OctoInformBeta

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

§3 (setting): the integral coefficient polynomial: Diagonal weights specialize to complex homogeneous polynomials

An integral family of coefficient polynomials.

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

Selecting coefficients of diagonal perturbations produces weighted polynomials in the matrix entries. Let RR be a commutative ring and II a finite index set. For a polynomial f∈R[s,t]f\in R[s,t], the notation [satb]f[s^at^b]f denotes its coefficient of bidegree (a,b)(a,b); for a nonnegative integer qq, the bidegree used below is (q,q)(q,q).

Introduce auxiliary row variables uiu_i and column variables viv_i, indexed by i∈Ii\in I. If εz\varepsilon_z denotes the standard basis vector at a variable zz, the edge multidegree αij\alpha_{ij} and the squarefree multidegree 𝟏\mathbf 1 are the elements of ℕI⊔I\mathbb N^{I\sqcup I} defined by (1). αij=εui+εvj,𝟏=∑i∈I(εui+εvi).\alpha_{ij}=\varepsilon_{u_i}+\varepsilon_{v_j}, \qquad \mathbf 1=\sum_{i\in I}(\varepsilon_{u_i}+\varepsilon_{v_i}).(1) The edge generating polynomial EM∈R[𝒖,𝒗]E_M\in R[\mathbf u,\mathbf v] of a matrix M∈RI×IM\in R^{I\times I} is defined by (2). EM(𝒖,𝒗)=∏(i,j)∈I×I(1+Mijuivj).E_M(\mathbf u,\mathbf v) =\prod_{(i,j)\in I\times I}(1+M_{ij}u_iv_j).(2)✦ For a tuple a=(ai)i∈Ia=(a_i)_{i\in I} in a commutative ring AA, the elementary symmetric polynomial eq(a)∈Ae_q(a)\in A is defined by (3). eq(a)=[zq]∏i∈I(1+aiz).e_q(a)=[z^q]\prod_{i\in I}(1+a_i z).(3)✦

For a nonnegative integer kk, write [k]={0,…,k−1}[k]=\{0,\ldots,k-1\}. The matrix variables 𝒙=(xij)(i,j)∈[k]×[k]\mathbf x=(x_{ij})_{(i,j)\in[k]\times[k]} are indexed by a row and a column; write X=(xij)X=(x_{ij}) for their matrix and ℂ[𝒙]\mathbb C[\mathbf x] for their polynomial ring over ℂ\mathbb C. Let 𝝑=(ϑi)i∈[k]\boldsymbol\vartheta=(\vartheta_i)_{i\in[k]} be independent parameter variables.

The Claim

Definition 1. For a matrix M∈RI×IM\in R^{I\times I}, weights θ=(θi)i∈I∈RI\theta=(\theta_i)_{i\in I}\in R^I, and a nonnegative integer qq, the diagonal coefficient Δq(M;θ)∈R\Delta_q(M;\theta)\in R is defined by (4). Δq(M;θ)=[sqtq]per⁡(M+diag⁡(s+θit)i∈I).\Delta_q(M;\theta) =[s^qt^q]\operatorname{per}\bigl(M+ \operatorname{diag}(s+\theta_i t)_{i\in I}\bigr).(4)✦ Its squarefree coefficient formula is (5), where [(𝒖,𝒗)𝟏][(\mathbf u,\mathbf v)^{\mathbf 1}] denotes the coefficient at multidegree 𝟏\mathbf 1. Δq(M;θ)=[(𝒖,𝒗)𝟏](EM(𝒖,𝒗)eq((uivi)i∈I)eq((θiuivi)i∈I)).\Delta_q(M;\theta) =[(\mathbf u,\mathbf v)^{\mathbf 1}] \left(E_M(\mathbf u,\mathbf v)\, e_q\bigl((u_iv_i)_{i\in I}\bigr)\, e_q\bigl((\theta_i u_iv_i)_{i\in I}\bigr)\right).(5)✦

For nonnegative integers k,qk,q, the universal integral coefficient family 𝒢k,q∈ℤ[𝝑][𝒙]\mathcal G_{k,q}\in\mathbb Z[\boldsymbol\vartheta][\mathbf x] is the polynomial defined by (6). 𝒢k,q(𝒙,𝝑)=Δq(X;𝝑).\mathcal G_{k,q}(\mathbf x,\boldsymbol\vartheta) =\Delta_q(X;\boldsymbol\vartheta).(6)✦ For every commutative ring KK and every θ∈K[k]\theta\in K^{[k]}, let ev⁡θ:ℤ[𝝑]→K\operatorname{ev}_{\theta}:\mathbb Z[\boldsymbol\vartheta]\to K be evaluation at θ\theta, with integers mapped canonically to KK, and extend it coefficientwise to polynomials in 𝒙\mathbf x. ev⁡θ(𝒢k,q)=Δq(X;θ)in K[𝒙].\operatorname{ev}_{\theta}(\mathcal G_{k,q})=\Delta_q(X;\theta) \quad\text{in }K[\mathbf x].(7)✦

For complex weights θ∈ℂ[k]\theta\in\mathbb C^{[k]}, the weighted coefficient polynomial Gk,q,θ∈ℂ[𝒙]G_{k,q,\theta}\in\mathbb C[\mathbf x] is defined by (8). Gk,q,θ(𝒙)=Δq(X;θ).G_{k,q,\theta}(\mathbf x)=\Delta_q(X;\theta).(8)✦ Let ℂ[𝒙]d\mathbb C[\mathbf x]_d denote the homogeneous component of total degree dd. Gk,q,θ∈ℂ[𝒙]max⁡{k−2q,0}.G_{k,q,\theta}\in\mathbb C[\mathbf x]_{\max\{k-2q,0\}}.(9)✦

In the paper, the result reads: for k≥1k\ge1 and 0≤q≤⌊k/2⌋0\le q\le\lfloor k/2\rfloor, put r=k−2qr=k-2q and D=diag⁡(θ0,…,θk−1)D=\operatorname{diag}(\theta_0,\ldots,\theta_{k-1}); the polynomial GD(X)=[sqtq]per⁡k(X+sIk+tD)=∑S,T⊆[k]S∩T=⌀,|S|=|T|=q(∏i∈Tθi)per⁡(XC,C)G_D(X)=[s^qt^q]\operatorname{per}_k(X+sI_k+tD) =\sum_{\substack{S,T\subseteq[k]\\S\cap T=\varnothing,\ |S|=|T|=q}} (\prod_{i\in T}\theta_i)\operatorname{per}(X_{C,C}), where C=[k]\(S∪T)C=[k]\setminus(S\cup T) and the empty permanent is 11, is integral in 𝜽,𝒙\boldsymbol\theta,\mathbf x and homogeneous of degree rr in 𝒙\mathbf x.

What Is Checked

The coefficient identity (5) is stated for matrices over every commutative ring and every finite index set. The specialization (7) allows every commutative ring as target, while the homogeneity assertion (9) is stated for all nonnegative k,qk,q and all complex weights.