Algebra, representation theory and topology

FI-modules and stability for representations of symmetric groups

Thomas Church, Jordan S. Ellenberg and Benson Farb

54 pages114 claims indexed66 referencesManuscript 2014-10-30

claude-fable-5-2 high: cef-moduli

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 7.1-7.3) I found no false theorem. All displayed theorems are eventual statements and survive, conditional on the internal results of Sections 3-4 and 6.4 (not re-audited here) and on external theorems of Johnson, Kawazumi and others, Jimenez Rolland, and Quillen/Hain, which could not be inspected. Three defects in the printed proofs, each with an elementary repair identified above as a new argument. (1) Section 7.1: kappa_j is defined by a Gysin map along a non-proper map with punctured-surface fibres. Under the natural compactified reading (Arbarello-Cornalba), the asserted FI-invariance 'V^{2i} = Q kappa_i = M(0)' fails, since f_* kappa_j = kappa_j minus the sum of psi_t^j over new points. Using the classes pulled back from M_g repairs the proof without changing the ring. (2) Sections 7.2 and 7.3: the claims that H_1(I_bullet^1;Q) = wedge^3(M(1)+M(1)) is a finitely generated FI-module and that H^*_Alb(I_bullet^1;Q) and gr(I_bullet^1) are of finite type are false at n = 1, 2: H_1(I_1^1;Q) = Q, and H_1(I_2^1;Q) is infinite dimensional, as the manuscript's own citation of Mess shows. Truncating to n >= 3 restores every step and leaves theorem:albanese1 and the gr(I_g^1) theorem intact, but Table 1 items 7 and 9 for the Torelli group need 'n >= 3' and refer to I_g^1, not the closed-surface group. (3) Citation and interface gaps: the deduction from Johnson's Theorem 1A to conjugacy of the induced maps, the missing citation for Johnson's abelianization theorem, the FI-structure on the dual factor in H_1(IA_n), the non-unital Lie-algebra use of thm:fisubalgebra, and the definition of 'graded pieces' of the Malcev Lie algebra. The IA_n computations (functoriality of IA_bullet, the decomposition M(2)+M(1,1)+M(2,1)+M(1,1,1), weight 3) and theorem:assoc:graded were verified directly. Chronology: every import actually needed that has a bibliography entry carries a pre-2013-07-14 date by that entry; the uncited Johnson abelianization theorem and the Quillen comparison are dated only from model memory and remain formally unresolved. No post-cutoff result was used to close any step. Retrospective note, kept separate: none of the post-cutoff works cited in the v4 introduction ([CEFN], [CE], [CP], Gan-Li, Putman-Sam) is used in Section 7. This is not a whole-paper verdict.

M1

Checked statement: Section 7.1: (a) M_{g,bullet} is a co-FI-space, so H^(M_{g,bullet};Q) is a graded FI-algebra; (b) the classes psi_s and kappa_j define a map of graded FI-modules t: V -> H^(M_{g,bullet};Q) with V^2 = M(1)+M(0), V^{2i} = M(0) for i>1, V^{odd} = 0; (c) hence R(M_{g,bullet}) is a graded FI-algebra of finite type, with R^j of weight <= j; (d) the characters and dimensions of R^j(M_{g,n}) are eventually polynomial of degree <= j; (e) the remark that H^*(M_{g,bullet};Q) is of finite type via Jimenez Rolland plus th:FIequiv.

Returned assessment: repairable_gap

The theorem is correct as stated, conditional on thm:fisubalgebra and thm:persinomial; the printed proof has one ill-posed definition and one unverified (reading-dependent) equivariance claim, both with a trivial repair. CHECKED DIRECTLY. (1) Co-FI structure and psi: for f: S -> T injective the forgetful map M_{g,T} -> M_{g,S} sends (X,phi) to (X,phi.f). On the open moduli space no stabilisation occurs, so the cotangent line at the s-th point of (X,phi.f) is literally the cotangent line at phi(f(s)); hence f_* psi_s = psi_{f(s)} and the span of the psi_s is a quotient of M(1). This matches the stated action sigma.psi_i = psi_{sigma(i)}. Rational cohomology of the orbifold or coarse space is the standard convention and harmless. (2) Finite type and weight: once the generators form a graded sub-FI-module that is a quotient of V, thm:fisubalgebra applies. V^0 = 0 because kappa_0 is a scalar; V is of finite type; wt(V^2) = 1 and wt(V^{2i}) = 0, so slope <= 1/2 in cohomological grading. The proof of thm:fisubalgebra via pr:gradedtensor works for any real slope, and the commutative ring R is a quotient of k{V}. So R^j, sitting in H^{2j}, is finitely generated of weight <= j. A direct count agrees: a monomial of grading j has at most j psi-factors, so it lies in a quotient of Sym^d(M(1)) tensor constants with d <= j, which has weight <= d and is generated in degree <= d by pr:tensor. (3) thm:persinomial then gives a character polynomial of degree <= wt <= j, and dimension P(n,0,...) of degree <= j. Sanity check: R^1 is spanned by kappa_1 and psi_1..psi_n, so dim <= n+1. (4) The remark is a correct use of the elementary direction of th:FIequiv (uniform representation stability plus finite dimensionality gives finite generation), provided Jimenez Rolland's consistent sequence is the one induced by the forgetful maps; her paper was not inspected. PROBLEMS are listed under issues: the kappa definition, the claim V^{2i} = Q kappa_i = M(0), and two slips.

Premises used

  • thm:fisubalgebra (Section 4.2): A a graded FI-algebra generated by a graded sub-FI-module V with V^0 = 0; V of finite type implies A of finite type; over a char-0 field, slope <= m passes to A. Used with k = Q, m = 1/2. Internal, not re-audited here.
  • thm:persinomial (Section 3.3): V finitely generated over a char-0 field implies chi_{V_n} = P_V for n >= stab-deg(V)+wt(V), with deg P_V <= wt(V). Finiteness of stab-deg comes from the proof of th:FIequiv, which uses th:noetherian (k Noetherian containing Q). Internal, not re-audited here.
  • pr:tensor and pr:tensorwt/pr:gradedtensor (additivity of generation degree and weight under tensor products). Internal.
  • Elementary: on the open moduli space M_{g,n}, the pullback of psi_s under a forgetful map is psi_{f(s)} (checked above).
  • External, recalled from memory, not inspected: Arbarello-Cornalba relation kappa_j^{AC} = kappa_j^{Mumford} + sum_i psi_i^j. Available_from: Arbarello-Cornalba, J. Algebraic Geom. 5 (1996), by recollection; date_basis: model memory; exact theorem number and version: unresolved. Used only to discuss which reading of kappa_j the authors intend; the repair does not need it.
  • External, for the remark only: Jimenez Rolland, Algebr. Geom. Topol. 11 (2011) 3011-3041, arXiv:1106.0947 (bibliography entry J1; pre-cutoff by its own date). Uniform representation stability of H^i(M_{g,n};Q) for g >= 2. Exact theorem number not given in the manuscript; not inspected.

Missing inputs

  • A well-posed definition of kappa_j on M_{g,n} for n >= 1 (compact-fibre universal curve, or pullback from M_g)
  • A proof or citation that the chosen kappa_j are fixed by the FI maps (pullback along forgetful maps)
  • Text of Jimenez Rolland [J1] to confirm her structure maps are the forgetful maps (remark and Table 1 item 3 only)

Scope changes

  • Theorem 7.1.1 should be read with kappa_j defined via the compact-fibre universal curve or as pullback from M_g; under either reading the statement is unchanged.
  • Table 1 item 3 (H^i(M_{g,n};Q) finitely generated) rests entirely on the external theorem of Jimenez Rolland via the remark, and is conditional on it.

Located concern

Section 7.1, paragraph 'Integration along the fiber yields a Gysin homomorphism (p_n)! ... define kappa_j := (p{n+1})!(psi{n+1}^{j+1})'; repeated in the proof with p_star: M_{g,S+star} -> M_{g,S}

Ill-posed definition. The fibre of M_{g,S+star} -> M_{g,S} is an |S|-punctured surface, non-compact when S is nonempty, and the map is not proper. Fibre integration on ordinary cohomology is not defined for it.

For a bundle with fibre F a punctured surface, H^2(F;Q) = 0, so the Leray edge map H^j(E) -> H^{j-2}(B;H^2(F)) is zero; psi^{j+1} is not vertically compactly supported. Only for S empty (M_{g,1} -> M_g, closed fibre) is the definition literally meaningful.

Define kappa_j on M_{g,S} either as the pullback of the Mumford-Morita-Miller class from M_g, or by integrating over the universal curve C_{g,S} -> M_{g,S} (compact fibres, with M_{g,S+star} as the complement of the sections). The latter needs a choice of extension of psi_star: c_1(omega) gives Mumford's class, c_1(omega(D)) gives the Arbarello-Cornalba class.

None on the theorem. Under either rigorous convention the subring generated by the psi_s and kappa_j is the same.

Located concern

Section 7.1, proof: 'These classes define a map of graded FI-modules t: V -> H^*(M_{g,bullet};Q)' with 'V^{2i} = Q kappa_i = M(0)'

Missing verification that kappa_i is fixed by the FI maps, and the claim depends on the convention. It is true for Mumford's kappa (pulled back from M_g), but false for the Arbarello-Cornalba kappa, which is what the displayed formula (p_{n+1})!(psi{n+1}^{j+1}) most naturally denotes.

With kappa^{AC}j(S) = kappa'j + sum{s in S} psi_s^j (kappa'j pulled back from M_g; relation recalled, not inspected) and f* psi_s = psi{f(s)}, one gets f_* kappa^{AC}_j(S) = kappa^{AC}j(T) - sum{t not in f(S)} psi_t^j. So Q kappa^{AC}_j is not a sub-FI-module and t is not an FI-map on V^{2j} = M(0). S_n-invariance still holds.

New elementary repair, not the printed proof: take as generators the psi_s and the classes kappa'j pulled back from M_g. Then f* kappa'j = kappa'j by functoriality of forgetful maps and f* psi_s = psi{f(s)}, so t: V -> H^* is an FI-map exactly as printed. If the AC convention is intended, note that kappa^{AC}_j - sum_s psi_s^j = kappa'_j lies in the same subring, so R is unchanged.

None. Finite type, weight <= j and eventual polynomiality of degree <= j all survive.

Located concern

Section 7.1, proof, last sentences: 'R^j(M_{g,bullet}) ... is finitely generated in degree <= i'

Index slip (i for j), and the generation-degree statement is not what thm:fisubalgebra provides; that theorem gives finite type and slope (weight), and it is weight <= j that thm:persinomial needs.

The statement of thm:fisubalgebra mentions only finite type and slope.

Write 'finitely generated of weight <= j'. Generation in degree <= j is also true, by pr:tensor applied to monomials with at most j psi-factors.

None.

Located concern

Section 7.1, proof: 'psi_s := c_1(L_i)'

Typo: should be c_1(L_s).

Index i is unbound in that sentence.

Replace L_i by L_s.

None.

M2

Checked statement: Section 7.2: definition of Albanese cohomology; theorem:albanese1 (dim H^i_Alb(I_g^1;Q) is a polynomial of degree <= 3i for g >> i), via I_bullet^1 as an FI-group up to conjugacy, H_1(I_bullet^1;Q) = wedge^3(M(1)+M(1)) finitely generated of weight 3, pr:cofisubalgebra and thm:persinomial; theorem:albanese2 (same for IA_n), via the FI-group IA_bullet and H_1(IA_bullet;Q) = wedge^2 M(1) tensor M(1) = M(2)+M(1,1)+M(2,1)+M(1,1,1).

Returned assessment: repairable_gap

Both theorem statements stand, conditional on the listed imports and internal results. The IA_n proof checks out. The surface-Torelli proof contains a false intermediate statement at n = 1, 2, repaired by truncation. DEFINITION. For finitely generated Gamma, H^(Gamma^ab;Q) is the exterior algebra on H^1, so H^Alb is the subalgebra generated by H^1. Fine. IA_n (theorem:albanese2). (i) f: IA_m -> IA_n is well defined: F_n = f_(F_m) * F(complement) and f_* phi is (transported phi) * id, an automorphism trivial on H_1. It is a homomorphism, and (f.g)* = f* . g_* by evaluating on generators. So IA_bullet is an honest FI-group. (ii) eq:H1IAn is consistent in small cases: n = 0, 1 give 0; n = 2 gives IA_2 = Inn(F_2) = F_2 with H_1 = Z^2 = wedge^2 Z^2 tensor (Z^2)^. (iii) FI-compatibility: with tau(phi)(x) = phi(x)x^{-1} in gamma_2/gamma_3 = wedge^2 H, one finds tau(f_ phi) = wedge^2(f_) . tau(phi) . p_f, where p_f sends x_{f(i)} to x_i and other x_j to 0. Under Hom(H,wedge^2 H) = H^ tensor wedge^2 H the H^* factor therefore carries the FI-structure x_i^* -> x_{f(i)}^, which is M(1). So H_1(IA_bullet;Q) = wedge^2 M(1) tensor M(1) as FI-modules. The manuscript's 'compatible with the maps f_' silently uses this self-duality of M(1). (iv) Decomposition: wedge^2 M(1) = M(1,1). In M(1,1) tensor M(1), the vectors with z in {x,y} give a 2-dimensional S_2-representation triv+sign, hence M(2)+M(1,1); those with x,y,z distinct give Ind_{S_2 x S_1}^{S_3}(sign) = V_(2,1)+V_(1,1,1), hence M(2,1)+M(1,1,1). Dimension check: 2C(n,2)+3C(n,3) = nC(n,2). Weight exactly 3. (v) IA_n is finitely generated (Magnus), so H^1 is finite dimensional and (H^1)^* = H_1 for every n. pr:cofisubalgebra with W = H^1 in grading 1 gives finite type and wt((H^i_Alb)^) <= 3i; thm:persinomial gives eventual character polynomials of degree <= 3i. SURFACE TORELLI (theorem:albanese1). (a) Existence of an embedding S_m^1 -> S_n^1 inducing a_i -> a_{f(i)}, b_i -> b_{f(i)} is not argued but is standard: compose any embedding with a mapping class realising a suitable element of Sp_{2n}Z. (b) Uniqueness up to I_n^1-conjugacy is attributed to Johnson [JohnsonCR, Thm 1A], which I could not inspect. What the proof needs is that I_n^1 acts transitively on separating curves inducing a given homology splitting. Bridging step (new, elementary): if h in I_n^1 carries iota(S_m^1) to iota'(S_m^1), isotope h near the boundary curve so that theta = iota'^{-1} h iota fixes the boundary pointwise. Then theta acts trivially on H_1(S_m^1) because iota_ = iota'* and h* = id, so theta lies in I_m^1, and iota'*(x) = c_h(iota(theta^{-1} x theta)); that is, iota'_ = c_{h iota_(theta)^{-1}} . iota_ with the conjugating element in I_n^1. (c) Naturality of the Johnson homomorphism under subsurface inclusion holds: pi_1(S_m^1) is a free factor, the extended map is the identity on complementary loops, and the symplectic identification H = H^* respects the orthogonal splitting. Also H_1(S_bullet^1;Q) = M(1)+M(1) with the chosen bases. (d) The isomorphism H_1(I_g^1;Q) = wedge^3 H holds only for g >= 3, as the manuscript itself says at eq:Johnsonhom. The proof nevertheless asserts it for the whole FI-module; see issues. After truncation to n >= 3, pr:schurfunctor gives wedge^3(M(1)+M(1)) generated in degree <= 3 and of weight <= 3, pr:cofisubalgebra (or th:noetherian directly) gives (H^i_Alb)^* finitely generated of weight <= 3i, and thm:persinomial gives a polynomial of degree <= 3i for large g. This agrees with the a priori bound O(g^{3i}).

Premises used

  • pr:cofisubalgebra (Section 4.2): A a graded co-FI-algebra over a char-0 field generated by W with W^0 = 0; W^* of finite type (resp. slope <= m) implies A^* of finite type (resp. slope <= m). Its proof uses k{W}^* = k{W^*} (needs each W_n^i finite dimensional), thm:fisubalgebra and th:noetherian. Internal, not re-audited.
  • pr:schurfunctor: wedge^3 of an FI-module generated in degree <= 1 and of weight <= 1 is finitely generated in degree <= 3 with weight <= 3. Internal.
  • thm:persinomial, with th:FIequiv and th:noetherian behind the finiteness of the stability degree. Internal.
  • External: D. Johnson, Math. Ann. 249 (1980) 243-263, Theorem 1A as cited (bibliography entry JohnsonCR; available 1980, pre-cutoff by its own date). Needed content: I-equivalence of bounding simple closed curves with the same homology splitting, for surfaces with one boundary component and ambient genus n >= 3. Exact statement and hypotheses not inspected.
  • External, no citation in the manuscript (the bibliography entry [Jo] is commented out): Johnson's theorem that I_g^1 is finitely generated for g >= 3 and H_1(I_g^1;Q) = wedge^3 H_1(S_g^1;Q) via the Johnson homomorphism. From memory: Ann. of Math. 118 (1983) and Topology 24 (1985); date_basis: model recollection; exact theorem and version: unresolved. Pre-cutoff if the recollection is right.
  • External: eq:H1IAn, H_1(IA_n;Z) = wedge^2 Z^n tensor (Z^n)^* via the Johnson homomorphism; manuscript cites 'e.g.' N. Kawazumi, arXiv:math/0505497v3 (2006; pre-cutoff by arXiv identifier). Exact theorem number not given; not inspected.
  • External: Magnus (1934), IA_n finitely generated; uncited standard fact.
  • External: Mess, Topology 31 (1992), used by the manuscript for H^1(I_2^1;Q) infinite dimensional; this is the fact that breaks the untruncated statement.
  • Elementary and checked here: Mod(S_1^1) = B_3 and I_1^1 = Z generated by the boundary twist; surjectivity of Mod(S_n^1) -> Sp_{2n}Z (standard, e.g. [FM]).

Missing inputs

  • Exact statement of Johnson's Theorem 1A (boundary and genus hypotheses) and the short deduction from transitivity on curves to conjugacy of the induced maps I_m^1 -> I_n^1
  • A bibliographic citation for Johnson's computation of H_1(I_g^1;Q), g >= 3
  • An explicit restriction to n >= 3 in the FI-module and co-FI-algebra statements for I_bullet^1
  • An explicit statement that H^* = (F_n^ab)^* is given the FI-structure x_i^* -> x_{f(i)}^* (isomorphic to M(1)) in the identification of H_1(IA_bullet;Q)

Scope changes

  • FI-module and co-FI-algebra statements for I_bullet^1 hold only after restricting to n >= 3 (set to 0 below 3). Theorem theorem:albanese1 as printed (g >> i) is unchanged.
  • Table 1 item 7 writes 'I_n = genus n Torelli group', while Section 7.2 treats the one-boundary group I_g^1. The closed-surface case is not addressed by the proof.
  • pr:cofisubalgebra as printed says 'If V has slope <= m'; this should read W (Section 4.2, outside the assigned range, noted because it is the interface used).

Located concern

Proof of theorem:albanese1: 'we obtain an isomorphism H_1(I_bullet^1;Q) = wedge^3(M(1)+M(1)). In particular, H_1(I_bullet^1;Q) is a finitely generated FI-module of weight 3'; also the preceding paragraph 'H_1(I_bullet^1;Q) = H^1(I_bullet^1;Q)^* is a finitely generated FI-module' and the sentence before the proof 'H^*_Alb(I_bullet^1;Q) is a graded FI-module of finite type'

Incorrect intermediate step. Johnson's isomorphism holds only for g >= 3; as statements about the whole FI-module (all n >= 0) the isomorphism and the finite generation are false.

n = 1: I_1^1 is infinite cyclic (kernel of B_3 -> SL_2 Z, generated by the boundary twist), so H_1(I_1^1;Q) = Q, whereas wedge^3(Q^2) = 0. n = 2: the manuscript itself records H^1(I_2^1;Q) = Q^infinity (Mess), whereas wedge^3(Q^4) has dimension 4. A finitely generated FI-module has every V_n finite dimensional (remark after prop:charfg), so H_1(I_bullet^1;Q) is not finitely generated. At n = 2, I_2^1 is not finitely generated, so the Albanese definition's hypothesis fails and (H^1)^* is not H_1. Hence the hypothesis 'W^* of finite type' of pr:cofisubalgebra fails for W = H^1(I_bullet^1;Q).

New elementary repair: truncate. For an FI-module V, the module V_{>=3} (V_n for n >= 3, zero otherwise) is a sub-FI-module. By Johnson (g >= 3) and naturality of the Johnson homomorphism, H_1(I_bullet^1;Q){>=3} = (wedge^3(M(1)+M(1))){>=3}. This is generated in degree 3, because wedge^3(M(1)+M(1)) is generated in degree <= 3 and every injection from a set of size <= 3 into a set of size >= 3 factors through a 3-element set. Its weight is <= 3. For n >= 3, (H^i_Alb(I_n^1;Q))^* embeds in wedge^i H_1(I_n^1;Q), and these form a sub-FI-module of (wedge^i of the truncated module)_{>=3}. That module is finitely generated of weight <= 3i (pr:schurfunctor), so the submodule is finitely generated by th:noetherian, and thm:persinomial applies.

theorem:albanese1 is unaffected: it is an eventual statement, with degree <= 3i. The untruncated claims in the proof, and Table 1 item 7 read literally for all n, are false at n = 2 and need 'n >= 3'.

Located concern

Proof of theorem:albanese1: 'Johnson [JohnsonCR, Theorem 1A] proved that the resulting injections I_m^1 -> I_n^1 are all conjugate in I_n^1'

Imported statement not verifiable from the supplied material. As far as I recall, Johnson's theorem concerns I-equivalence of curves, so the conjugacy of the induced homomorphisms needs a short extra argument the manuscript does not give. Existence of an embedding realising each f_* is also not argued.

No counterexample. The bridging argument in the 'argument' field shows the deduction is valid once transitivity on separating curves with fixed homology splitting is granted.

Insert the bridging argument (iota'* = c{h iota_(theta)^{-1}} . iota_, with theta in I_m^1) and one sentence on existence via surjectivity of Mod(S_n^1) -> Sp_{2n}Z. State Johnson's theorem with its exact hypotheses.

Conditional dependency; no change to the conclusion if Johnson's theorem is as recalled.

Located concern

Text before eq:Johnsonhom: 'Johnson proved that I_g^1 is finitely generated, and gave an isomorphism ...'; proof: 'Johnson's identification in eq:Johnsonhom comes from a natural isomorphism'

Load-bearing external theorem with no bibliographic citation in the rendered paper, and its naturality with respect to the chosen FI-structure is asserted, not shown.

The only Johnson entry in the bibliography is JohnsonCR; the entry [Jo] is commented out in the TeX and is not part of the rendered paper.

Cite Johnson's abelianization theorem precisely (g >= 3, one boundary component) and add the one-line naturality argument for subsurface inclusions sketched above.

None if the import is as recalled; chronology and exact version remain unresolved from the supplied record.

Located concern

Proof of theorem:albanese2: 'The isomorphism eq:H1IAn is provided by the Johnson homomorphism, which is compatible with the maps f_*, so we have an isomorphism of FI-modules H_1(IA_bullet;Q) = wedge^2 M(1) tensor M(1)'

Minor omission: the factor (F_n^ab)^* is naturally a co-FI-module. The claim needs the observation that the induced FI-structure on it is extension by zero on dual bases, which is isomorphic to M(1).

tau(f_* phi) = wedge^2(f_*) . tau(phi) . p_f with p_f the coordinate projection; verified in the argument field.

Add this sentence.

None; the stated isomorphism and decomposition are correct.

M3

Checked statement: Section 7.3: properties of gr(Gamma); functoriality (gr(pi_1 X) is a graded FI-algebra for a homotopy FI-space, gr(Gamma) for an FI-group up to conjugacy); theorem:assoc:graded (H_1(Gamma;k) finitely generated implies gr(Gamma) of finite type); the free-group example; example:pbgraded and theorem:malcevpn; the two theorems that dim gr(I_g^1)^i and dim gr(IA_n)^i are eventually polynomial.

Returned assessment: repairable_gap

theorem:assoc:graded, the free-group example, the pure-braid example and the IA_n dimension theorem are correct, conditional on internal results and on pr:CMhTop. The surface-Torelli theorem is true as an eventual statement but its proof inherits the false untruncated claim from Section 7.2. GENERALITIES. Conjugation by g acts trivially on Gamma_j/Gamma_{j+1} since [g,Gamma_j] lies in Gamma_{j+1}, so homomorphisms up to conjugacy induce well-defined maps on gr, and gr of an FI-group up to conjugacy is a graded FI-object whose structure maps are Lie homomorphisms. gr_1 = Gamma^ab tensor k = H_1(Gamma;k) naturally, by universal coefficients. gr is generated by gr_1 because Gamma_{j+1} = [Gamma,Gamma_j], so bracket: gr_1 tensor gr_j -> gr_{j+1} is onto. theorem:assoc:graded. Apply thm:fisubalgebra with V = gr_1 in grading 1 (V^0 = 0). The finite-type half of that theorem uses only pr:tensor and passage to quotients, so it holds over any field k, as stated. Formal mismatch: remark:FIalgebras requires algebras to be unital, and a Lie algebra with gr^0 = 0 is not; the proof of thm:fisubalgebra goes through verbatim with the non-unital free magma algebra (sum over j >= 1 of (V^{tensor j})^{P_j}), so this is cosmetic. Slope: over a char-0 field, wt(gr^i) <= i.wt(H_1). FREE GROUPS. The assignment S -> F_S with x_s -> x_{phi(s)} or 1 is a functor from FI# (composition checked on generators); gr_1 = M(1); finite type follows. Representation stability via th:FIequiv needs k of characteristic 0, which is implicit here. Since this is an FI#-module, co:fisharprange would give it without th:noetherian. PURE BRAIDS. R^2 is the interior of the closed disc, so pr:CMhTop applies (d = 2; ordered configurations in an annular collar are connected). H_1(P_bullet;Q) has basis A_ij indexed by 2-subsets, with A_ij -> A_{phi(i)phi(j)} or 0, so it is M of the trivial S_2-representation, generated in degree 2, weight 2. This is an independent check of the statement that the rendered arXiv text attributes to example:arnoldalgebra. Hence gr(P_bullet) is a graded FI#-module of finite type and each gr^i is uniformly representation stable by th:FIequiv (or co:fisharprange). theorem:malcevpn then follows provided 'grading-i piece of p_n' means gr^i of the Malcev Lie algebra, together with a natural, hence S_n-equivariant, isomorphism gr(P_n) tensor Q = gr(p_n). The manuscript attributes this to Hain [Ha]; it is a general fact I recall as due to Quillen, and I could not check the cited source. IA_n. H_1(IA_bullet;Q) = wedge^2 M(1) tensor M(1) is finitely generated (verified under M2), IA_bullet is an honest FI-group, so gr(IA_bullet) is of finite type with wt(gr^i) <= 3i, and thm:persinomial gives eventual polynomiality (degree <= 3i, not claimed in the text). SURFACE TORELLI. The proof says H_1(I_bullet^1;Q) = wedge^3(M(1)+M(1)) is finitely generated and concludes that gr(I_bullet^1) is of finite type. This fails at n = 1, 2 (see issues). After truncation to n >= 3 the argument is valid and the theorem holds with degree <= 3i.

Premises used

  • thm:fisubalgebra (finite type over any commutative ring; slope over a char-0 field; V^0 = 0), applied to Lie algebras in its evident non-unital form. Internal, not re-audited.
  • th:FIequiv (finitely generated FI-module over a char-0 field implies uniformly representation stable; uses th:noetherian) and thm:persinomial. Internal.
  • pr:CMhTop (Conf(M) is a homotopy FI#-space for M the interior of a compact connected oriented manifold of dimension >= 2 with nonempty boundary), applied to M = R^2. Internal, Section 6.4, not audited in this packet.
  • Results of Section 7.2 as assessed under M2: I_bullet^1 an FI-group up to conjugacy (conditional on Johnson Thm 1A); H_1(I_n^1;Q) = wedge^3 for n >= 3 only; H_1(IA_bullet;Q) = wedge^2 M(1) tensor M(1).
  • External: gr(F_n) is the free Lie algebra (Magnus-Witt; classical, uncited in the manuscript).
  • External: H_1(P_n;Q) of rank C(n,2) with basis A_ij (Arnol'd 1969, cited in the arXiv-only example:arnoldalgebra; classical, pre-cutoff).
  • External: natural isomorphism gr(Gamma) tensor Q = gr of the Malcev Lie algebra; manuscript cites R. Hain, J. Amer. Math. Soc. 10 (1997) (bibliography entry Ha; pre-cutoff by its own date). I recall the general statement as Quillen's (late 1960s); date_basis: model memory; exact theorem number: unresolved; not inspected.
  • External, context only, not load-bearing: Kohno, Invent. Math. 82 (1985); Habegger-Sorger preprint 2000; [CF] Conjecture 5.15 and Theorem 5.3.

Missing inputs

  • Restriction to n >= 3 for gr(I_bullet^1) and H_1(I_bullet^1;Q)
  • A precise definition of 'grading-i pieces of p_n' (the Malcev Lie algebra is filtered and complete, not a priori graded) and of the consistent-sequence maps in theorem:malcevpn; the exact statement of [CF] Conjecture 5.15
  • Exact source for gr(P_n) = gr(p_n) as S_n-representations
  • Statement that k has characteristic 0 wherever th:FIequiv or thm:persinomial is invoked in this subsection

Scope changes

  • gr(I_bullet^1) is of finite type only after restriction to n >= 3; the eventual-polynomiality theorem is unchanged.
  • Degree bounds are available but not claimed: deg <= 3i for gr(I_g^1)^i and gr(IA_n)^i, and weight <= 2i for gr(P_n)^i.
  • theorem:assoc:graded (finite type) holds over any field; the representation-stability and character corollaries need characteristic 0.

Located concern

Section 7.3, joint proof of the two gr-dimension theorems: 'the FI-modules H_1(I_bullet^1;Q) = wedge^3(M(1)+M(1)) and H_1(IA_bullet;Q) = ... are finitely generated. Theorem assoc:graded now implies that the graded FI-algebras gr(I_bullet^1) and gr(IA_bullet) are of finite type'

Incorrect intermediate step for I_bullet^1, inherited from the proof of theorem:albanese1: H_1(I_bullet^1;Q) is not finitely generated as an FI-module on all finite sets, and gr(I_bullet^1) is not of finite type.

gr(I_2^1)^1 = H_1(I_2^1;Q) is infinite dimensional (the manuscript cites Mess for H^1(I_2^1;Q) = Q^infinity), while every graded piece of a finite-type FI-module is finite dimensional at each n. Also H_1(I_1^1;Q) = Q is not wedge^3(Q^2) = 0.

New elementary repair: let L be the graded FI-object with L_n = gr(I_n^1) tensor Q for n >= 3 and 0 otherwise. It is a sub-FI-object of gr(I_bullet^1), closed under brackets pointwise, and generated as a Lie algebra by L^1 = H_1(I_bullet^1;Q){>=3} = (wedge^3(M(1)+M(1))){>=3}. The latter is finitely generated in degree 3 with weight <= 3. thm:fisubalgebra (non-unital form) gives L of finite type with wt(L^i) <= 3i, and thm:persinomial gives dim gr(I_g^1)^i = P_i(g) for large g, with deg P_i <= 3i.

The theorem 'dim gr(I_g^1)^i is polynomial in g for g >> i' is unaffected. The claim that gr(I_bullet^1) is of finite type, and Table 1 item 9 for the Torelli group read literally, need 'n >= 3'. The IA_n statement is unaffected.

Located concern

theorem:assoc:graded, proof; remark:FIalgebras; list of free algebras after thm:fisubalgebra (L(V))

Formal mismatch: thm:fisubalgebra is stated for unital, possibly non-associative graded algebras with a surjection k{V} -> A, but graded Lie algebras are not unital (gr^0 = 0).

k{V}^0 = k, while gr(Gamma)^0 = 0; no unital algebra map exists unless A = 0.

Remark that the proof of thm:fisubalgebra applies unchanged to the non-unital free algebra (sum over j >= 1), or allow non-unital algebras in remark:FIalgebras.

None.

Located concern

example:pbgraded: 'Hain [Ha] proved that there is an isomorphism of S_n-representations gr(P_n) = gr(p_n)'; theorem:malcevpn 'grading-i pieces of p_n'

Imprecise import and undefined terminology. The comparison of gr of a group with gr of its Malcev Lie algebra is a general fact and its attribution to [Ha] could not be checked. 'p_n^i' is undefined unless it means gr^i(p_n), or the degree-i part of the Drinfeld-Kohno graded Lie algebra with an S_n-equivariant identification.

No counterexample. With p_n^i := gr^i(p_n) and a natural isomorphism gr(Gamma) tensor Q = gr(Malcev(Gamma)), the theorem follows from th:FIequiv applied to gr(P_bullet)^i.

Define p_n^i := gr^i(p_n), cite the natural isomorphism precisely (naturality gives S_n-equivariance through the outer action), and state which maps p_n^i -> p_{n+1}^i form the consistent sequence (those of the FI-structure).

theorem:malcevpn holds conditionally on that import; representation stability of gr(P_n)^i tensor Q itself is established independently of it (given pr:CMhTop and th:FIequiv, or co:fisharprange).

Located concern

Free-groups example: 'given by phi(x_s) = x_{phi(s)} if i in A, and phi(x_s) = 1 if i in S-A'; opening of Section 7.3 'assume that k is a field'

Typos (i for s; phi used for both the bijection and the group map). The characteristic-0 hypothesis needed for th:FIequiv and thm:persinomial is left implicit.

th:FIequiv and thm:persinomial are stated only over fields of characteristic 0; Section 7.3 fixes only 'k a field' until the Torelli example sets k = Q.

Correct the indices and state k = Q (or char k = 0) where representation stability or character polynomials are concluded.

None.

Outside the audit

  • Proofs of th:noetherian, pr:stabdegstabrange, th:FIequiv, thm:persinomial, pr:tensor, pr:tensorwt, pr:schurfunctor, pr:gradedtensor, thm:fisubalgebra, pr:cofisubalgebra, th:charFIsharp, co:fisharprange (interfaces read; correctness assumed conditionally)
  • pr:CMhTop (Section 6.4) and the arXiv-only example:arnoldalgebra, beyond the direct check that H_1(P_bullet;Q) is M of the trivial S_2-representation
  • All external sources: Johnson 1980 Theorem 1A; Johnson's abelianization of the Torelli group; Kawazumi and the Farb and Cohen-Pakianathan computations of H_1(IA_n); Magnus; Mess; Jimenez Rolland [J1], [J2]; Hain [Ha]; Kohno; Habegger-Sorger; Arbarello-Cornalba; Quillen; [CF] Conjecture 5.15 and Theorem 5.3
  • Background assertions in Section 7 not used in proofs: results of Hain, Sakasai, Pettet, [CF2] lower bounds, the pseudo-nilpotence remark, the Drinfeld-Kohno presentation
  • Duke Math. J. published text (not collated); independent recount of TeX line numbers
  • Sections 1-6 other than as interfaces; Table 1 items 1, 2, 5, 6
  • Preserved source-access limitation: All external references (Johnson 1980, Johnson's abelianization papers, Kawazumi, Jimenez Rolland, Hain, Kohno, Quillen, Arbarello-Cornalba, [CF] Conjecture 5.15): not supplied
  • Preserved source-access limitation: Duke Math. J. published full text: not supplied
  • Preserved source-access qualification: Context read for interfaces only (not audited as proofs): remark:FIalgebras, def:finitegen, prop:charfg, pr:tensor, de:weight, pr:tensorwt, pr:fgweight, th:noetherian, pr:stabdegstabrange, th:FIequiv, thm:persinomial, pr:schurfunctor, th:charFIsharp, co:fisharprange, pr:gradedtensor, thm:fisubalgebra, pr:cofisubalgebra, pr:CMhTop, example:arnoldalgebra (arXiv-only, rendered because remove passes through), table:examples, cor:polydim:app
  • Preserved source-access qualification: Bibliography entries used only as date/version metadata: JohnsonCR, Ka, J1, J2, Ha, HS, Ko, Mess, Pe, Va, FM, CF, Ch, Mac
  • Preserved source-access qualification: Limits: no tools were run and no external source was opened. Line numbers are the packet-supplied section ranges and were not independently recounted. The journal text was not collated. Recollections of external theorems come from model memory and are marked as such.
Source record

Additional supporting archive items are identified in the original report. View the manuscript hashes and repository snapshot.

Cite this review

If this review or its supporting reports inform your work, please cite it.

Igor Rivin. Review of FI-modules and stability for representations of symmetric groups. Math Referee, .

Download .bib
View BibTeX
@misc{rivin2026review-human-church-ellenberg-farb-fi-modules,
  author = {Rivin, Igor},
  title = {{Review of FI-modules and stability for representations of symmetric groups}},
  year = {2026},
  month = {November},
  date = {2026-11-20},
  howpublished = {Math Referee},
  url = {https://mathreferee.dimensionreducers.ai/papers/human-church-ellenberg-farb-fi-modules/panel/},
  note = {Review report}
}