An extensive computational study is presented with the quest to investigate the nature of the ground-state geometry of the Si(001) surface, a subject of recent experimental controversy. We analyze for the first time in detail the possible sources of errors which would arise in any correlated calculation for a system size of interest here. For this purpose, we present a detailed analysis of the cluster model of the surface at the DFT and MCSCF level of theory. Estimates of errors arising from the use of pseudopotential, finite cluster size, and biased (method dependent) choice of ground-state geometry are given. The resulting error is estimated to be comparable to the energy scale of interest. On the other hand, the energy variation due to negative thermal expansion at low temperature is found to be qualitatively consistent with dimer symmetrization. (C) 2002 Elsevier Science B.V. All rights reserved.