OctoInformBeta

OpenAI preprint, September 2026 · §5. Centrality and rational stability

Lemma 5.1: Reduction under ε\varepsilon characterizes invertibility, linearly independent reduced columns extend, and elementary and single diagonal matrices generate GL⁡ι(R)\operatorname{GL}_\iota(R)

Invertible matrices and general linear group generation over RR..

Hover over a blue or green term in the text, or a node in the graph, to find it in the other. Select it to read its declaration both in Lean and in words.

The Setting

Let 𝔽2\mathbb{F}_2 be the field of two elements, and let RR be an 𝔽2\mathbb{F}_2-algebra equipped with the augmentation ε:R→𝔽2\varepsilon : R \to \mathbb{F}_2, an 𝔽2\mathbb{F}_2-algebra homomorphism. Entrywise application of ε\varepsilon reduces matrices over RR to matrices over 𝔽2\mathbb{F}_2. Under a nilpotency hypothesis on ker⁡ε\ker \varepsilon, the question is how invertibility and column extension for matrices over RR are determined over 𝔽2\mathbb{F}_2, and what generators suffice for the general linear groups over RR.

Definitions

A ring homomorphism φ:S→𝔽2\varphi : S \to \mathbb{F}_2 has the matrix nil kernel property if for every nonnegative integer nn and every n×nn \times n matrix AA over SS satisfying φ(Aij)=0\varphi(A_{ij}) = 0 for all indices ii and jj, the matrix AA is nilpotent.✦

Given a ring SS and a finite index set ι\iota, the general linear group GL⁡ι(S)\operatorname{GL}_\iota(S) is the group of invertible ι×ι\iota \times \iota matrices over SS; when ι={0,1,…,n−1}\iota = \{0, 1, \dots, n - 1\} for a nonnegative integer nn, we write GL⁡n(S)\operatorname{GL}_n(S) for GL⁡ι(S)\operatorname{GL}_\iota(S).

The diagonal matrix of units diag⁡(d)∈GL⁡ι(S)\operatorname{diag}(d) \in \operatorname{GL}_\iota(S) associated with a ring SS, a finite index set ι\iota, and a family d=(di)i∈ιd = (d_i)_{i \in \iota} of units di∈S×d_i \in S^\times of SS is the invertible matrix over SS whose (i,i)(i,i)-entry is did_i for each i∈ιi \in \iota and whose off-diagonal entries are 00.✦ At an index ii in a finite index set ι\iota and a unit u∈S×u \in S^\times of a ring SS, the single diagonal matrix Di(u)∈GL⁡ι(S)D_i(u) \in \operatorname{GL}_\iota(S) is the diagonal matrix of units whose jj-th diagonal entry is uu for j=ij = i and 11 for j≠ij \neq i.✦

Whenever i≠ji \neq j are distinct indices in a finite index set ι\iota and aa is an element of a ring SS, the elementary matrix eij(a)∈GL⁡ι(S)e_{ij}(a) \in \operatorname{GL}_\iota(S) is the invertible matrix given by (1), eij(a)=I+aEij,e_{ij}(a) = I + a E_{ij},(1)✦ Here II denotes the ι×ι\iota \times \iota identity matrix over SS and EijE_{ij} denotes the ι×ι\iota \times \iota matrix with entry 11 in row ii, column jj and 00 elsewhere.

Over a ring SS and a finite index set ι\iota, the elementary and diagonal subgroup Gι(S)G_\iota(S) of GL⁡ι(S)\operatorname{GL}_\iota(S) is the subgroup (2) generated by the elementary and single diagonal matrices: Gι(S)=⟨eij(a),Dk(u)∣i,j,k∈ι,i≠j,a∈S,u∈S×⟩.G_\iota(S) = \langle\, e_{ij}(a),\, D_k(u) \mid i, j, k \in \iota,\; i \neq j,\; a \in S,\; u \in S^\times \,\rangle.(2)✦ For a nonnegative integer nn, when ι={0,1,…,n−1}\iota = \{0, 1, \dots, n - 1\} we write Gn(S)G_n(S) for Gι(S)G_\iota(S).

The Claim

Theorem 1. Let RR be an 𝔽2\mathbb{F}_2-algebra with augmentation 𝔽2\mathbb{F}_2-algebra homomorphism ε:R→𝔽2\varepsilon : R \to \mathbb{F}_2 satisfying the matrix nil kernel property. If ι\iota is a finite index set and AA is an ι×ι\iota \times \iota matrix over RR with entrywise image A‾=(ε(Aij))i,j∈ι\bar{A} = (\varepsilon(A_{ij}))_{i,j \in \iota} over 𝔽2\mathbb{F}_2, then invertibility is characterized by (3): A∈GL⁡ι(R)⇔A‾∈GL⁡ι(𝔽2).A \in \operatorname{GL}_\iota(R) \iff \bar{A} \in \operatorname{GL}_\iota(\mathbb{F}_2).(3)✦

Given a finite index set ι\iota, an index set κ\kappa, an injection j:κ↪ιj : \kappa \hookrightarrow \iota, and a family (vk)k∈κ(v_k)_{k \in \kappa} of columns vk∈Rιv_k \in R^\iota, where for any ring SS we write SιS^\iota for the set of columns v:ι→Sv : \iota \to S, whose entrywise images v‾k=(ε(vk(i)))i∈ι\bar{v}_k = (\varepsilon(v_k(i)))_{i \in \iota} in 𝔽2ι\mathbb{F}_2^\iota are linearly independent over 𝔽2\mathbb{F}_2, there exists an invertible matrix U∈GL⁡ι(R)U \in \operatorname{GL}_\iota(R) satisfying (4): Ui,j(k)=vk(i)(k∈κ,i∈ι).U_{i,\, j(k)} = v_k(i) \qquad (k \in \kappa,\; i \in \iota).(4)✦

For each nonnegative integer nn, the identity (5) holds: GL⁡n(R)=Gn(R).\operatorname{GL}_n(R) = G_n(R).(5)✦

Over any finite index set ι\iota, generation takes the form (6): GL⁡ι(R)=Gι(R).\operatorname{GL}_\iota(R) = G_\iota(R).(6)✦

In the paper, the result reads: a square matrix over RR is invertible if and only if its reduction modulo ker⁡ε\ker \varepsilon (the paper’s II) is invertible; every tuple of columns whose reductions are linearly independent extends to an invertible matrix; and GL⁡n(R)\operatorname{GL}_n(R) (the paper’s HnH_n) is generated by elementary matrices and single-position diagonal matrices Di(u)D_i(u) (the paper’s di(u)d_i(u)) for u∈R×u \in R^\times.

What Is Checked

The claims are stated over an 𝔽2\mathbb{F}_2-algebra RR with augmentation 𝔽2\mathbb{F}_2-algebra homomorphism ε:R→𝔽2\varepsilon : R \to \mathbb{F}_2 satisfying the matrix nil kernel property, finite index sets ι\iota, index sets κ\kappa, injections j:κ↪ιj : \kappa \hookrightarrow \iota, κ\kappa-tuples of columns over RR whose entrywise images under ε\varepsilon are linearly independent over 𝔽2\mathbb{F}_2, nonnegative integers nn, and the elementary and diagonal subgroups Gn(R)≤GL⁡n(R)G_n(R) \le \operatorname{GL}_n(R) and Gι(R)≤GL⁡ι(R)G_\iota(R) \le \operatorname{GL}_\iota(R), built from diagonal matrices of units, single diagonal matrices, and elementary matrices.