Documentation

TauCeti.AlgebraicTopology.Cohomology.UniversalCoefficient

The universal coefficient theorem for singular cohomology #

Let k be a hereditary commutative ring, expressed by requiring every k-module to have projective dimension at most one. This holds for principal ideal domains, for instance ℤ (ModuleCat.hasProjectiveDimensionLE_one_of_isPrincipalIdealRing). Let R be a projective k-module, for instance k itself, and let M be any k-module. For a topological space X, the singular chain modules Cₙ(X; R) = ⨁ R are projective (SSet.projective_chainComplex_X). Their cycles are then projective and split off the terms (TauCeti.Algebra.Homology.HomologicalComplex.Projective.Dimension). These are the hypotheses of the algebraic universal coefficient sequence (TauCeti.Algebra.Homology.UniversalCoefficient.Basic), so

0 ⟶ Ext¹(Hₙ(X; R), M) ⟶ Hⁿ⁺¹(X; R, M) ⟶ Hom(Hₙ₊₁(X; R), M) ⟶ 0

is exact, where the second map is the Kronecker map TopCat.singularKronecker evaluating cohomology classes on homology classes. The sequence splits, and the Kronecker map is surjective in every degree; in degree 0 it is bijective (TopCat.singularKronecker_bijective_zero). Both maps are natural in X. The splitting TopCat.singularKroneckerSection, and the resulting decomposition TopCat.singularUniversalCoefficientEquiv, depend on a chosen retraction of the cycles in each degree and need not be natural in X.

With k = R = ℤ this is the classical sequence 0 ⟶ Ext(Hₙ(X; ℤ), M) ⟶ Hⁿ⁺¹(X; M) ⟶ Hom(Hₙ₊₁(X; ℤ), M) ⟶ 0 for an abelian group M. For an injective coefficient module the Ext¹ term vanishes and the Kronecker map is an isomorphism over any ring (TopCat.singularKroneckerEquiv).

Over an arbitrary commutative ring k the cycles of the singular chain complex need not split off, and there is no such short exact sequence in general. It survives, with vanishing Ext¹ term, when the homology is projective: if R and every Hₙ(X; R) are projective, the singular chain complex is a complex of projective modules with projective homology, so its cycles are projective and split off (ChainComplex.isSplitMono_iCycles_of_projective_homology), and the Kronecker map is an isomorphism Hⁿ(X; R, M) ≃ₗ[k] Hom(Hₙ(X; R), M) in every degree (TopCat.singularKroneckerEquivOfProjectiveHomology).

Over a hereditary ring, a single projective homology module already kills the Ext¹ term next to it: if Hₙ(X; R) is projective, the Kronecker map is an isomorphism in degree n + 1 (TopCat.singularKronecker_bijective_succ_of_projective_homology). So if Hₙ(X; R) is projective and Hₙ₊₁(X; R) = 0, then Hⁿ⁺¹(X; R, M) = 0 (TopCat.isZero_singularCohomology_succ_of_projective_homology), and likewise for pairs. For instance, integral cohomology vanishes above the top degree of a space whose integral homology is free in that degree and zero above it, while in the top degree it need not be torsion-free.

Main declarations #

References #

The universal coefficient theorem below the first non-projective homology, over any commutative ring k: for a projective coefficient module R, if the singular homology modules Hⱼ(X; R) are projective for j < n, the Kronecker map Hⁿ(X; R, M) →ₗ[k] (Hₙ(X; R) ⟶ M) is bijective, for every module M. For n = 1 the hypothesis is that H₀(X; R) is projective, which holds for free R (ModuleCat.free_singularHomology_zero).

Injective coefficient maps are injective on cohomology below the first non-projective homology: for a projective coefficient module R, if Hⱼ(X; R) is projective for j < n, an injective map g : M ⟶ M' of coefficient modules induces an injective map Hⁿ(X; R, M) → Hⁿ(X; R, M'), because both are Hom(Hₙ(X; R), -). For n = 1 and R = ℤ this makes H¹(X; ℤ) → H¹(X; ℝ) injective, so H¹(X; ℤ) is torsion-free.

The universal coefficient theorem with projective homology, over any commutative ring k: for a projective coefficient module R, if the singular homology modules Hₙ(X; R) are all projective, the Kronecker map Hⁿ(X; R, M) →ₗ[k] (Hₙ(X; R) ⟶ M) is bijective in every degree, for every module M. The singular chain modules are projective, so the cycles split off and the Ext¹ term of the universal coefficient sequence vanishes.

The universal coefficient isomorphism with projective homology: over any commutative ring, for a projective coefficient module R such that every Hₙ(X; R) is projective, the Kronecker map is a k-linear equivalence Hⁿ(X; R, M) ≃ₗ[k] (Hₙ(X; R) ⟶ M). This applies to R = k whenever every Hₙ(X; k) is a free k-module, as for spheres.

Instances For

    The Ext term of the universal coefficient sequence of singular cohomology over a hereditary commutative ring, the k-linear map Ext¹(Hₙ(X; R), M) →ₗ[k] Hⁿ⁺¹(X; R, M). It is the algebraic map TauCeti.ChainComplex.extToHomology of the singular chain complex: an extension class represented by a morphism β : Bₙ ⟶ M out of the singular boundaries goes to the class of the cocycle Cₙ₊₁(X; R) ⟶ Bₙ ⟶ M.

    Instances For

      The Ext term of the universal coefficient sequence of singular cohomology is the algebraic map TauCeti.ChainComplex.extToHomology of the singular chain complex.

      The characterization of TopCat.singularExtToCohomology: the class in Ext¹(Hₙ(X; R), M) of a morphism β : Bₙ ⟶ M out of the singular boundaries Bₙ = ker(Zₙ ⟶ Hₙ(X; R)) goes to the class of the cocycle Cₙ₊₁(X; R) ⟶ Bₙ ⟶ M.

      Injectivity in the universal coefficient sequence: over a hereditary commutative ring, the map Ext¹(Hₙ(X; R), M) →ₗ[k] Hⁿ⁺¹(X; R, M) is injective.

      Exactness in the middle of the universal coefficient sequence: over a hereditary commutative ring, a class in Hⁿ⁺¹(X; R, M) evaluates to zero on Hₙ₊₁(X; R) exactly when it comes from Ext¹(Hₙ(X; R), M).

      The splitting of the universal coefficient sequence of singular cohomology over a hereditary commutative ring: a k-linear right inverse Hom(Hₙ(X; R), M) →ₗ[k] Hⁿ(X; R, M) of the Kronecker map. It sends g to the class of the cocycle Cₙ(X; R) ⟶ Zₙ ⟶ Hₙ(X; R) ⟶ M, through a chosen retraction of the inclusion of the singular cycles, and need not be natural in X.

      Instances For

        Surjectivity in the universal coefficient sequence: over a hereditary commutative ring, every morphism Hₙ(X; R) ⟶ M is the evaluation of a class in Hⁿ(X; R, M).

        The splitting of the universal coefficient sequence of singular cohomology over a hereditary commutative ring: a k-linear equivalence Hⁿ⁺¹(X; R, M) ≃ₗ[k] Ext¹(Hₙ(X; R), M) × Hom(Hₙ₊₁(X; R), M) whose second component is the Kronecker map (TopCat.singularUniversalCoefficientEquiv_apply_snd) and whose inverse restricts to TopCat.singularExtToCohomology on the first factor (TopCat.singularUniversalCoefficientEquiv_symm_inl). It is built from the section TopCat.singularKroneckerSection and is not natural in X.

        Instances For
          @[simp]

          The second component of the splitting TopCat.singularUniversalCoefficientEquiv is the Kronecker map.

          Naturality of the universal coefficient sequence in the space: for a continuous map f : X ⟶ Y, pulling an extension back along f_* : Hₙ(X; R) ⟶ Hₙ(Y; R) and then including it in the cohomology of X is including it in the cohomology of Y and then pulling back along f.

          The universal coefficient theorem above a projective homology module, over a hereditary commutative ring: for a projective coefficient module R, if Hₙ(X; R) is projective, the Kronecker map Hⁿ⁺¹(X; R, M) →ₗ[k] (Hₙ₊₁(X; R) ⟶ M) is bijective, for every module M. The Ext¹(Hₙ(X; R), M) term of the universal coefficient sequence vanishes.

          Cohomology vanishes above a projective homology module followed by a zero one, over a hereditary commutative ring: for a projective coefficient module R, if Hₙ(X; R) is projective and Hₙ₊₁(X; R) = 0, then Hⁿ⁺¹(X; R, M) = 0 for every module M. With k = R = ℤ: if Hₙ(X; ℤ) is free abelian and Hₙ₊₁(X; ℤ) = 0, then Hⁿ⁺¹(X; M) = 0.

          The relative universal coefficient theorem below the first non-projective homology, over any commutative ring k: for a projective coefficient module R, if the relative singular homology modules Hⱼ(X, A; R) of a pair are projective for j < n, for instance zero, the relative Kronecker map Hⁿ(X, A; R, M) →ₗ[k] (Hₙ(X, A; R) ⟶ M) is bijective, for every module M.

          Relative cohomology vanishes up to the degree where relative homology does: for a projective coefficient module R, if Hⱼ(X, A; R) = 0 for j ≤ n, then Hⁿ(X, A; R, M) = 0 for every module M.

          The relative universal coefficient theorem above a projective homology module, over a hereditary commutative ring: for a projective coefficient module R, if Hₙ(X, A; R) is projective, the relative Kronecker map Hⁿ⁺¹(X, A; R, M) →ₗ[k] (Hₙ₊₁(X, A; R) ⟶ M) is bijective, for every module M. Nothing is assumed about the relative homology in the other degrees.

          Relative cohomology vanishes above a projective homology module followed by a zero one, over a hereditary commutative ring: for a projective coefficient module R, if Hₙ(X, A; R) is projective and Hₙ₊₁(X, A; R) = 0, then Hⁿ⁺¹(X, A; R, M) = 0 for every module M.

          The relative universal coefficient theorem with projective homology, over any commutative ring k: for a projective coefficient module R, if the relative singular homology modules Hₙ(X, A; R) of a pair are all projective, the relative Kronecker map Hⁿ(X, A; R, M) →ₗ[k] (Hₙ(X, A; R) ⟶ M) is bijective in every degree, for every module M.

          noncomputable def TopPair.singularKroneckerEquivOfProjectiveHomology {k : Type u_1} [CommRing k] (P : TopPair) (R M : ModuleCat k) [Module.Projective k ↑R] [hP : ∀ (n : ℕ), CategoryTheory.Projective (P.singularHomology R n)] (n : ℕ) :

          The relative universal coefficient isomorphism with projective homology: over any commutative ring, for a projective coefficient module R such that every relative homology module Hₙ(X, A; R) is projective, the relative Kronecker map is a k-linear equivalence Hⁿ(X, A; R, M) ≃ₗ[k] (Hₙ(X, A; R) ⟶ M).

          Instances For
            @[simp]

            The equivalence TopPair.singularKroneckerEquivOfProjectiveHomology is the relative Kronecker map.