The recent development of computational methods using very large plane-wave
basis sets, together with the large increase in computer performance, have
now made it possible to compute the optical properties of real surfaces wi
thin a parameter-free approach. In this work, we describe a scheme for such
a calculation, based on the state-of-the-art techniques for the computatio
n of the ground and excited electronic states. We outline the advantages an
d the drawbacks related to the introduction of various approximation scheme
s, as well as the role of the most important convergence parameters. We pre
sent examples and applications to GaAs(110) and Si(100) surfaces, comparing
the results with those of previous calculations, both semi-empirical and a
b-initio, and with experimental data.