An effective version of the primitive element theorem (Q2088991)

Let \(K\) be a field and \(F\) a finite separable extension of \(K\). The primitive element theorem asserts that there exists \(\gamma \in F\), such that \(F = K(\gamma )\). This follows from the cyclicity of the multiplicative group \(F^{\ast }\) in case \(K\) is finite, so we assume further that \(K\) is infinite. Then it is crucial for the proof of the theorem to observe that if \(F = K(\alpha , \beta )\), for some \(\alpha , \beta \in F\), then \(F = K(\alpha + u\beta )\), for all but finitely many \(u \in K\). More precisely, the set of those \(u \in K\) for which \(F \neq K(\alpha +u\beta )\) consists of at most \((d - 1)^2\) elements, where \(d = [F\colon K]\). As shown in the paper under review, the upper bound \((d - 1)^2\) can be improved to \(e(e + 1)/2\), where \(e = d/p(d)\) and \(p(d)\) is the least prime divisor of \(d\) (the author notes that the argument is due to an anonymous referee). The main results of the reviewed paper concern the special case where \(K\) equals the field \(\mathbb{Q}\) of rational numbers. With notation being as above, the author proves that the number of exceptional \(u \in \mathbb{Q}\) is at most equal to \(d\). On the other hand, he gives an example showing that this number can be greater than \((log \ d)^2/(log \ log \ d)^2\), for infinitely many \(d \in \mathbb{N}\).