Some conjectures on the maximal height of divisors of \(x^n - 1\) (Q618728)

Define the height of a polynomial to be its largest coefficient in absolute value. For a positive integer \(n\), let \(B(n)\) denote the maximal height of a polynomial \(f(x)\) having integer coefficients and dividing \(x^n-1\). The authors put 30 processors in action for several months to determine the value of \(B(n)\) for almost 300,000 positive integers \(n\). Let \(n\) have the prime factorization \(p_1^{e_1}\cdots p_s^{e_s}\), with \(p_1<p_2\dots <p_s\) primes. It is easy to show that \(B(n)\geq \min\{p_1^{e_1},\dots,p_s^{e_s}\}\). The authors observe that this bound is surprisingly good for the data they computed, at least when \(s\geq 2\). Indeed, it is known that \(B(p_1p_2)=\min\{p_1,p_2\}\) and \(B(p_1^2p_2)=\min\{p_1^2,p_2\}\). On basis of their computations the authors formulate five conjectures, for example that \(B(p_1^2p_2^2)\) is always attained by one of two specific divisors of \(x^{p_1^2p_2^2}-1\). The authors set up a database making their numerical computations available to those interested.