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

The reverse logarithmic Kodaira inequality and additivity

Let f:X→Yf : X \to Y be a surjective morphism with connected fibers between smooth connected projective varieties over C\mathbb{C}, and let EE and DD be reduced simple-normal-crossing divisors on XX and YY, respectively, with Supp⁡(f∗D)⊆Supp⁡(E)\operatorname{Supp}(f^*D) \subseteq \operatorname{Supp}(E). Assume stratum smoothness over V=Y∖Supp⁡DV = Y \setminus \operatorname{Supp} D, i.e., XX and every irreducible component of every nonempty intersection of components of EE, restricted to f−1(V)f^{-1}(V), are smooth over VV. For a very general y∈Vy \in V, set F=XyF = X_y, EF=E∣FE_F = E|_F, LX=KX+EL_X = K_X + E, LY=KY+DL_Y = K_Y + D, and LF=KF+EFL_F = K_F + E_F. Then κ(X,LX)≤κ(Y,LY)+κ(F,LF)\kappa(X, L_X) \leq \kappa(Y, L_Y) + \kappa(F, L_F), and if either term on the right is −∞-\infty, then H0(X,mLX)=0H^0(X, mL_X) = 0 for every m>0m > 0.

For the pairs f:(X,E)→(Y,D)f : (X, E) \to (Y, D) in Theorem 1.1, logarithmic additivity holds: κ(X,KX+E)=κ(Y,KY+D)+κ(F,KF+EF)\kappa(X, K_X + E) = \kappa(Y, K_Y + D) + \kappa(F, K_F + E_F), resolving Popa's logarithmic additivity conjecture positively in the projective reduced-SNC, stratum-smooth setting.

Under the hypotheses of Theorem 1.1, if 0≠s∈H0(X,mLX)0 \neq s \in H^0(X, mL_X) for some m>0m > 0, then κ(Y,LY)≥0\kappa(Y, L_Y) \geq 0. Moreover, if κ(Y,LY)=0\kappa(Y, L_Y) = 0 and ss vanishes identically on XyX_y for some y∈Vy \in V, there exist an integer a>0a > 0 and a nonzero τ∈H0(Y,aLY)\tau \in H^0(Y, aL_Y) with τ(y)=0\tau(y) = 0.

There are a dense smooth open Y∘⊂VY^\circ \subset V, a polarizable pure rational variation V\mathbb{V} on Y∘Y^\circ with an invariant lattice and quasi-unipotent boundary monodromy, and a nonzero algebraic map u:OY∘(−LY)→Edu : \mathcal{O}_{Y^\circ}(-L_Y) \to \mathcal{E}^d, where Ep=Gr⁡Fp(V⊗OY∘)\mathcal{E}^p = \operatorname{Gr}_F^p(\mathbb{V} \otimes \mathcal{O}_{Y^\circ}) and Fd+1V=0F^{d+1}\mathbb{V} = 0, such that on every higher log-smooth model (Y′,D′)(Y', D'), uu, all its Higgs iterates θj(u)\theta^j(u), and any homogeneous tensor replacement u∗u_* of degree a>0a > 0 and its iterates θj(u∗)\theta^j(u_*) have regular coefficients at every prime of Y′Y' in the Schmid-extended frames.

Given a nonzero algebraic morphism u:L→FpVCu : \mathcal{L} \to F^p\mathbb{V}_{\mathbb{C}} (Fp+1VC=0F^{p+1}\mathbb{V}_{\mathbb{C}} = 0) on a smooth irreducible complex quasi-projective variety B∘B^\circ, there exist a>0a > 0 and a homogeneous tensor replacement u∗:L⊗a→E0=Fp∗Au_* : \mathcal{L}^{\otimes a} \to E_0 = F^{p_*}\mathbb{A} (Fp∗+1A=0F^{p_*+1}\mathbb{A} = 0) descending to B∘B^\circ (after shrinking), and on smooth projective birational models a dominant connected-fiber morphism b:B→Sb : B \to S such that A\mathbb{A} descends to S∘S^\circ with irreducible connected representation factoring through a rational semisimple adjoint group GG, there is an algebraic compact-flag section juj_u making j=(E0,ju)j = (E_0, j_u) generically immersive with the full Lie-period map, and SS carries nef rational line bundles H\mathcal{H} and Hu\mathcal{H}_u with H−Hu\mathcal{H} - \mathcal{H}_u nef, ν(H)=rk⁡(dj)\nu(\mathcal{H}) = \operatorname{rk}(dj), KS+D0+cHuK_S + D_0 + c\mathcal{H}_u big for c≫0c \gg 0, and KS+b0HK_S + b_0\mathcal{H} big for b0≫0b_0 \gg 0.

For every component DiD_i of the boundary divisor D0D_0 on SS with nonidentity local monodromy on g\mathfrak{g}, ν(H∣Di)<ν(H)\nu(\mathcal{H}|_{D_i}) < \nu(\mathcal{H}); if ν(H)=0\nu(\mathcal{H}) = 0, then D0D_0 is empty.

If dim⁡S>0\dim S > 0, then KS+bHK_S + b\mathcal{H} is big for all sufficiently large rational bb.

For any dominant connected-fiber morphism h:W→Sh : W \to S of smooth connected projective varieties and any algebraic line subbundle ℓ⊂h∗E0\ell \subset h^*E_0 on a dense open of WW inducing the pulled-back compact marks h∗juh^*j_u, the Higgs-tail line P=−∑j≥0det⁡Uj=−∑i≥0(i+1)det⁡GiP = -\sum_{j \geq 0} \det U_j = -\sum_{i \geq 0} (i+1)\det G_i on a smooth connected unipotent level π:W^→W\pi : \widehat{W} \to W is nef and satisfies ν(P+π∗h∗H)=ν(P)\nu(P + \pi^*h^*\mathcal{H}) = \nu(P).

For the relative Iitaka diagram Y→xW→hSY \xrightarrow{x} W \xrightarrow{h} S of Lemma 6.1, after further smooth birational preparation, the divisor LW=KW+TWL_W = K_W + T_W is big over SS, satisfies H0(W,lLW)≃H0(Y0,l(KY0+DY0))H^0(W, lL_W) \simeq H^0(Y_0, l(K_{Y_0} + D_{Y_0})) in every sufficiently divisible degree ll, and decomposes as LW=KW+BW+MWL_W = K_W + B_W + M_W with 0≤BW≤10 \leq B_W \leq 1 having SNC support and MWM_W nef.

Let h:W→Sh : W \to S be a surjective morphism of smooth projective varieties with dim⁡S>0\dim S > 0, LL a rational divisor on WW, and HH a nef rational divisor on SS. If LL is big on the generic fiber of hh, L−h∗KSL - h^*K_S is pseudoeffective, KS+bHK_S + bH is big for some b≥0b \geq 0, and there exist a dominant generically finite π:W^→W\pi : \widehat{W} \to W, a nef PP on W^\widehat{W}, and Q>0Q > 0 with ν(P+π∗h∗H)=ν(P)\nu(P + \pi^*h^*H) = \nu(P) and Qπ∗L≥pePQ\pi^*L \geq_{\mathrm{pe}} P, then LL is big.

Methods

Logarithmic Hodge vector construction via cyclic root cover, weight-quotient projection, and Kovács-style connecting maps retaining the original base cotangent lattice ΩY′1(log⁡D′)\Omega^1_{Y'}(\log D'), tested for divisorial regularity by one-parameter semistable reduction.

Source excerpts

“The cyclic-cover construction and its Higgs iteration follow the Viehweg–Zuo method [36, §3]; here we must also retain every coefficient in the original base lattice.” · Show in paper

“One-parameter semistable reduction, after finite ramification, gives a projective semistable family with smooth total space g : Z_\Delta \to \Delta with reduced simple-normal-crossing special fiber V_0, and a horizontal divisor H such that V_0 + H has simple normal crossings.” · Show in paper

Homogeneous tensor replacement minimizing the number of retained rational simple monodromy factors, combined with Kempf's optimal one-parameter subgroup theorem to attach algebraic compact-flag sections juj_u on invisible factors.

Source excerpts

“Among all such homogeneous replacements, choose one with the fewest retained rational simple factors.” · Show in paper

“Kempf’s optimal one-parameter subgroup theorem [22, §3], in its origin-instability form [21, Theorem 6], applied to the affine representation and the closed origin, assigns to an unstable vector a unique optimal parabolic.” · Show in paper

Marked-period quotient and variation descent using Bakker–Brunebarbe–Tsimerman period-image algebraicity and generic Lie stabilizers.

Source excerpts

“The algebraicity theorem for pure integral period maps [2, Theorem 1.1] factors the period map through a dominant algebraic map to a quasi-projective algebraic image.” · Show in paper

Full-mass Monge–Ampère comparison for singular Hermitian metrics with two-sided logarithmic-logarithmic weight bounds near an SNC boundary to deduce nefness and identify numerical dimension with generic curvature rank.

Source excerpts

“The induced line metrics therefore have two-sided logarithmic-logarithmic weight bounds. A curvature computation on the open set alone would not identify their intersection numbers. The following full-mass argument makes that comparison.” · Show in paper

Flat-connection Ax–Schanuel estimate and orbit-filling calculation for local fibers of flag and projective-line maps.

Source excerpts

“The estimate is the semisimple case of the connection Ax–Schanuel framework of [3, Theorem A, Theorem 3.6, and Remark 3.7].” · Show in paper

Relative logarithmic Iitaka fibration over the marked-period quotient combined with extraction of finitely many source valuations and lc-trivial b-nefness to obtain an exact section-lattice comparison.

Source excerpts

“The next lemma extracts precisely those finitely many base valuations; subsequent source modifications add no new sections.” · Show in paper

Orbifold cotangent transfer on adapted covers and uniform pseudoeffective threshold via Campana–Păun orbifold cotangent positivity and ample perturbation.

Source excerpts

“We use [6, Theorem 1.3] in its orbifold tensor form: for a smooth projective SNC pair (Z, \Delta) with rational coefficients in [0, 1] and K_Z + \Delta pseudoeffective, every torsion-free quotient of an orbifold cotangent tensor power on an adapted cover has nonnegative degree against the pullback of every movable curve class on Z.” · Show in paper