Soft cohesive forces

We discuss dispersion forces, beginning with toy models that illustrate the limitations of various standard approaches. For metallic cohesion of very thin layers, we show that because the local density approximation (LDA) misses distant dispersion interactions, it also makes significant errors in the maximum cohesive force, a short-ranged property. Furthermore, perturbative methods fail for such large planar systems, and CI methods are impractical. For large planar and linear systems in the well-separated limit we show that insulating and metallic systems can exhibit very different dispersion forces, pairwise summation of atomic R-6 terms failing for the metallic cases. This could have implications for the interaction between nanotubes and between graphene planes: these planes are zero-gap insulators at large separation and weak metals at graphitic equilibrium. Graphitic cohesion and intercalation are fundamental to a hydrogen economy and to various nanotechnologies, yet our arguments strongly suggest that all standard methods are inadequate for these phenomena. We argue that nonlocal RPA-like correlation energy formulae contain all the required "seamless" physics of long- and short-ranged interaction, as needed for graphitic and other soft-matter systems. Indeed full calculations of this type are currently being attempted for graphite, and appear to be very delicate. We discuss recent efforts to approximate these calculations, and propose a new scheme. (C) 2004 Wiley Periodicals, Inc.