Headline corollary (the conjecture in sign form).
ProvedWeilDualitySign.DualEigensystem.rootSign_eq_neg_one_pow_degaether-catalogapplications
Headline corollary (the conjecture in sign form). An involutive duality with no
−q^{n/2} fixed point forces the root sign ε = (−1)^d.
theorem WeilDualitySign.DualEigensystem.rootSign_eq_neg_one_pow_deg(hno : ∀ i, E.σ i = i → E.α i ≠ -E.Q) :
E.rootSign = (-1 : K) ^ E.deg := by sorry
Formalization Note Transplanted verbatim from the Aether Catalog source Applications/WeilDualitySign/EigenvalueModel.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/EigenvalueModel.lean
import Mathlib
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.
# The Eigenvalue Model of the Functional-Equation Sign
## The problem
For a smooth projective variety `X / 𝔽_q` of dimension `n`, the middle-degree factor of
the zeta function is a polynomial
`P(T) = ∏_{i=1}^{d} (1 - α_i T)`, `|α_i| = q^{n/2}`,
and **Poincaré duality** acts on the multiset of Frobenius eigenvalues by
`α ↦ q^n / α`. Concretely there is a permutation `σ` of the index set with
`α_i · α_{σ(i)} = q^n = Q²`, `Q := q^{n/2}`.
Substituting `T ↦ 1/(Q² T)` in `P` produces the functional equation
`(Q²T)^d · P((Q²T)⁻¹) = ε · Q^d · P(T)`,
whose **sign** `ε = ±1` is the arithmetic invariant of interest (it is the root number
of the associated `L`-factor, and by the parity philosophy of
`Catalog/Applications/BSD/FunctionalEquation.lean` it governs the parity of the order of
vanishing at the central point).
## What this file proves
The whole sign is controlled by the *fixed points* of the duality involution. A fixed
point satisfies `α_i² = Q²`, hence `α_i = ±Q`; the mission conjecture is that forbidding
the value `α_i = −Q` at fixed points already forces
`∏ α_i = Q^d = q^{nd/2}`, hence `ε = (−1)^d`.
We prove this, and more: the exact sign law
`∏_i α_i = (−1)^{#{i : σ(i) = i ∧ α_i = −Q}} · Q^d` (`prod_alpha_eq_sign_mul_pow`)
together with its sharp converse (`prod_alpha_eq_pow_iff_even`, over any field where
`−1 ≠ 1`): the conjectured conclusion holds **iff** the number of `−Q`-fixed points is
*even*. The hypothesis "no `−Q` fixed point" is therefore sufficient but not necessary,
and the boundary is exactly a `ℤ/2` parity count of fixed points — a Lefschetz-style
statement. Explicit witnesses in degrees `1, 2, 4` (see `Witnesses.lean`) show every
branch is realised.
## Honest scope
`DualEigensystem` is an *axiomatised model*: it records exactly the duality structure
(`σ` an involution with `α_i α_{σ i} = Q²`) that the Weil conjectures supply, over an
arbitrary field. Nothing about the *existence* of such systems for actual varieties, nor
the Riemann-hypothesis bound `|α_i| = q^{n/2}`, is used or claimed — the results are
purely structural consequences of duality, and hence apply verbatim to any cohomological
setting with a perfect pairing into the Tate twist.
-- !-- Lab Notes -- !--
Hypothesis (Hypothesizer): the sign `ε` of a duality-symmetric eigenvalue system is
*not* extra data — it is a `ℤ/2` invariant of the fixed-point set of `σ`, namely the
parity of the number of fixed points carrying the "anti-diagonal" eigenvalue `−Q`.
Experiment (Experimenter): normalise `β_i := α_i / Q`, so duality reads
`β_i β_{σ i} = 1`. Kill the fixed points by hand (`β_i ↦ 1` there) and apply
`Finset.prod_involution` to the resulting function: every genuine 2-cycle cancels,
and `∏ β_i` collapses to the product over `Fix(σ)`, which is a product of `±1`.
Analysis (Analyst): the failed naive route is squaring — `(∏ α_i)² = Q^{2d}` follows in
one line from `σ` being a bijection, but it only determines `∏ α_i` up to sign, which
is precisely the content at stake. The involution/pairing argument is what breaks the
ambiguity, and it genuinely needs `σ ∘ σ = id`: a mere bijection with
`α_i α_{σ i} = Q²` is not enough (see the 4-cycle witness in `Witnesses.lean`).
Critique (Critic): "no `−Q` fixed point" is sufficient but *not* necessary — two `−Q`
fixed points cancel, so the honest theorem is the parity statement
`prod_alpha_eq_pow_iff_even`, which also needs `−1 ≠ 1` (in characteristic 2 the sign
question is empty, and indeed the iff carries that hypothesis).
Synthesis (PI): the conjecture is TRUE, and it is the shadow of the exact law
`ε = (−1)^{d + #neg-fixed}`; the `d = 1` witnesses show both signs occur, so no
hypothesis can be dropped.
-/
open Finset
open WeilDualitySign
open DualEigensystem
variable {K : Type*} [Field K] {ι : Type*} [Fintype ι] [DecidableEq ι]
variable (E : DualEigensystem K ι)
/-! ### Elementary structure -/
/-! ### The pairing cancellation -/
/-! ### The sign law -/
/-! ### The functional equation and its sign -/Formal statement
theorem WeilDualitySign.DualEigensystem.rootSign_eq_neg_one_pow_deg(hno : ∀ i, E.σ i = i → E.α i ≠ -E.Q) :
E.rootSign = (-1 : K) ^ E.deg := by sorrySource