OctoInformBeta

OpenAI preprint, September 2026 · §3. Quadratic signature inequalities · §3

F and F^∨ have the quadratic signature property: The weight qn−r(S)q^{n-r(S)} of a matroid of rank nn has it, and on 2n2n elements so does the same weight of its dual

For 0<q≤10<q\le 1, matroid rank weights have the slot signature property.

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

Take a matroid MM on a finite set Γ\Gamma with rank function rr, a natural number nn and a real number qq. The matroid weight of a set S⊆ΓS \subseteq \Gamma is the real number fq(S)=qn−r(S),f_q(S) = q^{\,n - r(S)},(1)✦ with the exponent cut off at 00 when r(S)>nr(S) > n. When MM has rank nn, every spanning set has weight 11 and a set loses a factor qq for each unit of rank it lacks. The dual matroid M*M^*, with rank function r*r^*, has the weight fq*(S)=qn−r*(S)f^*_q(S) = q^{\,n - r^*(S)} for the same nn and qq.

A multiaffine polynomial in variables ziz_i, i∈Γi \in \Gamma, is F(z)=∑S⊆Γf(S)zSF(z) = \sum_{S \subseteq \Gamma} f(S)\, z^S with zS=∏i∈Sziz^S = \prod_{i \in S} z_i, and it is identified with its coefficient function ff, a real function on the subsets of Γ\Gamma.

The question is whether fqf_q, for a matroid of rank nn and 0<q≤10 < q \le 1, passes a spectral test of the kind that defines Lorentzian polynomials. Differentiate its polynomial along pairs of elements, then in single variables, and take the Hessian at the origin of what remains: does every such Hessian have at most one positive eigenvalue? On a ground set of 2n2n elements, does the same hold for fq*f^*_q?

Derivatives Along Pairs

For a vector a∈ℝΓa \in \mathbb{R}^\Gamma, the directional derivative of a multiaffine polynomial FF along aa is the multiaffine polynomial DaF=∑i∈Γai∂F/∂ziD_a F = \sum_{i \in \Gamma} a_i\, \partial F / \partial z_i, and DafD_a f denotes its coefficient function. The iterated derivative of ff along a list (a1,…,ak)(a_1, \dots, a_k) of vectors of ℝΓ\mathbb{R}^\Gamma is the coefficient function D(a1,…,ak)f=Da1Da2⋯Dakf,D_{(a_1, \dots, a_k)} f = D_{a_1} D_{a_2} \cdots D_{a_k} f,(2)✦ which is ff itself for the empty list.

The indicator vector 𝟏E∈ℝΓ\mathbf{1}_E \in \mathbb{R}^\Gamma of a finite set E⊆ΓE \subseteq \Gamma has entry 11 at the elements of EE and 00 elsewhere.✦ The pair direction of a pair p=(x,y)p = (x, y) of elements of Γ\Gamma is 𝟏{x,y}\mathbf{1}_{\{x, y\}}; for x≠yx \neq y, the derivative along it is ∂zx+∂zy\partial_{z_x} + \partial_{z_y}.✦

The slot total of ff along a list P=(p1,…,pk)P = (p_1, \dots, p_k) of pairs pj=(xj,yj)p_j = (x_j, y_j) of elements of Γ\Gamma is the iterated derivative of ff along their pair directions, ∂Pf=D(𝟏{x1,y1},…,𝟏{xk,yk})f.\partial_P f = D_{(\mathbf{1}_{\{x_1, y_1\}}, \dots, \mathbf{1}_{\{x_k, y_k\}})} f.(3)✦ When the pairs are disjoint and avoid a set AA, the value ∂Pf(A)\partial_P f(A) is the sum of ff over the sets obtained from AA by adding one element of each pair.

The Slot Signature Property

The face Hessian of a real function gg on the subsets of Γ\Gamma at a set A⊆ΓA \subseteq \Gamma is the real symmetric matrix HA(g)H_A(g), indexed by Γ\A\Gamma \setminus A, with zero diagonal and with entries HA(g)ij=g(A∪{i,j})(i≠j).H_A(g)_{ij} = g(A \cup \{i, j\}) \qquad (i \neq j).(4)✦ It is the Hessian at the origin of ∏i∈A∂ziG\prod_{i \in A} \partial_{z_i} G, for GG the multiaffine polynomial with coefficient function gg, restricted to the variables outside AA.

A bilinear form BB on a real vector space WW has one positive direction when, for every e∈We \in W with B(e,e)>0B(e, e) > 0, B(v,v)≤0for every v∈W with B(v,e)=0.B(v, v) \le 0 \quad \text{for every } v \in W \text{ with } B(v, e) = 0.(5)✦ For a symmetric form this is the condition of at most one positive eigenvalue. A matrix has one positive direction when its bilinear form does.

A real function ff on the subsets of Γ\Gamma has the slot signature property when, for every list PP of pairs of elements of Γ\Gamma and every set A⊆ΓA \subseteq \Gamma, the face Hessian HA(∂Pf)H_A(\partial_P f) has one positive direction.✦

The Claim

Theorem 1. Let MM be a matroid of rank nn whose ground set is all of the finite set Γ\Gamma, and let 0<q≤10 < q \le 1. Then, for every list PP of pairs of elements of Γ\Gamma and every set A⊆ΓA \subseteq \Gamma, with fqf_q the matroid weight (1), HA(∂Pfq) has one positive direction.H_A(\partial_P f_q) \text{ has one positive direction.}(6)✦ If moreover Γ\Gamma has 2n2n elements, then for every such PP and AA, HA(∂Pfq) and HA(∂Pfq*) have one positive direction.H_A(\partial_P f_q) \text{ and } H_A(\partial_P f^*_q) \text{ have one positive direction.}(7)✦

In the paper, with VV standing for Γ\Gamma, the result reads: the polynomials F(z)=∑Sfq(S)zSF(z) = \sum_S f_q(S)\, z^S and F∨(z)=∑|S|=nfq(V\S)zSF^\vee(z) = \sum_{|S| = n} f_q(V \setminus S)\, z^S both have the quadratic signature property.

In (6) and (7), the slot signature property of the coefficient functions fqf_q and fq*f^*_q stands for the quadratic signature property of FF and F∨F^\vee, and fq*f^*_q agrees with S↦fq(Γ\S)S \mapsto f_q(\Gamma \setminus S) on the nn-element sets.

What Is Checked

The claim is stated over the matroid weights fqf_q and fq*f^*_q of a matroid of rank nn whose ground set is all of the finite set Γ\Gamma and of its dual, for 0<q≤10 < q \le 1 and, for the dual, a ground set of 2n2n elements, through the slot totals (3), the face Hessians (4) and the condition (5) of one positive direction.