The set of periods for crazy maps (Q5938642)

The paper studies the (least) periods that can occur in a class of skew product homeomorphisms \(F:\Sigma \times S^1\rightarrow \Sigma\times S^1\) of the form \(F(a,\theta) = (\sigma(a), f_{a(0)}(\theta))\); here \(\sigma:\Sigma\rightarrow \Sigma\) is the full \(N\)-shift for some \(N\geq 2\), and \(f_i\) are \(N\) homeomorphisms of the circle. A characterization of when \(F^k\) has a fixed point is developed, and this characterization is used to give a complete description of the periods that can occur when \(N=2\) and each \(f_i\) is a rotation. Unfortunately, there seems to be an error in this description in the case that each individual rotation is periodic-point free (Theorem 4.4(a)). The proof correctly identifies the condition on the two rotation numbers \(\alpha\), \(\beta\), for there to be periodic points, namely the existence of nonnegative integers \(j\), \(k\) with \(j+k>0\) and \(j\alpha + k\beta\) an integer, but mischaracterizes when this occurs. As an example of the periods that are missed, let \(f_0\) be any irrational rotation, and let \(f_1\) be the second iterate of the inverse of \(f_0\). If \(q_m\in \Sigma\) is the period \(3m\) sequence generated by the word formed by \(2m\) consecutive 0's followed by \(m\) consecutive 1's, then for this \(F\), \((q_m,\theta)\) has least period \(3m\) for every \(\theta\in S^1\).