OctoInformBeta

OpenAI preprint, September 2026 · §3. Selecting a coefficient by a permanent projection

Proposition 3.1: Affine substitutions and their homogenizations realize the coefficient of bidegree (q,q)(q,q) for a nonnegative integer qq, up to a nonzero scalar

Selecting diagonal permanent coefficients by projection.

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

A diagonal perturbation introduces two scalar parameters into a permanent. The question is whether extracting a coefficient in those parameters can be realized by substituting affine-linear forms into a larger permanent.

For a nonnegative integer ℓ\ell, write Iℓ={i∈ℤ:0≤i<ℓ}I_\ell=\{i\in\mathbb Z:0\le i<\ell\}. The matrix variables of size ℓ\ell are indexed by Iℓ×IℓI_\ell\times I_\ell, and ℂ[X]\mathbb C[X] denotes the polynomial ring in the entries of X=(xij)i,j∈IℓX=(x_{ij})_{i,j\in I_\ell}. The permanent polynomial per⁡ℓ(X)∈ℂ[X]\operatorname{per}_\ell(X)\in\mathbb C[X] of size ℓ\ell is defined by (1), where SIℓS_{I_\ell} is the permutation group of IℓI_\ell: per⁡ℓ(X)=∑σ∈SIℓ∏i∈Iℓxi,σ(i).\operatorname{per}_\ell(X) =\sum_{\sigma\in S_{I_\ell}}\prod_{i\in I_\ell}x_{i,\sigma(i)}.(1)✦

Given a finite set II, a matrix M∈RI×IM\in R^{I\times I} over a commutative ring RR, and a vector ϑ∈RI\vartheta\in R^I, define the diagonal permanent coefficient Δq(M;ϑ)∈R\Delta_q(M;\vartheta)\in R by (2), where [uqvq][u^qv^q] extracts the coefficient of bidegree (q,q)(q,q): Δq(M;ϑ)=[uqvq]per⁡(M+diag⁡i∈I(u+ϑiv)).\Delta_q(M;\vartheta) =[u^qv^q]\operatorname{per}\!\left( M+\operatorname{diag}_{i\in I}(u+\vartheta_i v)\right).(2) For a k×kk\times k variable matrix XX and θ∈ℂk\theta\in\mathbb C^k, the selector coefficient polynomial Gk,q,θ∈ℂ[X]G_{k,q,\theta}\in\mathbb C[X] is the specialization in (3): Gk,q,θ(X)=Δq(X;θ).G_{k,q,\theta}(X)=\Delta_q(X;\theta).(3)✦

For finite index sets U,VU,V and coefficients bu,cuv∈ℂb_u,c_{uv}\in\mathbb C, the affine substitution Φb,c:ℂ[yu:u∈U]→ℂ[xv:v∈V]\Phi_{b,c}:\mathbb C[y_u:u\in U]\to\mathbb C[x_v:v\in V] is the ℂ\mathbb C-algebra homomorphism determined by the first assignment in (4).✦ The homogeneous substitution Φ̃b,c:R[yu:u∈U]→R[xv:v∈V,z]\widetilde\Phi_{b,c}:R[y_u:u\in U]\to R[x_v:v\in V,z], with coefficients in a commutative ring RR, is the ring homomorphism determined by the second assignment, where zz is a new variable: Φb,c(yu)=bu+∑v∈Vcuvxv,Φ̃b,c(yu)=buz+∑v∈Vcuvxv.\Phi_{b,c}(y_u)=b_u+\sum_{v\in V}c_{uv}x_v, \qquad \widetilde\Phi_{b,c}(y_u)=b_uz+\sum_{v\in V}c_{uv}x_v.(4) For nonnegative integers m,km,k, U=Im2U=I_m^2, V=Ik2V=I_k^2, and complex coefficients, let L(X)L(X) and L̃(X,z)\widetilde L(X,z) be the m×mm\times m matrices with entries Φb,c(yij)\Phi_{b,c}(y_{ij}) and Φ̃b,c(yij)\widetilde\Phi_{b,c}(y_{ij}), respectively.

For nonnegative integers k,qk,q, the selector normalization ak,qa_{k,q} is the complex scalar given by (5): ak,q=(−2)2k(max⁡{k−q,0}!)4.a_{k,q}=(-2)^{2k}\bigl(\max\{k-q,0\}!\bigr)^4.(5)✦

The Claim

Theorem 1. For all nonnegative integers k,qk,q, ak,q≠0.a_{k,q}\ne0.(6)✦ Let k,q,mk,q,m be nonnegative integers with q≤kq\le k and m≥11k−2qm\ge11k-2q, and let θ∈ℂk\theta\in\mathbb C^k. There are complex coefficients b,cb,c such that per⁡m(L(X))=ak,qGk,q,θ(X).\operatorname{per}_m(L(X))=a_{k,q}G_{k,q,\theta}(X).(7)✦ Under the additional hypotheses k≥1k\ge1 and q≤⌊k/2⌋q\le\lfloor k/2\rfloor, the coefficients b,cb,c can be chosen to satisfy both identities per⁡m(L(X))=ak,qGk,q,θ(X),per⁡m(L̃(X,z))=ak,qzm−(k−2q)Gk,q,θ(X).\begin{aligned} \operatorname{per}_m(L(X)) &=a_{k,q}G_{k,q,\theta}(X),\\ \operatorname{per}_m(\widetilde L(X,z)) &=a_{k,q}z^{m-(k-2q)}G_{k,q,\theta}(X). \end{aligned}(8)✦

In the paper, the result reads: for k≥1k\ge1, 0≤q≤⌊k/2⌋0\le q\le\lfloor k/2\rfloor, r=k−2qr=k-2q, and θ∈ℂk\theta\in\mathbb C^k, every integer m≥11k−2qm\ge11k-2q admits an m×mm\times m matrix A(X)A(X) of affine-linear forms with per⁡m(A(X))=a0Gk,q,θ(X)\operatorname{per}_m(A(X))=a_0G_{k,q,\theta}(X), where a0=(−2)2k((k−q)!)4≠0a_0=(-2)^{2k}((k-q)!)^4\ne0, whose entrywise homogenization B(X,z)B(X,z) satisfies per⁡m(B(X,z))=a0zm−rGk,q,θ(X)\operatorname{per}_m(B(X,z))=a_0z^{m-r}G_{k,q,\theta}(X), with no distinctness or nonvanishing required of the θi\theta_i.

What Is Checked

The normalization assertion (6) is stated for all nonnegative k,qk,q; the projection assertions (7) and (8) concern complex matrix-variable polynomial rings, arbitrary θ∈ℂk\theta\in\mathbb C^k, and the affine and homogeneous substitutions of (4) under their stated bounds on k,q,mk,q,m.