A stable cohomotopy refinement of Seiberg-Witten invariants. I (Q1434280)

In this paper, the authors develop a refinement of the Seiberg-Witten invariants by using stable cohomotopy theory. In standard Seiberg-Witten theory, the moduli spaces of monopoles are used to produce a number (the Seiberg-Witten invariant) which encodes the information about the topology of the moduli space. More concretely, fix a metric and a \(K\)-theory orientation (i.e. an orientation and a \(spin^c\)-structure \(s\)). Then the Seiberg-Witten equations [\textit{E. Witten}, Math. Res. Lett. 1, 769--796 (1994; Zbl 0867.57029)] can be written as a map \(\mu\) from the spaces of connections and positive spinors to the spaces of self-dual \(2\)-forms and negative spinors. The Seiberg-Witten monopoles are the zeros of this map modulo the gauge group (the group of automorphisms of the equations). The stable cohomotopy refinement of the Seiberg-Witten invariants is defined by studying the homotopy-theoretic properties of the map \(\mu\) itself. This defines an invariant \([\mu]\) of \(X\) which lives in the equivariant stable cohomotopy group \[ \pi^{b^+}_{S^1, H} (\text{Pic}(X); \text{ind}(D)), \] where \(b^+\) is the dimension of a maximal positive definite subspace of the \(2\)-cohomology, \(\text{Pic}(X)=H^1(X;\mathbb{R})/H^1(X;\mathbb{Z})\) is the Picard torus, \(\text{ind}(D)\rightarrow Pic(X)\) is the complex index bundle associated to the Dirac operator \(D\), \(H\) stands for a suitable Hilbert space in which \(S^1\) acts (necessary to define the stable cohomotopy group). This invariant is better understood in the case \(b_1=0\), where \(Pic(X)\) is a point and \(ind(D)\) is a complex vector space of dimension \(d=(c_1(s)^2-\sigma(X))/8\), where \(\sigma(X)\) is the signature of \(X\). Then the equivariant cohomotopy group \(\pi^i_{S^1,H}(*;\mathbb{C}^d)\) equals the non-equivariant stable cohomotopy group \(\pi^{i-1}(\mathbb{C} P^{d-1})\). In this set-up, \([\mu]\) is obtained by taking finite dimensional approximations of \(\mu\), which define maps between spheres, behaving well under suspension. The seminal idea of this invariant comes from [\textit{M. Furuta}, Math. Res. Lett. 8, 279--291 (2001; Zbl 0984.57011)]. The usual Seiberg-Witten invariant can be obtained out of the stable cohomotopy Seiberg-Witten invariant, and in some cases, the new invariant is a torsion element in the stable cohomotopy group, whereas the Seiberg-Witten invariant is zero. On the other hand, it is a computable invariant. The more relevant application of this new invariant (to connected sums of \(4\)-manifolds) is to be found in [\textit{S. Bauer}, ibid., 21--40 (2004; Zbl 1051.57038)].