Documentation

TauCeti.RepresentationTheory.Homological.TateCohomology.Cup.TateNakayama

The Tate–Nakayama theorem #

Let G be a finite group, C a representation of G over ℤ and u ∈ H²(G, C). Suppose that for every subgroup S of G of prime-power order, H¹(S, C) = 0, the restriction of u generates H²(S, C), and H²(S, C) has |S| elements. Then for every representation M with Tor₁^ℤ(M, C) = 0, cup product with u is a bijection

H^r(G, M) → H^{r+2}(G, M ⊗ C)

in every integer degree r (TauCeti.TateCohomology.cup_bijective_of_forall_isPGroup_of_lTensor_injective). This is Nakayama's generalization (Cohomology of class field theory and tensor product modules I, Ann. of Math. 65 (1957)) of Tate's theorem, the case M = ℤ (TauCeti.TateCohomology.cup_bijective_of_forall_isPGroup). As in Rep.isZero_res_tensor_of_isZero_res, the vanishing of Tor₁^ℤ(M, C) is stated without a Tor functor: M ⊗ X → M ⊗ Y is injective for every short exact sequence 0 → X → Y → C → 0 of abelian groups. It holds for instance whenever M or C is torsion-free. Throughout this file, H^n(G, M) denotes Tate cohomology in the integer degree n.

Proof #

The class u sits in two short exact sequences, both split over ℤ: the splitting-module sequence 0 → C → C(u) → I_G → 0 (Rep.splittingModule) and the augmentation sequence 0 → I_G → ℤ[G] → ℤ → 0. The splitting module is cohomologically trivial under Tate's hypotheses (TauCeti.TateCohomology.isZero_res_splittingModule), and so is M ⊗ C(u): as an abelian group C(u) is I_G × C, so Tor₁^ℤ(M, C(u)) = Tor₁^ℤ(M, C) = 0, and the theorem of Nakayama and Rim applies (TauCeti.TateCohomology.isZero_res_tensor_splittingModule). Since M ⊗ ℤ[G] is induced, both tensored sequences have bijective connecting maps δ. As u dies in C(u), it is δ (δ z) for some z ∈ H⁰(G, ℤ) = ℤ/|G|, and the rule x ∪ δ y = (-1)^r δ (x ∪ y) for ℤ-split sequences (TauCeti.TateCohomology.cup_δ_of_leftInverse), applied twice, turns cup product with u into x ↦ δ (δ (x ∪ z)). Finally z is the class of an integer prime to |G|: a prime p ∣ |G| dividing it would make u restrict to zero on a subgroup of order p, whose H² it generates and which has p elements. Such an integer is invertible in H⁰(G, ℤ) = ℤ/|G|, so cup product with z is bijective, as cup product with the class of 1 is.

Main statements #

References #

theorem TauCeti.TateCohomology.isZero_res_tensor_splittingModule {k G : Type} [CommRing k] [Group G] [Finite G] [IsDomain k] [IsPrincipalIdealRing k] [CharZero k] (hk : ∀ (p : ℕ), Nat.Prime p → IsUnit ↑p ∨ (Ideal.span {↑p}).IsMaximal) (C : Rep.{0, 0, 0} k G) (u : ↑(groupCohomology C 2)) (h1 : ∀ (p : ℕ) [Fact (Nat.Prime p)] (S : Subgroup G), IsPGroup p ↥S → CategoryTheory.Limits.IsZero (groupCohomology (Rep.res S.subtype C) 1)) (hgen : ∀ (p : ℕ) [Fact (Nat.Prime p)] (S : Subgroup G), IsPGroup p ↥S → ∀ (x : ↑(groupCohomology (Rep.res S.subtype C) 2)), ∃ (r : k), r • (CategoryTheory.ConcreteCategory.hom (groupCohomology.map S.subtype (CategoryTheory.CategoryStruct.id (Rep.res S.subtype C)) 2)) u = x) (hfin : ∀ (p : ℕ) [Fact (Nat.Prime p)] (S : Subgroup G), IsPGroup p ↥S → Finite (k ⧸ Ideal.span {↑(Nat.card ↥S)})) (hcard : ∀ (p : ℕ) [Fact (Nat.Prime p)] (S : Subgroup G), IsPGroup p ↥S → Nat.card ↑(groupCohomology (Rep.res S.subtype C) 2) = Nat.card (k ⧸ Ideal.span {↑(Nat.card ↥S)})) (M : Rep.{0, 0, 0} k G) (hM : ∀ {X Y : Type} [inst : AddCommGroup X] [inst_1 : Module k X] [inst_2 : AddCommGroup Y] [inst_3 : Module k Y] (f : X →ₗ[k] Y) (g : Y →ₗ[k] ↑C), Function.Injective ⇑f → Function.Exact ⇑f ⇑g → Function.Surjective ⇑g → Function.Injective ⇑(LinearMap.lTensor (↑M) f)) (S : Subgroup G) [Fintype ↥S] (n : ℤ) :

Tensoring the splitting module. Let k be a principal ideal domain of characteristic zero in which every prime number is a unit or generates a maximal ideal, C a representation of a finite group G over k and u ∈ H²(G, C). Suppose that for every subgroup S of G of prime-power order, H¹(S, C) = 0, the restriction of u generates H²(S, C), and H²(S, C) has the finite order of k ⧸ |S|k. If Tor₁^k(M, C) = 0, in the form that M ⊗ X → M ⊗ Y is injective for every short exact sequence 0 → X → Y → C → 0 of k-modules, then M ⊗ C(u) is cohomologically trivial.

theorem TauCeti.TateCohomology.cup_bijective_of_forall_isPGroup_of_lTensor_injective {G : Type} [Group G] [Fintype G] (M C : Rep.{0, 0, 0} ℤ G) (u : ↑(groupCohomology C 2)) (h1 : ∀ (p : ℕ) [Fact (Nat.Prime p)] (S : Subgroup G), IsPGroup p ↥S → CategoryTheory.Limits.IsZero (groupCohomology (Rep.res S.subtype C) 1)) (hgen : ∀ (p : ℕ) [Fact (Nat.Prime p)] (S : Subgroup G), IsPGroup p ↥S → ∀ (x : ↑(groupCohomology (Rep.res S.subtype C) 2)), ∃ (m : ℤ), m • (CategoryTheory.ConcreteCategory.hom (groupCohomology.map S.subtype (CategoryTheory.CategoryStruct.id (Rep.res S.subtype C)) 2)) u = x) (hcard : ∀ (p : ℕ) [Fact (Nat.Prime p)] (S : Subgroup G), IsPGroup p ↥S → Nat.card ↑(groupCohomology (Rep.res S.subtype C) 2) = Nat.card ↥S) (hM : ∀ {X Y : Type} [inst : AddCommGroup X] [inst_1 : Module ℤ X] [inst_2 : AddCommGroup Y] [inst_3 : Module ℤ Y] (f : X →ₗ[ℤ] Y) (g : Y →ₗ[ℤ] ↑C), Function.Injective ⇑f → Function.Exact ⇑f ⇑g → Function.Surjective ⇑g → Function.Injective ⇑(LinearMap.lTensor (↑M) f)) (r r' : ℤ) (h : r + 2 = r') :

The Tate–Nakayama theorem (Nakayama, Ann. of Math. 65 (1957); Serre, Local Fields, IX §8). Let G be a finite group, C a representation of G over ℤ and u ∈ H²(G, C). Suppose that for every subgroup S of G of prime-power order

  • H¹(S, C) = 0,
  • the restriction of u generates H²(S, C), and
  • H²(S, C) has |S| elements.

Then for every representation M with Tor₁^ℤ(M, C) = 0, in the form that M ⊗ X → M ⊗ Y is injective for every short exact sequence 0 → X → Y → C → 0 of abelian groups, cup product with u is a bijection H^r(G, M) → H^{r+2}(G, M ⊗ C) in every integer degree r.