Some elements of finite order in \(K_{2}\mathbb Q\) (Q2641546)

Let \(K_2(\mathbb{Q})\) be the Milnor \(K_2\) group, and \(\Phi_n(x)\) the \(n\)th cyclotomic polynomial. This paper considers the set \(G_n(\mathbb{Q}) \subseteq K_2(\mathbb{Q})\) of symbols of the form \(\{a,\Phi_n(a)\}\), for all \(a \in \mathbb{Q}^\times\) (for which \(\Phi_n(a) \neq 0\)). It is known that \(G_n(\mathbb{Q})\) is a group for \(n = 1,2,3,4,6\), and conjecturally, for no other \(n\). This was confirmed for \(n = 5,7\), for \(n\) of the form \(2^a 3^b\), and for powers of regular primes. The techniques developed in this paper show that \(G_n\) is not a group for \(n = 15, 21, 33, 35, 60\) and \(105\). Let \(p\) denote the maximal prime divisor of \(n\), and let \(\Phi_n(x,y) = \Phi_n(x/y)y^{\phi(n)}\). The first step in the proof is to show that if \(G_n\) is closed under taking \(p\)-powers, then for every \(q \equiv 1 \pmod{n}\), there exist \(a,b,z \in \mathbb{Z}^3\), such that \(\Phi_n(a,b) = \varepsilon (zq)^p\) where \(\varepsilon \in \{1,p\}\) is uniquely determined by \(n\). To show that this diophantine equation has no solution (for a suitable \(q\)), one decomposes \(\Phi_n(a,b)\) over \(\mathbb{Z}[\zeta_p]\), and then reduces modulo \(q\).