Let V be a nonsingular projective surface defined over ޑ and having at least two elliptic fibrations defined over ;ޑ the most interesting case, though not the only one, is when V is a K3 surface with these properties. We also assume that V )ޑ( is not empty. The object of this paper is to prove, under a weak hypothesis, the Zariski density of V )ޑ( and to study the closure of V )ޑ( under the real and the p-adic topologies. The first object is achieved by the following theorem:Let V be a nonsingular surface defined over ޑ and having at least two distinct elliptic fibrations. There is an explicitly computable Zariski closed proper subset X of V defined over ޑ such that if there is a pointThe methods employed to study the closure of V )ޑ( in the real or p-adic topology demand an almost complete knowledge of V ; a typical example of what they can achieve is as follows. Let V c be 1. Introduction. Let V be a nonsingular projective surface defined over ޑ and having at least two elliptic fibrations defined over ;ޑ the most interesting case, though not the only one, is when V is a K3 surface with these properties. (By an elliptic fibration we mean a fibration by curves of genus 1; we do not assume that these curves have distinguished points, so they need not be elliptic curves in the number-theoretic sense.) We also assume that V )ޑ( is not empty. The object of this paper is to prove, under a weak hypothesis, the Zariski density of V )ޑ( and to study the closure of V )ޑ( under the real and the p-adic topologies. These results can be found in Section 2; Sections 3 and 4 contain applications and examples. Note that the geometric terminology in this paper is that of Weil.For the real and Zariski topologies, such results have been proved for a particular family of such surfaces in [Logan et al. 2010], and for one such surface already in [Swinnerton-Dyer 1968]; see Section 3. I am indebted to the referee for drawing