OctoInform Beta

OpenAI preprint, September 2026 · §2. Rectangular tests from subspace expansion · §2

§2 (displayed): the vanishing filtration (2.5): At a point aa, and the members divisible by (z−a)r(z-a)^r form a subspace of dimension the number of basis vectors of order ≥r\geq r

Bounded-degree polynomial spaces have bases of distinct vanishing orders.

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

Let KK be a field and K[z]K[z] the ring of polynomials in one variable zz over KK. For d≥0d \geq 0, Pd(K)P_d(K) is the subspace of K[z]K[z] of polynomials of degree below dd. Take a subspace VV of Pd(K)P_d(K) of dimension hh and a point a∈Ka \in K. The question is how the members of VV vanish at aa: to which orders, and, for each r≥0r \geq 0, how large the subspace of members divisible by (z−a)r(z-a)^r is.

Vanishing Orders and the Vanishing Filtration

The vanishing order ord⁡af\operatorname{ord}_a f of a nonzero polynomial f∈K[z]f \in K[z] at a point a∈Ka \in K is the multiplicity of aa as a root of ff: the largest m≥0m \geq 0 such that (z−a)m(z-a)^m divides ff.

The Taylor coefficients cm(f)c_m(f) of a polynomial f∈K[z]f \in K[z] at aa are its coefficients in the powers of z−az-a: f=∑m≥0cm(f)(z−a)m.f = \sum_{m \geq 0} c_m(f)\,(z-a)^m.(1)

The vanishing filtration of a subspace VV of K[z]K[z] at a point a∈Ka \in K is the family of subspaces Vr(a)V_r(a) of VV, one for each integer r≥0r \geq 0, where Vr(a)V_r(a) consists of the members of VV whose Taylor coefficients at aa of order below rr vanish, that is, the members of VV divisible by (z−a)r(z-a)^r: Vr(a)={f∈V:cm(f)=0 for 0≤m<r}.V_r(a) = \{\, f \in V : c_m(f) = 0 \text{ for } 0 \leq m < r \,\}.(2)✦

The Claim

Theorem 1. Let KK be a field and aa a point of KK.

(a) Let d≥0d \geq 0 and let VV be a subspace of Pd(K)P_d(K) of dimension hh. Then VV has a basis g1,…,ghg_1, \dots, g_h whose vanishing orders at aa are pairwise distinct: ord⁡agj≠ord⁡agk(1≤j<k≤h).\operatorname{ord}_a g_j \neq \operatorname{ord}_a g_k \qquad (1 \leq j < k \leq h).(3)✦

(b) Let VV be a subspace of K[z]K[z] with a basis g1,…,ghg_1, \dots, g_h whose vanishing orders at aa are pairwise distinct. Then for every integer r≥0r \geq 0, dim⁡KVr(a)=#{j∈{1,…,h}:ord⁡agj≥r}.\dim_K V_r(a) = \#\{\, j \in \{1, \dots, h\} : \operatorname{ord}_a g_j \geq r \,\}.(4)✦

In the paper, the result reads, with K=ℚK = \mathbb{Q} and a=ia = i: for a subspace V⊆Pd(ℚ)V \subseteq P_d(\mathbb{Q}) of dimension hh, row reduction at z=iz = i gives a basis of VV with vanishing orders di1<…<dihd_{i1} < \dots < d_{ih}, and dim⁡{f∈V:(z−i)r divides f}=#{j:dij≥r}\dim \{ f \in V : (z-i)^r \text{ divides } f \} = \#\{ j : d_{ij} \geq r \} for every r≥0r \geq 0.

What Is Checked

The claim is stated over subspaces of K[z]K[z] for an arbitrary field KK, their bases and the vanishing orders of their members at a point aa, and the vanishing filtration (2).

About that filtration the library proves that each Vr(a)V_r(a) contains 00 and is closed under sums and scalar multiples.