B6: local Tate duality for ℚ₂ — the dual module and the duality bundle #
This file provides the statement infrastructure for the paper's local Tate duality leaf
(B6): the μₙ-dual of a finite discrete G_ℚ₂-module, the evaluation cup pairing, and the
bundle TateDuality n packaging the invariant map inv : H²(G_ℚ₂, μₙ) ≃+ ℤ/n together with
perfectness of the cup pairing in the three degree pairs. The axiom itself
(GQ2.tateDualityAt : ∀ n [NeZero n], TateDuality n) lives in
GQ2/Foundations/Axioms.lean; everything here is definitions plus axiom-free,
bundle-parametrized stress tests.
Encoding decisions #
- Per-
nform, not the colimit. The literature states duality withμ = ⋃ₙ μₙandH²(G_k, μ) ≅ ℚ/ℤ. We state it pern(forn-torsion modules), which suffices for the paper (onlyn = 2and𝔽₂-modules are used, §§5–8) and avoids colimits of modules. Deviation flagged: cross-ncompatibility of theinv's (restriction alongμₙ ⊆ μₙₘ) is not asserted. - Pontryagin-dual encoding. For an
n-torsion finite abelianA,Hom(A, ℚ/ℤ) = Hom(A, ⅟n·ℤ/ℤ) ≅ Hom(A, ℤ/n); so "the dual ofH" is encoded as the plain hom-groupH →+ ZMod n, and noAddCircle/ℚ⧸ℤis needed. - The dual module
M′ = Hom(M, μₙ)isMuDual n M, adef(notabbrev) type synonym ofM →+ MuN n: Mathlib has a codomain-only action instance onA →+ B(DistribMulAction M (A →+ B),Mathlib/Algebra/GroupWithZero/Action/Hom.lean), so the Galois conjugation action(g • φ)(m) = g • φ(g⁻¹ • m)must live on a synonym to avoid an instance diamond. Continuity of the conjugation action is via open stabilizers (continuousSMul_iff_stabilizer_isOpen+isOpen_iInf_stabilizer: the joint action kernel onMandμₙis open and stabilizes everyφ). - Perfectness, single currying. For each degree pair —
(0,2),(1,1),(2,0), i.e. exactly the three cup shapes withM′in the left slot and the evaluation pairingmuDualPairing : M′ →+ M →+ μₙ— the clause asserts thatx ↦ inv ∘ (x ∪ ·) : Hⁱ(M′) → (H^{2−i}(M) →+ ZMod n)is bijective. Deviation flagged: the opposite currying (H^{2−i}(M) → Hom(Hⁱ(M′), ℤ/n)) is not asserted; for finite cohomology groups (B7) it follows by counting, and the paper consumes only the stated direction. - No normalization of
inv. The literature pinsinvdown (via the valuation map and Frobenius); the bundle only asserts existence of aninvmaking the pairings perfect, which is what the paper's dimension counts use. The explicitn = 2cup values enter through B7′ (the Hilbert symbol), not throughinv. Deviation flagged for review.
Citations #
NSW, Ch. VII §7.2, Theorem (7.2.6) (local Tate duality: the cup pairing
Hⁱ(G_k, M′) × H^{2−i}(G_k, M) → H²(G_k, μ) = ℚ/ℤ is non-degenerate for finite M);
Serre, Galois Cohomology II §5.2, Theorem 2; Milne, Arithmetic Duality Theorems, I.2.3.
Paper: §§5–8 (dimension counts over 𝔽₂), docs/literature-axioms.md B6.
The μₙ-dual of a discrete module #
The μₙ-dual module M′ = Hom(M, μₙ) of a discrete G-module M, with the conjugation
action (g • φ)(m) = g • φ(g⁻¹ • m) (G a local Galois group, e.g. G_ℚ₂ or a finite-index
subgroup G_K). A def (not abbrev): Mathlib's codomain-only action on M →+ MuN n must not
be found here (see module docstring). The type is group-free; only the action below depends
on G.
Equations
- GQ2.MuDual n M = (M →+ GQ2.MuN n)
Instances For
Equations
- One or more equations did not get rendered due to their size.
Equations
- GQ2.instFunLikeMuDualMuN n M = { coe := GQ2.instFunLikeMuDualMuN._aux_1 n M, coe_injective := ⋯ }
Extensionality for the dual module, keyed to the synonym's own head.
Evaluation of the zero dual (the synonym's FunLike head keeps Mathlib's
AddMonoidHom simp set from firing; these rfl-lemmas replace it).
Equations
- GQ2.instTopologicalSpaceMuDual n M = ⊥
The conjugation action of G on Hom(M, μₙ).
Equations
- One or more equations did not get rendered due to their size.
Continuity of the conjugation action (for finite M): the joint action kernel on M
and μₙ is an open subgroup fixing every φ, so all stabilizers are open.
The evaluation pairing Hom(M, μₙ) →+ M →+ μₙ — under the type synonym, literally the
identity. This is the μ fed to the cup products in the duality clauses.
Equations
- GQ2.muDualPairing n M = AddMonoidHom.id (M →+ GQ2.MuN n)
Instances For
Equivariance of the evaluation pairing — the hμ hypothesis of the cup products.
μₙ is n-torsion (needed to feed μₙ itself to the duality) #
μₙ is n-torsion: n • x = 0 additively, i.e. ζⁿ = 1.
The duality bundle #
B6 (local Tate duality), the bundle at a local Galois group G — per-n form (see the
module docstring for the encoding decisions and flagged deviations). G is a local Galois
group (G_ℚ₂, or a finite-index subgroup G_K for K/ℚ₂ finite; the axiom
GQ2.tateDualityAt supplies an instance for exactly these G). inv identifies H²(G, μₙ)
with ℤ/n, and for every finite discrete n-torsion G-module M the evaluation cup pairing
is perfect in the three degree pairs, in the sense that x ↦ inv ∘ (x ∪ ·) is a bijection onto
the Pontryagin dual H^{2−i}(G, M) →+ ZMod n.
Modules are quantified over Type (Type 0): every finite module is isomorphic to one there,
and all of the paper's coefficients (𝔽₂-modules, μₙ, duals) live there.
- inv : ContCoh.H2 G (MuN n) ≃+ ZMod n
The invariant map:
H²(G, μₙ) ≅ ℤ/n(unnormalized; see deviations). - perfect02 (M : Type) [AddCommGroup M] [TopologicalSpace M] [DiscreteTopology M] [DistribMulAction G M] [ContinuousSMul G M] [Finite M] : (∀ (x : M), n • x = 0) → Function.Bijective fun (c : ↥(ContCoh.H0 G (MuDual n M))) => self.inv.toAddMonoidHom.comp ((ContCoh.cup02 (muDualPairing n M) ⋯) c)
Perfectness in degrees
(0, 2):H⁰(M′) ≅ Hom(H²(M), ℤ/n). - perfect11 (M : Type) [AddCommGroup M] [TopologicalSpace M] [DiscreteTopology M] [DistribMulAction G M] [ContinuousSMul G M] [Finite M] : (∀ (x : M), n • x = 0) → Function.Bijective fun (c : ContCoh.H1 G (MuDual n M)) => self.inv.toAddMonoidHom.comp ((ContCoh.cup11 (muDualPairing n M) ⋯) c)
Perfectness in degrees
(1, 1):H¹(M′) ≅ Hom(H¹(M), ℤ/n). - perfect20 (M : Type) [AddCommGroup M] [TopologicalSpace M] [DiscreteTopology M] [DistribMulAction G M] [ContinuousSMul G M] [Finite M] : (∀ (x : M), n • x = 0) → Function.Bijective fun (c : ContCoh.H2 G (MuDual n M)) => self.inv.toAddMonoidHom.comp ((ContCoh.cup20 (muDualPairing n M) ⋯) c)
Perfectness in degrees
(2, 0):H²(M′) ≅ Hom(H⁰(M), ℤ/n).
Instances For
B6 at the base field ℚ₂ — the bundle over G_ℚ₂ = AbsGalQ2 (the k = ℚ₂ member of the
base-generalized family). An abbreviation, so every existing G_ℚ₂ consumer of TateDuality
is unchanged.
Equations
Instances For
G is a local dualizing group over ℚ₂ — the truth-side hypothesis gating the
base-generalized B6 axiom GQ2.tateDualityAt: G embeds topologically as a
open finite-index subgroup of G_ℚ₂, compatibly with the μₙ-action. Such
G are exactly the G_K = Gal(ℚ̄₂/K) for K/ℚ₂ finite (K = the fixed field of the image),
for which local Tate duality holds; G = G_ℚ₂ is the identity embedding.
Equations
- GQ2.IsLocalDualizingGroup G n = ∃ (ι : G →* GQ2.AbsGalQ2), Topology.IsOpenEmbedding ⇑ι ∧ ι.range.FiniteIndex ∧ ∀ (g : G) (x : GQ2.MuN n), g • x = ι g • x
Instances For
G_ℚ₂ itself is a local dualizing group (the identity embedding).
Stress tests (axiom-free: parametrized over an arbitrary bundle) #
Each consequence below takes D : TateDuality n, so it exercises the bundle's clauses without
consuming the axiom; #print axioms stays at the standard three.
Duality, (0,2) cardinality form: #H⁰(M′) = #Hom(H²(M), ℤ/n).
Duality, (1,1) cardinality form: #H¹(M′) = #Hom(H¹(M), ℤ/n).
Duality, (2,0) cardinality form: #H²(M′) = #Hom(H⁰(M), ℤ/n).
Injectivity extraction (the form used for dimension counts): a nonzero H¹(M′)-class cups
non-trivially against some H¹(M)-class.