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.

All papers

Claims

Findings

Schnell fiber spaces and good canonical models

Let f:X→Yf : X \to Y be a surjective morphism with connected fibers between smooth connected projective complex varieties, and let FF be its smooth geometric generic fiber with κ(F)=0\kappa(F) = 0. If there exist an ample Cartier divisor HH on YY and a positive integer m0m_0 such that m0KX−f∗Hm_0 K_X - f^* H is pseudo-effective, then κ(X)=dim⁡Y\kappa(X) = \dim Y.

Let XX be a smooth connected projective complex variety, let DD be an effective Cartier divisor on XX (or an effective rational divisor after clearing denominators), and let m0m_0 be a positive integer. If m0KX−Dm_0 K_X - D is pseudo-effective, then κ(X)≥κ(D)\kappa(X) \geq \kappa(D).

Let f:X→Yf : X \to Y be a surjective morphism with connected fibers between smooth connected projective complex varieties, and let FF be a very general smooth fiber. If HH is an ample Cartier divisor on YY and m0m_0 is a positive integer such that m0KX−f∗Hm_0 K_X - f^* H is pseudo-effective, then κ(X)=κ(F)+dim⁡Y\kappa(X) = \kappa(F) + \dim Y, and there exist positive integers rr and ℓ0\ell_0 such that H0(X,OX(ℓrKX−f∗H))≠0H^0(X, \mathcal{O}_X(\ell r K_X - f^* H)) \neq 0 for every integer ℓ≥ℓ0\ell \geq \ell_0.

Let f:X→Yf : X \to Y be a surjective morphism with connected fibers between smooth connected projective complex varieties, with smooth geometric generic fiber FF satisfying κ(F)=0\kappa(F) = 0. Assume XX has a projective good klt minimal model. If HH is an ample Cartier divisor on YY and m0m_0 is a positive integer such that m0KX−f∗H∈Eff‾(X)m_0 K_X - f^* H \in \overline{\mathrm{Eff}}(X), then κ(X)=dim⁡Y\kappa(X) = \dim Y.

Let f:X→Yf : X \to Y be a surjective morphism with connected fibers between smooth connected projective complex varieties. If its smooth geometric generic fiber FF satisfies κ(F)=0\kappa(F) = 0, then κ(X)≤dim⁡Y\kappa(X) \leq \dim Y.

Let q:W→Vq : W \to V be a birational morphism from a smooth projective complex dd-fold to a normal projective variety, DD a globally generated Cartier divisor on VV with image ZZ of dimension kk, AA a very ample Cartier divisor on VV, and h:W→Yh : W \to Y a surjective morphism to a smooth projective variety with y=dim⁡Y>ky = \dim Y > k. For every ample Cartier divisor HH on YY, the numerical one-cycle class C=(q∗D)k(q∗A)d−k−1C = (q^* D)^k (q^* A)^{d-k-1} satisfies: (1) B⋅C≥0B \cdot C \geq 0 for every pseudo-effective real divisor class BB on WW; (2) E⋅C=0E \cdot C = 0 for every qq-exceptional rational divisor EE; and (3) q∗D⋅C=0q^* D \cdot C = 0 and h∗H⋅C>0h^* H \cdot C > 0.

For a smooth connected projective complex variety TT, if KT∈Eff‾(T)K_T \in \overline{\mathrm{Eff}}(T), then κ(T)≥0\kappa(T) \geq 0.

Methods

Mixed intersection test on a common resolution: constructing a numerical one-cycle class C=(q∗D)k(q∗A)d−k−1C = (q^* D)^k (q^* A)^{d-k-1} from a globally generated divisor DD and very ample divisor AA on a good minimal model VV to test pseudo-effectivity, annihilate qq-exceptional divisors and q∗Dq^* D, and detect an ample divisor h∗Hh^* H from a separate base YY via local cotangent-space transversality and perturbation.

Source excerpts

“For every ample Cartier divisor H on Y , the numerical one-cycle class C = (q∗D)k(q∗A)d−k−1 satisfies the following properties: 1. B · C ≥ 0 for every pseudo-effective real divisor class B on W; 2. E · C = 0 for every q-exceptional rational divisor E; 3. q∗D · C = 0 and h∗H · C > 0.” · Show in paper

“To use this local intersection in the global intersection number, perturb the hyperplanes slightly in their complex parameter spaces. The transverse point persists for every sufficiently small perturbation, by the implicit function theorem.” · Show in paper

Pluricanonical section transfer from a good minimal model: pulling back sections of a globally generated multiple D=rKVD = r K_V along q:W→Vq : W \to V and multiplying by the canonical section of the effective qq-exceptional divisor rEr E to obtain an injection H0(V,D)↪H0(X,rKX)H^0(V, D) \hookrightarrow H^0(X, r K_X) that preserves section ratios on the complement of Supp(E)\mathrm{Supp}(E).

Source excerpts

“Pull back sections of D by q and multiply by the canonical section of the effective Cartier divisor rE. We obtain an injection H0(V, D) ,→ H0(W, rp∗KX) = H0(X, rKX). (2.5)” · Show in paper

Iterative divisor-to-fiber-space reduction: reducing the Campana–Peternell inequality to the zero-Kodaira fiber-space case by iteratively applying Stein factorization and base resolution to pluricanonical/twisted linear systems until reaching a fiber space with κ(Fj)=0\kappa(F_j) = 0.

Source excerpts

“As long as κ(Fi) > 0, the integer dim Yi strictly increases. It is bounded by dim X, so this process reaches an index j with κ(Fj) = 0.” · Show in paper