OctoInformBeta

OpenAI preprint, September 2026 · §3. Affine counting and integer patches

§3 (setting): Kernel geometries and equation locks in good patch pieces and covering, alongside coordinate and weighted Hilbert invariants

Bounded integer patches and Hilbert invariants for polynomial ideals.

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

Fix a natural number nn and a field KK, and consider a finite set of integer points in ℤn\mathbb{Z}^n of bounded coordinate height that satisfy polynomial equations over KK. Bounding the number of such points on an affine variety requires measuring ideals in K[x1,…,xn]K[x_1, \dots, x_n] through Hilbert functions and isolating smooth pieces on which low-degree polynomials vanish only when they belong to the ideal.

Preliminary Constructions

For any field KK and n∈ℕn \in \mathbb{N}, the canonical ring homomorphism ℤ→K\mathbb{Z} \to K maps each integer point z=(z1,…,zn)∈ℤnz = (z_1, \dots, z_n) \in \mathbb{Z}^n to z‾∈Kn\bar{z} \in K^n and induces a ring homomorphism ℤ[x1,…,xn]→K[x1,…,xn]\mathbb{Z}[x_1, \dots, x_n] \to K[x_1, \dots, x_n], written g↦g‾g \mapsto \bar{g}. An integer point z∈ℤnz \in \mathbb{Z}^n lies on an ideal I⊆K[x1,…,xn]I \subseteq K[x_1, \dots, x_n] if p(z‾)=0p(\bar{z}) = 0 for every polynomial p∈Ip \in I.✦

Given an index set σ\sigma, a field KK, and T∈ℕT \in \mathbb{N}, write VT⊆K[xi:i∈σ]V_T \subseteq K[x_i : i \in \sigma] for the KK-vector space of polynomials pp of total degree deg⁡p≤T\deg p \le T.✦ For n∈ℕn \in \mathbb{N}, a field KK, and an ideal I⊆K[x1,…,xn]I \subseteq K[x_1, \dots, x_n] with quotient homomorphism πI:K[x1,…,xn]→K[x1,…,xn]/I\pi_I \colon K[x_1, \dots, x_n] \to K[x_1, \dots, x_n]/I, an affine Hilbert polynomial PI∈ℚ[t]P_I \in \mathbb{Q}[t] of II is a rational polynomial satisfying (1) for all sufficiently large T∈ℕT \in \mathbb{N}: dim⁡KπI(VT)=PI(T).\dim_K \pi_I(V_T) = P_I(T).(1)✦ For every ideal I⊆K[x1,…,xn]I \subseteq K[x_1, \dots, x_n], the Hilbert function is eventually polynomial: there exist a rational polynomial P∈ℚ[t]P \in \mathbb{Q}[t] of degree at most nn and a nonnegative integer T0T_0 such that for all T≥T0T \ge T_0, the dimension of the image of the space of polynomials of total degree at most TT modulo II equals P(T)P(T).

Let σ\sigma be a finite index set, ω:σ→ℝ\omega \colon \sigma \to \mathbb{R} a weight vector, and T∈ℝT \in \mathbb{R}. For a finite index set σ\sigma, a weight vector ω:σ→ℝ\omega \colon \sigma \to \mathbb{R}, and T∈ℝT \in \mathbb{R}, the monomial weight wω(a)w_\omega(a) of an exponent vector a∈ℕσa \in \mathbb{N}^\sigma is ∑i∈σωiai\sum_{i \in \sigma} \omega_i a_i✦, and the weighted exponent bound Bω(T)B_\omega(T) is the element of ℕσ\mathbb{N}^\sigma with coordinates max⁡(⌊T/ωi⌋,0)\max(\lfloor T/\omega_i \rfloor, 0), taking T/0=0T/0 = 0. For a finite index set σ\sigma, a weight vector ω:σ→ℝ\omega \colon \sigma \to \mathbb{R}, and T∈ℝT \in \mathbb{R}, the set of weighted monomials Mω(T)M_\omega(T) is a finite subset of ℕσ\mathbb{N}^\sigma.

For a finite index set σ\sigma, a field KK, an ideal I⊆K[xi:i∈σ]I \subseteq K[x_i : i \in \sigma], a weight vector ω:σ→ℝ\omega \colon \sigma \to \mathbb{R}, and T∈ℝT \in \mathbb{R}, the space of weighted functions WI,ω(T)W_{I,\omega}(T) of II is the KK-subspace of K[xi:i∈σ]/IK[x_i : i \in \sigma]/I defined by (2), where xa=∏i∈σxiaix^a = \prod_{i \in \sigma} x_i^{a_i} for a∈ℕσa \in \mathbb{N}^\sigma: WI,ω(T)=span⁡K{xamod⁡I:a∈Mω(T)}⊆K[xi:i∈σ]/I.W_{I,\omega}(T) = \operatorname{span}_K \bigl\{ x^a \bmod I : a \in M_\omega(T) \bigr\} \subseteq K[x_i : i \in \sigma]/I.(2)✦

Given m∈ℕm \in \mathbb{N} and M∈ℝM \in \mathbb{R}, the integral kernel height bound Φ(m,M)\Phi(m, M) is the real number given by (3): Φ(m,M)=(1+m2M2)m!(1+mM2)m.\Phi(m, M) = (1 + m^2 M^2)\, m! \, (1 + m M^2)^m.(3)✦

The Definition

Definition 1. For n∈ℕn \in \mathbb{N}, a kernel geometry (S,H)(S, H) in ℤn\mathbb{Z}^n is a pair consisting of a finite subset S⊆ℤnS \subseteq \mathbb{Z}^n and a real height HH satisfying the bounds (4): H≥1and|zl|≤Hfor all z∈S and 1≤l≤n.H \ge 1 \qquad \text{and} \qquad |z_l| \le H \quad \text{for all } z \in S \text{ and } 1 \le l \le n.(4)✦

For n,D∈ℕn, D \in \mathbb{N}, a field KK, an ideal I⊆K[x1,…,xn]I \subseteq K[x_1, \dots, x_n], and a finite subset S⊆ℤnS \subseteq \mathbb{Z}^n, the pair (I,S)(I, S) has an equation lock at degree DD if every polynomial p∈K[x1,…,xn]p \in K[x_1, \dots, x_n] with deg⁡p≤D\deg p \le D satisfies (5): (p(z‾)=0 for every z∈S lying on I)⇔p∈I.\bigl(p(\bar{z}) = 0 \text{ for every } z \in S \text{ lying on } I\bigr) \iff p \in I.(5)✦

For n∈ℕn \in \mathbb{N}, a field KK, and an ideal I⊆K[x1,…,xn]I \subseteq K[x_1, \dots, x_n] with affine Hilbert polynomial PI∈ℚ[t]P_I \in \mathbb{Q}[t], the affine dimension dim⁡aff(I)\dim_{\mathrm{aff}}(I) of II is the natural number defined by (6): dim⁡aff(I)={deg⁡PIif PI≠0,0if PI=0.\dim_{\mathrm{aff}}(I) = \begin{cases} \deg P_I & \text{if } P_I \ne 0, \\ 0 & \text{if } P_I = 0. \end{cases}(6)✦

Let n∈ℕn \in \mathbb{N}, KK a field, and I⊆K[x1,…,xn]I \subseteq K[x_1, \dots, x_n] an ideal having affine Hilbert polynomial PI∈ℚ[t]P_I \in \mathbb{Q}[t], and let cI∈ℚc_I \in \mathbb{Q} be the coefficient of tdim⁡aff(I)t^{\dim_{\mathrm{aff}}(I)} in PIP_I; the affine degree deg⁡aff(I)\deg_{\mathrm{aff}}(I) of II is the rational number given by (7): deg⁡aff(I)=(dim⁡aff(I))!cI.\deg_{\mathrm{aff}}(I) = \bigl(\dim_{\mathrm{aff}}(I)\bigr)! \, c_I.(7)✦

Given a finite index set σ\sigma, a field KK, an ideal I⊆K[xi:i∈σ]I \subseteq K[x_i : i \in \sigma], a weight vector ω:σ→ℝ\omega \colon \sigma \to \mathbb{R}, and T∈ℝT \in \mathbb{R}, the weighted Hilbert function HI,ω(T)H_{I,\omega}(T) of II is the natural number defined by (8): HI,ω(T)=dim⁡KWI,ω(T).H_{I,\omega}(T) = \dim_K W_{I,\omega}(T).(8)✦

For n∈ℕn \in \mathbb{N}, an integer patch PP in nn variables is a pair of free coordinate indices and integral equations as in (9): P=(F,(gi)i∈{1,…,n}\F),F⊆{1,…,n},gi∈ℤ[x1,…,xn].P = \bigl(F, (g_i)_{i \in \{1, \dots, n\} \setminus F}\bigr), \qquad F \subseteq \{1, \dots, n\}, \quad g_i \in \mathbb{Z}[x_1, \dots, x_n].(9)✦

Given n∈ℕn \in \mathbb{N} and an integer patch P=(F,(gi)i∈{1,…,n}\F)P = (F, (g_i)_{i \in \{1, \dots, n\} \setminus F}) in nn variables, the patch Jacobian minor ΔP\Delta_P of PP is the polynomial in ℤ[x1,…,xn]\mathbb{Z}[x_1, \dots, x_n] defined by (10): ΔP=det⁡(∂gi∂xj)i,j∈{1,…,n}\F.\Delta_P = \det \biggl( \frac{\partial g_i}{\partial x_j} \biggr)_{i, j \in \{1, \dots, n\} \setminus F}.(10)✦

Over a field KK with n∈ℕn \in \mathbb{N}, an integer patch P=(F,(gi)i∈{1,…,n}\F)P = (F, (g_i)_{i \in \{1, \dots, n\} \setminus F}) in nn variables lies on an ideal I⊆K[x1,…,xn]I \subseteq K[x_1, \dots, x_n] if (11) holds: |F|=dim⁡aff(I),g‾i∈I for all i∈{1,…,n}\F,Δ‾P∉I.|F| = \dim_{\mathrm{aff}}(I), \qquad \bar{g}_i \in I \text{ for all } i \in \{1, \dots, n\} \setminus F, \qquad \bar{\Delta}_P \notin I.(11)✦

For n,D∈ℕn, D \in \mathbb{N} and H∈ℝH \in \mathbb{R}, an integer patch P=(F,(gi)i∈{1,…,n}\F)P = (F, (g_i)_{i \in \{1, \dots, n\} \setminus F}) in nn variables is (D,H)(D, H)-bounded if there exists d∈ℕd \in \mathbb{N} with d≤Dd \le D such that deg⁡gi≤d\deg g_i \le d for every i∈{1,…,n}\Fi \in \{1, \dots, n\} \setminus F and every coefficient cc of each gig_i satisfies (12): |c|≤Φ((d+nn),Hd).|c| \le \Phi\biggl(\binom{d + n}{n}, H^d\biggr).(12)✦

Given n,D,h,B∈ℕn, D, h, B \in \mathbb{N}, H∈ℝH \in \mathbb{R}, a field KK, an ideal I⊆K[x1,…,xn]I \subseteq K[x_1, \dots, x_n], and a finite subset S⊆ℤnS \subseteq \mathbb{Z}^n, a good patch piece (J,P)(J, P) for (I,S,D,h,B,H)(I, S, D, h, B, H) is a pair of an ideal J⊆K[x1,…,xn]J \subseteq K[x_1, \dots, x_n] and an integer patch PP in nn variables satisfying (13): J is prime,I⊆J,dim⁡aff(J)≤h,deg⁡aff(J)≤B,P lies on J,P is (B,H)-bounded,and (J,S) has an equation lock at degree D.\begin{gathered} J \text{ is prime}, \qquad I \subseteq J, \qquad \dim_{\mathrm{aff}}(J) \le h, \qquad \deg_{\mathrm{aff}}(J) \le B, \\ P \text{ lies on } J, \qquad P \text{ is } (B, H)\text{-bounded}, \qquad \text{and $(J, S)$ has an equation lock at degree $D$}. \end{gathered}(13)✦

Let n∈ℕn \in \mathbb{N}, KK a field, J⊆K[x1,…,xn]J \subseteq K[x_1, \dots, x_n] an ideal, and PP an integer patch in nn variables; the pair (J,P)(J, P) covers an integer point z∈ℤnz \in \mathbb{Z}^n if (14) holds: z lies on JandΔP(z)≠0.z \text{ lies on } J \qquad \text{and} \qquad \Delta_P(z) \ne 0.(14)✦

In the paper, the result reads: an ambient box S⊆ℤNS \subseteq \mathbb{Z}^N with |zl|≤CXωl|z_l| \le C X^{\omega_l} on an algebraic set Z⊆ℂNZ \subseteq \mathbb{C}^N of degree at most e0e_0 and dimension at most rr, subvarieties A⊆ZA \subseteq Z defined over ℚ\mathbb{Q}, bounded-height integer patches (A,G,Δ)(A, G, \Delta), and weighted filtered Hilbert functions HA(T)H_A(T), where nn stands for NN, II for ZZ, JJ for AA, (gi)i∈{1,…,n}\F(g_i)_{i \in \{1, \dots, n\} \setminus F} for GG, ΔP\Delta_P for Δ\Delta, BB for e0e_0, hh for rr, and HI,ω(T)H_{I,\omega}(T) for HA(T)H_A(T). Over an arbitrary field KK, the library replaces ℚ\mathbb{Q}-subvarieties A⊆Z⊆ℂNA \subseteq Z \subseteq \mathbb{C}^N and coordinatewise bounds CXωlC X^{\omega_l} with ideals I⊆J⊆K[x1,…,xn]I \subseteq J \subseteq K[x_1, \dots, x_n] and a uniform height bound H≥1H \ge 1.

What Is Checked

The definitions are stated over any natural number n∈ℕn \in \mathbb{N}, any finite index set σ\sigma, and any field KK, taking ideals in K[x1,…,xn]K[x_1, \dots, x_n] and K[xi:i∈σ]K[x_i : i \in \sigma], finite subsets S⊆ℤnS \subseteq \mathbb{Z}^n, integer patches PP and points z∈ℤnz \in \mathbb{Z}^n, bounds D,h,B∈ℕD, h, B \in \mathbb{N}, real parameters H,T∈ℝH, T \in \mathbb{R}, and weights ω:σ→ℝ\omega \colon \sigma \to \mathbb{R}. The library also proves that for every ideal II of polynomials in nn coordinates over a field KK, the Hilbert function is eventually polynomial: there exist a rational polynomial PP of degree at most nn and a nonnegative integer T0T_0 such that for all T≥T0T \ge T_0, the dimension of the image of the space of polynomials of total degree at most TT under the quotient map modulo II equals P(T)P(T).