Relative cohomological index theories (Q1089648)

In this paper, connected with previous works [the authors and \textit{P. H. Rabinowitz}, Trans. Am. Math. Soc. 274, 345-360 (1982; Zbl 0506.58010); the second author, Contemp. Math. 12, 157-175 (1982; Zbl 0531.57036)], the authors set up the general theory of relative cohomological indices. More precisely, they define two different indices, \(index_{\Lambda}(X,A)\) and \(index^{\delta}_{\Lambda}(X,A)\), of actions of a compact Lie group G on a pair (X,A) of paracompact G-spaces. Here \(\Lambda\) is a subring of \(H^*(BG)\) and \(\delta\) : \(H^*_ G(A)\to H^*_ G(X,A)\) is the coboundary operator. From the definitions it follows at once that \(index^{\delta}_{\Lambda}(X,A)\leq index_{\Lambda}(X,A)\) and \(index^{\delta}_{\Lambda}X=index_{\Lambda}X\) when \(A=\emptyset\). These general notions are first compared with previous cohomological index theories and then basic properties, such as monotonicity, additivity, lower semicontinuity and piercing property, are illustrated for both of them. Special attention is devoted to the description of the conditions under which \(index_{\Lambda}(X,A)\), hence \(index^{\delta}_{\Lambda}(X,A)\), is finite, and to the presentation of some useful computations of \(index_{\Lambda}\) and \(index^{\delta}_{\Lambda}\) in many cases of interest in the applications. After this long discussion it is clear that each of these two index theories has its own advantages over the other. Finally the authors illustrate how their relative index theory \(index^{\delta}_{\Lambda}\) can be successfully applied in the Lyusternik-Shnirel'man method. Hence they are able to obtain abstract critical point existence theorems for invariant \(C^ 1\) functionals, and to cover those cases in which the action of G is not necessarily free. The authors establish various sufficient conditions under which the functionals above possess unbounded sequences of critical values. In particular, Theorem 5.5 contains as a corollary one of the main results of the first paper quoted above, which can now be proved by a direct application of the \(\delta\)-index theory instead of invoking the generalization of the Borsuk-Ulam theorem proved there.