Co-Euler structures on bordisms (Q492232)

This paper extends the results of \textit{D. Burghelea} and \textit{S. Haller} [Geom. Topol. 10, 1185--1238 (2006; Zbl 1204.58027)] about co-Euler structures to the setting of manifolds with boundary. Let \(M^m\) be a closed Riemannian manifold. In the case where \(M\) is orientable and the Euler characteristic \(\chi(M)\) vanishes, co-Euler structures on \(M\) consist of pairs of a metric \(g\) together with a primitive \(\alpha\) of the Euler form \(\mathrm{Pf}(g)\), modulo the following equivalence relation: two primitives \(\alpha,\alpha'\) of \(\mathrm{Pf}(g),\mathrm{Pf}(g')\) are said to be equivalent if their difference equals the transgression \(e(g,g')\) modulo exact \(m-1\) forms (the transgression form itself is well-defined modulo exact forms). The author extends this definition to the case of manifolds with boundary with vanishing relative Euler class, using the complex of relative differential forms on \((M,\partial M)\). The definition is then extended to the more general case where \(M\) is not orientable (using forms with coefficients in the orientation bundle) and non-vanishing Euler class (here he needs to choose a base point, obtaining the so-called base-pointed co-Euler structures). He computes the variation of co-Euler structures with respect to first-order variations of the metric, finding (like in the boundary-less case studied by Burghelea and Haller) that it is similar to the variation of analytic torsion. The co-Euler structure is then bundled together with the analytic torsion to construct a topological invariant independent of the Riemannian metric.