Recursos
Resumen
You will be able to take a compact Kähler manifold, read off its algebraic dimension and whether it is simple, and say which part of the proof of log abundance handles it. You will also see why an analytically nef K_X + Δ then becomes semiample.
The gap the theorem closes
On a Kähler space, nef is analytic. The class c₁(K_X + Δ) lies in the closure of the Kähler cone, which means there are metrics whose curvature is at least −εω [0:09]. Semiample asks for more: an actual multiple whose global sections generate it at every point. The paper assumes logarithmic Iitaka subadditivity for smooth projective varieties with reduced SNC boundaries. From that it proves that for a log canonical pair (X, Δ), with X normal, irreducible and compact Kähler, nef implies semiample [0:52].
The engine: a decomposition with its negative part
The induction proves a stronger statement, G_n [1:28]. Let X be smooth compact Kähler, let B be an SNC boundary with coefficients in [0,1], and let J = K_X + B be pseudo-effective. Then after a modification μ:
- μJ ∼ℚ P + R, where P is semiample and R = N(μJ), the negative part (the sum over prime divisors D of the minimal multiplicities ν_D times D).
- This is an identity of line bundles, not only of Chern classes. That extra information is what lets sections come back down.
Worked case [2:31]: take a log resolution whose boundary is the strict transform of Δ plus all exceptional divisors at coefficient one. The adjoint is pseudo-effective, so G_n applies. Because the pullback is nef, R = 0 and P is the pullback itself. Normality and the projection formula then give a globally generated multiple of K_X + Δ.
Routing a space through the induction
Ask about a(X) first, then about simplicity [3:27]:
- a(X) = n: X is Moishezon and Kähler, hence projective. The projective good-model theorem applies, and this is where Assumption 1.1 is used.
- 0 < a(X) < n: the algebraic reduction gives a fibration, which is the fibration case (Prop 2.11).
- a(X) = 0, simple: P is torsion and J ∼ℚ N(J). The starting divisor comes from meromorphic nonvanishing (Theorem 6.1): some power of K_X has a nonzero meromorphic section [4:14].
- a(X) = 0, not simple: Campana's covering family gives a generically finite cover X′ with a fibration. P is torsion there, and Prop 2.9 brings J ∼ℚ N(J) back down to X. This is the practice case [4:57].
Next, read Sections 5 and 7 of the paper, and see how Section 6's meromorphic section gets the simple case started [5:44].