claude-fable-5-2 high: cef-foundations
Reviewer record, not the publication verdict. This preserves a completed model response, including unaccepted proposals and its stated source limitations. The curated report and repair note specify which conclusions were adopted. Seven initial source-label failures were reconciled in explicit derivatives; native statuses and original access declarations remain in the provenance record.
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.
Within the assigned scope (Sections 2-4 and Theorems 1.3, 1.5, 1.13, 1.7) I found no false numbered theorem.
Verified as printed: the Noetherian theorem for Noetherian k containing Q; the equivalence of finite generation with uniform representation stability plus finite-dimensionality; the FI#-classification over any ring; the growth and character equivalences of Theorem 4.1.7; and the finite-type results for graded FI/co-FI-algebras.
The one substantive defect is in Proposition 3.3.3. Its multiplicity argument cancels a common summand, which is valid only for finite multiplicities, while the proposition and Definition 3.3.2 allow infinite ones. Theorems 1.13 and 3.3.4 are unaffected because finitely generated modules have finite-dimensional V_n. Corollary 4.1.8 inherits the gap for infinite-dimensional FI#-modules. I give a new repair (common kernels of coinvariant maps), which yields stability for n >= stab-deg+|lambda| at any multiplicity, and a direct argument for Corollary 4.1.8; both are new and should be checked independently.
Lesser points: two incorrect incidental statements in Section 2 (the rank of M(m_1) tensor M(m_2), and 'for any W' in the polynomial-dimension remark); an unreferenced Kronecker bound behind Proposition 3.2.2 (new short proof supplied); an omitted 'Noetherian' in Section 4.2; and ordering/typo issues in the proof of Theorem 4.1.7.
Conditional items: Theorem 3.3.4 depends on Macdonald's Example I.7.14, and all monotonicity statements depend on Church's Theorem 2.8 and Proposition 2.3; neither source was supplied.
Chronology and limits: every import actually used ([FH] 1991, [Mac] 1995, [MacL] 1998, [We] 1994, [Ch] arXiv 2011) predates the 2013-07-14 cutoff according to the bibliography, but versions and exact statements were not inspected. No later literature was used to justify any step. I cannot remove knowledge of later developments from my training, and none is relied on here. No tools were run, and the journal text was not collated.
F1
Checked statement: Section 2: FI/FB-modules, M(W) as left adjoint with M(W)_S = direct sum of W_T over T in S (eq:MWdecomp, eq:MWinduced), M(m), span and Lemma 2.3.2 (lemma:Mspan), finite generation (def:finitegen, prop:charfg), tensor products (pr:tensor), H_0 (def:H0, remark:H0), non-semisimplicity (rem:nonsemisimple), and the incidental numerical claims.
Returned assessment: repairable_gap
The constructions and every numbered statement in Section 2 check out; two incidental sentences are wrong as printed (see issues) and do not affect any numbered result.
(1) eq:MWdecomp: the comma category of pairs (T, f: T into S) has only trivial automorphisms because f is injective, and its isomorphism classes correspond to subsets f(T) of S, so the pointwise left Kan extension is the direct sum of W_T over subsets T of S. Exactness of M(-), eq:MWinduced and the count sum_a dim(W_a) binom(n,a) follow. The counit from the sum of M(V_n) to V is onto via the T = S summand.
(2) def:Mm: M(m) = k[Hom_FI(m,-)] is isomorphic to M(k[S_m]); maps M(m) -> V correspond to elements of V_m by Yoneda.
(3) lemma:Mspan and prop:charfg are correct as printed. V_S is a finitely generated k-module whenever V is finitely generated.
(4) pr:tensor: tensor is right exact and commutes with direct sums, so the reduction to M(m_1) tensor M(m_2) is valid, including for the 'generated in degree <= m_1+m_2' clause with infinitely many generators. A basis pair (f_1, f_2) is pushed forward from T = im f_1 union im f_2, which has at most m_1+m_2 elements. Generation in bounded degree plus finitely generated U_n gives finite generation.
(5) def:H0: Hom_FI(V, W) for W extended by zero equals the FB-maps V -> W that kill images of proper injections, i.e. Hom_FB(H_0 V, W). In remark:H0, the direction from finitely generated H_0 to finitely generated V is a graded Nakayama induction on n and is valid. H_0(M(W)) = W holds.
(6) rem:nonsemisimple: for s: W -> V, the relation f_* s_0 = s_1 f_* = 0 with f_* an isomorphism on V forces s_0 = 0. Correct.
(7) The identities for wedge^3 V(1)_n and V(1)_n tensor V(1)_n hold for n >= 4 (at n = 4: 9 = 1+3+2+3). The descriptions of M(3) and M(1,1) are correct.
(8) Harmless imprecisions: the Young symmetrizer c_lambda is idempotent only after rescaling; 'every representation of S_n is self-dual' needs finite dimension.
Premises used
- Pointwise left Kan extension formula [MacL, Eq. X.3.10] (S. Mac Lane, Categories for the Working Mathematician, 2nd ed. 1998; available before the 2013-07-14 cutoff by publication date; exact text not supplied). The formula was also re-derived directly above.
- Functor categories into an abelian category are abelian with pointwise (co)kernels [We, A.3.3] (Weibel 1994; standard; not supplied)
- Classification of irreducible S_n-representations in characteristic 0 (standard, e.g. [FH] 1991)
Missing inputs
- rem:projective (arXiv-only) cites [Lue, Cor. 9.40] for 'every projective FI-module is a direct sum of M(P_n)'; that source was not supplied, so this sentence is unaudited
Scope changes
- The polynomial-dimension remark after eq:MWinduced holds only for W with finite-dimensional total space.
Located concern
Proof of pr:tensor (Proposition 2.3.6), parenthetical '(of rank binom(n,m_1) binom(n,m_2))', approx. lines 775-782
Incorrect numerical statement (incidental; not a step of the proof). M(m)_n has a basis of injections m -> n, so its rank is n!/(n-m)!, not binom(n,m).
Take m_1 = 2, m_2 = 0, n = 2. Then U_2 = M(2)_2 tensor M(0)_2 = k[S_2] has rank 2, while binom(2,2) binom(2,0) = 1. The paper's own def:Mm says the basis is 'ordered sequences of m distinct elements'.
Replace by rank (n!/(n-m_1)!) (n!/(n-m_2)!) = m_1! m_2! binom(n,m_1) binom(n,m_2).
None. Only finiteness of the rank is used.
Located concern
Sentence after eq:MWinduced: 'This shows that for any W, the dimension of M(W)_n is given for all n >= 0 by a single polynomial in n', approx. lines 690-700
False as stated for arbitrary FB-modules W. It requires W_a = 0 for large a and each W_a finite-dimensional.
Let W_a be the trivial representation k for every a >= 0. Then dim M(W)_n = sum_a binom(n,a) = 2^n, which is not a polynomial. This agrees with thm:polyFIsharp, where polynomial dimension is equivalent to H_0 finitely generated.
Insert the hypothesis that the direct sum of all W_a is finite-dimensional.
None on numbered results; the correct statement is thm:polyFIsharp (iii) => (vi).
F2
Checked statement: Sections 3.1-3.2: Phi_a, stability/injectivity/surjectivity degree, lemma:injsurjdescend, pr:stabdegprojective, pr:surjdeg, weight, pr:tensorwt, lemma:branching, pr:Mmuweight, pr:fgweight, pr:stabdegMlambda, lem:coinvariants, pr:narrow, and the proof of Theorem 1.3 (th:noetherian) for Noetherian k containing Q.
Returned assessment: repairable_gap
The Noetherian theorem and its whole proof chain are correct as printed, under exactly the stated hypotheses; the only gaps are an unreferenced 'well-known' fact behind pr:tensorwt and an unstated trivial case in pr:narrow.
(a) Phi_a. T is well defined because any two injections id + f differ by post-composition with S_{n+1}. Phi_a is right exact in general and exact when every n! is invertible, i.e. when k contains Q.
(b) pr:stabdegprojective. The orbit decomposition eq:PhiaMW is correct: the stabilizer of U + k is S_k x S_{n-k}, and S_{n-k} acts trivially. T is the identity on matching summands, so inj-deg(M(W)) = 0. For W = k[S_m] the summand with U empty and k = m is (k[S_m])_{S_m} = k, which first appears at n = m, so surj-deg(M(m)) = m exactly.
(c) pr:surjdeg and the counterexample to its converse are correct. If n >= d then a+n+1 >= d+1, with equality only when a = 0 and n = d, where the coinvariants vanish by hypothesis. An example is the sign representation placed in degree d+1 >= 2.
(d) lemma:branching is Pieri's rule plus Frobenius reciprocity (imported). pr:Mmuweight: a horizontal strip of size n-m on mu forces nu_1 >= n-m, so |lambda| <= m, and mu[n] occurs for n >= m+mu_1. pr:fgweight follows.
(e) lem:coinvariants. (i) A horizontal strip of size k <= mu_1 can always be removed (take the bottom box of the k rightmost columns), and none of size k > mu_1 exists. (ii) Removing a box from every column gives lambda. (iii) I checked both inclusions: nu is obtained from lambda[n] by removing a horizontal r-strip if and only if nu/lambda is a horizontal c-strip with nu_1 <= r+c, and the last condition is vacuous when r >= lambda_1. (iv) holds for n >= a, which is how it is used.
(f) pr:stabdegMlambda. New summands in eq:PhiaMVlambda need k <= lambda_1, and for a >= |lambda|-lambda_1 a nonzero summand with k = lambda_1 appears exactly at n = lambda_1. Correct.
(g) pr:narrow. Correct; the proof only uses surj-deg(V). The case m <= s is not addressed but is trivial (see issues).
(h) Proof of th:noetherian.
- Phi_a(V) is a finitely generated graded k[T]-module: each Phi_a(V)n is a quotient of the finitely generated V{a+n}, and T is onto for n >= a by pr:surjdeg.
- Phi_a(W) is a submodule by exactness (uses Q in k) and is finitely generated by the Hilbert basis theorem (uses k Noetherian); homogeneous generators can be chosen.
- Phi_a(W~) is a k[T]-submodule containing the x_i, so it equals Phi_a(W), hence Phi_a(W/W~) = 0.
- Over k containing Q, k[S_n] is a product of matrix algebras over k, so (W/W~)_n is a direct sum of N_lambda tensor_Q V(lambda)_n.
- Restriction of scalars to Q preserves generation in degree <= a, so only |lambda| <= a occurs (pr:fgweight).
- For n >= a, (V(lambda)n){S_{n-a}} is nonzero by lem:coinvariants(i), which forces N_lambda = 0.
- For n < a, W~_n = W_n because generators of W_n were included; these exist since k is Noetherian. No use is made of pr:tensorwt, pr:narrow or Section 3.3, so there is no circularity.
Premises used
- Pieri/branching rule for S_n [FH] (Fulton-Harris 1991; before cutoff; text not supplied; standard)
- Hilbert basis theorem for k[T]
- Semisimplicity of k[S_n] for k containing Q, with irreducibles defined over Q
- prop:charfg, lemma:Mspan (F1)
Missing inputs
- A reference or proof for the Kronecker-coefficient bound invoked for pr:tensorwt (see issue 1)
Scope changes
- pr:narrow in fact gives lambda_1 <= surj-deg(V).
- th:noetherian is proved exactly for commutative Noetherian k containing Q; the proof uses Q in k in two places (exactness of Phi_a and the isotypic decomposition) and does not extend as written.
Located concern
pr:tensorwt (Proposition 3.2.2) and the sentence preceding it
Missing proof. The bound '|nu| <= |lambda|+|mu| whenever V(nu)_n occurs in V(lambda)_n tensor V(mu)_n' is called well known, with no proof or citation.
No counterexample; the statement is true. It is used later for slopes (pr:gradedtensor, pr:schurfunctor, thm:fisubalgebra).
NEW self-contained argument using only the paper's earlier results. By lemma:branching(i), V(lambda)_n is a constituent of M(lambda)_n for n >= |lambda|+lambda_1, hence of M(|lambda|)n, since M(V_lambda) is a summand of M(k[S{|lambda|}]). By semisimplicity, V(lambda)_n tensor V(mu)_n is then a subrepresentation of (M(|lambda|) tensor M(|mu|))_n. That FI-module is generated in degree <= |lambda|+|mu| by pr:tensor, hence has weight <= |lambda|+|mu| by pr:fgweight. Neither result depends on pr:tensorwt.
None once repaired; th:noetherian does not depend on it.
Located concern
Proof of pr:narrow (Proposition 3.2.8), second paragraph, use of Phi_{m-s-1}
Small gap: Phi_{m-s-1} is only defined when m >= s+1; the case m <= s is not mentioned.
For m <= s the index m-s-1 is negative.
If m <= s, any mu partitioning m has mu_1 <= m <= s trivially.
None.
Located concern
Sentence before lem:coinvariants ('(Res V_lambda)_{S_k}'), lem:coinvariants(iv), and proof of th:noetherian ('wt(V^Q) <= a by Proposition pr:fgweight states that ...')
Typos and an implicit hypothesis: 'V_lambda' should be 'V_n'; (iv) needs n >= a; one sentence is garbled.
Read directly from the source.
Editorial.
None.
F3
Checked statement: Section 3.3: remark on consistent sequences, definition:repstab, Proposition 3.3.3 (pr:stabdegstabrange: uniform representation stability for n >= wt(V)+stab-deg(V) for any FI-module over a characteristic-0 field), Theorem 1.13 (th:FIequiv), and Theorems 1.5/3.3.4 (thm:persinomial) with their exact ranges.
Returned assessment: repairable_gap
Theorems 1.13 and 3.3.4 are correctly proved, with their stated ranges, apart from the monotonicity clause, which rests on unsupplied results of [Ch]. The printed proof of the multiplicity part of Proposition 3.3.3 is valid only when the multiplicities involved are finite, whereas the statement allows arbitrary FI-modules.
(1) Remark on consistent sequences. The k = 2 reduction is valid: Sym{n+1..n+k} is generated by adjacent transpositions, and (n+i, n+i+1) acts trivially on the image of V_{n+i-1} in V_{n+i+1}, which is carried equivariantly into V_{n+k}.
(2) pr:stabdegstabrange, conditions I and II. Correct for all V, including infinite-dimensional V_n. Both factorizations of T exist; exactness of coinvariants and lem:coinvariants(iv) (with n >= d, resp. n+1 >= d) give K_n = 0 and C_{n+1} = 0 for n >= s+d. If wt or stab-deg is infinite the statement is vacuous.
(3) pr:stabdegstabrange, condition III. The printed argument cancels the first summand of eq:repstabcoinvariants from an abstract isomorphism. That is legitimate only when the multiplicities are finite. definition:repstab explicitly allows multiplicities up to infinity, and the proposition places no finiteness hypothesis on V. With that hypothesis added (for example V_n finite-dimensional), the printed proof is complete and the range n >= s+d is justified: pr:narrow gives lambda_1 <= s, so lem:coinvariants(iii) applies for n >= s+m. I did not find a counterexample to the unrestricted statement; a new repair under 'issues' indicates it is true.
(4) th:FIequiv, forward direction. With M a finite sum of M(i)'s mapping onto V and kernel K, th:noetherian gives K generated in degree <= r. The diagram chase is correct: if x in Phi_a(M)n has Tx in Phi_a(K){n+1} and n >= r, then Tx = Ty with y in Phi_a(K)_n, and injectivity on M gives x = y. So inj-deg(V) <= r and stab-deg(V) <= max(g,r). With wt(V) <= g the range is n >= max(g,r)+g. Here V_n is finite-dimensional, so the gap in (3) does not arise.
(5) th:FIequiv, converse. Condition II for n >= N gives V = span(V_{<=N}); finite-dimensionality then gives finite generation. Correct.
(6) Monotonicity. The claims that finitely generated FI-modules are monotone 'in some stable range' rest entirely on [Ch, Theorem 2.8] and [Ch, Proposition 2.3], which were not supplied. They are conditional.
(7) thm:persinomial. Given P_lambda with chi_{V(lambda)n} = P_lambda for n >= |lambda|+lambda_1 and deg P_lambda = |lambda| (imported from [Mac, Ex. I.7.14]), stability of multiplicities for n >= s+d together with |lambda| <= d and lambda_1 <= s (pr:narrow) gives chi{V_n} = sum of c_lambda P_lambda for all n >= s+d. The sum is finite. Uniqueness: a polynomial in X_1..X_r vanishing on S_n for infinitely many n vanishes on arbitrarily large boxes of N^r (pad with one cycle of length > r), hence is zero. The range n >= stab-deg+wt is exactly what the proof gives. Finiteness of stab-deg(V) comes from step (4), so this theorem depends on th:noetherian.
Premises used
- th:noetherian (F2), pr:surjdeg, pr:stabdegprojective, pr:fgweight, pr:narrow, lem:coinvariants (F2)
- [Mac, Example I.7.14]: existence of P_lambda of degree |lambda| computing chi_{V(lambda)_n} for n >= |lambda|+lambda_1 (Macdonald, 2nd ed., 1995; available before the 2013-07-14 cutoff by publication date; exact wording not supplied, so the range and degree are taken on trust)
- [Ch, Theorem 2.8, Proposition 2.3, Definition 1.2] (Church, arXiv:1103.2441, 2011; Invent. Math. 2012; before cutoff by the bibliography's dating; not supplied)
- Definition of uniform representation stability from [CF] (arXiv:1008.1368, 2010), as restated in definition:repstab
Missing inputs
- Text of [Ch] Theorem 2.8 and Proposition 2.3 (needed for every monotonicity assertion)
- Text of [Mac] Example I.7.14 (exact range n >= |lambda|+lambda_1 and degree |lambda|)
Scope changes
- pr:stabdegstabrange condition III: as printed, proved only for V with finite multiplicities; the new repair gives all V, in the range n >= stab-deg(V)+|lambda|.
- pr:stabdegstabrange is vacuous unless wt(V) and stab-deg(V) are both finite.
- thm:persinomial is conditional on the exact statement of [Mac, Ex. I.7.14].
Located concern
Proof of pr:stabdegstabrange (Proposition 3.3.3), final two paragraphs around eq:repstabcoinvariants, approx. lines 1090-1105
Incorrect intermediate step for the stated generality. The proof passes from 'A + B_n has constant isomorphism class and A is constant' to 'B_n is constant'. This cancellation fails when multiplicities are infinite, yet the proposition covers every FI-module over a characteristic-0 field and definition:repstab allows multiplicities up to infinity.
As abstract S_m-representations, the direct sum of infinitely many copies of U with X is isomorphic to its direct sum with X' whenever X and X' are sums of at most countably many copies of U, so cancellation is not a valid inference. FI-modules of finite weight and stability degree with infinite-dimensional V_n exist, e.g. an infinite direct sum of copies of M(0), or M(W) with W_1 infinite-dimensional. This is a gap in the proof, not a counterexample to the statement; I found none.
Either (a) add the hypothesis that each V_n is finite-dimensional (enough for th:FIequiv and thm:persinomial), or (b) use the following NEW argument, not in the paper and to be checked independently.
Step 1. For m <= n let K be the common kernel of the m forgetful maps d_j: M(m)_n -> M(m-1)n, and R = {m,...,n}. Then K has no nonzero Sym(R)-invariants. Take x = sum of c_f f in K invariant under Sym(R), and induct on r = number of j with f(j) in R (r >= 1 always). Choose j with f(j) in R and g = f restricted off j. The coefficient of g in d_j(x) is (n-m+2-r) c_f plus terms c{f'} with r-1 coordinates in R; the latter vanish by induction, and none exist when r = 1. Hence c_f = 0. By lem:coinvariants(i), K has no constituent V(mu)_n with |mu| < m.
Step 2. By semisimplicity and Hom_{S_n}(M(m)n, V) = V^{H_m} with H_m = Sym{m+1..n}: for |mu| < m, V(mu)n^{H_m} is the sum over j of the invariants of H^(j) = Sym({j} union {m+1..n}). Dually, the common kernel of the m natural quotient maps q_j: (V_n){H_m} -> (V_n){H^(j)} is exactly the image of the isotypic components with |lambda| = m, which by lem:coinvariants(i),(ii) is the direct sum of c_{lambda,n} copies of V_lambda. This holds for any multiplicities.
Step 3. T commutes with the q_j. It is an isomorphism on Phi_m(V){n-m} for n-m >= s and injective on Phi{m-1}(V)_{n-m+1}. So T restricts to an S_m-isomorphism between the common kernels at n and n+1.
Conclusion: c_{lambda,n} = c_{lambda,n+1} as cardinals for all n >= s+|lambda|, without pr:narrow or finiteness.
Proposition 3.3.3 is established by the printed proof only when multiplicities are finite; conditions I and II are unaffected. Theorems th:FIequiv and thm:persinomial are unaffected (finitely generated V has finite-dimensional V_n). co:fisharprange inherits the gap for infinite-dimensional FI#-modules (see F4).
Located concern
Proof of th:FIequiv, paragraph citing [Ch, Theorem 2.8] and [Ch, Proposition 2.3]; the monotonicity sentences in the introduction and in the Schur-functor corollary
Monotonicity is asserted purely by citation, and the range ('some stable range') is not made explicit.
The cited statements are not in the supplied source.
Collate against [Ch] (arXiv:1103.2441) and state the range in which monotonicity is obtained.
Monotonicity claims are conditional. The 'if and only if' of th:FIequiv and thm:persinomial do not use monotonicity.
F4
Checked statement: Section 3.4: pr:Vlambda, Murnaghan's theorem (thm:Murnaghan), pr:schurfunctor and its corollaries. Section 4: FI#-modules, th:charFIsharp, thm:polyFIsharp (hence thm:intro:FIsharp), co:fisharprange, graded FI-modules, pr:gradedtensor, thm:fisubalgebra, pr:cofisubalgebra, def:tensorbullet, pr:Vtensorbullet, and the arXiv-only remarks in this range.
Returned assessment: repairable_gap
All numbered results in this range are correct; the defects are an inherited gap in co:fisharprange, a missing 'Noetherian' in one sentence, and expository slips in thm:polyFIsharp.
(1) pr:Vlambda. Among constituents V_mu of M(lambda)_n, a horizontal strip of size n-|lambda| has at most one box per column, so n-mu_1 <= |lambda|, with equality only for mu = lambda[n]. Thus V(lambda)_n is a well-defined multiplicity-one subspace, and by restriction no other constituent can receive V(lambda)_m. Injectivity of M(lambda)d -> M(lambda)n plus irreducibility gives generation in degree d = |lambda|+lambda_1. inj-deg = 0 (submodule of M(lambda), lemma:injsurjdescend). surj-deg <= lambda_1 via M(lambda[d]) and pr:stabdegMlambda, and it is sharp because Phi{|lambda|}V(lambda){lambda_1} = V_lambda is nonzero while the previous term is 0.
(2) thm:Murnaghan. Follows from pr:Vlambda, pr:tensor and th:FIequiv. The index set is finite because finitely generated modules have finite-dimensional V_n. It depends on th:noetherian through th:FIequiv.
(3) pr:schurfunctor. S_lambda(V) is the image of the rescaled Young symmetrizer on V^{tensor d}, so finite generation and degree come from pr:tensor and the weight bound from pr:tensorwt (see the F2 repair). The Schur-functor corollaries follow; their monotonicity clause is conditional on [Ch] (F3).
(4) ex:MWFIsharp. Functoriality holds: T lies in phi^{-1}(B intersect C) iff T lies in A and phi(T) lies in C.
(5) th:charFIsharp. I checked each step: E is an FI#-endomorphism (terms with C not in A factor through V_{C intersect A} = 0); E is idempotent; ker E vanishes in degrees <= n; F: M(V_n) -> V is an FI#-map; E^M = id; and the kernel/cokernel argument (E is simultaneously 0 and id) gives M(V_n) = EV. The induction yields V as the direct sum of M(V^(i)_i), which is pointwise finite, and naturality of each splitting gives natural isomorphisms M H_0 = id and H_0 M = id. Valid over any commutative ring k.
(6) thm:polyFIsharp. All implications hold. (ii)=>(i) is immediate, (i)=>(iii) is by the equivalence, (iii)=>(iv) by V_n = sum of W_i^{binom(n,i)}. For (iv)=>(iii), W_i must be known finitely generated before choosing a maximal ideal with W_i tensor k/m nonzero; this is true (see issues). (iii)=>(vi) and (iii)=>(vii) are verified, including the count of sigma-stable subsets by products of binomial(X_i, n_i(lambda)). thm:intro:FIsharp is (v) <=> (vi).
(7) co:fisharprange. stab-deg(V) <= wt(V) is correct. The stated range then follows from pr:stabdegstabrange, so the F3 gap is inherited when H_0(V) is infinite-dimensional; a direct argument closes it (see issues).
(8) Section 4.2. pr:gradedtensor is correct. thm:fisubalgebra is correct over any k: k{V}^i is a finite sum because V^0 = 0, and A is a quotient. pr:cofisubalgebra is correct: finite type forces each W^i_S finite-dimensional, so (W^{tensor j})^* = (W^)^{tensor j}, and A^ is a sub-FI-module of k{W^*}, finitely generated by th:noetherian (field of characteristic 0).
(9) def:tensorbullet. Functoriality of eq:tensorbullet uses naturality of the unit and augmentation (M(0) has all structure maps the identity) and augmentation composed with unit equal to the identity for indices in B but not in C; both hold. pr:Vtensorbullet: the multiset decomposition is preserved by FI-maps; each summand is generated in degree <= j + sum of m_{l_i} by a pigeonhole count; there are finitely many multisets per grading. Correct over any k.
(10) arXiv-only material. The three decompositions in eq:dlmn are correct by orbit counting (the second-row count for M(2) tensor M(2) is 6 pairs with free S_3-action, giving the regular representation). rem:finiteprojresolution is sound given rem:projective. The statements about the coefficients d^nu_{lambda,mu} were not audited.
Premises used
- pr:tensor, prop:charfg, remark:H0 (F1)
- pr:stabdegMlambda, pr:Mmuweight, pr:tensorwt, pr:fgweight, lemma:branching, lem:coinvariants, lemma:injsurjdescend, th:noetherian (F2)
- th:FIequiv, pr:stabdegstabrange, thm:persinomial (F3)
- Pieri restriction rule [FH, Exercise 4.44] and Schur functors as images of Young symmetrizers [FH, Lecture 6] (1991; before cutoff; not supplied; standard)
- Nakayama's lemma, used to complete (iv)=>(iii)
Missing inputs
- [Ch] Theorem 2.8 / Proposition 2.3 for the monotonicity clause of the Schur-functor corollary
- [Lue, Cor. 9.40] for the 'projective = FI#' statements in the arXiv-only remarks
Scope changes
- co:fisharprange: printed proof covers FI#-modules with finite-dimensional H_0(V); the general case needs the repair above.
- The 'subquotient of finite type' assertion requires k Noetherian containing Q.
- Monotonicity in the Schur-functor corollary is conditional on [Ch].
Located concern
co:fisharprange (Corollary 4.1.8) and its proof
Inherits the infinite-multiplicity gap of pr:stabdegstabrange: the corollary is stated for every FI#-module over a characteristic-0 field, with no finiteness hypothesis.
V = M(W) with W_1 infinite-dimensional has weight 1 and infinite multiplicities, which the printed proof of condition III in pr:stabdegstabrange does not cover.
NEW direct argument. By lemma:branching(i), V(nu)_n occurs in M(lambda)_n (with multiplicity one) iff lambda/nu is a horizontal strip and n >= |nu|+lambda_1. Since |nu| <= |lambda| <= wt(V) and lambda_1 <= wt(V), the multiplicity of V(nu)_n in the direct sum of M(lambda)^{c_lambda} is the same cardinal for all n >= 2 wt(V). Conditions I and II follow from the valid part of pr:stabdegstabrange, or directly from the FI#-splittings. Alternatively use the repair given under F3.
Statement true; printed proof complete only when H_0(V) is finite-dimensional.
Located concern
Paragraph after the definition of 'finite type' in ss:graded: 'When k contains Q, Theorem th:noetherian implies that any subquotient of V is of finite type'
Hypothesis omitted: th:noetherian needs k Noetherian as well as containing Q.
th:noetherian is stated and proved only for Noetherian k; the proof uses that submodules of finitely generated k-modules are finitely generated and that k[T] is Noetherian.
Insert 'Noetherian'.
None for the later uses in this section, which are over fields of characteristic 0.
Located concern
Proof of thm:polyFIsharp, (iii)=>(iv) and (iv)=>(iii)
A typo ('c_i < infinity for all i >= d' should read 'i <= d') and a step used out of order: choosing a maximal ideal m with W_i tensor k/m nonzero needs W_i finitely generated (false for general modules, e.g. Q over Z), which is only noted afterwards and only for i <= d.
Q tensor_Z F_p = 0 for every prime p, so the existence of m is not automatic for a nonzero module.
First observe that V_i is a finitely generated k-module for every i: it is a quotient of V_N for large N via the FI#-retraction, and V_N is finitely generated by (iv). Then W_i = H_0(V)_i is a finitely generated quotient of V_i, and Nakayama gives the required m.
None.
Located concern
List after thm:fisubalgebra (free Lie algebra L(V)); statement of pr:cofisubalgebra ('If V has slope <= m')
Minor inconsistencies: remark:FIalgebras requires algebras to be unital, but L(V) is not; 'V' should be 'W' in pr:cofisubalgebra.
Read directly from the source.
Allow non-unital algebras (the proof is unchanged: drop the j = 0 summand); fix the typo.
None.
Outside the audit
- Sections 5-7 (applications) and all introduction claims about them
- Exact statements of [Mac, Ex. I.7.14], [Ch, Thm 2.8, Prop 2.3], [Lue, Cor. 9.40, Lemma 16.10b], [FH] Pieri formulas, [Snowden, Thm 2.3]
- Introduction claim that the character polynomial of a finitely generated FI-module is integer-valued
- arXiv-only statements about the coefficients d^nu_{lambda,mu} (support bounds, relation to Kronecker and Littlewood-Richardson coefficients) and the 'leading term' of S_lambda(M(1))
- Sufficiency direction of the remark that eq:consistentseqcondition allows promotion of a consistent sequence to an FI-module (only the k = 2 reduction was checked)
- Comparison with the published Duke Math. J. text
- Preserved source-access limitation: Journal (Duke Math. J.) full text: not supplied, not collated
- Preserved source-access limitation: All cited external sources ([Mac], [FH], [MacL], [We], [Ch], [CF], [Lue], [Li], [Snowden]): not supplied; exact statements not inspected
- Preserved source-access limitation: Exact TeX line numbers: no tool was available to count lines, so locators are given by label, with the packet's line ranges quoted as approximate
Additional supporting archive items are identified in the original report. View the manuscript hashes and repository snapshot.