Sister Beiter and Kloosterman: a tale of cyclotomic coefficients and modular inverses (Q740467)

For any natural number \(n\) let \(\zeta_{n}\) denote a fixed primitive \(n\)-th root of unity in \(\mathbb{C}\); for example, one can take \(\zeta_{n}=\exp(2 \pi i / n)\). The \(n\)-th cyclotomic polynomial may be defined as the minimal polynomial of \(\zeta_{n}\) over \(\mathbb{Q}\), or alternatively, by \(\Phi_{n}(x) = \prod (x-\zeta_{n}^{j})\) where the product ranges over integers \(j\) coprime to \(n\) in the range \(1 \leq j \leq n\). Thus \(\Phi_{n}(x)\) is a monic polynomial with integer coefficients and of degree \(\varphi(n)\), where \(\varphi\) denotes Euler's totient function. A well-known and fundamental relation among cyclotomic polynomials is \(x^{n}-1 = \prod_{d \mid n} \Phi_{n}(x)\) where the product ranges over the positive divisors of \(n\). Thus for \(p\) prime we have \(\Phi_{p}(x) = 1+ x + \cdots + x^{p-1}\). Moreover, for \(n>1\) odd we have \(\Phi_{2n}(x)=\Phi_{n}(-x)\). Let \(a_{n}(k)\) denote the coefficient of the degree \(k\) term of \(\Phi_{n}(x)\), so that we may write \(\Phi_{n}(x) = \prod_{k=0}^{\varphi(n)} a_{n}(k)x^{k}\). The height of \(\Phi_{n}(x)\) is defined to be \(A(n) = \max \{ |a_{n}(k)| : k \geq 0 \}\). It is straightforward to show that \(A(n)=A(N)\) where \(N\) is the product of the odd prime factors of \(n\). If \(n\) has at most two distinct odd prime factors then \(A(n)=1\) and so, in particular, \(A(n)=1\) for \(n<105\). In fact, \(A(3 \cdot 5 \cdot 7)=2\) with \(a_{105}(7)=-2\). Thus the easiest situation in which we can expect non-trivial behaviour of \(A(n)\) is in the ternary case where \(n=pqr\) with \(2<p<q<r\) odd primes. It is known that \(A(pqr) < 3p/4\), which is clearly only depends on \(p\). Thus for a fixed prime \(p\) the maximum \(M(p) := \max \{ A(pqr) : p < q < r \}\), where \(q\) and \(r\) range over all the primes satisfying \(p < q < r\), exists. A major open problem involving ternary cyclotomic coefficients is to find a finite procedure to determine \(M(p)\). \textit{H. Möller} [Math. Z. 119, 33--40 (1971; Zbl 0196.07201)] gave a construction showing that \(M(p) \geq (p+1)/2\) for \(p > 5\). In the other direction, \textit{Sister M. Beiter} [Am. Math. Mon. 75, 370--372 (1968; Zbl 0157.08804)] conjectured that \(M(p) \leq (p+1)/2\). It has since been shown that \(M(3)=2\), \(M(5)=3\) and \(M(7)=4\). Therefore Beiter's conjecture becomes \(M(p)=(p+1)/2\) and holds for \(p \leq 7\). However, \textit{Y. Gallot} and \textit{P. Moree} [J. Reine Angew. Math. 632, 105--125 (2009; Zbl 1230.11030)] showed that this conjecture is in fact false for every \(p \geq 11\) and also provided a construction of counter-examples. Furthermore, they showed that if \(\varepsilon>0\) then \(2p(1-\varepsilon)/3 \leq M(p) \leq 3p/4\) for sufficiently large \(p\). The article under review investigates the abundance of counter-examples to Beiter's conjecture and sharpens the above lower bound for \(M(p)\). The techniques used involve Kloosterman sums \(K(a,b;p) := \sum_{1 \leq x \leq p-1} \exp(2 \pi i (ax+b\bar{x})/p)\), where \(p\) is prime and \(\bar{x}\) denotes the inverse of \(x\) modulo \(p\), and other tools from the study of the distribution of modular inverses.