) are needed. As far as checked in the author text, this is all the consumers use: Lemma 4.2 (y in the fake center plaque of x) and Lemma 6.2 (q'_n and z'_n on a common fake center leaf).
No effect on Theorems 0.1 and 0.3 after either repair. As printed, the cs/cu clauses of Prop. 3.1(iii) are not established and can fail for an admissible metric.
Located concern
BW journal p. 467, paragraph beginning 'When 1 < k < 2'; p. 465 item (vii) last sentence; p. 488 reference [BW06b]
Unverifiable import in the C^(1+delta) branch; missing justification, not a demonstrated error. The whole of (vii) in this branch is delegated to the 'main result' of [BW06b], said to carry over to F_r. The note is not supplied, its version at the cutoff is unknown (bibliography says Preprint, 2006), and the paper does not state its exact theorem. The introduction (p. 455) describes the holonomy result as Lipschitz, while (vii) asserts a C1 subfoliation.
No counterexample. One point was checked: strong bunching (7)-(8) does carry over to F_r, because the rates of F_r(p,.) are uniformly close to those of T_pf and the inequalities are strict on compact M. Not checkable: applicability of the note to a bundle map over a discrete base with non-compact fibers, and whether it yields C1 foliation charts or only Lipschitz holonomy between center leaves.
State the exact theorem of the note with its hypotheses. Either verify that it gives C1 subfoliation in the uniform non-compact setting, or weaken (vii) in this branch to what the consumers need: uniformly Lipschitz fake-stable holonomy between fake center leaves inside a fake cs leaf (Lemma 6.1), and Fubini-type absolute continuity with bounded Jacobians of the fake unstable foliation inside fake cu leaves (end of Section 7). Bi-Lipschitz holonomies between center leaves would give both.
Theorem 0.3 and Corollary 0.4 remain conditional on the unsupplied note. Theorem 0.1 and Corollary 0.2 are unaffected.
Located concern
BW journal p. 466 last paragraph to p. 467 fourth paragraph; PSW correction p. 105
Missing justification at the interface in the C2 branch; plausible. The section theorem and [PSW97]/[PSW00] are applied to F_r on the disjoint union of cu leaves in all tangent spaces, justified only by the remark that compactness is used solely for uniform estimates. The uniformity in p claimed after (vii) is not derived. Minor slips: p. 467 asks for 'some a > 1' although the stated criterion requires a <= k-1 = 1 (a = 1 is what is used, and it holds by compactness). The supplied correction says the graph transform 'covers the identity on W' and writes 'm(A) approximately m(T^c f)', whereas the bundle map covers f|W and the base contraction is m(Tf|TW) = m(T^c f); its final displayed inequality is the correct bunching condition.
Checked: bunching arithmetic for both F_r and F_r^(-1); the fiber contraction |T^cs f|/m(T^u f) equals |T^c f|/m(T^u f) in a box norm on T_pM adapted to the splitting at p (uniformly equivalent norms); the Frobenius step on the C1 leaf via the linear projection chart. Not checked: the exact hypotheses of the C1 section theorem over a non-separable, non-compact C1 base.
Record the uniform hypotheses actually used: F_r uniformly C2 on fibers, a leafwise C1 diffeomorphism, pointwise fiber contraction times inverse base contraction uniformly < 1, uniform moduli of continuity. Cite the section theorem in a form that allows a non-compact base with uniform estimates. Replace 'a > 1' by a = 1.
None expected; the C2 branch of (vii) is conditional on unsupplied but pre-cutoff sources.
Located concern
BW journal p. 467, second paragraph; p. 465 item (vi)
Sketch-level step. Hoelder continuity of the fake center foliation is inferred from the center being the intersection of cs and cu. That is immediate for the bundle; for the foliation (charts or holonomies) it needs a short argument. The Hoelder foliation statements for cs, cu, s, u rest on [PSW97], not supplied.
NEW REPAIR sketch: label a point by its cs-leaf and cu-leaf labels (Hoelder by the charts of the two foliations) together with the linear projection to E^c(p). The inverse solves a contraction fixed-point problem for two small-slope graphs depending Hoelder-continuously (in C0) on the labels, hence is Hoelder.
Add the argument above or a precise citation. Note that the checked consumers (Lemma 7.1 in the author text) use only Hoelder continuity of the tangent distributions and continuity of the foliations.
None expected.
Located concern
BW journal p. 464 (iii)(a); p. 465 (iii)(b) and Step 1; p. 467 last line; p. 468 Remark and following paragraph
Typographical and notational defects in the statement and proof. In (iii) the fake leaves through q_n, q_j, q_{-n}, q_{-j} carry subscript p but must be based at p_n, p_j, p_{-n}, p_{-j}. Page 467 writes f(exp_p(p,v)). Page 468 lists the fake s foliation twice and omits the fake u foliation, and the Remark writes W^cu_p and W^c_p for the genuine foliations. F_r is defined for all r in (0,r_0] although exp^(-1) f exp on the r-ball needs r Lip(f) <= 2 r_0. The trace of a fake leaf in B(p,r) need not be connected, so 'leaf' should mean a connected component (harmless for (ii), since images of connected plaques are connected).
Read directly from the images; the intended meaning is clear from Step 2 and from the use in Lemmas 4.2-4.3.
Correct the subscripts; restrict r to r <= 2 r_0 / Lip(f); define plaques as connected components.
None.
G3
Checked statement: Section 4 (journal pp. 468-470): Lemma 4.1 (Hoelder cocycle distortion), the sets S_{n,sigma}(p) and T_{n,sigma,tau}(p), Lemmas 4.2, 4.3, 4.4, the standing scale conventions on p. 468, and their interfaces with the choice of sigma, tau and the definition of juliennes in Section 5 (author text PDF pp. 29-31; journal pp. 471-476 not supplied as images).
Returned assessment: repairable_gap
Section 4 is correct in substance; the gaps are interface and citation defects with easy repairs.
CHECKED. (1) Lemma 4.1. Since alpha is positive and continuous on compact M, log alpha is theta-Hoelder with some constant H, and the telescoping estimate is correct. The index range i = 1..n in the backward statement matches alpha_{-n}(p) = prod_{i=1..n} alpha(p_{-i})^(-1). H depends only on alpha. (2) Lemma 4.2. Take x in W^s(p,1) and y in the fake center plaque of radius sigma_n(p) at x. Lemma 1.1 gives d(x_j,p_j) <= nu_j(p) d(x,p) < 1. By induction, using local invariance, coherence and the one-step center bound, d(x_j,y_j) <= gammahat_j(p)^(-1) sigma_n(p) = (sigma/gammahat)j(p) sigma{n-j}(p_j) < 1 for j <= n. So the hypotheses of Prop. 3.1(iii)(a) (x_i within r_1 of p_i, y_i in the r_1-plaque at x_i) hold at each step, the image of the connected plaque stays in the stated plaque, and f^j(S_n) lies in B(p_j,2). The second conclusion holds with kappa = max(sup nu, sup sigma/gammahat) < 1 and C = 2. (3) Lemma 4.3. For y in T_n with y_n in the fake or genuine unstable plaque of radius tau_n(p) at x_n, with x in S_n: Prop. 3.1(iii)(b) at base point p_n (fake) or Lemma 1.1 (genuine) gives d(y_j,x_j) <= nuhat_{n-j}(p_j) tau_n(p) <= tau_j(p). The hypotheses hold because x_{n-i} lies in B(p_{n-i},2), inside B(p_{n-i},r_1) and B(p_{n-i},R). Hence the inclusion in B(p_j,3) for tau <= 1, and in B(p_j, C kappa^j) for tau < 1, with kappa = max(sup nu, sup sigma/gammahat, sup tau) and C = 3. (4) Uniformity. C and kappa are independent of n and also of p (compactness). The lemmas quantify only over n, but the uniformity in p holds and is what Section 5 uses. (5) Lemma 4.4. For x, y in T_n, d(x_i,y_i) <= 2C kappa^i, so the sum of d^theta over i < n is bounded by B = (2C)^theta/(1-kappa^theta), independent of n and p. Lemma 4.1 gives the constant e^(HB). (6) Interface with Section 5 (author text PDF p. 30). Continuous sigma, tau with nu < tau < sigmagamma and sigma < min(gammahat,1) exist exactly because nu/gamma < min(gammahat,1), i.e. nu < gammagammahat and nu < gamma. Then tau < sigmagamma < gammagammahat < 1, so the strict hypothesis tau < 1 of Lemmas 4.3-4.4 holds. Also taunuhat < sigmagammanuhat < sigmagammagammahat <= sigma is correct, using nuhat < gammahat and gammagammahat < 1. With S_n = S_{n,sigma}(p) and T_n = T_{n,sigma,tau}(p), both the fake and genuine unstable juliennes f^(-n) of the tau_n-plaques at y_n lie in T_n by construction. The fake center-unstable julienne lies in the fake cu leaf of x by backward local invariance (valid since iterates stay in B(p_j,3)) and coherence. (7) Scale conventions. Each property used is scale-invariant and survives shrinking the unit, so the successive rescalings (unit ball small; R large; r_1 much bigger than 1) are consistent and not circular. All rates are evaluated at the base orbit p_j for every point of B(p_j,r_1); this is legitimate because r_1 is chosen from the modulus of continuity of the rate functions before rescaling.
NEEDS AN ADDITIONAL SOURCE OR ARGUMENT.
- Lemma 4.2 inherits Prop. 3.1 (see G2).
- Lemma 4.4 is applied later to a function not covered by its statement (first issue below).
- The single-point intersection properties assumed on p. 468 (fake cs with fake u, fake cs with genuine local u, and so on) are routine local product structure, asserted without proof. 'Much bigger than 1' must absorb a constant depending on the angles between the bundles.
UNCHECKED. Figure 1; the use of these lemmas in Sections 6-8 beyond the interfaces named above.
Premises used
- Lemma 1.1 (author text PDF pp. 9-10; journal p. 458 not supplied as image), one-step contraction (12) on genuine local stable/unstable leaves
- Proposition 3.1 (ii), (iii), (iv) (journal pp. 464-465), subject to G2
- Compactness of M; continuity and strictness of sigma < min(gammahat,1), tau < 1, nu < 1
- Center bunching nu < gamma*gammahat and (1) for the existence of sigma, tau (author text PDF p. 30)
Missing inputs
- Journal pp. 458 and 471-486 as images, to confirm the wording of Lemma 1.1, the choice of sigma, tau, and Lemma 7.1 in the published edition
- Hoelder continuity of the genuine bundle E^u for C^(1+delta) f (imported; needed when Lemma 4.4 is applied with alpha = Jac(Tf|E^u) for the genuine juliennes)
Scope changes
- Lemmas 4.1 and 4.4 should be read for uniformly Hoelder families on the disjoint union of balls along the orbit of p when applied to fake-foliation Jacobians
- Lemma 4.2 should be read with the center-leaf form of Prop. 3.1(iii)(a)
Located concern
BW journal p. 468 (Lemma 4.1 statement) and p. 470 (Lemma 4.4): alpha is a function M -> R; consumer: Lemma 7.1, author text PDF p. 38 (journal page not supplied as image)
Interface mismatch; missing justification, repairable. Lemmas 4.1 and 4.4 are stated for a positive Hoelder function on M. Lemma 7.1 applies Lemma 4.4 with alpha = Jac(Tf restricted to the tangent distribution of the fake unstable foliation). That distribution is defined only on the disjoint union N of the balls B(p_j,r) and depends on the ball index j, so alpha is not a function on M, and alpha_n(y) is a product of values of different functions alpha_{p_i} at y_i.
The proof of Lemma 4.1 uses only |log alpha(x_i) - log alpha(y_i)| <= H d(x_i,y_i)^theta for x_i, y_i in the same ball B(p_i,r), with H independent of i. This holds for the fake Jacobian provided: Prop. 3.1(vi) is uniform in p (asserted on p. 465), Tf is delta-Hoelder, the Jacobian is bounded away from 0 (f is a diffeomorphism of compact M), and the fake distribution is Tf-invariant along the first n iterates of T_n (local invariance, since f^i(T_n) lies in B(p_i,3)).
NEW REPAIR (restatement): formulate Lemmas 4.1 and 4.4 for a family alpha_{p_j} of positive functions on B(p_j,r), j >= 0, whose logarithms are theta-Hoelder with a constant independent of j and p, with alpha_n(y) = prod_{i<n} alpha_{p_i}(y_i). The proofs are unchanged.
None after the restatement. The mismatch affects the fake-foliation branch; for the genuine juliennes alpha = Jac(Tf|E^u) is a function on M, Hoelder when f is C^(1+delta), and the lemma applies as printed.
Located concern
BW journal p. 470, first line of the proof of Lemma 4.2 ('By part (iii.a) of Proposition 3.1')
Citation gap. Prop. 3.1(iii)(a) is stated for the fake stable and fake cs foliations, while Lemma 4.2 concerns a fake center plaque. The displayed inclusion also needs local invariance (ii) for beta = c and coherence (iv), plus the inductive check that y_i stays in the r_1-plaque at x_i. Through the cs clause the lemma inherits the metric-normalisation problem recorded under G2.
The induction was carried out in the argument above and closes because gammahat_j(p)^(-1) sigma_n(p) < sigma_{n-j}(p_j) <= 1 < r_1. Only the cone about E^c and inequality (3) are needed for points on a common fake center leaf, so the lemma is true for every adapted metric in the sense of p. 457.
Cite (ii), (iv) and a center-leaf version of (iii)(a) (repair B under G2), and state the induction explicitly.
None.
Located concern
BW journal p. 470, proof of Lemma 4.2 and statements of Lemmas 4.3-4.4; p. 469 Lemma 4.2
Minor imprecisions. (a) The strict inequality gammahat_j^(-1)(p) sigma_n(p) < gammahat_j^(-1)(p) sigma_j(p) is an equality at j = n. (b) T_{n,sigma,tau} is introduced for tau <= 1, while the exponential bound and Lemma 4.4 need tau < 1; this is stated correctly in the lemmas. (c) The independence of C and kappa from p is not stated. (d) The symbol C is reused for different constants in Lemmas 4.2, 4.3 and 4.4.
Direct reading of p. 470; the uniformity in p was verified above from compactness.
Replace '<' by '<=' and add 'uniformly in p in M' to Lemmas 4.2-4.4.
None.
Located concern
Author text PDF p. 30 ('In what follows we will use only the first of these inequalities') and PDF p. 39 (end of Section 7); journal pages 475 and about 482 not supplied as images, so this is not verified against the printed edition
Bookkeeping inaccuracy at the Section 4/5 interface; not an error in the theorem. Section 5 says only nu < gammagammahat is used for W^s-saturation. That inequality does suffice to choose sigma and tau and for the Lipschitz fake-stable holonomy of Lemma 6.1. However, the last step of Section 7 (regularity of the fake cu-juliennes via absolute continuity of the fake unstable foliation inside the fake cu leaf) invokes Prop. 3.1(vii) for unstable inside center-unstable, which in the C2 branch is derived from the second inequality nuhat < gammagammahat (journal p. 467).
Journal p. 467: 'The center bunching hypothesis nuhat < gamma*gammahat implies ... E^u is a C1 bundle over X, the leaves of the center-unstable foliation.' Author text PDF p. 39: 'Proposition 3.1, part 7. implies that [the fake unstable foliation] C1 subfoliates [the fake center-unstable foliation]; in particular ... absolutely continuous with bounded Jacobians.'
State that both bunching inequalities are used in each half of Theorem 5.1; or prove leafwise absolute continuity of the fake unstable foliation inside fake cu leaves without bunching (not attempted here).
None for Theorems 0.1, 0.3, 5.1, which assume both inequalities. It would matter only for a one-sided variant assuming a single inequality.
Outside the audit
- Journal pp. 458-463 (Lemma 1.1 printed form, Section 2) and pp. 471-486 (Sections 5-8): not supplied as images; consulted only through the author text for interfaces
- Line-by-line proofs of Propositions 2.3-2.7, 5.3, 5.4, 5.5 and Theorem 5.1
- [PSW97] Theorems A and B (body), [HPS77], [KH95], [SSS92], [Has94], [Wil98]: not supplied
- [BW06b] (holonomy note) and [BW05]: not supplied; dates and versions relative to the cutoff unresolved
- [BP74] Proposition 5.1 and the absolute continuity theorems of [BP74], [PS72], Pesin theory, [AV08]: not supplied
- The single-point intersection (local product) assumptions on journal p. 468 beyond a plausibility check
- Figure 1 (dimensions of the neighborhoods T_n) and Section 9 examples on p. 487
Additional supporting archive items are identified in the original report. View the manuscript hashes and repository snapshot.