Representation theory and topology

Representation theory and homological stability

Thomas Church and Benson Farb

91 pages66 claims indexed80 referencesManuscript 2013-06-10

gpt-6.2-astra max: cf-lie-challenges

Reviewer record, not the publication verdict. The full response includes unaccepted proposals and source limitations. The curated report states what was adopted. Initial packets used a conservative 2010 cutoff; focused packets used the verified 2012 journal received date. The focused Fable answer needed a metadata-only derivative; its original failure and partial source access remain recorded.

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.

The two main defects have different statuses. The coefficient problem is resolved by making explicit the morphisms that the chain square presupposes; without that interpretation even dimensions can alternate. The simple-stability problem persists with perfectly functorial coefficients and the canonical inclusions: an entire new Schur type cannot be generated by zero. The finite uniform strong theorem (and the multiplicity-only symplectic deduction) has a viable, degreewise proof on its intended inputs. The all-rank polynomial conclusions survive in a changed sense, with explicit low-rank hypotheses or arguments, not with printed Condition II. This is a non-blind adjudication of the supplied proposals. Historical cutoff used: 2012-01-19, journal_received, from the supplied recorded first-page date evidence, not the Sol fallback or the 2013 publication year. Model knowledge cannot literally be time-filtered; no later theorem is used to close a proof and no external search or computation is claimed.

LC1

Checked statement: Whether the explicitly listed conditions on coefficient maps in Definition 5.2 suffice for Theorem 5.3, in particular (1) ⇒ (5).

Returned assessment: claim_false

There are two readings of “graded L_n-modules with maps.” If these are morphisms of modules over the Lie inclusions (by restriction of scalars), the objection to functoriality is defeated. This is a natural intended reading, supported by the use of the chain square. But the list of conditions on p.286, read as conditions on linear maps, does not imply that reading. Equivariance of a G_n-action with respect to a module action at rank n is not compatibility of two module actions at different ranks. On the linear-map reading the problem is not just a failure to specify which maps to take on homology: the asserted stable multiplicities can be false.

Premises used

  • All objects in the test are over Q. Strong stability of a coefficient grade includes consistency as a sequence of group representations; that does not test a Lie-module equation.
  • The degree-zero chain group is M_n and the image in it is L_n M_n, by the displayed differential. No external result is needed.

Missing inputs

  • An explicit module-morphism convention for the inter-rank maps, if this is to be inferred rather than added.

Scope changes

  • The negative verdict is for the explicit linear-map hypotheses. If “maps” already means morphisms of the indicated pairs, the alternating-action example is inadmissible.
  • This defect does not apply to the symplectic, trivial-coefficient, multiplicity-only statement of 5.4.

Located concern

CF-lie 2505–2509 versus 2684–2696; journal-lie-p37, Definition 5.2

The chain square needs a hypothesis on the pair of transition maps.

The Sol test works. Take the abelian L_n=Qz in degree 1, M_n^0=Qu, M_n^1=Qv, with all transitions identities on the indicated spaces, all group actions trivial. Let zu=v at even n and zu=0 at odd n, and let zv=0. This is a graded Lie-module action at each n, equivariant at that n. The Lie piece and both nonzero coefficient pieces are strongly stable, with type-preserving maps. Yet H_0(L_n;M_n)^1=Qv/(Qz)(Qu) is 0 at even n and Q at odd n. In particular the identity transitions on the coefficients do not commute with the differential.

Take i_n:L_n→L_{n+1} to be a degree-preserving, G_n-equivariant Lie homomorphism (injective as in 5.1), and f_n:M_n→M_{n+1} a degree-preserving, G_n-equivariant linear map with f_n(xm)=i_n(x)f_n(m). Use the larger group action restricted along G_n→G_{n+1}; require the group actions at each rank to respect the bracket and the module action as at 2454–2458 and 2471–2474. Then Λ^i i_n ⊗ f_n commutes term by term with the bracket part and the action part of the differential. These are graded equivariant chain maps and induce the asserted graded equivariant homology maps. This does not by itself make those maps injective.

Literal insufficiency, with a counterexample to the universal coefficient clause; a repairable missing categorical specification, not a counterexample with compatible coefficient maps. Trivial coefficients and the natural adjoint coefficients already have the compatibility.

LC2

Checked statement: The all-rank simple-stability claim in Theorem 5.6 and its application to the actual maps of Corollaries 5.7–5.9, using Definition 5.5 as printed.

Returned assessment: claim_false

The journal image says “for all n ≥ 1 it satisfies Conditions I and II.” II is generation of the entire target by the larger-group orbit of the image. Its quantifier is not the one in SIII, which starts separately at the row length of each partition. Below that length the Schur space is zero, not a one-dimensional trivial representation. Read SIV, charitably, as a test on actual irreducible copies in the source. If a positive-multiplicity shape first exists at rank r>1, SIV sends each old irreducible at r−1 into a copy of its own polynomial continuation. The sum of their spans misses the new shape. In fact, starting at rank one, all-rank II plus SIV precludes all such births. The decomposition at 2782–2787 is not equivalent to Definition 5.5. The criticism in prior-lie is sustained; the part of sol-lie concerning canonical Schur maps cannot vindicate literal II, as its detailed argument also recognizes.

Premises used

  • Simple stability uses polynomial S_λ(Q^n) and excludes negative types. It is not the result of moving the threshold in strong stability down to 1 while retaining the same GL labels.
  • In the usual §2 eventual sense of Condition IV, E_n=Q^n with the standard inclusions is type-preserving (in particular for n≥2); it satisfies polynomial SIV at every rank. The trivial coefficient in degree 0 satisfies even the printed simple definition and has compatible maps.
  • No external homology computation is needed for any of the witnesses. Tirao is an attribution, not a premise of the disproof: available_from=2002; date_evidence=CF-lie 4653–4654; date_basis=journal publication year; version=J. Lie Theory 12, 309–323, Theorem 2.9 as cited at 2885–2892 for free k-step nilpotent Lie algebra homology with trivial coefficients. The antecedent's own exact statement about maps has not been supplied.

Missing inputs

  • None for the rank-1-to-2 tests. The primary Tirao paper would be needed to audit the attribution or the decomposition table independently.

Scope changes

  • If “type-preserving” in 5.6 is instead strengthened to §2 constant-tail IV at every rank, E_1 has signature (1), whose constant-tail continuation is det of GL_2, not E_2. The abelian example is then not an example under that changed hypothesis. That reading also fails to provide the advertised general interface for polynomial examples at their boundary ranks. It does not change the direct falsifications of 5.7–5.9. The same distinction applies to det of GL_2=Λ²E_2 continued to GL_3 versus the polynomial Λ²E_3.
  • Read in Schur notation, the table on p.291 itself illustrates the problem: S_(4,2) at rank 2 continues under SIV to S_(4,2) at rank 3, not to a span including the new three-row types. This observation is conditional on the displayed decomposition; its correctness is not needed for the counterexamples.
  • The results at issue are over Q. No integral or change-of-coefficients stability conclusion follows.

Located concern

CF-lie 2765–2779, 2791–2807, 2856–2860, 2888–2893, 2931–2935

One all-rank orbit-generation defect affects the theorem, its asserted closure under Schur operations, and all three corollaries.

For 5.6 take the abelian L_n=E_n in internal degree 1 with its standard group action, M_n=Q trivial. Condition (1) of the simple analogue holds. The differential is zero and H_2(L_n;Q)^2=Λ²E_n. The map 0=Λ²E_1→Λ²E_2≠0 is injective but cannot satisfy II. Thus even (1) ⇒ (2) fails, without using the coefficient-map omission. For 5.7 take m=2: the degree-two free Lie words give Λ²E_n (the bracket is alternating, and the two-step algebra E_n ⊕ Λ²E_n with bracket v∧w realizes all these brackets). For 5.8 take k=1, i=2: the truncation is abelian and gives the same exterior square. For 5.9 take k=1, i=1: the adjoint action is zero, so H_1=E_n ⊗ E_n. At 1→2 the orbit of e_1⊗e_1 spans Sym²E_2 by polarization, not the alternating summand. These use the prescribed inclusions, not pathological maps.

Allow the first nonzero evaluation of a Schur type. Ask for orbit generation on types existing at the source, not generation of a summand by a zero source, and use polynomial continuations in the map condition. State separately any eventual full-span result. See LC3 for the proof obligations of such a change.

The literal conclusions are false; I is not the failing condition in these witnesses. Constant Schur multiplicities in the nonvanishing range, compatibility of the old copies, and eventual strong stability in these bounded-degree examples can all survive. These are not three independent errors in the computations of Lie homology.

LC3

Checked statement: Validity and scope of the intended eventual strong/uniform proof and of a replacement all-rank theorem allowing Schur births, including the reverse coefficient inference and SIV versus IV′.

Returned assessment: repairable_gap

In the eventual, degreewise finite-dimensional setting, the compatible-coefficient five-way equivalence has a repair, relative to the finite uniform representation-theory inputs below. The reverse proof is a character calculation; the forward proof also needs a proof about maps on a quotient. Theorem 5.4 is only a reverse character calculation. Neither needs any pointwise-to-uniform compactness principle. For an all-rank replacement, let B mean polynomial SIII, injectivity and polynomial SIV for old copies, with no demand to span a shape at its birth from the preceding rank. There is a forward homology theorem for B, and the canonical examples satisfy B. The reverse proofs for B separate into recovering characters and controlling the Lie maps. Trivial and adjoint coefficients require no tensor inversion. A general coefficient requires a detection argument at each tested rank. Even after recovering all the characters one needs all-rank control of the Lie maps in the polynomial convention, not just eventual IV. The repairs below specify where one can prove these assertions; they are not a validation of the printed sketch.

Premises used

  • Throughout the repairs the representations in a fixed Lie or coefficient grade are finite-dimensional split rational algebraic representations in characteristic zero, semisimple with the highest-weight labelling of the manuscript. No assertion about arbitrary abstract linear actions of GL_n(Q) is needed. Lie grades are strictly positive, coefficient grades nonnegative, and the maps of pairs are as in LC1.
  • Internal operation inputs: CF-lie 1367–1401, Theorem 3.1, uniform finite-dimensional tensor products and fixed Schur functors for SL, GL, Sp, and the strong versions of these operations for SL and GL with strong inputs. Only finite sums, products and exterior powers in each fixed slice are needed. For the cancellation repair it suffices that products of each of finitely many pairs of fixed labels have rank-independent finite expansions at sufficiently large rank. The §3 proofs of the operation statements are not in the supplied lines. In particular the Sp conclusion here is conditional on the uniform symplectic operation input, not on an independently checked plethysm formula or range.
  • Internal branching input: multiplicity one of the matching smaller type in restriction, CF-lie 1116–1134, for the SL/GL strong argument; absolute simplicity gives scalar endomorphisms (an endomorphism is determined on the generating highest line). The Cartan constituent in a tensor product comes from the tensor of highest vectors. The finite-representation character-domain argument is at 1866–1873: by highest weights the character is injective and the weight-lattice group ring is a Laurent polynomial domain.
  • Classical background used in the polynomial repairs: the Schur/highest-weight model over characteristic zero. For E of dimension n, a polynomial simple of degree d is S_λ(E), |λ|=d, obtained by a Young projector on E^{⊗d}, zero for ℓ(λ)>n and otherwise irreducible. Equivalently evaluating the Specht module of λ on E^{⊗d} over Q S_d gives this Schur space. Finite-dimensional rational GL highest weights are the weakly decreasing integral signatures; polynomial ones have nonnegative last entry. These are background inputs, not new stability theorems. External background record: available_from=1991; date_evidence=CF-lie 4475–4477; date_basis=publication year of the cited book; version=Fulton–Harris, Representation theory. A first course, GTM 129, 1991 (classical Schur functor and highest-weight theory; one may extend scalars on the split Q-forms). The book pages were not inspected. No later edition or later functor-category theorem is used.

Missing inputs

  • Full proofs of the quoted §3 operation/branching inputs and their external formula pages are not supplied. The finite uniform and strong inputs used here are much narrower than a blanket endorsement of §3.
  • An amended all-rank statement must specify the coefficient class grade by grade, the small-rank map condition on the Lie algebra, and the treatment of ranks/grades with zero coefficients. Merely replacing the word “strongly” or deleting II does not specify these.

Scope changes

  • Ordinary representation stability includes eventual map conditions and pointwise multiplicity limits; uniform multiplicity stability alone has no map conditions. Strong stability adds type control to uniform representation stability. In the finite slices the uniform thresholds and finite multiplicities are essential to the strong multiplicity-complex proof.
  • The B proofs are new arguments for a specified modified property. A natural polynomial coefficient functor with compatible action and canonical maps is one useful source of the needed all-rank structure, not an assumption implicit in arbitrary sequences of group modules. Homogeneous polynomial degree control on a Lie map is also useful: for equal total polynomial degrees the old central weight forces its image into last-coordinate weight zero. With SIII and injectivity that gives the whole required zero-weight part.
  • No full external proof of the manuscript's §3 inputs, or of any cited antecedent, is certified by these deductions.

Located concern

CF-lie 2550–2659, 2733–2753 and 1855–1885

Which reverse implications actually need cancellation, and why the printed projection argument is not enough.

For example, the claimed multiplicative projection for SL at 3→2 kills the two-row column in E_3⊗E_3=Sym²E_3 ⊕ Λ²E_3, whereas E_2⊗E_2 contains Λ²E_2=1. Killing the column is not multiplication in the lower-rank ring. This invalidates that proof step, not finite uniform cancellation itself.

Use C_i^d=(Λ^i L ⊗ M)^d, the direct sum over q+Σ p r_p=d, Σ r_p=i, of (⊗_p Λ^{r_p}L^p) ⊗ M^q. Thus there are no doubled ordered off-diagonal pairs in C_2. The complex is finite in a fixed d and Σ(−1)^i[C_i^d]=Σ(−1)^i[H_i^d]. For (4) ⇒ (1) take the least a with M_n^a eventually nonzero; strong stability in the finitely many grades below a makes them eventually zero. In degree a+k the only unknown involving L^k is L^k⊗M^a in C_1; all other nonzero summands use smaller Lie grades. Uniform stability of the other terms by induction and of the homology determines uniform multiplicities of this product by subtraction. New finite-uniform cancellation argument: let W=M^a be eventually nonzero, X=V⊗W uniformly multiplicity stable, and W uniformly multiplicity stable. If X is eventually zero, so is V. Otherwise fix a label β occurring in stabilized W. For every constituent α of V, α+β occurs in X. The stabilized support of X is finite, so there are only finitely many candidates α, independently of rank: in SL normalize to last entry zero; in GL write a signature as a zero-last-entry partition plus t(1,…,1), where t is the repeated tail. Both the partition part and t of α are determined by subtracting β from a candidate sum. On these finite lists use fixed-label stable product coefficients. Transport the equations for any two sufficiently large ranks to one common large rank by the stabilized labels. Their right sides agree and the factor [W] is the same and nonzero. Cancellation in the character domain there gives identical multiplicities of V. This proves the needed finite uniform lemma without the projection. Finally, assumed IV on L, with finite equal multiplicities and matching restriction multiplicity one, gives strong stability: the inclusions of matching multiplicity spaces are isomorphisms. For (2) ⇒ (1) the factor is just Q. For (3) ⇒ (1) in total degree k the unknown L^k is C_0, and all C_i for i>0 use strictly smaller Lie grades, including the adjoint coefficient grade. In degree 1 it is H_0^1=L^1, not H_1^1. Euler subtraction suffices. The same unit-coefficient calculation for Sp proves precisely 5.4, using only uniform multiplicities in the operations, no transition-map conclusion.

A complete reverse deduction given the stated uniform operation inputs; only the general one-coefficient implication needs nontrivial uniform cancellation. This repair does not prove nonuniform cancellation or cancellation with infinite multiplicities. No converse symplectic homology theorem with maps has been obtained.

Located concern

CF-lie 2661–2730, especially 2698–2705 and 2717–2729; journal-lie-p40 p.289; cf. Proposition 2.14 at CF-lie 1162–1174

Passing IV′ to a homology image does not establish injectivity of the induced map, an assumption of the invoked criterion.

Commuting inclusions of complexes only guarantee that cycles go to cycles and boundaries to boundaries; an old nonboundary could in principle become a boundary. Also [C_i]=[ker ∂_i]+[im ∂_i] is an identity at one rank, not a functorial splitting, and H_i is ker ∂i / im ∂{i+1}, not the quotient with both indices shifted on p.289.

New chain-map argument. Fix total degree d. Strong closure for the finite constructions gives strong chain terms when L and M are degreewise strong. Take one threshold for the finitely many terms in this slice, not for the entire graded algebra. Write each chain term as ⊕λ S{λ,n} ⊗ A^i_{λ,n}, with S denoting the appropriate group simple. Fix a matching inclusion u_{λ,n}:S_{λ,n}→S_{λ,n+1}. IV confines each old isotype to the matching isotype, and matching branching multiplicity one makes the block map u_{λ,n} ⊗ T^i_{λ,n}. Injectivity forces T injective; finite constant multiplicity makes T an isomorphism. At each rank the differential on λ is 1 ⊗ D^i_{λ,n}. By the chain square, the T's intertwine D, hence are chain isomorphisms on the multiplicity complexes. On homology the block map is u_{λ,n} tensored with the induced isomorphism there. This proves injectivity, orbit generation, uniform multiplicity constancy and IV, including for arbitrary diagonal copies of a simple. It also works directly for adjoint coefficients under (1), without calling an identically zero L an admissible nonzero module.

With compatible coefficients this gives (1) ⇒ (5), and (1) ⇒ (3). Taking the admissible unit Q gives (5) ⇒ (2), (2) ⇒ (4); together with the reverse deductions this closes the intended 5.3, relative to the indicated representation inputs. Injectivity on homology is a result of the argument, not a consequence of the word “subquotient.”

Located concern

CF-lie 1154–1215 versus 2797–2804; journal-lie-p41 p.290, first proof paragraph

The eventual highest-weight test and an all-rank polynomial test are different. The asserted all-rank SIV ⇔ IV′ is false.

The Sol example passes the other tests in the assertion. Set F_n=E_n ⊕ Λ²E_n with canonical maps except at 1→2, where e_1↦(e_1,e_1∧e_2). The two components both have GL_1-weight 1. This is injective; its GL_2-span is both summands (they are inequivalent simples and both projections are nonzero). Later the canonical images span. SIII holds; P_2 fixes both components and the later right-column groups fix the canonical images. Nevertheless the rank-one copy of E_1 does not generate just E_2. This is a failure of IV′ ⇒ SIV even with I and the printed II, separate from the obstruction to II in LC2.

For the eventual criterion in the finite uniform setting, decompose the image of the given highest vector into larger-torus weight components after observing that it is killed by all positive roots under IV′. Its nonzero components are larger highest vectors, not arbitrary weights. For sufficiently large rank in the stabilized finite support only the matching highest label restricts to the given highest weight (use a long constant tail for GL and a normalized zero tail for SL). The image therefore generates one matching copy. Conversely the matching branch contains the larger highest line, and P, normalized by G_n, fixes its smaller-group span. At small rank distinct polynomial highest labels can have the same restriction. Use polynomial SIV directly, or add control of total polynomial degree: for polynomial highest weights a prescribed first n entries and the same total degree force the appended entry to be zero. Internal Lie grading alone is not such a degree hypothesis.

The finite-support eventual use of Proposition 2.14 can be justified. Its proposed all-rank extension cannot be used to upgrade a polynomial multiplicity calculation to the required map statement.

Located concern

CF-lie 2770–2787, 2797–2822, 2831–2840, 2876–2893, 2927–2935; prior-lie 306–308; sol-lie L2

What the birth-allowing repair proves in the forward direction and for the applications.

The slot-module proposal gives no source vectors at all in a newly nonzero Schur block. Its reasoning about evaluation of a characteristic-zero multilinear complex supports the polynomial conclusion, not generation of that block one rank earlier.

New direct proof for B. In a polynomial module at rank n+1 take the weight-zero subspace for scaling the last coordinate, denoted F[0]. For S_λ(E_{n+1}) it is precisely S_λ(E_n), possibly zero: the Young projector preserves the number of last-coordinate inputs in tensor words, and the zero-count tensors are exactly E_n^{⊗|λ|}. In a B sequence the old image equals F_{n+1}[0]. Indeed SIV puts each old copy in its matching polynomial type, and the smaller central-scalar degree |λ| inside that matching degree-|λ| module forces last-coordinate count zero. I and SIII give equality by dimension in each old isotype. Conversely, an equivariant identification with this zero-weight space gives SIII and SIV at the step: each old isotype there comes only from the matching larger type, via its canonical branch tensored with a multiplicity space. This also explains why generation on old types needs no extra surjectivity assumption. All last-coordinate weights are nonnegative, so the zero-weight part of a tensor product or exterior power is the same construction on the zero-weight parts. For compatible B inputs grade by grade, the old fixed-degree chain complex is thus identified with the zero-weight subcomplex of the new one. Differentials preserve weights, so its homology is precisely the zero-weight part of new homology. This proves (1) ⇒ (5) for B coefficients, and works for Q and for the adjoint under (1). For the canonical examples there is a direct check of the input: degree-m free Lie words use m vector inputs, and the defining relations preserve the count of the new basis letter. The zero-count words are just the old free Lie piece, included injectively by the linear retraction killing the new generator. These are polynomial representations (quotients of sums of tensor powers). The same check holds for truncation by all degrees >k, the ideal interpretation of 2879. Apply the chain argument with trivial or adjoint coefficients. At fixed i there are only degrees ≤ki or ≤k(i+1), respectively, so the conclusion holds for the finite sum over degrees. Equivalently, numbering the input slots gives the multilinear word complexes in sol-lie; evaluation against tensor powers is exact since finite symmetric-group module splittings can be averaged over Q.

This proves constant Schur multiplicities as soon as defined, injectivity and generation of the matching type by each old copy for 5.7–5.9. At ranks beyond the bounded degrees all the shapes have appeared and their polynomial continuations agree with the eventual GL labelling, giving strong stability there. It does not prove printed simple stability. For an arbitrary birth-allowing sequence one must not infer a common threshold without a support bound.

Located concern

CF-lie 2505–2509, 2576–2647, 2804–2812; journal-lie-p41 p.290, appeal to the proof of Theorem 3.5

The reverse all-rank argument needs both low-rank detection and a separate hypothesis or argument about the Lie maps.

The SIII cancellation assertion as worded at 2804–2807 has a small-rank counterexample. Take U_n=Λ²E_n, V_1=Q and V_n=Q² trivial for n≥2. Then V_n⊗U_n=U_n^{⊕2} at every rank, including rank 1. Both the product and U satisfy SIII, but the trivial multiplicity in V changes. U_1=0: this is not a nonzero-at-every-rank tensor factor. Nor is U alone an admissible total coefficient under 5.2. It is a possible individual coefficient grade, so it defeats an unqualified inference using just the eventually nonzero grade. It does not refute every corrected compatible-coefficient equivalence.

New all-rank deductions, with hypotheses. Work with finite-dimensional pieces and B for the coefficient and homology conditions, requiring polynomial SIII grade by grade. For maps in a reverse conclusion assume the Lie inclusions satisfy the matching polynomial copy condition at every rank once the pieces are polynomial, or supply it from functorial/degree control. For (2) ⇒ (1) and (3) ⇒ (1), the unit and adjoint Euler subtractions above work at every rank: polynomial classes specialize by setting discarded torus variables to zero, sending S_λ to the smaller Schur space or zero. This specialization commutes with tensor and exterior operations by the zero-weight calculation. Subtraction therefore recovers compatible polynomial classes [L_n^k] inductively, which is SIII; the separate map input gives B. For (4) ⇒ (1) a sufficient single-grade hypothesis for the original style of induction is that the bottom coefficient grade is nonzero at every tested rank. Once V_m⊗U_m is known polynomial with SIII and U has SIII, choose at large m a constituent μ of U already present at a smaller rank, hence μ_m=0 in its last coordinate. The Cartan constituent α+μ forces every α of V_m to be polynomial. Specialize the large-rank product identity to rank n and cancel [U_n] in the rank-n domain, valid if U_n≠0. This recovers the specialization of V_m even at small n. Without the nonvanishing, this particular proof stops. A stronger new repair uses the whole coefficient if, as in 5.2, M_n≠0 at every rank; it need not be the same grade that detects every rank. Here are the details so this is not an assumption of all-rank tensor inversion. Let a be the smallest grade occurring anywhere in M, and choose a shape in M_b^a at some rank b. SIII carries it to every rank m>b with last coordinate zero. At such m, Euler subtraction in degree a+k, inductively in k, shows L_m^k⊗M_m^a is polynomial: all other terms and homology are polynomial, and the irreducible basis detects any nonpolynomial constituent. The Cartan test with the chosen shape shows L_m^k is polynomial. Now put M_n(t)=Σ_q[M_n^q]t^q and P_n(t)=∏{p≥1}(Σ{r≥0}(−1)^r[Λ^r L_n^p]t^{pr}). Slice-wise Euler gives M_n(t)P_n(t)=Σ_{i,d}(−1)^i[H_i^d(n)]t^d, with finite coefficients in each degree. For m>max(b,n), specialize this equation from m to n; all its coefficients at m are polynomial. M and the homology specialize correctly by SIII. Compare with the actual rank-n equation in R_n[[t]], a domain. Since M_n(t)≠0, cancel it to get P_n equal to the specialization of P_m. The coefficient of t^k is −[L^k] plus exterior operations in smaller grades. Induction gives [L_n^k]=specialization([L_m^k]) for every k. Comparing two ranks to a common m proves SIII, including at low ranks. With the stated all-rank polynomial map input and the Lie injections, this gives B. If a total coefficient were allowed to be zero at a tested rank, this detection argument would not work there.

Thus there are complete conditional replacements: forward for compatible B pieces; reverse on characters directly for trivial/adjoint, and for a detecting coefficient with the extra cancellation argument; reverse on maps with all-rank polynomial type control. A common everywhere-nonzero bottom grade is sufficient but is not logically necessary under the whole-graded argument. As before (5) gives (2) and (4) via Q, and zero adjoint cases can be treated directly. None of this promotes eventual constant-tail IV to all-rank SIV, makes merely strongly stable coefficient grades polynomial functors, or proves a five-way equivalence by deleting II alone.

Outside the audit

  • Omitted parts of §2–§3, including the full tensor, branching and plethysm proofs and external formula pages. Only their identified inputs, not their full proof histories, enter the conditional deductions.
  • The original proofs and precise formulations in Tirao and the other cited Lie-homology antecedents, and the numerical entries of the displayed H_3 table.
  • Other surface-group, Heisenberg, braid and modular applications; no extension from multiplicity statements to their stabilization maps has been audited here.
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 Representation theory and homological stability. Math Referee, .

Download .bib
View BibTeX
@misc{rivin2026review-human-church-farb-homological-stability,
  author = {Rivin, Igor},
  title = {{Review of Representation theory and homological stability}},
  year = {2026},
  month = {November},
  date = {2026-11-21},
  howpublished = {Math Referee},
  url = {https://mathreferee.dimensionreducers.ai/papers/human-church-farb-homological-stability/panel/},
  note = {Review report}
}