Optical properties of real surfaces from microscopic calculations of the dielectric function of finite atomic slabs

We present a critical analysis of the procedures and approximations which are generally used to obtain the optical properties of a real solid surface starting from microscopic calculations of the dielectric function of a finite atomic slab. We show that relatively thin slabs-often the only ones accessible to modem ab initio methods-are indeed adequate provided that the proper procedure is followed for the extraction of the actual surface contributions, and we point out some possible sources of artifacts that should be avoided.