Recursos
Resumen
Pick a smooth projective n-fold X, an ample line bundle L (it may have no sections at all) and a point x. Then explain why K_X + (n+1)L has a section that does not vanish at x, and why n+1 is the smallest number that works.
Why n+1, and why it was hard
- Global generation means that at every point some section is nonzero. It does not mean very ampleness. [0:08]
- Earlier bounds in arbitrary dimension (Angehrn–Siu ~n²/2, then n·log log n, then ~1.78n) all came in above n+1. The usual proof cuts a positive-dimensional center down to a point, and cutting costs positivity. [0:22]
- The bound is sharp. On Pⁿ with O(1), K + (n+1)L = O is generated, while on P² the bundle K + 2L = O(−1) has no sections. [1:00]
The proof chain (by contradiction)
Assume every section vanishes at x. Then [1:45]:
- Strict ordinary-order excess. The mean vanishing order per degree is strictly greater than n/(n+1). [2:45]
- Exponential average. Minimize (1/N) Σ exp(A(v)/t − v(sᵢ)/m) over bases and weighted monomial orders, with t < n+1. The discrepancy pushes the value up and vanishing pulls it down. The minimum is below 1, and it is attained. [2:11]
- Only point centers. Moving along the tangent directions of a positive-dimensional center W gives an upper bound, and Brunn–Minkowski on lattice sums gives a lower bound that contradicts it. The estimate does not depend on the degree or singularities of W. [3:15]
- Lift. A rigid divisor S lies over x. Kawamata–Viehweg vanishing kills H¹(T, D), a fiber value lifts, and it descends to s(x) ≠ 0. That contradicts the assumption. [3:56]
All m ≥ n+1. Apply the theorem to X × Pʳ with r = m − n − 1. [4:36]
Take it with you: when you read an effective positivity proof, ask where it spends its positivity.