By OpenAISourceGitHub

Manuscripts from openai/math

Fork
Clone
git clone https://openversions.com/git/index/openai-math.git
Cite
Version link
https://openversions.com/index/openai-math/commit/dfd8c46870b48511e697cbb43a3c8c91d375fa2f
Version identifier (commit hash)
dfd8c46870b48511e697cbb43a3c8c91d375fa2f

Fork to create your own copy with credit to the source. Clone to get a Git copy on your computer.

Latest version

Checks complete

Sync OpenAI Math manuscripts from 4eaee87548250b2682affda7d05d947c05554bae

Committed

All papers

Claims

Findings

The abelian Zilber–Pink conjecture

For every number field KK, every abelian variety A/KA/K, and every irreducible subvariety X⊆AK‾X \subseteq A_{\overline{K}}, there are only finitely many maximal atypical subvarieties of XX in S=⟨X⟩spS = \langle X \rangle_{\mathrm{sp}}; equivalently, there are finitely many proper atypical subvarieties Y1,…,Ym⊊XY_1, \dots, Y_m \subsetneq X such that every atypical subvariety is contained in one of the YjY_j.

Let AA be an abelian variety over Q‾\overline{\mathbb{Q}}, and let X⊆AX \subseteq A be an irreducible subvariety contained in no proper torsion coset. Then X∩A[dim⁡X+1]X \cap A^{[\dim X + 1]} is not Zariski dense in XX.

Let AA be an abelian variety over Q‾\overline{\mathbb{Q}}, let Γ⊆A(Q‾)\Gamma \subseteq A(\overline{\mathbb{Q}}) be a subgroup satisfying dim⁡Q(Γ⊗ZQ)<∞\dim_{\mathbb{Q}}(\Gamma \otimes_{\mathbb{Z}} \mathbb{Q}) < \infty, and let X⊆AX \subseteq A be an irreducible closed subvariety of dimension dd contained in no translate of a proper algebraic subgroup of AA. Then X(Q‾)∩(A[d+1](Q‾)+Γ)X(\overline{\mathbb{Q}}) \cap (A^{[d+1]}(\overline{\mathbb{Q}}) + \Gamma) is not Zariski dense in XX, and if d=1d = 1, this intersection is finite.

For sequence data as in Definition 2.2, let V⊆Lie(W)V \subseteq \mathrm{Lie}(W) be the image of an idempotent in End(A)R\mathrm{End}(\mathcal{A})_{\mathbb{R}}, let Φ\Phi be the orthogonal projector with kernel VV, and put qi=(yiℓ/Hiℓ)ℓ∈I∈H(A)q_i = (y_{i\ell}/\sqrt{H_{i\ell}})_{\ell \in I} \in \mathcal{H}(\mathcal{A}). Then codimLie(W)V>∑ℓ∈Imℓ\mathrm{codim}_{\mathrm{Lie}(W)} V > \sum_{\ell \in I} m_\ell and ∥Φ(qi)∥→0\|\Phi(q_i)\| \to 0 cannot both hold.

The auxiliary sequence theorem (Theorem 2.3) implies Theorem 1.2 for every abelian variety over Q‾\overline{\mathbb{Q}}.

Theorem 2.3 holds for all-high sequence data.

For the bounded-degree, bounded-field-degree subproducts Zi=∏ℓ=0sZiℓZ_i = \prod_{\ell=0}^s Z_{i\ell} with fixed positive-dimensional low factor Z0Z_0, dense factorwise height bounds hLℓ=O(Hiℓ)h_{L_\ell} = O(H_{i\ell}), fixed high cycle classes, and positive tangent integrals for Φ\Phi, the homomorphisms and scalings in (6.1) satisfy lim inf⁡i→∞(E‾id+1⋅Zi)/(Mi2Bi)>0\liminf_{i\to\infty} (\overline{E}_i^{d+1} \cdot Z_i)/(M_i^2 B_i) > 0.

Under the hypotheses of Proposition 7.1 (smoothness and bounded-degree/height conormal equations at yiℓy_{i\ell}, biℓ≪Mib_{i\ell} \ll M_i, biℓ2Hiℓ≍Mi2b_{i\ell}^2 H_{i\ell} \asymp M_i^2, hE‾i(yi)=o(Mi2)h_{\overline{E}_i}(y_i) = o(M_i^2), and small sections sis_i satisfying (7.2)), there exists η>0\eta > 0 independent of ii such that indni,bi(si;yi)≥η\mathrm{ind}_{n_i, b_i}(s_i; y_i) \ge \eta for all sufficiently large ii, with no bound on the degree of the field of yiy_i required.

For the minimal products of Proposition 5.3 with d0>0d_0 > 0, the small sections of Lemma 6.2 cannot satisfy the positive index conclusion of Proposition 7.1 along an infinite subsequence.

Every irreducible subvariety of every abelian variety over Q‾\overline{\mathbb{Q}} contains only finitely many optimal subvarieties.

Methods

Height-scale decomposition of a sequence of algebraic points on an abelian variety via Poincaré reducibility into at most one bounded scale and finitely many successively larger unbounded scales with nondegeneracy for every fixed nonzero homomorphism.

Source excerpts

“Lemma 2.4 (Height decomposition). Let A be a nonzero abelian variety and xi ∈ A(Q). After passing to a subsequence, there is an isogeny α : A −→ Y ℓ∈I Wℓ with nonzero factors, and scales Hiℓ ≥ 1, such that the coordinates piℓ of α(xi) have norms O(√Hiℓ) and successive scale ratios tend to zero. There is at most one bounded scale, placed first and equal to one. Every other scale tends to infinity and satisfies (2.2) with Aℓ = Wℓ for every fixed nonzero homomorphism on Wℓ .” · Show in paper

Uniform tangent positivity for product subvarieties via Ax's analytic-subgroup theorem, diagonal Grassmannian degeneration to a product subgroup, and finiteness of geodesic-optimal directions in fixed parents.

Source excerpts

“Lemma 3.2 (Product tangent obstruction). Let Aℓ , ℓ ∈ I, be complex abelian varieties, let Zℓ ⊆ Aℓ be subvarieties, and put Z = Q ℓ∈I Zℓ and A = Q ℓ∈I Aℓ . Let g : Lie(A) −→ E be a complex linear map to a Hermitian vector space. If PZ(g) = 0, there are abelian subvarieties Cℓ ⊆ Aℓ such that, for every z = (zℓ)ℓ ∈ Z, X ℓ∈I dimzℓ (Zℓ ∩ (zℓ + Cℓ)) > rank ( g |L ℓ∈I Lie(Cℓ) ) . (3.6)” · Show in paper

“Lemma 3.3 (Finite directions for local rank inequalities). Fix subvarieties Yℓ ⊆ Aℓ of complex abelian varieties. For each ℓ, there is a finite set Hℓ of abelian subvarieties of Aℓ with the following property.” · Show in paper

Real-algebraic quantifier elimination applied to multivariate Newton expansions of top intersection polynomials PZ(g)P_Z(g) to obtain uniform degree-dependent lower bounds over a fixed rational box B\mathcal{B}.

Source excerpts

“Lemma 4.1. Let B be a fixed closed box with rational endpoints in the coordinates of an integral basis of End(A). There are positive integers θ, ω1, depending only on the fixed ambients, polarizations, and box, with the following property. For every irreducible subproduct Z ⊆ A such that PZ is strictly positive everywhere on B, PZ(g) ≥ θ −1D−ω1 Z (g ∈ B). (4.1)” · Show in paper

Dyadic balancing of height scales using multipliers viℓ=u2kv_{i\ell} = u 2^k to match the height scales of all factors while keeping polynomial coordinate-system coefficient heights quadratic in viℓv_{i\ell} for Dill's generalized Vojta–Rémond inequality.

Source excerpts

“We want the multiplied points in every factor to have the same height scale: viℓ√Hiℓ should be close to one number Ri . At the same time, the multiplication maps must have polynomial formulas whose coefficient heights grow at most quadratically in viℓ. Both requirements are met by a dyadic family. Choose an integer N > 2/η and put U = {N, N + 1, . . . , 2N}. The numbers u2 k , with u ∈ U and k ≥ 0, approximate every real number at least N to relative error at most η. Taking Ri ≥ N maxℓ √Hiℓ, choose viℓ = u2 k , u ∈ U, k ≥ 0, (4.4)” · Show in paper

Minimal admissible subproduct descent over fields of uniformly bounded degree without assuming any degree bound on the marked points.

Source excerpts

“Call a sequence of subproducts Zi = Ys ℓ=0 Ziℓ, yiℓ ∈ Ziℓ ⊆ Yℓ , admissible if the following bounds hold uniformly in i: the geometric degrees of its factors are bounded; all factors are geometrically integral over a field ki ⊇ k with [ki : Q] bounded; and each Ziℓ contains a Zariski dense set of algebraic points of height at most CHiℓ for one constant C. We also allow passage to a subsequence in making such a choice. This definition places no condition on [ki(yiℓ) : ki ].” · Show in paper

Two-measure comparison on the fixed low factor Z0rZ_0^r combined with uniform metric variation via the arithmetic Siu inequality and fundamental height inequalities.

Source excerpts

“We first compute two distinct limiting measures geometrically. We then show that a vanishing subsequence of the arithmetic intersection quotients would force those same measures to agree.” · Show in paper

Uniform local weighted-index estimation using conormal-equation analytic radii and place-by-place metric variation bounds independent of the field degree of the marked point, followed by field norms along finite projections and Faltings' product theorem.

Source excerpts

“For a first nonzero Taylor coefficient, Cauchy’s inequality and the product formula will compare the cost of differentiation, P ℓ |αℓ |Hiℓ, with the small-section gain niM2 i . The relation b 2 iℓHiℓ ≍ M2 i will turn this comparison into the asserted weighted index.” · Show in paper

“We use Faltings’ product theorem [10, Theorem 3.1] in Evertse’s explicit form: for m ≥ 2, 0 < ϵ ≤ 1, and Dℓ/Dℓ+1 ≥ (mN/ϵ) N , (8.5) a common irreducible component of the index loci at levels σ and σ + ϵ is a product of subvarieties of the individual projective factors.” · Show in paper