Cup products in low degrees on the explicit model #
A cup product on continuous cochains is relative to a G-equivariant biadditive pairing
μ : M →+ N →+ P, that is one with μ (g • m) (g • n) = g • μ m n, which is furthermore
jointly continuous. This file builds the six low-degree shapes
(p, q) ∈ {(0,0), (0,1), (1,0), (0,2), (1,1), (2,0)}, p + q ≤ 2,
of the general inhomogeneous formula
(a ⌣ b)(g₁, …, g_{p+q}) = μ (a (g₁, …, g_p)) ((g₁ ⋯ g_p) • b (g_{p+1}, …, g_{p+q})), namely
(0,0): m ⌣ n = μ m n, (0,1): (m ⌣ b) g = μ m (b g),
(1,0): (a ⌣ n) g = μ (a g) (g • n), (0,2): (m ⌣ b) (g, h) = μ m (b (g, h)),
(1,1): (a ⌣ b) (g, h) = μ (a g) (g • b h), (2,0): (a ⌣ n) (g, h) = μ (a (g, h)) ((g * h) • n).
Main definitions #
TauCeti.ContCohomology.pairingLeftandTauCeti.ContCohomology.pairingRight: partial application of the pairing at an invariant element, as an equivariant additive homomorphism. This is what makes the five shapes with a degree-0factor instances of the coefficient mapsTauCeti.ContCohomology.explicitCoeff0,explicitCoeff1andexplicitCoeff2.TauCeti.ContCohomology.explicitCup00,explicitCup01,explicitCup10,explicitCup02,explicitCup11,explicitCup20: the six shapes, each biadditive by construction, with the_mktheorems fixing the cochain formula on classes of cocycles.
Main statements #
TauCeti.ContCohomology.cup01_mem_Z1,cup10_mem_Z1,cup02_mem_Z2,cup11_mem_Z2andcup20_mem_Z2: a cocycle cupped with a cocycle is a cocycle, the cochain-level heart of each shape.TauCeti.ContCohomology.cup11_mem_B2_leftandcup11_mem_B2_right: the(1,1)cup descends through coboundaries in each variable. These are the only descent statements proved by hand: the other five shapes descend because they are coefficient maps, whose descent isTauCeti.ContCohomology.cochainsMap1_mem_B1andcochainsMap2_mem_B2.TauCeti.ContCohomology.explicitCup00_comm,explicitCup01_eq_cup10_flip,explicitCup02_eq_cup20_flipandexplicitCup11_eq_neg_flip: graded commutativitya ⌣_μ b = (-1)^{pq} (b ⌣_{μᵒᵖ} a)in each bidegree withp + q ≤ 2, whereμᵒᵖ n m = μ m nis Mathlib'sAddMonoidHom.flip. With the sign1in the three shapes that have a degree-0factor the identity already holds on cochains, and those three are stated at cochain level as well (cup01_eq_cup10_flip,cup02_eq_cup20_flip), a degree-0class being invariant. In bidegree(1,1)it holds only on classes, with the sign-1.TauCeti.ContCohomology.explicitCup11_comm_of_neg_eq_self: the2-torsion specialization, in which the(1,1)cup is symmetric.TauCeti.ContCohomology.explicitCup_assoc000and its nine siblings: associativity(x ⌣_{μ₁} y) ⌣_{μ₂} z = x ⌣_{ν₂} (y ⌣_{ν₁} z), one theorem for each tridegree(p, q, r)withp + q + r ≤ 2, the three digits of the name being that tridegree. Each holds already on cochains.TauCeti.ContCohomology.explicitCup_assoc000_muland its nine siblings: the same ten identities for a topologicalG-ringR, all four pairings its multiplication.
Implementation notes #
The translation factor g • in the (1,0), (1,1) and (2,0) formulas is the one the general
inhomogeneous formula carries, and it is kept in the statements even though a degree-0 class is
invariant and the factor is therefore invisible in the (1,0) and (2,0) shapes. Keeping it is
what makes those two shapes the specializations of the general formula that the roadmap's
associativity instances need, rather than the (0,1) and (0,2) shapes read backwards.
The (1,1) shape is the only one that is not a coefficient map, and the only one whose descent
through coboundaries needs a homotopy. Its two primitives are
a ∈ B¹, a = d⁰ m : (a ⌣ b) = d¹ (x ↦ μ m (b x)),
b ∈ B¹, b = d⁰ n : (a ⌣ b) = d¹ (x ↦ -μ (a x) (x • n)),
recorded here because the graded-commutativity statement refers to them.
The (1,1) graded-commutativity homotopy is fixed once, here, and every sign below is read off
it: for continuous 1-cocycles a and b,
(a ⌣_μ b) + (b ⌣_{μᵒᵖ} a) = d¹ (g ↦ -μ (a g) (b g)),
which is TauCeti.ContCohomology.cup11_add_cup11_flip_eq_d1.
Each cocycle proof opens with a change. It only beta-reduces: groupCohomology.IsCocycle₁ and
IsCocycle₂ are predicates on a function, so with the cup cochain supplied as a lambda the goal
is stated with a redex and the identity to be proved is unreadable until it is contracted. No
change below alters the goal by more than beta.
Continuity of the cup cochains is where joint continuity of μ is used. It is automatic when M
and N are discrete, which is the case in every arithmetic application, but it is carried as a
hypothesis rather than derived so that the shapes are available at the general topological
coefficients the explicit complex is built for.
Associativity is stated relative to four G-equivariant biadditive pairings
μ₁ : A →+ B →+ D, μ₂ : D →+ C →+ E, ν₁ : B →+ C →+ F and ν₂ : A →+ F →+ E: these four are
what it takes to type the two composites, and the coefficient identity
μ₂ (μ₁ a b) c = ν₂ a (ν₁ b c) is the hypothesis that identifies them. The
(1,0) and (0,0) shapes are in the family because the instances explicitCup_assoc110 and
explicitCup_assoc100 need them on their right-hand sides.
This implements the "six low-degree shapes", "graded commutativity" and "associativity"
milestones of Layer 8 of the human-authored roadmap at
TauCetiRoadmap/ProfiniteCohomology/README.md, whose §3 fixes the six formulas, whose Layer 8
fixes the ten associativity instances and the G-ring specialization, and whose
Suggested.lean fixes the names explicitCup00, explicitCup01, explicitCup10,
explicitCup02, explicitCup11 and explicitCup20.
References #
- J. Neukirch, A. Schmidt, K. Wingberg, Cohomology of Number Fields, 2nd ed., I §4: the cup product on inhomogeneous cochains and its low-degree formulas, and (1.4.4) for graded commutativity.
- K. Brown, Cohomology of Groups, V §3: the cochain-level cup product, (3.5) for associativity and (3.6) for graded commutativity.
Partial application of an equivariant pairing at an invariant element of the first
factor. Invariance is what makes μ m : N →+ P equivariant, and hence what makes the (0, q)
cup shapes coefficient maps.
Equations
- TauCeti.ContCohomology.pairingLeft μ hequiv m = { toFun := ⇑(μ ↑m), map_smul' := ⋯, map_zero' := ⋯, map_add' := ⋯ }
Instances For
Partial application of an equivariant pairing at an invariant element of the second factor.
Equations
- TauCeti.ContCohomology.pairingRight μ hequiv n = { toFun := fun (m : M) => (μ m) ↑n, map_smul' := ⋯, map_zero' := ⋯, map_add' := ⋯ }
Instances For
The underlying additive homomorphism of TauCeti.ContCohomology.pairingLeft.
The underlying additive homomorphism of TauCeti.ContCohomology.pairingRight is the flip of
the pairing.
The defining formula for TauCeti.ContCohomology.pairingLeft.
The defining formula for TauCeti.ContCohomology.pairingRight.
The pairing with an invariant first argument absorbs the action from the second.
The pairing with an invariant second argument absorbs the action from the first.
The opposite pairing μᵒᵖ n m = μ m n is equivariant. Mathlib's AddMonoidHom.flip is
the μᵒᵖ of the graded-commutativity statements below, and this is the hypothesis it has to be
fed to be cupped against.
A jointly continuous pairing is continuous in the second variable.
A jointly continuous pairing is continuous in the first variable.
The opposite pairing of a jointly continuous pairing is jointly continuous, being its composite with the swap homeomorphism.
The (0,0) cup product, m ⌣ n = μ m n: the pairing of two invariant elements is
invariant. No topology is involved, H⁰ being a subgroup and not a quotient.
Equations
- One or more equations did not get rendered due to their size.
Instances For
The (0,0) cup is the pairing itself.
The (0,1) cup of an invariant element with a continuous 1-cocycle is a continuous
1-cocycle.
The (0,1) cup product, (m ⌣ b) g = μ m (b g). For invariant m this is the
coefficient map induced by μ m.
Equations
- One or more equations did not get rendered due to their size.
Instances For
The (0,1) cup is the coefficient map induced by the pairing at an invariant element.
The cochain formula for the (0,1) cup on the class of a continuous 1-cocycle.
The (1,0) cup of a continuous 1-cocycle with an invariant element is a continuous
1-cocycle. The translation factor g • is the one the general inhomogeneous formula carries;
it disappears only because n is invariant.
The (1,0) cup product, (a ⌣ n) g = μ (a g) (g • n). For invariant n this is the
coefficient map induced by m ↦ μ m n.
Equations
- One or more equations did not get rendered due to their size.
Instances For
The (1,0) cup is the coefficient map induced by the pairing at an invariant element.
The cochain formula for the (1,0) cup on the class of a continuous 1-cocycle.
The (0,2) cup of an invariant element with a continuous 2-cocycle is a continuous
2-cocycle.
The (0,2) cup product, (m ⌣ b) (g, h) = μ m (b (g, h)).
Equations
- One or more equations did not get rendered due to their size.
Instances For
The (0,2) cup is the coefficient map induced by the pairing at an invariant element.
The cochain formula for the (0,2) cup on the class of a continuous 2-cocycle.
The (2,0) cup of a continuous 2-cocycle with an invariant element is a continuous
2-cocycle.
The (2,0) cup product, (a ⌣ n) (g, h) = μ (a (g, h)) ((g * h) • n). This is the last
of the six shapes: no explicit cup goes above total degree 2.
Equations
- One or more equations did not get rendered due to their size.
Instances For
The (2,0) cup is the coefficient map induced by the pairing at an invariant element.
The cochain formula for the (2,0) cup on the class of a continuous 2-cocycle.
The (1,1) cup of two continuous 1-cocycles is a continuous 2-cocycle. This is the
cochain-level heart of the only shape that is not a coefficient map: the translation factor g •
on the second cocycle is what turns the two 1-cocycle identities into the 2-cocycle
identity.
The (1,1) cup descends through coboundaries in the first variable: if a = d⁰ m then
a ⌣ b = d¹ (x ↦ μ m (b x)).
The (1,1) cup descends through coboundaries in the second variable: if b = d⁰ n then
a ⌣ b = d¹ (x ↦ -μ (a x) (x • n)).
The (1,1) cup product, the descent of the cochain formula
(a ⌣ b) (g, h) = μ (a g) (g • b h). This is the only one of the six shapes that is not a
coefficient map.
Equations
- One or more equations did not get rendered due to their size.
Instances For
The cochain formula for the (1,1) cup on the classes of two continuous 1-cocycles.
Graded commutativity in bidegree (0,0). The sign (-1)^{pq} is 1, and the two
cochains, here two elements of P, are literally equal.
The three shapes with a degree-0 factor commute already on cochains, because a degree-0
class is invariant. Neither statement mentions a topology or an action on the second factor.
Graded commutativity in bidegree (0,1), at cochain level. The (1,0) cup of a cochain
b with the invariant m along the opposite pairing is the (0,1) cup of m with b: the
translation factor g • the (1,0) formula carries acts on m, which is invariant.
Graded commutativity in bidegree (0,2), at cochain level, the degree-2 counterpart of
TauCeti.ContCohomology.cup01_eq_cup10_flip.
Graded commutativity in bidegree (0,1). The sign (-1)^{pq} is 1, and by
TauCeti.ContCohomology.cup01_eq_cup10_flip the identity already holds on cochains.
Graded commutativity in bidegree (0,2). The sign (-1)^{pq} is 1, and by
TauCeti.ContCohomology.cup02_eq_cup20_flip the identity already holds on cochains.
The homotopy behind graded commutativity in bidegree (1,1). The two cochains
a ⌣_μ b and b ⌣_{μᵒᵖ} a need not be equal in general; their sum is the coboundary of the
1-cochain g ↦ -μ (a g) (b g).
The sum of the two (1,1) cup cochains is a coboundary, by
TauCeti.ContCohomology.cup11_add_cup11_flip_eq_d1; its primitive is continuous because μ is
jointly continuous.
Graded commutativity in bidegree (1,1), a ⌣_μ b = -(b ⌣_{μᵒᵖ} a), the sign
(-1)^{pq} now being -1. This is an identity of classes and not of cochains: the two cochains
differ by the coboundary exhibited in
TauCeti.ContCohomology.cup11_add_cup11_flip_eq_d1.
Graded commutativity in bidegree (1,1) for 2-torsion coefficients: when every element
of P is its own negative the (1,1) cup is symmetric on classes. This is the form the
𝔽₂-valued arithmetic applications use, where every sign is 1.
Associativity #
Typing the two sides of an associativity statement needs four G-equivariant biadditive pairings
μ₁ : A →+ B →+ D, μ₂ : D →+ C →+ E, ν₁ : B →+ C →+ F and ν₂ : A →+ F →+ E. The coefficient
identity μ₂ (μ₁ a b) c = ν₂ a (ν₁ b c) is not part of that: it is the hypothesis under which
the two well-typed composites are equal. The ten instances below are the ten
tridegrees (p, q, r) with p + q + r ≤ 2, each named explicitCup_assoc followed by the three
digits p, q, r. Each holds already on cochains; the classes are equal because their
representatives are.
Where the right-hand side cups y and z together first, the left-hand side carries the
translation factor of the general inhomogeneous formula on y and on z separately while the
right-hand side carries a single one on y ⌣_{ν₁} z. The equivariance of ν₁ is what merges
them, and that is what the proofs of explicitCup_assoc100, explicitCup_assoc200,
explicitCup_assoc110 and explicitCup_assoc101 use it for.
Associativity of the cup product in tridegree (0,0,0), where all three classes are
invariant elements and the identity is the coefficient identity itself.
The tridegrees (0,0,r): the first two factors are invariant, so μ₁ is used only through
the (0,0) cup and no continuity of it is needed.
Associativity of the cup product in tridegree (0,0,1).
Associativity of the cup product in tridegree (0,0,2), the degree-2 counterpart of
TauCeti.ContCohomology.explicitCup_assoc001.
The tridegrees (0,q,0): the outer factors are invariant.
Associativity of the cup product in tridegree (0,1,0).
Associativity of the cup product in tridegree (0,2,0), the degree-2 counterpart of
TauCeti.ContCohomology.explicitCup_assoc010.
The tridegrees (p,0,0): the last two factors are invariant, so ν₁ is used only through the
(0,0) cup. The translation factor of the general formula sits on ν₁ b c, and it is
hequiv₃ that moves it onto the two factors separately.
Associativity of the cup product in tridegree (1,0,0).
Associativity of the cup product in tridegree (2,0,0), the degree-2 counterpart of
TauCeti.ContCohomology.explicitCup_assoc100.
Associativity of the cup product in tridegree (0,1,1). The (1,1) cup appears on the
right-hand side and its translation factor is common to both sides, the outer factor being
invariant.
Associativity of the cup product in tridegree (1,1,0). This is the instance that forces
the (1,0) shape into the family: the right-hand side cups the invariant z onto y in
bidegree (1,0) before pairing with x. The two translation factors match by mul_smul and the
equivariance of ν₁.
Associativity of the cup product in tridegree (1,0,1), the last of the ten instances
with p + q + r ≤ 2.
Associativity for a topological G-ring #
The specialization the applications use: one coefficient ring R acted on by ring
automorphisms, all four pairings its multiplication AddMonoidHom.mul, and the coefficient
identity mul_assoc. The continuity and equivariance a pairing has to come with are
continuous_mul and smul_mul', which apply to AddMonoidHom.mul as they stand. Each of the
ten statements below is the corresponding general statement instantiated there.
Associativity of the ring cup product in tridegree (0,0,0).
Associativity of the ring cup product in tridegree (0,0,1).
Associativity of the ring cup product in tridegree (0,1,0).
Associativity of the ring cup product in tridegree (1,0,0).
Associativity of the ring cup product in tridegree (0,0,2).
Associativity of the ring cup product in tridegree (0,2,0).
Associativity of the ring cup product in tridegree (2,0,0).
Associativity of the ring cup product in tridegree (0,1,1).
Associativity of the ring cup product in tridegree (1,1,0).
Associativity of the ring cup product in tridegree (1,0,1).