Definition 6.1. A 2-dimensional topological field theory valued in a symmetric monoidal -category is a symmetric monoidal functor .
6.1. Topological Field Theory from perfect stacks
Let denote the -category whose objects are compact oriented -manifolds, and whose morphisms consist of the classifying spaces of oriented topological surfaces with fixed incoming and outgoing components . This -category has a symmetric monoidal structure given by disjoint union of 1-manifolds. We will abuse notation by denoting the object consisting of the disjoint union of copies of by .
Fix a base commutative derived ring .
Definition 6.2. Let denote the -category of stable, presentable -linear -categories, with 1-morphisms consisting of continuous exact functors.
Note that has a symmetric monoidal structure, the tensor product of presentable stable categories as studied in [L4, 4.2], and the unit of the monoidal structure is the -category of -modules. Since the empty -manifold is the unit of , a topological field theory sends its endomorphisms, which are closed surfaces , to endomorphisms of the unit of the target -category . When , the unit is and is a -module.
Proposition 6.3. For a perfect stack , there is a 2d TFT with and , for closed surfaces .
Proof. We define on objects by assigning .
To define on morphisms, observe first that since is perfect, is perfect, and for , the mapping stack is also perfect (special cases of Corollary 3.25). Now for each , consider the correspondence
Since all of the stacks involved are perfect, pullback and pushforward of quasi-coherent sheaves along this correspondence defines a colimit-preserving functor
Thus applying this construction in families, we obtain a map of spaces
Suppose now that is obtained by sewing two surfaces and . Then we have a diagram of correspondences
Since all of the stacks involved are perfect, base change provides a canonical equivalence of functors
Similar diagrams define the higher compositions.
To complete the construction, note that comes equipped with a canonical symmetric monoidal structure. Namely, by Theorem 4.7, there is a canonical equivalence
and it clearly extends to a symmetric monoidal structure. ∎
As an example, take (in characteristic zero). Then is the -category of sheaves on the adjoint quotient , while is the cohomology of the structure sheaf of the moduli stack of -local systems on .
More generally, the particular topological field theory operations we are considering are not sensitive to the structure of manifolds, and the same construction provides a field theory living over “bordisms” of spaces. In what follows, by a space we will mean a space homotopy equivalent to a finite simplicial set.
Definition 6.4. Let denote the bordism -category of spaces, with 1-morphisms from to given by spaces with maps , and composition of and given by the homotopy pushout of spaces
We endow with a symmetric monoidal structure given by disjoint union.
The proof of the previous proposition immediately extends to give the following:
Proposition 6.5. For any perfect stack , there is a symmetric monoidal functor
with and the functor given by pullback and pushforward along the correspondence
The 2d topological field theory above is a categorified analogue of the 2d TFTs defined by string topology on a compact oriented manifold, or the topological B-model defined by a Calabi-Yau variety [C]. Namely, we assign to the circle the -category of quasi-coherent sheaves, rather than the complex of chains, on the loop space. This has the advantage that we may construct maps for arbitrary correspondences without the assumption that is a Calabi-Yau or satisfies Poincaré duality. On the other hand, we have gone up a level of categoricity, pushing off the problems of duality and orientation to the definition of operations between for cobordisms between surfaces or invariants for 3-manifolds. Recall [BK] that an extended 3-dimensional topological field theory assigns a braided (in fact ribbon) category to the circle. This category must however satisfy a strong finiteness and nondegeneracy condition (modularity) coming from its extension to three-manifolds. It would be interesting to see what conditions need to be imposed on the sheaves we consider and on the stack in order to extend in such a fashion. Note that Corollary 4.8 shows that for perfect satisfies an analogue of the Calabi-Yau or Frobenius property: namely it is self-dual as a -module.
Original source: arXiv:0805.0157v5