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-configurations

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 the core rational results survive: the finite-type theorem (theorem:coho:config2), the spectral-sequence induction and ranges in th:config, thm:confncharpoly with range n >= 2i, and co:vakilwood with range n >= i+|mu| all check out, conditional on the cited results of Totaro and Church [Ch, Section 3.3], which were not supplied. Found: (a) one direction of lemma:homologyfacts(ii) is false but unused; (b) the general-ring step of thm:ConfMFIsharpfg is an invalid inference, repaired here by a new universal-coefficient argument (the field and Z cases are fine modulo an unproved E_2 description with those coefficients); (c) item 3 of thm:intro:openbetti is false for all n>=0 by an explicit counterexample (max(n, binom(n,2))) and holds only eventually; (d) several statements omit hypotheses (connectedness, finite-dimensional cohomology) or say 'degree i' where 'degree at most i' is meant. Chronology: every import used in these sections (Totaro 1996, Church arXiv 2011 / Invent. 2012, Forster 1964, classical Kunneth and universal coefficients) predates the 2013-07-14 cutoff according to the bibliography of the supplied file; versions were not inspected. No later literature was used to close any proof obligation; I cannot remove later knowledge from model weights, and the repairs and counterexamples above are self-contained elementary arguments labelled as new. No tools were run, and the journal text was not collated.

C1

Checked statement: def:coFIMbullet, pr:cohoMbullet, remark:FIspaces, the co-FI-space Conf(M), and theorem:coho:config2: for M a connected oriented manifold of dimension at least 2 with finite-dimensional total rational cohomology, H^*(Conf(M);Q) is a graded FI-algebra of finite type.

Returned assessment: repairable_gap

The theorem follows from the printed argument once two small points are patched and Totaro's E_2 description is granted. Checked directly: (1) M^bullet is a co-FI-space, and formula eq:MbulletFIsharp is strictly functorial on FI-sharp: for composable partial injections both g_f_(phi) and (gf)_(phi) equal phi on psi(B cap C) and the basepoint elsewhere. (2) pr:cohoMbullet: over a field with each H^i(M;k) finite-dimensional the cohomology Kunneth isomorphism holds for singular cohomology, and the FI structure inserts the unit in the new tensor factors. Here V=H^(M;k) is constant, so V^i is a sum of b_i copies of M(0) and one may take the free cover in pr:Vtensorbullet to be V itself. FI maps send basis tensors to plus or minus basis tensors with the same multiset of nonunit labels, so each summand is generated in degree at most the number of nonunit factors, which is at most i. The Koszul sign is harmless. (3) theorem:coho:config2: the Leray spectral sequences of Conf_S(M) into M^S are natural for the commuting restriction squares and multiplicative, so each page is a bigraded FI-algebra and the filtration on H^i is by sub-FI-modules. Granting Totaro's presentation, E_2^{0,d-1} is spanned by the G_ab with G_ab=(-1)^d G_ba, so it is M of the trivial S_2-representation for d even and M of the sign representation for d odd. The positive-degree generators form a finite-type graded FI-module, thm:fisubalgebra gives E_2 of finite type, th:noetherian over Q passes this to each E_r and to E_infinity (first-quadrant, so each entry stabilises), and extensions of finitely generated FI-modules are finitely generated. Not verified: the Totaro import, and the Dold-Thom, Barratt-Priddy-Quillen and Sagave-Schlichtkrull statements in remark:FIspaces, which are remarks and not used in any proof.

Premises used

  • pr:Vtensorbullet (finite type of V^{tensor bullet} when V^0 = M(0) and V is of finite type)
  • thm:fisubalgebra (requires generating sub-FI-module V with V^0=0)
  • th:noetherian (Noetherian property over rings containing Q)
  • Kunneth theorem for singular cohomology over a field under degreewise finiteness
  • External: B. Totaro, Configuration spaces of algebraic varieties, Topology 35 (1996), Theorem 1, for oriented manifolds of dimension d: E_2 of the Leray spectral sequence of Conf_n(M) in M^n is generated by H^*(M^n) and classes G_ab of degree d-1. available_from 1996; date_evidence: bibliography entry [To] in the supplied TeX; date_basis: journal publication; version: journal article; text not supplied, so exact hypotheses and coefficient ring are not verified
  • Functoriality and multiplicativity of the Leray spectral sequence (standard, classical)

Missing inputs

  • Text of Totaro's Theorem 1 (hypotheses, coefficients, naturality in the index set) was not supplied
  • Dated, versioned statements for the Dold-Thom, Barratt-Priddy-Quillen and Sagave-Schlichtkrull results quoted in remark:FIspaces

Scope changes

  • Finite-type conclusion holds under: connected, oriented, dimension at least 2, total rational cohomology finite-dimensional; Table 1 entry 1 must be read with these hypotheses.
  • Conditional on Totaro's Theorem 1 as cited.

Located concern

theorem:coho:config2, proof paragraph beginning 'E_2^{,0} simeq H^(M^bullet;Q) is of finite type'

Missing step. thm:fisubalgebra requires V^0=0, but the stated generating set (E_2^{0,d-1} together with E_2^{p,0} for all p>=0) contains E_2^{0,0}=M(0) in total grading 0.

thm:fisubalgebra's hypothesis is 'V subset A with V^0=0'; with V^0 nonzero the sum over j of V^{tensor j} is not finite in a fixed grading.

Use only the positive-degree generators E_2^{p,0} for p>=1 and E_2^{0,d-1}. This needs d-1>=1, which holds. Since M is connected E_2^{0,0}=Q.1 and the free algebra k{V} is unital, so these generate E_2.

None after the one-line repair.

Located concern

pr:cohoMbullet, sentence 'it is easy to see that the proof of that proposition is not affected by this sign'

Missing proof: pr:Vtensorbullet is applied to an FI-module that is only isomorphic to V^{tensor bullet} up to Koszul signs, and the sign-twisted version is asserted rather than proved.

No argument is printed for the signed version.

New elementary argument (given in the argument field): for constant V the basis of tensors with a fixed multiset of nonunit labels is permuted up to sign by FI maps, so each summand is generated in degree at most the size of the multiset.

None.

Located concern

Section 6.2 text and theorem:coho:config2 proof

Typographical slips: 'f_(varphi)=varphi o f : S into T' should have target M; 'the Leray spectral sequence for f' : X to Y' should read X' to Y'; 'i_S o f_ = f_* o i_T' has S and T interchanged; pr:CMhTop calls Conf(M): co-FI to Top an 'FI-space'.

Direct reading of the source.

Correct the symbols.

None.

Located concern

Table 1 entry 1 (table:examples) and cor:polydim:app versus theorem:coho:config2

Hypothesis mismatch: the table lists H^i(Conf_n(M);Q) for any connected oriented manifold as finitely generated, but the theorem needs dim_Q H^*(M;Q) finite and dim M at least 2.

If H^i(M;Q) is infinite-dimensional then V_1=H^i(M;Q) is not a finitely generated k-module, so by the remark after prop:charfg the FI-module is not finitely generated. For M=R, Conf_n(R) has n! components, so b_0 is not eventually polynomial.

Carry the hypotheses of theorem:coho:config2 into Table 1 and cor:polydim:app.

Restricts the scope of the introductory claims only.

C2

Checked statement: lemma:homologyfacts (i),(ii); th:config (weight at most i, inj-deg at most i+2-d, surj-deg at most i for d>=3); the deduction of thm:confncharpoly; the dimension-2 remark; co:vakilwood.

Returned assessment: repairable_gap

The spectral-sequence induction and all stated ranges check out; one direction of lemma:homologyfacts(ii) is false but unused, and there are hypothesis and wording slips. Lemma (i): after applying the exact functor Phi_a, take y in ker g_n with Ty=f(x'). If T is onto on X (n>=B) write x'=Tx, so T(y-fx)=0 and injectivity on Y (n>=C) gives y=fx. If y' is in ker g_{n+1} and T is onto on Y (n>=D) write y'=Ty; then T(gy)=0 and injectivity on Z (n>=E) gives gy=0. This gives exactly (max(B,C), max(D,E)). Lemma (ii): the 'if' direction follows by induction on extensions (short five lemma after Phi_a). Induction in th:config, with D=d-1>=2 and rows only at heights qD: d_r is nonzero only for r=kD+1 and maps (p,qD) to (p+kD+1,(q-k)D). Base: E_{D+1}=E_2 has type (0,p+2q); the three terms have surj-degs p+2q+1-D, p+2q, p+2q+D-1, so E_{2D+1} has type (p+2q+1-D, p+2q), which is (a_2),(b_2). Step: surj-deg(X) <= (p-kD-1)+2(q+k)+(k-2)(D-2) = p+2q+3-2D <= p+2q+1-D since min(k-2,q+k-1)=k-2 and D>=2; with (a_k) for Y this gives (a_{k+1}). If k>q>0 then Z=0 and (b_k) for Y suffices. If k<=q then inj-deg(Z) <= p+kD+1+2(q-k)+1-D = p+2q+(k-1)(D-2) and surj-deg(Y) <= p+2q+(k-2)(D-2), giving (b_{k+1}). Limit with p+qD=i: p+2q+1-D <= i+2-d; for q>0, p+2q+(q-1)(D-2)=p+qD+2-D <= i; for q=0, E_infinity^{p,0} is a quotient of E_2^{p,0}. Negative bounds occur only where the bound is read as 0 or the term vanishes. Weight: subquotient of E_2^{p,qD}, weight at most p+2q <= i. thm:confncharpoly: stab-deg+weight <= 2i and degree <= weight <= i, via thm:persinomial; this needs finite generation from theorem:coho:config2. co:vakilwood: stab-deg <= i makes T an S_a-equivariant isomorphism (V_n){S{n-a}} to (V_{n+1}){S{n+1-a}} for n >= i+a with a=|mu|; taking S_mu-invariants and using the transfer identification gives the isomorphism for n >= i+|mu|. Dimension-2 remark: weight <= p+2q = 2i-p <= 2i. Not verified: Church's Section 3.3 description of E_2 and the quoted earlier range of [Ch, Theorem 5].

Premises used

  • Exactness of Phi_a over rings containing Q (Definition def:Phia discussion)
  • pr:stabdegprojective: M(W) has inj-deg 0 and surj-deg at most m if W_i=0 for i>m
  • pr:Mmuweight and subquotient monotonicity of weight
  • lemma:injsurjdescend
  • thm:persinomial with range n >= stab-deg + weight (depends on pr:stabdegstabrange, pr:narrow; not re-audited here)
  • theorem:coho:config2 for finite generation
  • External: T. Church, Homological stability for configuration spaces of manifolds, Invent. Math. 188 (2012), Section 3.3: E_2^{p,q(d-1)} is a direct sum of M(W_k) for k <= p+2q. available_from: arXiv:1103.2441 (March 2011, by arXiv number); date_evidence: bibliography entry [Ch] in the supplied TeX; date_basis: arXiv identifier and journal year; version not checked; text not supplied
  • External: Totaro 1996 as in C1

Missing inputs

  • Text of [Ch, Section 3.3], including whether it assumes finite-dimensional rational cohomology and whether the decomposition is natural for all FI maps
  • Statement of [Ch, Theorem 5] to confirm the quoted earlier range

Scope changes

  • th:config and co:vakilwood are verified conditional on Totaro's theorem and [Ch, Section 3.3]; unconditional status without finite-dimensional cohomology is unresolved.
  • thm:confncharpoly holds for connected oriented manifolds of dimension at least 3 with finite-dimensional rational cohomology (e.g. compact), with 'degree at most i'.
  • Only one implication of lemma:homologyfacts(ii) is valid.

Located concern

lemma:homologyfacts (ii)

Incorrect statement: the 'only if' direction of the claimed equivalence is false.

Over Q let V=M(0) with filtration 0 < W < V, where W_S=Q for S nonempty and W_emptyset=0. V has type (0,0). Phi_0(W)_0=0 maps to Phi_0(W)_1=Q, not onto, so surj-deg(W)=1. Phi_0(V/W)_0=Q maps to Phi_0(V/W)_1=0, not injective, so inj-deg(V/W)=1.

State only the implication: graded pieces of type (A,B) imply V of type (A,B).

None: the proof of th:config uses only the valid direction.

Located concern

th:config statement and first paragraph of its proof

Hypothesis ambiguity: th:config omits the assumption dim_Q H^*(M;Q) finite made in theorem:coho:config2, while the proof 'continues with the notation' of that proof and imports the E_2 description from Totaro and [Ch, Section 3.3].

The cohomological Kunneth isomorphism H^(M^n;Q)=H^(M;Q)^{tensor n} fails without finite type; whether the cited description covers that case cannot be checked from the supplied source.

Either add the finiteness hypothesis, or prove the M(W) decomposition of E_2 directly. New sketch, not verified in detail: R^q i_* Q is a sum of orientation-twisted constant sheaves on diagonals; H^p(M^m;Q) is the dual of H_p(M^m;Q), which by the homological Kunneth theorem is a finite sum of terms involving at most p coordinates nontrivially.

With finite-dimensional cohomology (in particular M compact, the case used for thm:confncharpoly) no effect; the general case is unresolved pending the import.

Located concern

thm:confncharpoly, last sentence ('agrees with a polynomial of degree i')

Wording: the Betti-number polynomial has degree at most i, not necessarily i. The theorem statement also relies on the preceding paragraph for the hypotheses connected, oriented, compact.

For d>=3 the row q=0 is the only contribution in total degree 1, so H^1(Conf_n(M);Q)=H^1(M^n;Q). For M=S^3 this is 0 for all n, so the polynomial is 0, not of degree 1.

Write 'degree at most i' and state the hypotheses in the theorem.

Cosmetic.

Located concern

Remark after th:config (dimension 2)

Typo: the direct sum is indexed by i but should run over p with p+q=i.

Direct reading.

Index the sum by p.

None.

C3

Checked statement: Definition of homotopy FI-sharp-space; pr:CMhTop; thm:ConfMFIsharpfg: for M connected, oriented, of dimension d>=2, the interior of a compact manifold with nonempty boundary, and k Noetherian of finite Krull dimension, H^i(Conf(M);k) is a finitely generated FI-sharp-module generated in degree at most i (d>=3) or 2i (d=2).

Returned assessment: repairable_gap

The Q case, the field case and the Z case go through (the last two conditional on an E_2 description with those coefficients); the general-ring step is invalid as printed, but I give a new argument below that recovers the statement. pr:CMhTop: Conf_{T-B}(R) is connected because R is a connected manifold of dimension d>=2, so Psi_B^T is well defined up to homotopy; identities and co-FI morphisms (B=T) give the original structure. The composition homotopy is only sketched; it is easy if Phi is the standard inward push along a larger collar, since then the constant points lie in the connected collar region M minus Phi(Phi(M)) and can be moved into R before homotoping Phi o Phi to Phi through Phi o Phi_t. Q case: from th:config and the dimension-2 remark, weight <= i (resp. 2i); for an FI-sharp-module V = sum of M(W_m), wt(M(mu))=|mu| forces W_m=0 beyond that bound, and finite generation is theorem:coho:config2. Field case: dim (E_2^{p,q(d-1)})_n = O(n^{p+2q}), vector-space subquotients and finite filtrations give dim H^i = O(n^i) (resp. O(n^{2i})), and thm:polyFIsharp (v) implies (iii) with W_j=0 for j>i. Z case: valid because for finitely generated abelian groups the minimal number of generators does not increase under subquotients and is subadditive in extensions. General k: the sentence 'Since F is a field, this bound passes to subquotients' is not a valid inference, because tensoring with F over k is not left exact; see issues. New repair (my argument, not in the paper): X=Conf_n(M) has finitely generated integral homology in each degree, so its integral cochains are homotopy equivalent to a sum of elementary complexes. Hence, as k-modules, H^i(X;k) = k^{b_i} + sum_j k/m_j k + sum_l k[m'_l], where m_j are the invariant factors of the torsion of H^i(X;Z), m'_l those of H^{i+1}(X;Z), and k[m] is the m-torsion ideal. For Noetherian k only finitely many primes p (set P_k, each lying in an associated prime) have k[p] nonzero, and the ideals k[m] form a finite set, so each needs at most c_k generators. The number of l with p dividing m'_l is at most dim H^i(X;F_p). Therefore gens_k H^i(X;k) <= gens_Z H^i(X;Z) + c_k * sum over p in P_k of dim H^i(X;F_p) = O(n^i) (resp. O(n^{2i})), using the Z and F_p cases. thm:polyFIsharp (iv) implies (iii) then finishes. This uses neither Forster's theorem nor finite Krull dimension.

Premises used

  • th:charFIsharp and thm:polyFIsharp ((iv) implies (iii), (v) implies (iii), with W_j=0 for j above the growth exponent)
  • th:config, theorem:coho:config2 and the dimension-2 remark for k=Q
  • Existence of collars and connectivity of ordered configuration spaces of connected manifolds of dimension at least 2
  • Structure theory of finitely generated abelian groups; universal coefficients via elementary complexes (classical)
  • External: Totaro 1996 / Church 2012 Section 3.3 description of E_2, needed here with coefficients in an arbitrary field and in Z (not shown in the paper; texts not supplied)
  • External: O. Forster, Math. Z. 84 (1964), Satz 1, as quoted in the paper; available_from 1964; date_evidence: bibliography entry [Fo] in the supplied TeX; date_basis: journal publication; version: journal article; text not supplied; not needed for the new repair

Missing inputs

  • A proof or precise citation that (E_2^{p,q(d-1)})n with coefficients in F_p or Z is generated by O(n^{p+2q}) elements. Plausible sketch, not verified against the sources: R^q i* k is a sum over partitions pi with n-q blocks of c(pi) orientation-twisted constant sheaves on diagonals isomorphic to M^{|pi|}; H^p(M^m;k) needs O(m^p) generators using a finite CW model with one 0-cell; the weighted partition count is the Stirling number c(n,n-q)=O(n^{2q}).
  • Exact statement of Forster's Satz 1

Scope changes

  • Printed proof covers k a field and k=Z (conditional on the E_2 description with those coefficients); the general Noetherian case needs the new repair argument.
  • With the repair the finite Krull dimension hypothesis is not needed, but generator counts depend on k.

Located concern

thm:ConfMFIsharpfg, final paragraph ('Since F is a field, this bound passes to subquotients ...')

Incorrect intermediate step for general Noetherian k. E_infinity and H^i(Conf_n(M);k) are subquotients and extensions of E_2 as k-modules, and tensoring with F=k/m over k does not preserve inclusions. The parenthetical claim of a bound independent of F coming from the k=Z generator count is also false.

(a) The paper's own example: k=C[x,y], B=k, A=(x,y), F=k/(x,y): dim_F of A tensor F is 2, of B tensor F is 1. (b) Under the theorem's hypotheses: let k_r=Z[x_1..x_r]/(2x_1,..,2x_r), Noetherian of Krull dimension r, and let M be a connected oriented open manifold as in the theorem with H_1(M;Z) containing Z/2 and H^1(M;Z)=0 (e.g. the 7-manifold in C4). Then H^1(Conf_1(M);k_r)=Hom(H_1(M),k_r)=k_r[2]=(x_1..x_r), and tensoring with F=k_r/(2,x_1..x_r) gives dimension r, while (E_2^{1,0})_1 for k=Z is H^1(M;Z)=0. Reading E_2 with F coefficients does not help either: H^1(M;F_2) has dimension 1 < r.

Use the new universal-coefficient argument in the argument field; the bound is O(n^i) with a constant depending on k (through c_k and P_k). This is a new repair, not the printed proof.

The theorem statement survives for every Noetherian k, conditional on the F_p and Z cases; the printed proof is valid as written only for fields and for Z.

Located concern

thm:ConfMFIsharpfg, sentences 'We recall from the proof of Theorem th:config that each entry E_2^{pq} ... is a finitely generated FI-sharp-module' and 'From the proof of Theorem th:config, each entry E_2^{p,q(d-1)} ... generated in degree <= p+2q'

Missing proof: the proof of th:config established (by citation) the structure of E_2 only with Q coefficients; it is reused for arbitrary fields and for Z, where the Kunneth formula has Tor terms.

No argument for other coefficients appears in the source.

Supply the sheaf-theoretic count sketched under missing_inputs.

The positive-characteristic and integral cases are conditional on this input.

Located concern

pr:CMhTop, last paragraph of proof

Sketchy homotopy and typos: the restriction of (g o f)_*(varphi) to U minus psi(B cap C) is the chosen configuration for psi(B cap C) in U, not q_B^T; 'u=psi(phi^{-1}(s))' should be u=psi(phi(s)). For an arbitrary Phi merely isotopic to the identity the constant points are not obviously movable in the complement of the moving points.

Direct computation: g_f_(varphi) equals Phi(Phi(varphi(s))) on psi(B cap C), Phi(q_B^T) on psi(C minus B), and the chosen configuration on U minus psi(C).

Fix Phi to be the standard collar push; then M minus Phi(Phi(M)) is a connected collar of dimension at least 2, the constant configuration can be moved into R inside it, and Phi o Phi_t has image in M minus closure(R).

None after the repair.

Located concern

thm:ConfMFIsharpfg statement

Typo: 'finitely generated in dimension <= i' should read 'degree <= i'.

Direct reading.

Replace 'dimension' by 'degree'.

None.

C4

Checked statement: thm:intro:openbetti: for M the interior of a compact manifold with nonempty boundary, each of (1) the rational Betti number b_i, (2) the mod-p Betti number, (3) the minimum number of generators of H^i(Conf_n(M);Z), (4) the minimum number of generators of the p-torsion part of H^i(Conf_n(M);Z) is a polynomial in n for all n>=0, of degree i (dim>=3) or 2i (dim 2).

Returned assessment: claim_false

Items 1, 2 and 4 follow from thm:ConfMFIsharpfg and the classification; item 3 is false as stated, and the statement also needs the hypotheses connected, oriented, dimension at least 2 and 'degree at most'. By th:charFIsharp, V=H^i(Conf(M);k) is a sum of M(W_j) with W_j=0 for j>i (resp. 2i), so V_n is a sum of binom(n,j) copies of W_j as k-modules. Items 1 and 2 (k=Q, F_p): dim V_n = sum_j dim(W_j) binom(n,j) for all n>=0. Item 4 (k=Z): the number of generators of the p-primary torsion equals the number of p-primary cyclic summands, which is additive over direct sums, giving sum_j t_{j,p} binom(n,j) for all n>=0. Item 3: the minimal number of generators of a finitely generated abelian group A is the maximum over primes p of dim(A tensor F_p), which is not additive over direct sums (Z/2+Z/3=Z/6). So for V_n it is the maximum over p of the polynomials P_p(n)=sum_j (r_j+t_{j,p}) binom(n,j): eventually a polynomial, but not for all n>=0 unless one prime dominates everywhere. The deduction 'just as in the proof of thm:polyFIsharp' gives only the upper bound sum_j c_j binom(n,j). The counterexample under issues satisfies every hypothesis. Items 2 and 4 inherit the conditional status of the F_p and Z cases of thm:ConfMFIsharpfg.

Premises used

  • thm:ConfMFIsharpfg for k=Q, F_p, Z
  • th:charFIsharp (V is a sum of M(W_j), W_j=H_0(V)_j) and eq:MWdecomp
  • thm:polyFIsharp (iii) implies (vi)
  • Structure theorem for finitely generated abelian groups; Kunneth and universal coefficient theorems over Z (classical)
  • For the counterexample: existence of regular neighbourhoods of finite complexes in Euclidean space, and H^i(Conf_n(M);Z)=H^i(M^n;Z) for i<=d-2 because the fat diagonal has codimension d (also visible from the Leray spectral sequence, whose rows 0<q<d-1 vanish)

Missing inputs

  • Inputs listed under C3 for the F_p and Z cases of thm:ConfMFIsharpfg

Scope changes

  • Items 1, 2, 4 hold for all n>=0 for connected oriented M of dimension at least 2 (items 2 and 4 conditional on the F_p and Z cases of thm:ConfMFIsharpfg), with degree at most i or 2i.
  • Item 3 holds only for sufficiently large n in general.

Located concern

thm:intro:openbetti item 3

False as stated: the minimum number of generators of H^i(Conf_n(M);Z) need not be a polynomial in n for all n>=0.

Let X=RP^2 wedge (S^2 union_3 e^3), so reduced H_1(X;Z)=Z/2, H_2(X;Z)=Z/3, all else 0. Embed X in R^7 and let M be the interior of a regular neighbourhood: connected, oriented, 7-dimensional, interior of a compact manifold with nonempty boundary, homotopy equivalent to X. For i=3<=d-2=5, H^3(Conf_n(M);Z)=H^3(X^n;Z). By Kunneth, H_2(X^n;Z)=(Z/3)^n + (Z/2)^{binom(n,2)} and H_3(X^n;Z) is torsion, so H^3(X^n;Z)=(Z/3)^n + (Z/2)^{binom(n,2)}. Its minimal number of generators is max(n, binom(n,2)) = 0,1,2,3,6,10,15,... for n=0,1,2,... This equals binom(n,2) for n>=3; the only polynomial with those values is binom(n,2), which gives 0 at n=1 and 1 at n=2, not 1 and 2.

Replace 'for all n>=0' by 'for all sufficiently large n' in item 3 (the maximum of finitely many polynomials is eventually one of them), or replace item 3 by exact-for-all-n statements: the rank and each p-primary generator count (item 4) are polynomial for all n>=0, and the minimal number of generators is the maximum over p of rank plus p-count.

Item 3 fails in the stated range; items 1, 2, 4 are unaffected.

Located concern

thm:intro:openbetti hypotheses

Omitted hypotheses: the statement does not say connected, oriented, or dimension at least 2; connectedness and dimension at least 2 are essential.

M = disjoint union of two open discs (interior of the compact manifold D^2 disjoint union D^2): Conf_n(M) is the disjoint union over subsets S of Conf_S(R^2) x Conf_{n-S}(R^2), so b_0=2^n. M=(0,1): b_0(Conf_n)=n!.

State 'connected, oriented, of dimension at least 2' as in thm:ConfMFIsharpfg.

Literal statement false without connectedness; fine with the standing hypotheses of Section 6.

Located concern

thm:intro:openbetti, parenthetical 'of degree i if dim M>=3, and of degree 2i if dim M=2'

Should be 'degree at most'.

M=R^3, i=1: H^1(Conf_n(R^3);Q)=0 for all n, so the polynomial is 0.

Write 'degree at most i (resp. 2i)'.

Cosmetic.

Located concern

thm:intro:openbetti items 1 and 2

Notation typos: 'dim_Q Conf_n(M;Q)' and 'dim_{F_p} Conf_n(M;F_p)' should be dimensions of H^i(Conf_n(M);Q) and H^i(Conf_n(M);F_p).

Direct reading.

Insert H^i.

None.

Outside the audit

  • Totaro, Topology 35 (1996): statement, coefficient generality and naturality of the E_2 description
  • Church, Invent. Math. 188 (2012), Section 3.3 and Theorems 1, 5, 2.8: decomposition of E_2, earlier stable ranges, monotonicity
  • Forster 1964, Satz 1
  • Dold-Thom, Barratt-Priddy-Quillen and Sagave-Schlichtkrull statements quoted in remark:FIspaces
  • Sections 2-4 beyond the specific statements used (in particular pr:narrow, pr:stabdegstabrange, th:FIequiv and the full proof of thm:persinomial were only spot-checked as dependencies)
  • Sections 5 and 7 and all other introductory claims
  • Journal full text (not collated with the arXiv v4 TeX)
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}
}