A combination of low-energy electron diffraction, x-ray and ultraviolet photoelectron spectroscopy, and scanning-tunneling microscopy studies, in conjunction with ab initio calculations leads us to suggest a model for the carbon (C)-induced Si(001)c(4x4) atomic structure. This surface superstructure is obtained in a defined range of C2H4 exposures at 600 degreesC. Experimental probes show that the c(4x4) superstructure involves C atoms in both surface and subsurface sites. This is reflected in well-marked features in photoemission valence- and core-level spectra. Surface carbon atoms are stabilized in Si-C heterodimers, with a surface density of about 0.25 monolayer (ML) [i.e., two C atoms per c(4x4) unit cell of eight atoms]. In the subsurface region, carbon atoms substitute for Si atoms in well-defined sites of the third or fourth layers of the Si substrate. The subsurface C density increases with C2H4 exposure time up to a limit value of about 0.5 ML, within the c(4x4) surface structure. Further exposure disrupts the c(4x4) reconstruction and leads to a (2x1) low-energy electron diffraction pattern. Interaction with atomic hydrogen shows that the surface contains a mixture of heterodimers (Si-C) and homodimers (Si-Si), with an 1:1 proportion. These assignments are supported by first-principle calculations, which yield valence band and core level states in fairly good agreement with the experiment. Furthermore, total energy calculations strongly favor C incorporation in surface Si-C dimers and in third and fourth layer sites, and rule out C incorporation in sites of the second Si layer. The most stable c(4x4) surface configuration, suggested by our calculations, consists of alternate Si-C and Si-Si dimer lines. In such a configuration, surface carbon atoms in Si-C dimers induce a surface stress that leads to charge redistribution and atomic relaxation of the adjacent Si-Si dimers, consistent with scanning-tunneling microscopy images. Additional C atoms (in excess of those accommodated in surface sites) are forced in selected compressive (alpha) sites of the third and fourth layers. This model is discussed with respect to the previous models published in the literature
