In this paper we prove that the set of S-integral points of the smooth cubic surfaces in A 3 over a number field k is not thin, for suitable k and S. As a corollary, we obtain results on the complement in P 2 of a smooth cubic curve, improving on Beukers' proof that the S-integral points are Zariski dense, for suitable S and k. With our method we reprove Zariski density, but our result is more powerful since it is a stronger form of Zariski density. We moreover prove that the rational integer points on the Fermat cubic surface x 3 + y 3 + z 3 = 1 form a non-thin set and we link our methods to previous results of Lehmer, Miller-Woollett and Mordell. 1 2 SIMONE COCCIA Theorem 1.5 ([9]). The K3 surface defined in P 3 by x 4 + y 4 = z 4 + w 4is not (uni)rational over C and has the Hilbert property over Q.In the recent preprint [10] J. L. Demeio provided further examples of varieties with the HP, such as K3 surfaces or quotients of varieties under the action of a finite group. Corvaja and Zannier formulated analogous definitions and conjectures for the integer case, which are the object of this paper.Definition 1.6. We say that an algebraic variety X/k has the S-integral Hilbert propertyEquivalently, X has the IHP if given finitely many k-covers π i : Y i → X, each without rational sections defined overk, the setRemark 1.7. One typical assumption is that the covers are given by finite maps π : Y → X where Y is a normal absolutely irreducible variety. In fact, consider a cover π : Y → X where Y is an absolutely irreducible variety. By removing a closed set, we can assume that π is a dominant morphism. We can mutually decompose Y and X in affine open sets such that π restricts to π : V → U . The map π is still a dominant morphism and so it defines an inclusion π * :one obtains a normal absolutely irreducible affine varietyṼ = Spec(A) equipped with a finite morphismπ :Ṽ → U and a birational isomorphism f : V Ṽ such that π =π • f . Gluing these affine spectra we obtain the desired finite map, which differs from π only on a closed set, thus proving that this reduction is possible.
The original version of Hilbert's Irreducibility theorem then becomes:Theorem 1.8. The affine line A 1 has the IHP over Z.The natural analogue of Problem 1.4 is then: Problem 1.9. Any smooth quasi-projective variety topologically simply connected and with a Zariski-dense set of S-integral points has the IHP.In order to ask a similar question for non-simply connected varieties, one is led to the following definition.Definition 1.10 (Weak integral Hilbert property). A normal quasi-projective algebraic variety X/k has the weak integral Hilbert property (WIHP) if, given finitely many covers π i : Y i → X each ramified above a non-empty divisor, the set X(O S ) \ π i (Y i (k)) is Zariski dense. Problem 1.11. If a normal quasi-projective algebraic variety has a Zariski dense set of S-integral points then it has the WIHP.Remark 1.12. Clearly for simply connected varieties the IHP is equivalent to the WIHP.Remark 1.13. It could be possible that we should...