Equivariant holonomy for bundles and abelian gerbes (Q455691)

From the text: This paper generalizes Bismut's equivariant Chern character to the setting of abelian gerbes. In [Commun. Math. Phys. 98, No. 2, 213--237 (1985; Zbl 0591.58027)], \textit{J.-M. Bismut} introduced the equivariant Chern character of vector bundles \(E\rightarrow M\) with connection over a manifold. This is done via analytic means, which results in a periodically closed form on the free loop space of the base of the bundle. The idea for constructing such a class was to provide a twisted counterpart for the idea, suggested by Witten and carried out by \textit{M. F. Atiyah} [Astérisque 131, 43--59 (1985; Zbl 0578.58039)], that the index of the Dirac operator, given by a path integral over the loop space, can be calculated by a stationary phase approximation of an invariant symplectic form on the loop space of the manifold. Bismut's equivariant Chern character is the contribution to the integrand of this path integral over the loop space necessary to take into account the effects of the connection on the auxillary bundle used for twisting the Dirac operator. Thus, localization techniques, which gave rise to \(\int_M \widehat{A}(TM)\) in the untwisted case, would now yield \(\int_M \widehat{A}(TM)\wedge \text{Ch}(E)\), as expected by the classical index theorem of the twisted Dirac operator. More recently, in a paper of \textit{F. Han} [``Supersymmetric QFT, super loop spaces and Bismut-Chern character'', Preprint (2007), \url{arXiv:0711.3862}], the Bismut Chern character has appeared in the passage from a \(1|1\) supersymmetric (SUSY) field theory on \(M\), determined by a connection on a vector bundle on \(M\), to a \(0|1\) supersymmetric field theory on \(LM\). This work relies on constructing a SUSY field theory out of a vector bundle and a superconnection as described by \textit{F. Dumitrescu} [Pac. J. Math. 236, No. 2, 307--332 (2008; Zbl 1155.58001); addendum ibid. 256, No. 1, 253--256 (2012; Zbl 1244.53029)]. Naively, a bundle and a connection can be thought of as a 1-dimensional field theory in which to 0-dimensional objects (points) one assigns a vector space, the fibre of the bundle above the point, and to 1-dimensional objects, paths in \(M\), one assigns a map between the vector spaces assigned to the end points. In [F. Dumitrescu, loc. cit.], this scheme is generalized to \(0|1\) dimensional objects, super points in \(M\), and \(1|1\) dimensional objects, super paths in \(M\). Furthermore, by thinking the space of maps \(\mathbb{R}^{1|1}\rightarrow M\) as the space of maps \(\mathbb{R}^{0|1}\rightarrow \{\mathbb{R}^{1|0}\rightarrow M\}\), one can get from a \(1|1\) theory on \(M\) to a \(0|1\) theory on the space of paths in \(M\) and similarly \(LM\), as described by F. Han [loc. cit.]. Further work is on the way as part of the general Stolz and Teichner program [\textit{S. Stolz} and \textit{P. Teichner}, ``Supersymmetric field theories and generalized cohomology'', in: H. Sati (ed.) et al., Mathematical foundations of quantum field theory and perturbative string theory. Proceedings of Symposia in Pure Mathematics 83, 279--340 (2011; Zbl 1257.55003)] of understanding the concordance classes of supersymmetric field theories to construct \(2|1\) supersymmetric field theories associated to gerbes; in much the same way, \(1|1\) theories were constructed out of a bundle endowed with a superconnection (see [F. Dumitrescu, loc. cit.]). One can then interpret such a \(2|1\) theory on \(M\) as a \(0|1\) theory on \(M^{\mathbb{T}}\), the space of maps from the torus \(\mathbb{T}\) to \(M\), using the adjunction \(\{\mathbb{R}^{2|1}\rightarrow M\}\simeq\{\mathbb{R}^{0|1}\rightarrow \{\mathbb{R}^{0|2}\rightarrow M\}\}\). Inspired by such connections to supersymmetric and non-supersymmetric field theories, the authors address the naturally arising question of the extension of the holonomy of an abelian gerbe to a torus equivariant class on the mapping space of the torus \(\mathbb{T}\) on \(M\). \textit{E. Getzler, J. D. Jones} and \textit{S. Petrack} [Topology 30, No. 3, 339--371 (1991; Zbl 0729.58004)] reinterpreted Bismut's construction using the Hochschild complex of matrix-valued forms, given an imbedding of the bundle and connection into a trivial bundle. This work shed light on the Bismut class and gave a specific Hochschild cocycle manufactured from the connection and its curvature. This class was later shown by \textit{W. Zhang} in [Lectures on Chern-Weil theory and Witten deformations. Nankai Tracts in Mathematics. 4. Singapore: World Scientific (2001; Zbl 0993.58014)] to be independent of the imbedding and connection. The contents of the present work is as follows. In the first part (Sections 2 and 3), the authors review and provide a conceptual and more fundamental, pre-trace, interpretation of this construction in terms of holonomy, its covariant derivative and other naturally constructed higher degree forms and their covariant derivative. More explicitly, the authors think of holonomy \(\text{hol}\in \Omega^0(LM,\mathcal{E})\) as a section of the pullback of the endomorphism bundle on the free loop space via the map that sends a loop to its base point. This bundle on the loop space has an induced connection with respect to which one can take the covariant derivative of the holonomy as a section. The result is a 1-form on the free loop space with values in this pullback bundle. The authors then show that there is a naturally defined 2-form with values in this bundle whose contraction with the natural velocity vector field on the loop space is the covariant derivative of the holonomy. One then asks for the covariant derivative of this naturally defined 2-form, and the answer turns out to be a 3-form which itself is given by the contraction of a naturally defined 4-form on the loop space. This process continues ad infinitum to produce an infinite sequence of naturally defined even degree forms on the free loop space, starting with the holonomy 0-form. Theorem 2.18. For all \(k\geq 0\), we have forms \(\text{hol}_{2k}\in \Omega^{2k}(LM,\mathcal{E})\), where \(\text{hol}_0=\text{hol}\) is the holonomy, such that \(\nabla^* \text{hol}_{2k}=-\iota_t \text{hol}_{2k+2}\in \Omega^{2k+1}(LM,\mathcal{E})\), where \(\iota_t\) is contraction by the natural vector field on \(LM\) given by the circle action. As a corollary of this theorem, the authors obtain a result previously obtained by Bismut [loc. cit.] and Getzler, Jones and Petrack [loc. cit.]. Corollary 2.19. For \(\text{Ch}^{(u)}(E;\nabla):=\sum _{k\geq 0}u^{-k} \text{hol}_{2k}\), where \(u\) is a formal variable of degree 2, we have \((\nabla^*+u.\iota_t)(\text{Ch}^{(u)}(E;\nabla))=0\). This last result is derived in Section 3 using a local version of the Hochschild complex, which is suitable for the local data of a bundle with connection (Theorem 3.14 and Corollary 3.15). It is this local version that the authors generalize to the case of abelian gerbes with connection. The second part of the work (Sections 4 and 5) comprise an analogous discussion for the 2-holonomy of an abelian gerbe. Namely, starting with an abelian gerbe \(M\), and given a fixed closed oriented surface \(\Sigma\), one can consider 2-holonomy as a real valued function on the mapping space \(M^{\Sigma}\). The authors again use a local version of (higher) Hochschild complexes, which are suitable for the local data of an abelian gerbe with connection. They remark that the situation is in some sense simpler than in the case of the bundle, since 2-holonomy is now complex valued. This, specify the authors, is due to the fact that they are considering abelian gerbes, which are higher analogues of 1-dimensional vector bundles. Section 4 considers the case of the torus \(\Sigma=\mathbb{T}\), where 2-holonomy is regarded as a real valued zero form on the mapping space, \(\text{hol}\in\Omega^0(M^{\mathbb{T}})\), so that one can ask for its exterior derivative. The authors show that the result is a 1-form on \(M^{\mathbb{T}}\) which is the contraction of a naturally defined 2-form on \(M^{\mathbb{T}}\) with respect to a Killing vector field on the torus. There are many such 2-forms, one for each choice of the Killing vector field on the torus. Similar to the bundle case above, the authors give a hierarchical construction which extends 2-holonomy to an equivariant closed element on the torus mapping space. More precisely, the authors obtain the following theorem and corollary. Theorem 4.17. For all \(k,l\geq 0\), we have forms \(\text{hol}_{2k,2l}\in \Omega^{2k+2l}(M^{\mathbb{T}})\), where \(\text{hol}_{0,0}=\text{hol}\) is 2-holonomy, such that \(d(\text{hol}_{2k,2l})=-\iota_{\mathbf{t}}(\text{hol}_{2k+2,2l})=-\iota_{\mathbf{u}}(\text{hol}_{2k,2l+2})\), where \(\iota_{\mathbf{t}}\) and \(\iota_{\mathbf{u}}\) are contraction by the two natural vector fields \(M^{\mathbb{T}}\) given by the two circle actions of the torus \(\mathbb{T}=S^1\times S^1\). Corollary 4.18. For \(a+b=1\), we have that \(\text{Ch}(\mathcal{G},a,b):=\sum_{k\geq 0,l \geq 0} a^k.b^l.\text{hol}_{2k,2l}\in \Omega(M^{\mathbb{T}})^{inv(\mathbf{t}+\mathbf{u})}\) is a closed element, that is, \((d+\iota_{\mathbf{t}}+\iota_{\mathbf{t}})(\text{Ch}(\mathcal{G},a,b))=0\). The authors remark that, in this description of the equivariant Chern character, they omitted the use of any formal variables. The reason for this is that, without formal variables, they can determine an interesting relationship between the equivariantly closed classes \(\text{Ch}(E;\nabla):=\sum_{k\geq 0}\text{hol}_{2k}\) and \(\text{Ch}(\mathcal{G},a,b)\) for a suitable setting. In fact in Subsection 4.4, the authors recall a well -known construction starting from an abelian gerbe on \(M\) to induce a line bundle \(E\) with connection \(\widehat{A}\) on the free loop space \(LM\). Using this construction, they show that the equivariant forms can be identified. In the last section, Section 5, the authors consider the case of a general surface \(\Sigma\). They show how to produce the gerbe holonomy \(\text{hol}\in\Omega^0(M^{\Sigma})\) via local Hochschild methods, and calculate its De Rham differential (Proposition 5.12) as \(d_{\text{DR}}(\text{hol})=i \cdot \text{It}(H)\wedge \text{hol}\), where \(H\) denotes the 3-curvature of the gerbe. However, unlike in the case \(\Sigma=\mathbb{T}\), we do not have any natural candidates to complete holonomy to an equivariantly closed form.