Smith equivalence of representations for odd order cyclic groups (Q1079237)

If a compact Lie group G acts linearly on a sphere with exactly two fixed points, then the local behavior of the action is the same near either fixed point. Over the past quarter century there has been a considerable amount of work to determine the extent to which an analogous result holds for differentiable actions. In the last few years several families of counterexamples were constructed for actions of even order cyclic groups, but these results did not indicate whether similar examples could exist for odd order cyclic groups. The main result of this paper is the existence of counterexamples in the latter case, and this is done using the authors' machinery for equivariant surgery. The proof begins with a construction of specific equivariant surgery problems from cyclic actions on surfaces with exactly one fixed point, and the next step is an analysis of the obstructions to modifying these manifolds equivariantly into homotopy spheres. These obstructions split into character-theoretic invariants arising from the Atiyah-Singer G- signature formula and related torsion invariants arising from projective class groups and work of Alexander, Conner, and Hamrick. Finiteness properties and elementary combinatorial arguments are used to dispose of the latter, and the former are studied by number-theoretic methods. In order to complete this analysis it is necessary to know that certain sorts of number-theoretic expressions vanish. This is verified provided the cyclic group's order is the product of some very large sequence of odd primes; the existence of a suitable group order with close to 3 million digits is mentioned. \{Reviewer's note: In the last paragraph of page 283 the authors ask if their smooth actions on homotopy spheres are in fact smooth actions on the standard sphere. This can be shown as follows: By construction the homotopy spheres have stable framings that are framed cobordant to disjoint unions of products of framed surfaces. The framed bordism class of the homotopy sphere with framing represents an element in the stable homotopy groups of spheres, and the relation to framed surfaces implies this element is a sum of monomials in 2-dimensional classes. Since every two elements in the stable 2-stem have zero product, it follows that the framed homotopy spheres in question represent zero. But the homotopy spheres in question are even-dimensional, and the results of Kervaire and Milnor imply that an even - (but not 4)-dimensional homotopy sphere is standard if it bounds a framed manifold.\}