Lefschetz fixed point formula for manifolds with cylindrical ends (Q1182220)

Lefschetz number \(L(f)\) is a topological characteristic of a mapping \(f: M\to M\) that can be used to compute fixed points of \(f\). In the paper under review, the authors generalize this result to an important particular case of manifolds with the boundary, namely, to manifolds with cylindric ends. By this, they mean a manifold \(M\) with such a boundary \(N=\partial M\) that some neighbourhood of \(N\) is diffeomorphic to \(N\times (-\varepsilon,0)\). For such manifolds, one can also use cohomologies to define Lefschetz number \(L(f)\) of mappings \(f\) for which \(f(M)\subset M\). However, it turns out that, unlike the case when \(\partial M=\phi\), this number \(L(f)\) cannot be expressed in terms of fixed points of \(f\). So, how to compute \(L(f)\)? Two ways are proposed. First, we can extend \(M\) to a manifold without a boundary by adding an infinite cylinder \(N\times (0,\infty)\), and extend \(f\) by demanding that the extension of \(f\) is linear in the coordinate \(t\in (0,\infty)\) for all \(t\geq t_ 0\) for some \(t_ 0\), and some other reasonable demands. Then, \(L(f)\) can be expressed in terms of fixed points of this extension. Thus, \(L(f)\) does not depend on how exactly we extend \(f\). The second approach is to find the expression of \(L(f)\) in terms of fixed points of the original mapping \(f\) and topological characteristics of \(N\). In the paper, such a formula is given for the case when \(N\) is a union of finitely many disjoint spheres. There are several typos: p. 100, after the formula (ND), the graph should be \(\{(x,f(x))\mid x\in M\}\); p. 100, condition 3), should contain the inequality \(\text{dist}(\hat f(z),z)>\varepsilon\); p. 101, part A of the proof, the first displayed inequality should be \(f^ n(y,t)\leq 0\).