Cheeger-Chern-Simons theory and differential string classes (Q295537): Difference between revisions

The classical Chern-Weil theory of characteristic classes has two well-known refinements: the Cheeger-Simons differential cohomology theory and the theory of Chern-Simons secondary characteristic classes. In this paper, the author provides a common generalization -- what he calls the Cheeger-Chern-Simons theory -- of these refinements. Two applications of this unified theory are also offered. The first is the Cheeger-Chern-Simons variant of the Hopf theorem on the structure of the cohomology ring of a compact Lie group and the second is a treatment of so-called string structures on a compact Riemmanian spin-manifold. In more details: Section 2 deals with the generalization. Recall that given a principal \(G\)-bundle \(\pi: P\rightarrow X\) over a smooth manifold \(X\) with compatible connection \(\theta\) and an \(\mathrm{Ad}\)-invariant homogeneous polynomial \(\lambda\) of degree \(k\) on the Lie algebra of \(G\), one can construct a closed \(2k\)-form \(\mathrm{CW}_\theta(\lambda)\) having integer periods on \(X\). The Chern-Weil characteristic class is the corresponding de Rham cohomology class \([\mathrm{CW}_\theta (\lambda)]\) which actually takes its value in \(H^{2k}(X;\mathbb Z)\). On the one hand, the \textit{Cheeger-Simons construction} provides us with a unique lift \(\widehat{\mathrm{CW}}_\theta(\lambda)\) of \(\mathrm{CW}_\theta(\lambda)\) into the absolute differential cohomology group \(\widehat{H}^{2k}(X;\mathbb Z)\) such that the associated curvature of \(\widehat{\mathrm{CW}}_\theta(\lambda)\) is \(\mathrm{CW}_\theta(\lambda)\) and its associated characteristic class is \([\mathrm{CW}_\theta(\lambda)]\). On the other hand, the \textit{Chern-Simons construction} gives rise to a \((2k-1)\)-form \(\mathrm{CS}_\theta(\lambda)\) on \(P\) satisfying \(\mathrm d\mathrm{CS}_\theta(\lambda)=\pi^*\mathrm{CW}_\theta(\lambda)\). (Depending on the connection and \(X\), \(\mathrm{CS}_\theta(\lambda)\) may or may not descend to a \((2k-1)\)-form on \(X\).) The author's first result is the construction of a unique element \(\widehat{\mathrm{CCS}}_\theta(\lambda,u)\) (where \(u\in H^{2k}(BG; \mathbb Z)\) is a universal characteristic class) in the relative differential cohomology group \(\widehat{H}^{2k}(\pi ;\mathbb Z)\) associated to the fibration \(\pi :P\rightarrow X\) such that its corresponding curvature and covariant derivative are \(\mathrm{CW}_\theta(\lambda)\) and \(\mathrm{CS}_\theta(\lambda)\), respectively (see Theorem 2.2). This is the \textit{Cheeger-Chern-Simons construction}. Section 3 and Section 4 goes through some technicalities. Namely, on the one hand, the author studies the generalization of classical transgression theory for singular cohomology of a principal fibration i.e., the existence of a map \(T: H^k(X;\mathbb Z)\rightarrow H^{k-1}(G;\mathbb Z)\) to transgression to loop space as well as to transgression in differential cohomology (see Theorem 3.11). As an application, he offers a differentio-cohomological refinement of the Hopf theorem on the singular cohomology ring of a compact Lie group (see Theorems 3.3 and 3.14). On the other hand, Section 4 is devoted to the theory of differential trivializations in differential cohomology. Roughly speaking this means to find those differential cohomology classes whose associated characteristic classes are trivial (see Definition 4.1). Finally in Section 5, the author applies his technique to a recent problem in Riemannian geometry motivated by string theory, namely finding so-called \textit{string structures} on a compact Riemannian spin manifold \((X,g)\). Let \(P\) denote a lift of the frame bundle of \((X,g)\) to a \(\mathrm{Spin}_n\)-principal bundle. Applying the loop functor one obtains the \(\mathcal L\mathrm{Spin}_n\)-bundle \(\mathcal LP\) over \(\mathcal LX\). In general all the positive mass representations of \(\mathcal L\mathrm{Spin}_n\) are projective only hence in order to construct vector bundles over \(\mathcal LX\) one has to pass to the central-extended-principal \(\widehat{\mathcal L\mathrm{Spin}}_n\)-bundle over \(\mathcal LX\). The only obstruction against this lift is a cohomology class in \(H^3(\mathcal LX;\mathbb Z)\). This obstruction class can be loop-transgressed to \(\frac{1}{2}p_1(P)\in H^4(X;\mathbb Z)\) hence one can work over \(X\) (instead of its loop space). In this \(X\)-picture \(\frac{1}{2}p_1(P)\) is the obstruction to lift the \(\mathrm{Spin}_n\)-bundle \(P\) to a principal bundle whose structure group is the \(3\)-connected covering group \(\mathrm{String}_n\) (this group can be modeled as an infinite dimensional Frèchet Lie group). A \textit{string structure} on \(X\) is a lift of \(P\) to a \(\mathrm{String}_n\)-principal bundle and roughly speaking such lifts can be studied by considering \(\frac{1}{2}p_1(P)\)-differential trivializations (which are called \textit{string classes}).