On cyclotomic numbers and the reduction map for the \(K\)-theory of the integers (Q5952135)

For any even integer \(n \geq 2\) the \(K\)-theory group \(K_{2n+1}(\mathbb Z)\) has rank \(1\). It is known by a result of \textit{D. Arlettaz} and \textit{G. Banaszak} [J. Reine Angew. Math. 461, 63-79 (1995; Zbl 0813.19001)] that in this case the reduction map \[ K_{2n+1}(\mathbb Z) \rightarrow K_{2n+1}(\mathbb F_p) \] is nontrivial for infinitely many primes \(p\). The author considerably improves this result in the following way: Let \(l\) be an odd prime and \(k\) a fixed positive integer. Let \(M_k\) denote the set of primes \(p\) for which the \(l\)-torsion part of \(K_{2n+1}(\mathbb F_p)\) is isomorphic to \(\mathbb Z/l^k\) and for which the reduction map \(K_{2n+1}(\mathbb Z) \rightarrow K_{2n+1}(\mathbb F_p)\) is nontrivial. Then the set \(M_k\) has a positive Dirichlet density, which is equal to \((l^k-1)/l^{2k}\). This result holds for almost all primes \(l\), more precisely for those \(l\), which do not divide the index of the subgroup of \(K_{2n+1}(\mathbb Z)\) generated by the Beilinson element. If Vandiver's conjecture and the Quillen-Lichtenbaum conjecture hold for the prime \(l\), then in fact this index is prime to \(l\). The proof uses the fact that under the \(l\)-adic Chern character the Beilinson element is mapped to the cyclotomic element in étale cohomology constructed by \textit{C. Soulé} [Astérisque 147-148, 225-257 (1987; Zbl 0632.12014)].