The non-fixed indices carrying the central eigenvalue also come in duality pairs.
ProvedWeilDualitySign.DualEigensystem.even_card_nonfixed_centralaether-catalogapplications
The non-fixed indices carrying the central eigenvalue also come in duality pairs.
theorem WeilDualitySign.DualEigensystem.even_card_nonfixed_central:
Even ((univ.filter (fun i => ¬ E.σ i = i ∧ E.α i = E.Q)).card) := by sorry
Formalization Note Transplanted verbatim from the Aether Catalog source Applications/WeilDualitySign/CentralParity.lean; the statement is byte-identical to the source declaration, elaborated with autoImplicit disabled in the platform environment.
Preamble
-- Thm stub generated from Applications/WeilDualitySign/CentralParity.lean
import Mathlib
import Definitions.Def_Applications_WeilDualitySign_CentralParity
import Definitions.Def_Applications_WeilDualitySign_EigenvalueModel
/-
Copyright (c) 2026 Harmonic. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
# Cycle 2: the sign is the parity of the central order of vanishing
`EigenvalueModel.lean` computed the functional-equation sign of a duality eigensystem as
`ε = (−1)^{d + #neg-fixed}`.
That formula refers to the duality permutation. This file eliminates `σ` from the answer
entirely and proves the **intrinsic** law
`ε = (−1)^{m₊}`, `m₊ := #{ i : α_i = Q }`,
where `m₊` is the multiplicity of the "central" eigenvalue `α = q^{n/2}`. Since
`P(T) = ∏ (1 - α_i T)` vanishes at the central point `T = Q⁻¹` to order exactly `m₊`
(`charPoly_factor_central`, `centralFactor_ne_zero`), this says:
> **the sign of the functional equation is the parity of the order of vanishing at the
> central point** —
the exact function-field analogue of the parity statement proved analytically for
`Λ(2 - s) = w Λ(s)` in `Catalog/Applications/BSD/FunctionalEquation.lean`
(`analyticRank_parity`). There the input was Taylor symmetry of an analytic function;
here it is a purely combinatorial pairing of Frobenius eigenvalues. The two theorems are
the archimedean and the finite-field faces of one statement.
The bridge from cycle 1 is a `ℤ/2` count (`even_deg_add_negFixed_add_centralOrder`):
duality 2-cycles contribute evenly to *both* `d` and `m₊`, `+Q` fixed points contribute
oddly to both, and `−Q` fixed points contribute `1` to `d` and `1` to `#neg-fixed`. The
hypothesis `−1 ≠ 1` is necessary and sharp: in characteristic `2` the two fixed-point
types coincide and the bookkeeping collapses.
-- !-- Lab Notes -- !--
Hypothesis (Hypothesizer): `ε` should not depend on `σ` at all — `σ` is auxiliary data,
while the multiset `{α_i}` is intrinsic. The only intrinsic `ℤ/2` invariant available
is the multiplicity of the self-dual eigenvalue `+Q`.
Experiment (Experimenter): count. Split `ι` into `Fix(σ)` and its complement `R`. On
`R`, `σ` is a free involution, so `|R|` is even; and `{α_i = Q}` is `σ`-stable
(`α_i = Q ⟹ α_{σ i} = Q²/Q = Q`), so its intersection with `R` is also even. The two
even contributions cancel mod `2` and leave `m₊ ≡ |Fix₊|` and `d + #neg ≡ |Fix₊|`.
Analysis (Analyst): the same "free involution ⟹ even" principle is used twice, so it is
isolated as `even_card_of_free_involution`; it is the combinatorial engine of the whole
project (`Finset.prod_involution` over `ℤ` with constant value `−1`).
Critique (Critic): in characteristic `2` the statement `ε = (−1)^{m₊}` degenerates
(`ε = 1` always), and the counting argument genuinely fails, so the theorems carry
`hchar : (-1 : K) ≠ 1`. The central factorisation, by contrast, needs no such
hypothesis and is stated in full generality.
Synthesis (PI): cycle 1's sign law plus this parity bridge give a self-contained,
σ-free statement of the conjecture: *no `−q^{n/2}` self-dual eigenvalue* forces
`ε = (−1)^d`, and *always* `ε = (−1)^{ord_{T = q^{-n/2}} P}`.
-/
open Finset
open scoped Classical
open WeilDualitySign
open DualEigensystem
variable {K : Type*} [Field K] {ι : Type*} [Fintype ι] [DecidableEq ι]
variable (E : DualEigensystem K ι)
/-! ### The central eigenvalue set is duality-stable -/Formal statement
theorem WeilDualitySign.DualEigensystem.even_card_nonfixed_central:
Even ((univ.filter (fun i => ¬ E.σ i = i ∧ E.α i = E.Q)).card) := by sorrySource