0:005:17

Recursos

Resumen

Take a rational Hodge class on an abelian variety with good reduction, and a divisor that shows up only after reduction. This class shows why their pairing is a rational number, and which step of the proof makes it one.

The problem

An algebraic cycle specializes to the reduction, and its intersection with any complementary product of divisors is a rational degree that every cohomology theory agrees on. A Hodge class is not known to be a cycle. Deligne's absolute-Hodge theorem gives it compatible ℓ-adic and crystalline realizations. It does not say that its pairing with divisors that do not lift is rational [0:06].

The theorem

Let A have dimension d, γ ∈ H^{2r}(A(ℂ), ℚ(r)) be a rational Hodge class, and D₁, …, D_{d−r} be divisors on A₀. Then the trace of γ₀ · c₁(D₁)⋯c₁(D_{d−r}) is one rational number q. You get the same q in every ℓ-adic realization (ℓ ≠ p) and in crystalline cohomology extended to C_w. This holds for every prime, including p = 2 [0:36].

The route

  • Move only along rational pullbacks. The Lefschetz pairing property says every pairing with a rational polynomial in divisors is one rational number. Lemma 2.3 shows that rational pullback preserves it, via the projection formula [1:25].
  • Transport to CM. Corollary 3.4 gives γ₀ = (a_CM j)*κ₀ for a CM variety C. The similitude scalars from each realization are normalized to one rational number using divisor intersections [1:53].
  • Reduce to balanced Weil classes. These live on pieces of rank 2s over a CM field E, with signature (s, s) at every embedding. Their pullbacks span the CM Hodge class [2:47].
  • Use the dichotomy (Prop. 5.2). Let 𝒰 be the set of Rosati-fixed endomorphisms that twist the E-action by complex conjugation. If 𝒰 has no invertible element, every pairing is 0. If it has one, the class is a rational polynomial in divisors in every realization [3:15], [3:46].
  • The hard branch: a filtration that preserves E, u and the polarization, realized on an auxiliary variety B that need not be CM. There, coefficient extraction gives a rational divisor identity [4:06].

When you face a new case: transport to CM, split into balanced Weil classes, and test 𝒰 for an invertible element. With the CM Hodge theorem added, γ₀ is even represented by a rational algebraic cycle on the special fiber. The proof says nothing about lifting that cycle to the generic fiber, or about the Hodge conjecture in characteristic 0 [4:46].

Milne's rationality conjecture for abelian varieties

Ver clases