Topological quantum field theory on non-Abelian gerbes (Q864250)

A non-Abelian \(G\)-gerbe is a natural generalization of a principal bundle; as for principal bundles, in order to investigate the geometry of a gerbe it is convenient to endow it with a connection. The author shows how (under the technical hypotesis that the gerbe is fully decomposed) the infinitesimal symmetries of a gerbe can be described in terms of a BRST operator acting on the space of fields naturally associated to it. More precisely, these fields are the cocycle data \((\lambda_{ij},g_{ijk})\) and the connection data \((m_i,\gamma_{ij},B_i)\); the curvature triple of the given connection is denoted \((\nu_i,\delta_{ij},\omega_i)\). All these fields are \(G\)- or \(\text{ Aut}(G)\)-valued differential forms. Symmetries of these data consist of local gauge symmetries \(h_i\), shifts of the connection by affine data \((\pi_i,\eta_{ij},\alpha_i,E_i)\), and reduced gauge symmetries \(\rho_i\). Infinitesimal symmetries can be naïvely encoded into a BRST-type differential operator \(Q\), by introducing suitable ghost field \(c_i\), \(\varphi_i\), \(\phi_i\) and \(\sigma_i\). The operator \(Q\) actually fails to be nilpotent since \(Q^2\eta_{ij}\neq 0\). However, \(Q\) is nilpotent on all the other fields of the theory, and the obstruction to its nilpotency when acting on the field \(\eta_{ij}\) is precisely the obstruction to extending the ghost fields to global sections. Hence, one can expect to obtain a truly nilpotent operator by introducing suitable new fields controlling the behaviour of the ghost fields on the double intersections of the gerbe charts. It is indeed so, and this phenomenon has an extremely neat interpretation in terms of the universal gerbe, a natural generalization of the universal bundle of pricipal bundles theory. Namely, the connection data for the universal gerbe are of the form \((\mu_i,V_{ij},A_i)\), where, expanding each field in ghost number, \(\mu_i=m_i+c_i\), \(V_{ij}=\gamma_{ij}+a_{ij}\) and \(A_i=B_i+E_i+\phi_i\); note that the lowest components are precisely the connection data \((m_i,\gamma_{ij},B_i)\) of the underlying gerbe. The corrsponding curvature triple is \((F_i,\Delta_{ij},\Omega_i)\), where \(F_i=\nu_i+\pi_i+\varphi_i\), \(\Delta_{ij}=\delta_{ij}+\eta_{ij}+b_{ij}\) and \(\Omega_{i}=\omega_i+\alpha_i+\rho_i+\sigma_i\). Again, the lowest components are precisely the curvature tripe for the given connection on the underlying gerbe. What is remarkable is that all the above introduced fields and ghost fields naturally appear as components of the universal gerbe connection or curvature, together with two additional fields \(a_{ij}\) and \(b_{ij}\), wich precisely control the behaviour of the ghost fields on the double intersections. The exterior derivative \(q\) on the moduli space \({\mathcal H}\) of fully decomposed \(G\)-gerbes is a manifestly nilpotent (of order two) differential on the space of fields, which coincides with the naïve operator \(Q\) on-shell. More precisely, any path integral with a \(q\)-invariant measure and a \(q\)-invariant integrand localizes in the fields \(a_{ij}\) and \(b_{ij}\) and the effective BRST operator \(q\) acts on the gauge-fixed locus as \(Q\). The above universal gerbe construction is reminiscent of the universal bundle construction in topological Yang-Mills theory, and indeed \(q\)-observables, i.e., \(q\)-cohomology classes, can be seen as gerby generalizations of the Donaldson-Witten invariants. Moreover, the trace-part of the \(q\)-cohomology gives the standard Čech-de Rham cohomology of the gerbe.