The Inverse-Similarity Problem for Real Orthogonal Matrices (Q3130166)

This is an accessible presentation of some recent results [\textit{G. R. Goodson}, \textit{A. Del Junco}, \textit{M. Lemańczyk} and \textit{D. J. Rudolph}, Ergodic Theory Dyn. Syst. 16, No. 1, 97-124 (1996; Zbl 0846.28008); \textit{G. R. Goodson} and \textit{M. Lemańczyk}, Transformations conjugate to their inverses have even essential values, Proc. Am. Math. Soc. 124, No. 9, 2703-2710 (1996)] which reinforces the importance of canonical forms and matrix decompositions in the undergraduate linear algebra curriculum. Given an orthogonal matrix \(U\) and \(UA= AU^T\), where \(A\) (nonsingular) is a similarity between \(U\) and \(U^T\) one derives necessary and sufficient conditions for every real orthogonal similarity \(A\) to be an involution (i.e. \(A^2=I\)); the cases where there exist such similarities that are not involutions are also investigated. Finally, one shows that the eigenvalues corresponding to eigenvalues of \(U\) in the orthogonal complement of the subspace \(\{x: A^2x=x\}\) have even multiplicity. This result is generalized to the infinite-dimensional situation.