On the roots of certain Dickson polynomials (Q1708079)

Let \(\mathbb F_q\) denote the finite field of \(q\) elements and \(D_n(x)\) the \(n\)th Dickson polynomial of the first kind. M. Freedman (in a private communication) asked is it true that for every divisor \(m >1\) of the \(j\)th Fermat number \(F_j=2^{2^j}+1\) that there exists \(\alpha\in\mathbb F_{2^{2^j}}^*\) with \(D_m(\alpha )=D_m(\alpha^{-1})=0\). The authors consider a more general question: if \(m\mid 2^n+1\), does there exist \(\alpha\in\mathbb F_{2^n}^*\) with \(D_m(\alpha )=D_m(\alpha^{-1})=0\)? They show the answer is yes if \(m=(2^k+1)\ell\) and that the answer is no if \(m>3\) is prime, \(m\equiv 3\pmod4\) and \(2\) is a primitive element of \(\mathbb F_m\). They show that the answer to Freedman's question is no already for \(j=5\) and \(m=341\). They also consider the question: does there exist \textit{any} proper divisor \(m\) of the Fermat number \(F_j\) such that there exists \(\alpha\in \mathbb F_{2^{2^j}}^*\) with \(D_m(\alpha )=D_m(\alpha^{-1})=0\)? They show the answer is yes if \(j=5\) and \(8\leq j\leq 18\). The proof depends on a connection with the notion of \textit{mad numbers} introduced by Blokhuis and Brouwer.