Cohomology of a smooth projective variety over a finite field: trace formula, Poincaré duality and Deligne's bound on
OpenWeilConjectures.exists_frobenius_cohomology_weight_boundLet be a finite field of characteristic and let be a non-singular projective variety of dimension over : the structure morphism is smooth of relative dimension , geometrically irreducible and projective. For let .
Theorem. There exist a field of characteristic zero and finite-dimensional -vector spaces with endomorphisms such that:
- Lefschetz trace formula. For all ,
- Poincaré duality. For every there is a perfect bilinear pairing with .
- . is one-dimensional and is the identity.
- Weights (Deligne). For every , every eigenvalue of is algebraic over , and every complex conjugate of satisfies
Precisely: there is a nonzero such that the characteristic polynomial divides in and all complex roots of have absolute value at most .
With for a prime , and the action of the Frobenius endomorphism, items 1 and 2 are Grothendieck's trace formula and Poincaré duality, item 3 holds because is geometrically connected, and item 4 follows from Deligne's theorem that the eigenvalues of are algebraic integers all of whose complex conjugates have absolute value exactly ; one may take to be the product of the minimal polynomials over of the eigenvalues, with multiplicities.
Only the upper bound is asked for in item 4. Together with items 1 to 3 it implies the equality , the integrality of the polynomials and the Riemann hypothesis for in the form of WeilConjectures.riemann_hypothesis; this implication is the Proved node WeilConjectures.weil_riemann_hypothesis_of_cohomology_weight_bound. Items 1 and 2 alone are the node WeilConjectures.exists_frobenius_cohomology_poincare_duality.
Formalization Note The hypotheses are exactly those of WeilConjectures.riemann_hypothesis; is pointCount p f m and is Nat.card k. The spaces are given in coordinates: is d i and is a square matrix Φ i of size d i; the indices are not constrained. The pairing on is given by its Gram matrix G with a two-sided inverse H, and the compatibility with Frobenius reads . Item 3 is d 0 = 1 ∧ Φ 0 = 1. In item 4, "complex root of " is aeval z W = 0 for z : ℂ and is the real power (Nat.card k : ℝ) ^ ((i : ℝ) / 2). The field K lives in Type.
import Mathlib import Definitions.Def_WeilConjectures_projectiveSpace import Definitions.Def_WeilConjectures_zetaFunction open AlgebraicGeometry CategoryTheory Polynomial open WeilConjectures
theorem WeilConjectures.exists_frobenius_cohomology_weight_bound (p : ℕ) [Fact p.Prime]
(k : Type) [Field k] [Finite k] [CharP k p] (n : ℕ) (X : Scheme.{0}) (f : X ⟶ Spec (.of k))
[SmoothOfRelativeDimension n f] [GeometricallyIrreducible f]
(hproj : IsProjectiveOver k f) :
∃ (K : Type) (_ : Field K) (_ : CharZero K) (d : ℕ → ℕ)
(Φ : (i : ℕ) → Matrix (Fin (d i)) (Fin (d i)) K),
(∀ m : ℕ, 0 < m → (pointCount p f m : K) =
∑ i ∈ Finset.range (2 * n + 1), (-1 : K) ^ i * ((Φ i) ^ m).trace) ∧
(∀ i ≤ 2 * n, ∃ (G : Matrix (Fin (d i)) (Fin (d (2 * n - i))) K)
(H : Matrix (Fin (d (2 * n - i))) (Fin (d i)) K),
G * H = 1 ∧ H * G = 1 ∧
(Φ i).transpose * G * Φ (2 * n - i) = (Nat.card k : K) ^ n • G) ∧
(d 0 = 1 ∧ Φ 0 = 1) ∧
(∀ i ≤ 2 * n, ∃ W : ℚ[X], W ≠ 0 ∧ (Φ i).charpoly ∣ W.map (algebraMap ℚ K) ∧
∀ z : ℂ, aeval z W = 0 → ‖z‖ ≤ (Nat.card k : ℝ) ^ ((i : ℝ) / 2)) := by sorry