The connecting maps and the explicit low-degree cup products #
The Leibniz rule δ (x ⌣ y) = δ x ⌣ y + (-1)^p (x ⌣ δ y) is not a statement until one says
which short exact sequences of coefficients the two connecting maps belong to, and how the
pairing relates them. Cupping with a fixed class in the second variable and cupping with a fixed
class in the first variable are two different constructions, each attached to a short exact
sequence in its own variable, so this file proves two families of identities rather than one sum
rule.
In the first variable the input is a short exact sequence 0 → A' → A → A'' → 0 of discrete
G-modules, a topological G-module B, a short exact sequence 0 → C' → C → C'' → 0, and
three G-equivariant biadditive pairings
μ : A →+ B →+ C, μ' : A' →+ B →+ C', μ'' : A'' →+ B →+ C''
with μ (incl a') b = incl (μ' a' b) and μ'' (proj a) b = proj (μ a b), that is a map of short
exact sequences after pairing with B. For x ∈ H^p(G, A'') and y ∈ H^q(G, B) the identity is
δ (x ⌣ y) = δ x ⌣ y in H^{p+q+1}(G, C').
In the second variable the sequence is 0 → B' → B → B'' → 0, the fixed module is A, the
pairings are μ : A →+ B →+ C, μ' : A →+ B' →+ C' and μ'' : A →+ B'' →+ C'', and for
x ∈ H^p(G, A) and y ∈ H^q(G, B'') the identity carries the sign of the degree it moves past:
δ (x ⌣ y) = (-1)^p (x ⌣ δ y) in H^{p+q+1}(G, C').
Six instances have all three of p, q and p + q + 1 at most 2, so six theorems exhaust what
the low-degree model can state.
The third family moves a connecting map from one variable to the other. Its input is a pair
of short exact sequences 0 → A₁ → A → A₂ → 0 and 0 → B₂ → B → B₁ → 0 of discrete
G-modules, compatibly paired into one topological G-module C by μ : A →+ B →+ C,
μ₁ : A₁ →+ B₁ →+ C and μ₂ : A₂ →+ B₂ →+ C with
μ (incl a₁) b = μ₁ a₁ (proj b), μ a (incl b₂) = μ₂ (proj a) b₂,
so that the sub-object A₁ is orthogonal to the sub-object B₂ and pairs with the quotient
B₁, while the quotient A₂ pairs with the sub-object B₂. No nondegeneracy is assumed; the
motivating instance is a short exact sequence and its dual sequence under an evaluation pairing.
For x ∈ H^p(G, A₂) and
y ∈ H^q(G, B₁) the two connecting maps are adjoint up to the Leibniz sign:
δ x ⌣ y = (-1)^(p+1) (x ⌣ δ y) in H^{p+q+1}(G, C),
because δ x ⌣ y + (-1)^p (x ⌣ δ y) is the coboundary of the cup of lifts. Three bidegrees
have p + q + 1 ≤ 2, so three theorems exhaust this family too. These are the identities that
make the duality maps H^i(G, M) → H^{2-i}(G, M')^∨ of a Demushkin group commute with the long
exact sequences, the step of Tate's argument that Serre records.
Main statements #
TauCeti.ContCohomology.explicitDelta0_explicitCup00_left,explicitDelta1_explicitCup01_leftandexplicitDelta1_explicitCup10_left: the three first-variable identities, in bidegrees(0,0),(0,1)and(1,0).TauCeti.ContCohomology.explicitDelta0_explicitCup00_right,explicitDelta1_explicitCup01_rightandexplicitDelta1_explicitCup10_right: the three second-variable identities, in bidegrees(0,0),(0,1)and(1,0), the last with its sign.TauCeti.ContCohomology.explicitCup10_explicitDelta0_eq_neg_explicitCup01_explicitDelta0,explicitCup11_explicitDelta0_eq_neg_explicitCup02_explicitDelta1andexplicitCup20_explicitDelta1_eq_explicitCup11_explicitDelta0: the three adjointness identities for a pair of compatibly paired short exact sequences, in bidegrees(0,0),(0,1)and(1,0).
References #
- J. Neukirch, A. Schmidt, K. Wingberg, Cohomology of Number Fields, 2nd ed., (1.4.3) and (1.4.5): the compatibility of the cup product with the connecting homomorphisms.
- J. S. Milne, Arithmetic Duality Theorems, 2nd ed., I §0, the cup-product properties (0.1.1)-(0.1.6), stated with the same sign conventions.
- J.-P. Serre, Structure de certains pro-p-groupes (d'après Demuškin), Séminaire Bourbaki 8 (1962/63), exposé 252, §9.1: Tate's duality argument, which uses the adjointness identities to compare the long exact sequences of a finite module and of its dual.
δ⁰ passes through the (0,0) cup in the first variable. For an invariant x of A''
and an invariant y of B, the class δ⁰ (x ⌣ y) ∈ H¹(G, C') is δ⁰ x ⌣ y, the (1,0) cup
against the pairing μ' of the sub-objects.
δ¹ passes through the (0,1) cup in the first variable. For an invariant x of A''
and a class y ∈ H¹(G, B), the class δ¹ (x ⌣ y) ∈ H²(G, C') is the (1,1) cup δ⁰ x ⌣ y.
δ¹ passes through the (1,0) cup in the first variable. For a class x ∈ H¹(G, A'')
and an invariant y of B, the class δ¹ (x ⌣ y) ∈ H²(G, C') is the (2,0) cup
δ¹ x ⌣ y.
δ⁰ passes through the (0,0) cup in the second variable. For an invariant x of A
and an invariant y of B'', the class δ⁰ (x ⌣ y) ∈ H¹(G, C') is the (0,1) cup
x ⌣ δ⁰ y; the sign (-1)^p is 1 because x has degree 0.
δ¹ passes through the (0,1) cup in the second variable. For an invariant x of A
and a class y ∈ H¹(G, B''), the class δ¹ (x ⌣ y) ∈ H²(G, C') is the (0,2) cup
x ⌣ δ¹ y.
δ¹ passes through the (1,0) cup in the second variable, with a sign. For a class
x ∈ H¹(G, A) and an invariant y of B'', the class δ¹ (x ⌣ y) ∈ H²(G, C') is
-(x ⌣ δ⁰ y), the sign (-1)^p at p = 1.
A pair of compatibly paired short exact sequences #
The connecting maps of 0 → A₁ → A → A₂ → 0 and of 0 → B₂ → B → B₁ → 0 are adjoint under
pairings that make A₁ orthogonal to B₂. Every coefficient module of the two sequences is
discrete here, so the joint continuity of each pairing is automatic and is not taken as a
hypothesis, and the equivariance of μ₁ and μ₂ follows from that of μ through the two
compatibilities, so only μ is assumed equivariant; the common target C is any topological
G-module.
The pairing of the sub-object A₁ with the quotient B₁ is equivariant. Lift b to
B and read μ₁ through μ on SA.incl a.
The pairing of the quotient A₂ with the sub-object B₂ is equivariant. Lift a to
A and read μ₂ through μ on SB.incl b.
The two δ⁰ are anti-adjoint under the (1,0) and (0,1) cups. For invariants x of A₂
and y of B₁, the class δ⁰ x ⌣ y ∈ H¹(G, C) is -(x ⌣ δ⁰ y).
δ⁰ and δ¹ are anti-adjoint under the (1,1) and (0,2) cups. For an invariant x
of A₂ and a class y ∈ H¹(G, B₁), the class δ⁰ x ⌣ y ∈ H²(G, C) is -(x ⌣ δ¹ y).
δ¹ and δ⁰ are adjoint under the (2,0) and (1,1) cups. For a class
x ∈ H¹(G, A₂) and an invariant y of B₁, the class δ¹ x ⌣ y ∈ H²(G, C) is x ⌣ δ⁰ y;
the sign (-1)^(p+1) is 1 because x has degree 1.