Diogo Andrade
Back to home

Why You Should Learn About Factorization Homology

Let me expand on some additional curiosities, examples and observations about Factorization Homology:

1 - Actions of the Mapping Class Group

Take (S,⊗)(\mathcal{S},\otimes) to be a target symmetric monoidal (∞,1)(\infty,1)-category, and consider A ⁣:Disk2or→SA\colon \mathcal{D}\mathsf{isk}_2^{\mathsf{or}}\rightarrow \mathcal{S} to be an (oriented) E2\mathbb{E}_2-algebra in S\mathcal{S}. Then consider, for any genus gg oriented surface Σg\Sigma_{g}, the Lie group Diff+(Σg)\mathsf{Diff}^{+}(\Sigma_g), of orientation-preserving diffeomorphisms of Σg\Sigma_g. Note that this corresponds to the automorphisms of Σg\Sigma_g in Mfld2or\mathsf{M}\mathsf{fld}_{2}^{\mathsf{or}}, and therefore factorization homology furnishes a map

Diff+(Σg)=homMfld2or(Σg,Σg)→homS(∫ΣgA,∫ΣgA) \mathsf{Diff}^{+}(\Sigma_g)=\mathsf{hom}_{\mathcal{M}\mathsf{fld}_2^{\mathsf{or}}}(\Sigma_g,\Sigma_g) \rightarrow \mathsf{hom}_{\mathcal{S}}\Big(\int_{\Sigma_{g}}A,\int_{\Sigma_{g}}A\Big)

Taking connected components[1], we get an action of the mapping class group on π0\pi_0 of the object on the right hand side, and in the case that S⊗=LinCatk×\mathcal{S}^{\otimes}=\mathsf{LinCat}_{\mathbb{k}}^{\times}, this recovers well-known actions on certain invariants of quantum groups.

2 - Pushforward Property

Factorization homology enjoys a Fubini-like property. This is often called the pushforward property. In its "simpler" guise it states that, for manifolds Mm,NnM^{m},N^{n} (with superscript indicating dimension), we get the following equivalence

∫M×NA≃∫M(∫N×RmA) \int_{M\times N}A \simeq \int_{M}\bigg(\int_{N\times \mathbb{R}^m}A\bigg)

To make sense of it, one should note that ∫N×RmA\int_{N\times \mathbb{R}^m}A is an Em\mathbb{E}_m-algebra, much like we saw that ∫M×RA\int_{M\times \mathbb{R}} A was an E1\mathbb{E}_1-algebra. In its fancier, and evidently more general form, it says that given a map f ⁣:Mm→Nnf\colon M^m \rightarrow N^n which fibers over the interior and boundary of NN, and for AA an Em\mathbb{E}_m-algebra, we get an equivalence

∫Nf∗A≃∫MA. \int_{N}f_{\ast}A \simeq \int_{M}A.

Where (f∗A)(U):=∫f−1(U)A(f_\ast A)(U):=\int_{f^{-1}(U)}A. To recover the Fubini-looking identity we choose πM ⁣:M×N→M\pi_{M}\colon M\times N \rightarrow M, in this case f−1(Dm)=Dm×Nf^{-1}(\mathbb{D}^m)=\mathbb{D}^m\times N, as expected.

3 - Factorization homology as ordinary homology

Factorization homology, in its version using ChZ⊕\mathsf{Ch}_{\mathbb{Z}}^{\oplus} as a target symmetric monoidal category, recovers "usual" homology of CW-complexes à la Eilenberg-Steenrod. Note that in this case ⊕\oplus corresponds to the coproduct in the category, and as such forces any En\mathbb{E}_n-algebra V⊕V→VV\oplus V \rightarrow V to be simple addition of chains.

We define the following chain complex

C∗(−;V):=∫−V \mathsf{C}_\ast(-;V):=\int_{-}V

and we call it "formal singular chains with coefficients in VV". Consider the ⊗\otimes-excision axiom:

C∗(X1⊔X1∩X2X2;V)≃C∗(X1;V)⨂C∗(X1∩X2;V)C∗(X2;V) \mathsf{C}_\ast(X_{1}\underset{X_1 \cap X_2}{\sqcup} X_2;V)\simeq \mathsf{C}_\ast(X_1;V)\underset{\mathsf{C}_\ast (X_1 \cap X_2;V)}{\bigotimes} \mathsf{C}_\ast(X_2;V)

In this case, the module structure has to be also given by addition i.e. this is the right C∗(X1∩X2;V)\mathsf{C}_\ast(X_1\cap X_2;V)-module structure map

C∗(X1;V)⊕C∗(X1∩X2;V)→(−+−)C∗(X1;V). \mathsf{C}_\ast(X_1;V)\oplus \mathsf{C}_\ast(X_1\cap X_2;V)\xrightarrow{(-+-)} \mathsf{C}_\ast(X_1;V).

This means that the relative tensor product of modules is given by the coequalizer

coeq(C∗(X1;V)⊕C∗(X1∩X2;V)⊕C∗(X2;V)⟶⟶C∗(X1;V)⊕C∗(X2;V)) \mathsf{coeq}\Big(\mathsf{C}_\ast(X_1;V)\oplus \mathsf{C}_\ast(X_1\cap X_2;V) \oplus \mathsf{C}_\ast(X_2;V) \substack{\longrightarrow\\[-1em] \longrightarrow \\} \mathsf{C}_\ast(X_1;V)\oplus \mathsf{C}_\ast(X_2;V)\Big)

where the top arrow is the right module structure map, and the bottom one the left-module structure map. We can recast in more familiar terms as the cokernel

coker(C∗(X1∩X2;V)→(x,−x)C∗(X1;V)⊕C∗(X2;V)) \mathsf{coker}\Big(\mathsf{C}_\ast(X_1\cap X_2;V)\xrightarrow{(x,-x)} \mathsf{C}_\ast(X_1;V)\oplus \mathsf{C}_\ast(X_2;V)\Big)

. This is an injective map, and so all put together we get a short exact sequence of chain complexes

0→C∗(X1∩X2;V)→(x,−x)C∗(X1;V)⊕C∗(X2;V)⟶C∗(X1∪X2;V)⟶0 0 \rightarrow \mathsf{C}_\ast(X_1\cap X_2;V) \xrightarrow{(x,-x)} \mathsf{C}_\ast(X_1;V)\oplus \mathsf{C}_\ast(X_2;V) \longrightarrow \mathsf{C}_\ast(X_1\cup X_2;V)\longrightarrow 0

and this can be seen as the Mayer-Vietoris sequence which "universally" characterizes singular homology. From this, we can then conclude that

∫XV \int_X V

are actual singular chains in XX with coefficients in VV, and ⊗\otimes-excision translates to Mayer-Vietoris.

4 - Where to look for En\mathbb{E}_n-algebras?

  1. Fix a base ring RR, letting the target symmetric monoidal (∞,1)(\infty,1)-category be ChainR⊗RL\mathcal{C}\mathsf{hain}_{R}^{\otimes_{R}^{\mathbb{L}}} whose objects are chain complexes of RR-modules, and whose symmetric monoidal structure is given by the derived tensor product ⊗RL\otimes_R ^{\mathbb{L}}. A commutative RR-algebra is an En\mathbb{E}_n-algebra in ChainR⊗RL\mathcal{C}\mathsf{hain}_{R}^{\otimes_{R}^{\mathbb{L}}} for any nn. More generally, a cdga (a commutative dg algebra) is an En\mathbb{E}_n-algebra for all nn.

  2. The Hochschild cochain complex of any associative algebra (or dg algebra, or A∞A_{\infty}-algebra) is an example of an E2\mathbb{E}_2-algebra in Chaink⊗k\mathcal{C}\mathsf{hain}_{\mathbb{k}}^{\otimes_\mathbb{k}}.

  3. nn-fold loop spaces are En\mathbb{E}_n-algebras in Top×.\mathcal{T}\mathsf{op}^{\times}.

  4. The free En\mathbb{E}_n-algebra on one generator (in Top×\mathcal{T}\mathsf{op}^{\times}) is Conf(Rn)\mathsf{Conf}(\mathbb{R}^n) - the unordered configuration space of Rn\mathbb{R}^n.

  5. En\mathbb{E}_n-algebras are also often constructed as deformations of commutative or cocommutative objects. The braided monoidal category of representations of the quantum group Uq(g)\mathsf{U}_q(\mathfrak{g}) is obtained by deforming a cocommutative Hopf algebra (the universal enveloping algebra of g\mathfrak{g}).


  1. here you could have easily said that the proof was too small to fit in the margin and get done with it, couldn't you? it would have been so easy... back

Originally written for the Seminar on ∞-categories.