On the Hilbert property and the fundamental group of algebraic varieties (Q2396180)

Throughout this review, \(k\) is always a number field, and a \textit{variety} over \(k\) is an integral scheme, quasi-projective over \(k\). A \textit{cover} of a variety \(X\) over \(k\) is a dominant rational map \(\pi: Y\to X\) of finite degree, where \(Y\) is a variety over \(k\). We say that a variety \(X\) over \(k\) has the \textit{Hilbert property} if, for any finite collection of covers \(\pi_i: Y_i\to X\), each of degree \(>1\), the set \(X(k)\setminus\bigcup_i \pi_i(Y_i(k))\) is Zariski-dense in \(X\). This is motivated by the Hilbert Irreducibility Theorem, which says that \(\mathbb P^1_k\) has the Hilbert property. The Hilbert property trivially fails to hold for a variety \(X\) if \(X(k)\) is not Zariski-dense, and also fails if \(X\) has nontrivial Albanese variety. We say that a variety \(X\) over \(k\) is \textit{algebraically simply connected} if any cover of its normalization, not having rational sections over \(\bar k\), is ramified (somewhere). The paper under review then shows that another necessary condition for a variety to have the Hilbert property is that it be algebraically simply connected. The proof of this condition relies on the following fact. Let \(\pi: Y\to X\) be a cover over \(k\), without rational sections over \(\bar k\). Let \(T\) be a subset of \(Y(k')\cap\pi^{-1}(X(k))\) for some number field \(k'\supseteq k\). Then there exist finitely many covers \(\pi_i: Y_i\to X\) over \(k\), each without rational sections over \(\bar k\), such that \(\pi(T)\subseteq\bigcup_i\pi_i(Y_i(k))\). This fact can be combined with the (traditional) Chevalley-Weil theorem to give the following variant of the latter, which does not involve passing to a larger number field. Let \(X\) and \(Y\) be geometrically irreducible projective varieties over \(k\), and let \(\pi: Y\to X\) be a finite étale morphism of degree \(>1\). Then there exist finitely many geometrically irreducible covers \(\pi_i: Y_i\to X\) such that \(X(k)\subseteq\bigcup_i \pi_i(Y_i(k))\). Such varieties \(X\) fail to satisfy the Hilbert property, so this variant may be viewed as the opposite of the Hilbert property. The paper then gives an example of a smooth projective surface (an Enriques surface) which has Zariski-dense rational points (over \(\mathbb Q\)), is not algebraically simply connected, and has trivial Albanese, yet does not satisfy the Hilbert property. The paper also discusses a further weakening of the Weak Approximation property, and shows that varieties with this condition are algebraically simply connected, as already noted [\textit{Kh. P. Minchev}, Dokl. Akad. Nauk BSSR 33, No. 1, 5--8 (1989; Zbl 0693.14001)]. It proposes some related questions/conjectures, one of which would imply that all \(k\)-unirational varieties have the Hilbert property. Relevant to this, the paper shows that the K3 surface \(x^4+y^4=z^4+w^4\) is not geometrically (uni)rational, yet has the Hilbert property over \(\mathbb Q\).