Chern characters for supersymmetric field theories (Q6138571)

This paper is a valuable contribution in the Stolz-Teichner program to construct a cocycle model for topological modular forms (TMF) in terms of 2-dimensional supersymmetric quantum field theory (SQFT). Let \(M\) be a smooth manifold. \textit{S. Stolz} and \textit{P. Teichner} [Proc. Symp. Pure Math. 83, 279--340 (2011; Zbl 1257.55003)] have constructed a category called ``\(\mathsf{2|1-EBord}(M)\)'' of 2|1-dimensional super-Euclidean bordisms equipped with a smooth map to \(M\). Define \(\mathsf{2|1-EFT}(M)\) to be the groupoid of symmetric monoidal functors from \(\mathsf{2|1-EBord}(M)\) to a certain category \(\mathcal{V}\) of topological vector spaces. This should be thought of as the groupoid of ``unextended'' aka ``not fully local'' 2D SQFTs with background fields in \(M\). The Stolz-Teichner program suggests to extend from 1- to 2-categories, with appropriately chosen source and target. The resulting 2-groupoid of ``extended'' aka ``fully local'' 2D SQFTs over \(M\) should be a cocycle model (in the \(M\) variable) for TMF. For this proposal to be true, a necessary condition is that there should be a Chern map to ``complexified TMF'', or in other words to ordinary cohomology with coefficients in the algebra \(\mathrm{MF}\) of weakly holomorphic modular forms. This paper constructs such a map. Specifically, inside (the endomorphisms of the empty 1-manifold in) the category \(\mathsf{2|1-EBord}(M)\) is a groupoid that I will call \(\mathcal{G}(M)\), whose elements are closed super tori for which the map to \(M\) depends only on the fermionic directions of the super torus. Clearly, one can restrict an SQFT along \(\mathcal{G}(M) \hookrightarrow \mathsf{2|1-EBord}(M)\) to get a function on this groupoid. Since an element of \(G\) is a torus with extra data, a function on \(\mathcal{G}\) should end up being some sort of modular form. This restriction map will end up being the Chern map. At this point, there are many technical hurdles to overcome. Most interesting to this reviewer is the heavy use of super geometry: the groupoid \(\mathcal{G}(M)\) is really a super Lie stack (a stack on the site of super manifolds); its precise construction, and the notion of functions thereon, relies on this extra geometry. Stacks like to be presented as quotients, and the stack I have called \(\mathcal{G}(M)\) arises as the quotient of a certain super moduli space \(\mathcal{L}_0^{2|1}(M)\), in which the super torus is \(\mathbb{R}^{2|1}/\mathbb{Z}^2\) with some super Euclidean structure, by a certain super Lie group \(\mathsf{Euc}_{2|1}\) that forgets the presentation as \(\mathbb{R}^{2|1}/\mathbb{Z}^2\). Note that these are all smooth objects, without any holomorphicity built in. Rather, the super geometry and super stack structures end up imposing a sort of ``derived holomorphicity''. Recall that if \(\mathcal{F}(-)\) is a sheaf on the site of manifolds, then two elements of \(\mathcal{F}(M)\) are called \textit{concordant} when they are the restrictions along \(M \times \{0\}\) and \(M \times \{1\}\) of an element of \(\mathcal{F}(M \times \mathbb{R})\); the Stolz-Teichner conjecture states that \(\mathrm{TMF}(M)\) is supposed to be about concordance classes of SQFTs over \(M\). The main technical results in the paper are the following. First the assignment \(M \mapsto C^\infty(\mathcal{G}(M)) := C^\infty(\mathcal{L}_0^{2|1}(M))^{\mathsf{Euc}_{2|1}}\) is a sheaf in the \(M\)-variable. Second, there is a surjective map from concordance classes of elements of this sheaf to \(\mathrm{H}(M; \mathrm{MF})\). The paper also proves completely analogous results for 1|1-dimensional field theories and for \(\mathrm{KU}\) in place of \(\mathrm{TMF}\).