OctoInform Beta

OpenAI preprint, September 2026 · The second preprint: One Rational Matrix Hitting Point for Noncommutative Formulas · §4 of the second preprint

Lemma 4.1 [Wronskian], One Rational Matrix Hitting Point: A space of polynomials of bounded degree has a basis of distinct orders at 00; in characteristic zero every basis has a nonzero Wronskian

Distinct vanishing orders fix the Wronskian’s order in characteristic zero.

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 field KK and a space VV of polynomials in zz over KK, all of degree below a fixed bound dd, of dimension kk. Every nonzero member of VV vanishes at 00 to some order, and two questions arise: whether VV has an echelon basis, one whose members vanish at 00 to pairwise distinct orders a1<…<aka_1 < \dots < a_k, and how far the Wronskian of a basis of VV vanishes at 00. The Wronskian lemma, Lemma 4.1 of the second preprint, OpenAI’s One Rational Matrix Hitting Point for Noncommutative Formulas (called the paper below), answers both, the second in characteristic zero.

Orders and Wronskians

For d≥0d \ge 0, the space Pd(K)P_d(K) is the KK-vector space of polynomials in zz over KK of degree below dd. The order ord⁡0f\operatorname{ord}_0 f at 00 of a nonzero polynomial f∈K[z]f \in K[z] is the least mm such that the coefficient of zmz^m in ff is nonzero.

The Wronskian W(f1,…,fk)W(f_1, \dots, f_k) of polynomials f1,…,fkf_1, \dots, f_k in zz over a commutative ring is the determinant of the k×kk \times k matrix whose entry in row ii and column jj is the formal derivative fj(i−1)f_j^{(i-1)} of fjf_j of order i−1i-1: W(f1,…,fk)=det⁡(fj(i−1))1≤i,j≤k=det⁡(f1⋯fkf1′⋯fk′⋮⋮f1(k−1)⋯fk(k−1)).W(f_1, \dots, f_k) = \det\bigl(f_j^{(i-1)}\bigr)_{1 \le i, j \le k} = \det \begin{pmatrix} f_1 & \cdots & f_k \\ f_1' & \cdots & f_k' \\ \vdots & & \vdots \\ f_1^{(k-1)} & \cdots & f_k^{(k-1)} \end{pmatrix}.(1)✦

The Claim

Theorem 1. Let KK be a field.

  1. Every subspace VV of Pd(K)P_d(K) is spanned by linearly independent polynomials f1,…,fhf_1, \dots, f_h with pairwise distinct orders at 00: V=span⁡K(f1,…,fh),ord⁡0fi≠ord⁡0fj(i≠j).V = \operatorname{span}_K(f_1, \dots, f_h), \qquad \operatorname{ord}_0 f_i \neq \operatorname{ord}_0 f_j \quad (i \neq j).(2)✦

  2. If b1,…,bkb_1, \dots, b_k is a basis of a subspace VV of Pd(K)P_d(K), then VV has a basis b1′,…,bk′b'_1, \dots, b'_k with ord⁡0bi′≠ord⁡0bj′(i≠j).\operatorname{ord}_0 b'_i \neq \operatorname{ord}_0 b'_j \quad (i \neq j).(3)✦

  3. If KK has characteristic zero and f1,…,fk∈K[z]f_1, \dots, f_k \in K[z] are nonzero polynomials with pairwise distinct orders at 00, then W(f1,…,fk)≠0,ord⁡0W(f1,…,fk)=∑j=1kord⁡0fj−(k2).W(f_1, \dots, f_k) \neq 0, \qquad \operatorname{ord}_0 W(f_1, \dots, f_k) = \sum_{j=1}^{k} \operatorname{ord}_0 f_j - \binom{k}{2}.(4)✦

  4. If b1,…,bkb_1, \dots, b_k and b1′,…,bk′b'_1, \dots, b'_k are bases of a subspace VV of K[z]K[z] and P=(Pij)P = (P_{ij}) is the change-of-basis matrix, bj′=∑iPijbib'_j = \sum_i P_{ij} b_i, then W(b1′,…,bk′)=det⁡(P)W(b1,…,bk).W(b'_1, \dots, b'_k) = \det(P)\, W(b_1, \dots, b_k).(5)✦

  5. If KK has characteristic zero and b1,…,bkb_1, \dots, b_k is a basis of a subspace VV of Pd(K)P_d(K), then W(b1,…,bk)≠0.W(b_1, \dots, b_k) \neq 0.(6)✦

  6. If f1,…,fk∈K[z]f_1, \dots, f_k \in K[z] are nonzero polynomials with pairwise distinct orders at 00, c1,…,ck∈Kc_1, \dots, c_k \in K, r≥0r \ge 0 is an integer, and zrz^r divides ∑jcjfj\sum_j c_j f_j, then cj=0for every j with ord⁡0fj<r.c_j = 0 \quad \text{for every } j \text{ with } \operatorname{ord}_0 f_j < r.(7)✦

In the paper, the result reads, with KK for its field FF: for a field KK of characteristic zero and a subspace VV of K[[z]]K[[z]] of finite positive dimension kk, there is a basis b1,…,bkb_1, \dots, b_k of VV with orders 0≤a1<…<ak0 \le a_1 < \dots < a_k at 00, for which ord⁡0W(b1,…,bk)=∑jaj−(k2)\operatorname{ord}_0 W(b_1, \dots, b_k) = \sum_j a_j - \binom{k}{2}, and in particular the Wronskian of any basis of VV is nonzero, its order does not depend on the basis, and every nonzero member of VV has order at most aka_k.

The box works with polynomials rather than power series, and reaches the paper’s last two assertions through (5) and through (7) with r=ak+1r = a_k + 1.

What Is Checked

The library states the claims over polynomials in zz over a field KK, with their order at 00 and the Wronskian (1) of a family of them, and over subspaces of Pd(K)P_d(K), or of K[z]K[z] in (5), with their bases and change-of-basis matrices.