By OpenAISourceGitHub

Manuscripts from openai/math

Fork
Clone
git clone https://openversions.com/git/index/openai-math.git
Cite
Version link
https://openversions.com/index/openai-math/commit/dfd8c46870b48511e697cbb43a3c8c91d375fa2f
Version identifier (commit hash)
dfd8c46870b48511e697cbb43a3c8c91d375fa2f

Fork to create your own copy with credit to the source. Clone to get a Git copy on your computer.

All papers

Claims

Findings

Rational homology and Quillen's conjecture

For every finite group GG and every prime pp, if Op(G)=1O_p(G) = 1, then H~∗(Ap(G);Q)≠0\widetilde{H}_*(\mathcal{A}_p(G); \mathbb{Q}) \neq 0, where reduced homology is augmented so that H~−1(∅;Q)=Q\widetilde{H}_{-1}(\varnothing; \mathbb{Q}) = \mathbb{Q}.

Let WW have dimension h≥2h \ge 2 and one of the four permitted structures (no-form over Fu\mathbb{F}_u, hermitian over Fu2\mathbb{F}_{u^2}, or symmetric over Fu\mathbb{F}_u with uu odd and either both or one line type allowed), and suppose WW admits an allowed frame. Then Z(W)\mathcal{Z}(W) consists of cycles of degree h−2h-2 supported on full flags ending at sign groups of rank h−1h-1, and dim⁡Z(W)≥Fh∣F(W)∣(h−1)!\dim \mathcal{Z}(W) \ge F_h |\mathfrak{F}(W)|(h-1)! with explicit positive constants FhF_h (in particular Fh≥(17/25)hF_h \ge (17/25)^h for u≥5u \ge 5 in the first three cases, and F3≥3/8F_3 \ge 3/8 in the two-type symmetric case).

Let pp be an odd prime, qq odd with p∣q+1p \mid q+1, and L=PSUn(q)L = \mathrm{PSU}_n(q) with n≥5n \ge 5. If L≤H≤Aut(L)L \le H \le \mathrm{Aut}(L) and H/LH/L is elementary abelian of pp-power order, then H~mp(H)−1(Ap(H);Q)≠0\widetilde{H}_{m_p(H)-1}(\mathcal{A}_p(H); \mathbb{Q}) \neq 0.

Let GG have least order among finite groups with O2(G)=1O_2(G) = 1 and H~∗(A2(G);Q)=0\widetilde{H}_*(\mathcal{A}_2(G); \mathbb{Q}) = 0. Then GG has no simple component isomorphic to PSLn(q)\mathrm{PSL}_n(q) or PSUn(q)\mathrm{PSU}_n(q) with n≥5n \ge 5 and qq of characteristic at least five.

A minimal-order counterexample to (2) at p=2p = 2 cannot have a simple component isomorphic to PSp2n(q)\mathrm{PSp}_{2n}(q), where n≥3n \ge 3 and qq has characteristic at least five.

A minimal-order counterexample to (2) at p=2p = 2 has no simple component L=PΩ(V)L = \mathrm{P}\Omega(V), where VV is a nondegenerate symmetric space over Fq\mathbb{F}_q of characteristic at least five and dimension NN is odd with N≥5N \ge 5, even with N≥8N \ge 8, or N=6N = 6, except possibly when N=6N = 6, q=5q = 5, and the determinant of the form is a square.

A minimal-order counterexample to the rational homological implication at p=2p = 2 has no simple component isomorphic to PSL4(5)\mathrm{PSL}_4(5).

Under the inductive hypothesis that every proper subgroup T<GT < G with Op(T)=1O_p(T) = 1 satisfies H~∗(ΔAp(T);Q)≠0\widetilde{H}_*(\Delta\mathcal{A}_p(T); \mathbb{Q}) \neq 0, if AA is a faithful configuration for a simple component LL of GG (p∣∣L∣p \mid |L|) satisfying the maximality conditions (i) and (ii) and there is a cycle α∈C~r−1(ΔAp(LA);Q)\alpha \in \widetilde{C}_{r-1}(\Delta\mathcal{A}_p(LA); \mathbb{Q}) with nonzero coefficient on a saturated flag ending at AA, then H~∗(ΔAp(G);Q)≠0\widetilde{H}_*(\Delta\mathcal{A}_p(G); \mathbb{Q}) \neq 0.

Let MM be a finite group, H≤MH \le M, and pp a prime. If α∈C~m−1(Ap(M))\alpha \in \widetilde{C}_{m-1}(\mathcal{A}_p(M)) is a cycle supported on full flags ending at rank-mm elementary abelian groups, and E∈E(α)E \in \mathcal{E}(\alpha) achieves the maximum index [E:E∩H]=pc[E : E \cap H] = p^c (0≤c<m0 \le c < m) among all supported top groups, then there is a nonzero cycle β∈C~m−c−1(Ap(H))\beta \in \widetilde{C}_{m-c-1}(\mathcal{A}_p(H)) supported on full flags of rank m−cm-c with a nonzero coefficient at a full flag ending at E∩HE \cap H.

Methods

Homology propagation from a faithful component extension H=LAH = LA to GG by deleting faithful elementary overgroups of AA, taking the product shuffle α⊠β\alpha \boxtimes \beta with a nonbounding cycle in CG(H)C_G(H), and detecting nonbounding via the saturated residue RσR_\sigma.

Source excerpts

“We first delete certain vertices without changing the homotopy type, then construct a cycle in the remaining poset and detect it by Lemma 3.4.” · Show in paper

Construction of frame cycles by imposing plane relations on coned apartment chains and bounding the resulting coefficient-space dimension via multilinear Lie superbrackets, supporting-subspace-graded PBW, and a marked-factor inequality on normalized formal series.

Source excerpts

“The proof has three stages. Boundary cancellation produces the cycles, and merger coefficients identify the space whose dimension must be estimated. The supporting-subspace grading then permits a PBW count without forgetting where a coefficient is supported. Finally, the relation kernel and its marked-factor inequality reduce the estimate to explicit plane counts.” · Show in paper

Decoration of frame cycles by adjoining commuting elementary generators (ordinary or quaternion four-group) and cancelling new boundary faces via linear equations bounded by partial-datum completion counts.

Source excerpts

“First, we adjoin commuting elementary generators and cancel the additional boundary faces. Second, we pass from a group of automorphisms to a specified subgroup while retaining a nonzero full-flag coefficient.” · Show in paper

Galois / semilinear descent of compatible line or plane frames for involutory or order-pp field and diagram automorphisms.

Source excerpts

“We use the following fixed-vector form of Galois descent. Let K/k be a cyclic finite-field extension of degree d, with generator σ, and let B be an additive map on a finite-dimensional K-space V satisfying B(av) = σ(a)B(v) and Bd = id.” · Show in paper

Equivalent-poset construction for a centerless component (PSL4(5)\mathrm{PSL}_4(5)) when all faithful elementary 2-extensions have nontrivial centralizer 2-core, using Mayer–Vietoris and mod-5 Euler characteristic to inject join homology into H~∗(A2(G);Q)\widetilde{H}_*(\mathcal{A}_2(G); \mathbb{Q}).

Source excerpts

“Apply Theorem 9.3 to this centerless component of even order. In its notation, no point of F is comparable with a pure L-point (U, 1).” · Show in paper