OctoInformBeta

OpenAI preprint, September 2026 · §6. From curve defects to matrix codimension

Proposition 6.1: A gradient quotient over an infinite perfect field of characteristic two

A dimension bound for diagonal permanent coefficient critical loci.

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

For a diagonal affine pencil, the question is how large the critical locus of a fixed coefficient of its permanent can be.

Let II be a finite index set and RR a commutative ring. For a square matrix M∈Mat⁡I(R)M\in\operatorname{Mat}_I(R) and weights θ∈RI\theta\in R^I, the diagonal matrix DθD_\theta is given by Dθ=diag⁡(θi)i∈ID_\theta=\operatorname{diag}(\theta_i)_{i\in I}, and II denotes the identity matrix in an affine pencil. The bidegree (q,q)(q,q), for q∈ℕq\in\mathbb N, is the pair of exponents assigning exponent qq to each of the variables ss and tt.✦ The diagonal permanent coefficient Fθ,q(M)F_{\theta,q}(M) of MM with weights θ\theta and order qq is the element of RR defined by Fθ,q(M)=[sqtq]per⁡(M+sI+tDθ).F_{\theta,q}(M) =[s^qt^q]\operatorname{per}(M+sI+tD_\theta).(1)✦

Let KK be a field and let kk be a nonnegative integer; write N=k+1N=k+1 for the matrix size. The generic matrix X=(xij)1≤i,j≤NX=(x_{ij})_{1\le i,j\le N} has entries in the coordinate ring A=K[xij:1≤i,j≤N]A=K[x_{ij}:1\le i,j\le N]. For weights θ∈KN\theta\in K^N, the gradient ideal Jθ,qJ_{\theta,q} of the polynomial Fθ,q(X)F_{\theta,q}(X) in (1) is the ideal of AA generated by all its first partial derivatives: Jθ,q=(∂Fθ,q(X)∂xij:1≤i,j≤N).J_{\theta,q} =\left(\frac{\partial F_{\theta,q}(X)}{\partial x_{ij}} :1\le i,j\le N\right).(2)✦ The critical locus of Fθ,q(X)F_{\theta,q}(X) is the common zero locus of the generators in (2).

The Claim

Theorem 1. Let KK be an infinite perfect field of characteristic two. Suppose that the matrix size N=k+1N=k+1 satisfies N≥4N\ge4, that q∈ℕq\in\mathbb N satisfies 3q≤N−13q\le N-1, and that θ1,…,θN\theta_1,\ldots,\theta_N are pairwise distinct elements of KK. For the gradient ideal in (2), dim⁡Krull(A/Jθ,q)≤N2−⌊(q+22)18⌋.\dim_{\mathrm{Krull}}(A/J_{\theta,q}) \le N^2-\left\lfloor\frac{\binom{q+2}{2}}{18}\right\rfloor.(3)✦

In the paper, the result reads: for K=𝔽¯2K=\overline{\mathbb F}_2, matrix size N≥2N\ge2 corresponding to the paper’s kk, q=⌊(N−1)/3⌋q=\lfloor(N-1)/3\rfloor, T=(q+22)T=\binom{q+2}{2}, pairwise distinct θ1,…,θN\theta_1,\ldots,\theta_N, and GDθ(X)=[sqtq]per⁡N(X+sI+tDθ)G_{D_\theta}(X)=[s^qt^q]\operatorname{per}_N(X+sI+tD_\theta), one has dim⁡{X∈KN×N:∇GDθ(X)=0}≤N2+N−T\dim\{X\in K^{N\times N}:\nabla G_{D_\theta}(X)=0\}\le N^2+N-T.

What Is Checked

The claim (3) is stated over the coordinate ring of a generic N×NN\times N matrix and the gradient ideal of its diagonal permanent coefficient, with distinct diagonal weights over an infinite perfect field of characteristic two and with N≥4N\ge4 and 3q≤N−13q\le N-1.