Maximum gap in cyclotomic polynomials (Q1981578)

For a polynomial \(f(x)=c_1x^{e_1}+\ldots+c_tx^{e_t},\) with non-zero coefficients \(c_i\) and ordered \(e_1<\ldots<e_t,\) the authors define the maximum gap \(g(f)\) as \(\max_{0\le i<t}(e_{i+1}-e_i).\) The subject of the paper is the behavior of \(g(\Phi_n),\) with \(\Phi_n\) the \(n\)-th cyclotomic polynomial. \textit{H. Hong} et al. [J. Number Theory 132, No. 10, 2297--2315 (2012; Zbl 1273.11051)] showed that \(g(\Phi_{p_1p_2})=p_1-1\) for primes \(p_2>p_1.\) The cyclotomic polynomials involved here are called binary and are related to numerical semigroups generated by two elements. In that setting the reviewer gave a conceptual reproof of the latter result [Am. Math. Mon. 121, No. 10, 890--902 (2014; Zbl 1325.11024)]. The authors establish a major generalization by showing that \(g(\Phi_{mp})=\varphi(m)\) for square-free odd \(m\) and prime \(p>m.\) The proof techniques involves partitioning the cyclotomic polynomial \(\Phi_{mp}\) into blocks, choosing representative blocks, and cleverly showing that all representative blocks are divisible by the \(m\)-th inverse cyclotomic polynomial, which is defined by \((X^m-1)/\Phi_m(x)\) (for basic properties see the paper by the reviewer [J. Number Theory 129, No. 3, 667--680 (2009; Zbl 1220.11037)]). More precisely, they write \(\Phi_{mp}(x)=\sum_{i=0}^{\varphi(m)-1}f_{m,p,i}(x)x^{ip}\) with deg \(f_{m,p,i}<p\) and study the blocks \(f_{m,p,i}\) and some related block notions.