In the last decade two principal factors have stimulated the progressive regain of interest for non-reactive metal/oxide interfaces. On one hand, the efforts invested by the community of model catalysis in analyzing the reactivity properties of supported metal nano-clusters have resulted in an abundance of high quality experimental data. They have also risen several precise questions on the direct and indirect role played by the substrate in the determination of catalytic properties of deposited metal particles, and have thus reiterated the interrogations concerning the nature of interactions at the metal/oxide interfaces. On the other hand, conceptual improvements of first-principles calculations, such as implementations of various GGA functionals, have added enormously to the reliability of these methods, and have enlarged considerably their field of application. However, most ab initio simulations are sooner or later confronted to the constrains on the computational cost, inherent of this kind of approaches. It concerns principally the limited size of systems which can be treated in practice, a factor which turns out to be particularly limiting in realistic studies of interfaces, where the mismatch of lattice parameters is at the origin of incommensurate interface structures, long-range reconstructions, or/and complex structural deformations and dislocations. In this context, the goal of this paper is two-fold. First, we give an overview of essential metal/oxide interface characteristics as obtained from ab initio calculations on model systems, with a special focus on late transition metals, such as Pd, and on a highly ionic oxide substrate, such as MgO. It includes results on the relation between the strength of interfacial interaction and the type of deposited metal (transition, noble), the nature of metal deposit (isolated atoms, constituted interface), and the character of the substrate (non-polar, polar). We also comment on effects due to defects at the oxide surface. Secondly, we describe an effective approach to simulate non-reactive deposition of nano-scale metal objects on a surface of highly ionic oxide. The core of this approach is a many-body potential energy surface (PES) constructed on the basis of results of ab initio calculations for model metal/oxide interface structures. We present its application to studies on the deposition of late transition metals (Pd, Ag) on the MgO(100) surface, including the determination of substrate-induced change of equilibrium atomic structure and morphology of metal nano-clusters, the analysis of stress release at the interface, and the investigation of the role of the oxide substrate on the melting properties of supported clusters. (c) 2006 WILEYNCH Verlag GmbH & Co. KGaA, Weinheim.