Polynomial minimum root separation (Q5950981)

The problem considered in this paper is the important following one: given an univariate integer polynomial \(P\) study the minimum distance between distinct roots of \(P\), say \(\text{sep} (P)\), in terms of the (naïve) height of \(P\), i.e. the maximum of the absolute values of the coefficients of this polynomial. This problem has been studied in several directions: lower bounds on \(\text{sep}(P)\) (obtained by suitable Liouville estimates), explicit examples for which \(\text{sep}(P)\) is small (such as \((aX-1)(X^n-aX+1)\) and \(X^n-2(aX-1)^2\) for large \(a\) and \(n\geq 3\)). The main contribution of this paper is a massive experimental study. According to this work, the author considers that this leads to conjecture that the ``worst'' case should be approximatively the square root of the best theoretical lower bound presently known. In my opinion, the difficulty with extensive computational studies is that interesting special examples may be too far or to rare to be found. In any case, the problem remains open and these experiences bring interesting information.