Abstract. We argue that many phenomena associated with metal/non-metal interfaces and similar situations with a large dielectric constant mismatch can be understood in terms of the image interactions due to charges in the nonmetal. The effects are additional to the traditional interactions, and are especially significant when no reactions between the phases occur. The image-charge concept allows us to rationalise much apparently unrelated information concerning: (a) the systematics of wetting and non-wetting of oxides by liquid metals; ( b ) the systematics of strong metal-support interaction in catalysis; (c) the spatial variation of stoichiometry in oxides grown on metals; (d) the dependence on thickness of the observed changes in the wetting by water of oxide grown on silicon; (e) some features of radiationenhanced adhesion; and U, a number of correlations of behaviour with non-metal properties in which the precise choice of metal is not critical.The idea of an image charge is long established. If there is a planar boundary dividing space into two regions of different dielectric constant E ' , E" (e.g. metal/non-metal, vacuum/solid, liquid/solid) then the polarisation energy of any charge will be affected by the existence of the boundary. In a simple case, the effect on a charge in region I can be represented by imagining region I extended to fill all space, but including interaction with an image charge whose magnitude depends on both the actual magnitudes of the dielectric constants ( E ' in I, E" in 11) and whose position is the mirror image of the charge, regarding the boundary as a reflecting plane (Smythe 1939, Landau andLifshitz 1960). Thus, at distance z from the boundary, a charge Q causes an image term
U ( Z ) = ( Q * /~z E ' ) ( E ' -&")/(E' + E")thus lowering the energy near a metal (Ell infinite (Landau and Lifshitz 1960, p 40)) and increasing the energy near a vacuum (E" unity). This simple expression does not cover all cases we shall need, though the other cases can be obtained by standard means (see Smythe 1939). Nevertheless, the equation and its generalisations have been applied widely in situations as varied as electron motion near interfaces, electron-phonon coupling in inelastic tunnelling, and many instances of classical electrostatics. Our application is different, and appears to be far reaching in a wide range of interface behaviour. We note that the change in polarisation energy due to the boundary (which we may call loosely the image energy) affects the interfacial energy too, i.e. the energy needed to cleave the two dielectrics at the boundary depends on charges present near the interface. This simple feature allows us to rationalise a large number of observations. A full comparison will be given in later papers; for the moment we give an overview.