Equivariant Pontrjagin classes and applications to orbit spaces. Applications of the G-signature theorem to transformation groups, symmetric products and number theory (Q2553115)

If a finite group \(G\) acts smoothly, effectively, and in an orientation preserving way an a closed manifold \(X\), the orbit space \(X/G\) is a rational homology manifold. In this situation it is interesting to compare characteristic classes of the action and of the orbit space. For a rational homology manifold M, by Thom and Milnor one has a rational \(L\)-class \(L(M)\) (derived from their rational Pontrjagin classes) which determines the signature sign\((M)\) of \(M\) in an easy way. On the other hand, for \(g\in G\), the \(G\)-signature theorem of Atiyah and Singer determines the equivariant signature sign\((g,X)\) in the same way via an equivariant \(L\)-class \(L'(g,X) \in H^*(X^g)\) in the cohomology of the fixed point set \(X^g\) of \(g\). Thus one is led to compare these \(L\)-classes, and the result is (Theorem 1. p. 4): \(\pi^*L(X/G)=\sum_{g\in G} j_! L' (g,X)\). Here, \(\pi: X\to X/G\) is the orbit map, \(j:X^g\to X\) the inclusion, and \(j_!: H^*(X^g)\to H^*(X)\) the Umkehr homomorphism. As \(\pi^*\) is injective (complex coefficients), the Atiyah-Singer \(L\)-classes determine \(L(X/G)\) completely. Next this central result is extended to apply to an oriented rational homology \(G\)-manifold \(X\) (\(G\) finite abelian). As in this case \(j\), \(L'(g,X)\) is not defined, in the above formula it has to be replaced by an equivariant \(L\)-class \(L(g,X)\) whose definition is analogue to the definition of the \(L\)-class for rational homology manifolds by Milnor. Furthermore, if \(h\) is a \(G\)-equivariant automorphism of \(X\) having finite order and inducing \(h'\) on \(X/G\), then \(\pi^*L(h',X/G)=\sum_{g\in G}L(gh,X)\). In the last section of Chapter I, Bott's formula for the \(L\)-class of the orbit space of \(\mathbb{P}^n\mathbb{C}\) by a finite linear group action is verified. In Chapter II, \(L\)-classes of symmetric products of smooth manifolds are calculated. If a smooth oriented \(n\)-manifold \(X\) is even-dimensional, then the symmetric group \(S_n\) acts smoothly an \(X'\) and this action preserves the orientation. Hence \(X(m) = X^m/S_m\), the \(m\)-th symmetric product of \(X\), is a rational homology manifold, and the results of Chapter I apply. It turns out that the restriction of the \(L\)-class of \(X(m + 1)\) to \(X(m)\) is \(Q_s (\eta_m) L(X(m))\) where \(\eta_m\) is derived from a universal class \(\eta\in H^{2s}(X(\infty))\) by restriction to \(X(m)\), and \(Q_s\) is a power series depending only an \(s = n/2\). This is a remarkable formula, not only because of its simplicity: It can be restated as \(L(X(m))\) = restriction of \(Q_s(\eta^{m+1}\cdot G\) where \(G\in H^{**}(X(\infty))\), and so reveals the dependence of the \(L\)-class of \(X(m)\) on \(s\) and \(m\). A similarly simple formula is derived for the equivariant \(L\)-classes of an action an \(X(m)\) that is defined by an action on \(X\). Applications to be mentioned are a new proof of a formula by Hirzebruch for the signature of \(X(m)\), a new proof of a formula by Macdonald for the \(L\)-dass of the symmetric product of a Riemann surface, and the computation of the equivariant \(L\)-class of \(X(m)\) for \(X\) an \(n\)-sphere. The final chapter deals with problems of elementary number theory suggested by the \(G\)-signature theorem. Some of these occurred in the first two chapters already, and now they are treated systematically. Among other results there is a proof of a generalization of Rademacher's generalized reciprocity law. This is an identity involving cotangent sums and the Hirzebruch \(L\)-polynomials. Two more proofs of this are given by constructing appropriate group actions and then applying the \(G\)-signature theorem in the first case, the formula of Chapter I involving equivariant classes for rational homology-manifolds in the second case. The final section presents some new formulae for the signature of Brieskorn varieties, for the signatures of some automorphisms of Brieskorn varieties, and the so-called \(\alpha\)-invariant of a certain type of \(S^1\)-action on Brieskorn manifolds. No doubt, this is the right form of publication for the subject, because it is possible to present proofs in some detail here. This way the material should be accessible also to readers with less experience in the field. The sections collecting what is needed from other sources will be helpful in this case. Also, there is a good chance that some of the obvious enthusiasm of the author about the subject might be communicated to the reader.