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

Every complex K3 surface is Oka

Every complex K3 surface XX is Oka; equivalently, for every integer m≥1m \ge 1, every nonempty compact convex set K⊂CmK \subset \mathbb{C}^m, every open neighborhood UU of KK, every holomorphic map f:U→Xf : U \to X, and every ϵ>0\epsilon > 0, there exists a holomorphic map F:Cm→XF : \mathbb{C}^m \to X such that sup⁡z∈KdX(F(z),f(z))<ϵ\sup_{z \in K} d_X(F(z), f(z)) < \epsilon for any smooth Hermitian distance dXd_X.

If YY is a complex surface with a complete compatible distance such that the local operation LL holds at every point of YY, then YY has the convex approximation property (CAP) in every source dimension.

Every smooth projective complex K3 surface SS has property LL everywhere.

There is a nonempty open subset of the full marked K3 period domain such that every surface with period in this subset has property LL at every point.

For every primitive positive integral vector v∈Λv \in \Lambda, there is a nonempty relatively open subset Ov⊂Dv\mathcal{O}_v \subset \mathcal{D}_v such that every marked K3 surface with period in Ov\mathcal{O}_v has property LL at every point.

Given an infinite chain of holomorphic symplectic charts ϕj:{−2h<Y<Lj+2h}×Δr→M\phi_j : \{-2h < Y < L_j + 2h\} \times \Delta_r \to M with Lj≥1L_j \ge 1 and exact symplectic transitions FjF_j uniformly close to degree-one shears Jj(w,p)=(w+aj(p),p)J_j(w, p) = (w + a_j(p), p), there exist b>0b > 0 and e∗>0e_* > 0 independent of LjL_j and the number of blocks such that for e<e∗e < e_*, the charts sew into a periodic holomorphic symplectic immersion Φ:C×Δb→M\Phi : \mathbb{C} \times \Delta_b \to M.

For every complex K3 surface SS, every x∈Sx \in S, and every nonzero vector v∈TxSv \in T_x S, there is a holomorphic immersion f:C→Sf : \mathbb{C} \to S such that f(0)=xf(0) = x, f′(0)=vf'(0) = v, and f(C)‾=S\overline{f(\mathbb{C})} = S in the ordinary complex topology; if SS is projective, then f(C)f(\mathbb{C}) is Zariski dense.

Every complex Enriques surface is Oka, every connected minimal compact complex surface of Kodaira dimension zero is Oka, and a connected minimal compact complex surface of class VII is Oka if and only if it is a Hopf surface or an Enoki surface.

Every projective complex K3 surface and every complex Enriques surface is elliptic in Gromov's sense.

Methods

Local approximation property L from two transverse complete strip maps using polynomial coordinate selection and bounded additive Cousin splitting.

Source excerpts

“Proposition 2.2 (Two complete directions). Suppose there are holomorphic maps σ1 : C × ∆b −→ X, σ2 : ∆b × C −→ X which are locally biholomorphic at (0, 0) and satisfy σ1(z, 0) = σ2(z, 0) for small z. If σ2 is locally biholomorphic at (0, c), then L holds at x = σ2(0, c).” · Show in paper

“The following is the additive Cousin splitting underlying holomorphic spray gluing; compare [11, Lemma 4.6]. We give its integral construction to retain bounds independent of the number and radii of the passive parameters.” · Show in paper

Polynomial automorphism change of source coordinates to concentrate the exterior of a polydisc into one long planar variable with thin bounded parameters, enabling polydisc approximation.

Source excerpts

“We instead deform the source by a polynomial automorphism. It will be nearly the identity on the old polydisc, while the rest of the enlarged source has coordinates (s, ξ) with s planar and ξ bounded.” · Show in paper

Induction on dimension and removal of defining real faces of a convex polytope one at a time via real-parameter complexification, real-face approximation, and Hölder-trace nonlinear gluing across a real interface.

Source excerpts

“Unlike the buffered overlap in Lemma 3.2, the complex neighborhood on which this accuracy persists may depend on the approximant. We therefore solve the additive jump problem using only its Hölder trace. The nonlinear correction will use the same quadratic iteration as open-overlap gluing, now in this trace norm.” · Show in paper

“Write K′ = ∩_{j=1}^M {ℓj(z) ≤ bj}, where the ℓj’s are real linear forms, and remove these constraints one at a time inside ∆_R^n.” · Show in paper

Exact symplectic sewing of infinite annular chains with uniform bounds independent of chain length, using Fourier splitting and transverse recentering to enforce exactness.

Source excerpts

“Theorem 6.1 (Annular-chain sewing).” · Show in paper

“Lemma 6.4 (Period adjustment). Suppose symplectically normalized charts, together with corresponding model charts, are available for centers varying in larger transverse discs. ... The transition into a newly added block can be made exact by a transverse translation of that block of size bounded by its transition error.” · Show in paper

Semistable degeneration of two quadrics and core-neck sewing of regular rational pencils to produce an unrestricted open region of K3 periods with property L.

Source excerpts

“Choose two general smooth quadrics Q+, Q− ⊂ P^3. Their intersection E is a smooth elliptic quartic. For a general quartic form G consider the family Q+Q− = αG.” · Show in paper

Isotrivial genus-one fibration central model with conservation of imaginary symplectic action to keep recentered annuli away from singular fibers for periods containing a rational direction.

Source excerpts

“Lemma 8.2. For every primitive v ∈ Λ of square 2d > 0, there is a marked projective K3 surface X0 whose period contains v and which admits a genus-one fibration π : X0 → P^1 with all smooth fibers of j-invariant zero.” · Show in paper

Ratner's orbit closure theorem, Borel density, and Torelli theory to transfer property L across period space according to the rational dimension of the period plane.

Source excerpts

“Ratner’s orbit closure theorem [27], applied to H · eΓ in G/Γ, gives H · eΓ = S · eΓ for a connected closed subgroup H ⊂ S ⊂ G such that S ∩ Γ is a lattice in S.” · Show in paper