0:006:01

Recursos

Resumen

You can follow, link by link, why no curve can push the multipoint Seshadri constant below the volume bound once there are many general points, and why each step of the proof still holds in characteristic p.

The question and the ceiling

At the r points, the constant is an infimum of L·C divided by the sum of multiplicities. Blowing up the points gives the volume bound ε ≤ (L^n/r)^{1/n} = w [0:20]. The paper claims that for every r ≥ r0(X, L) the constant at the geometric generic r-tuple equals w exactly, in dimension n ≥ 3 and also for n = 2.

The chain of the proof

  • Jets give lower bounds for every curve at once [1:12]. If sections of NL separate jets of order ⌊Nθ⌋, every curve has ratio ≥ θ. So you need jets for every θ < w.
  • Leading exponents turn sections into lattice sets S_N with S_N + S_M ⊆ S_{N+M}. A complete-intersection flag puts them in the simplex Δ(1/D, …, 1/D, D^{n-1}V), and their count is exactly the full count [1:44].
  • Reshaping two intercepts [2:38]: (A, B) → (w, AB/w), with the count and the product of intercepts unchanged. A node bounds the exponents by a quadrilateral of the same area. A shear and slice compressions then make it the triangle you want.
  • Why y² + xy − x³ [3:25]: its tangent cone y(y + x) has two distinct lines in every characteristic, because −1 ≠ 0 even when p = 2. The textbook node y² = x² + x³ fails: in characteristic 2 its cone is (y + x)², so it becomes a cusp [3:50].
  • Interior filling and the Chinese remainder theorem [4:20]: gradedness and full density fill every interior lattice point. Then the CRT interpolates jets at r points (1, …, 1, c_i) in any characteristic.

How to use it

When you read a positive-characteristic proof, test each link [5:16]. Ask whether it uses only algebra (formal expansions, Hensel lifting, the CRT) or quietly needs characteristic zero. One open question: since r0 comes from requiring w < ℓ/100, how small could it really be?

Multipoint Seshadri constants in positive characteristic

OpenAI Math Preprints

Clase 94 / 100

Ver clases