Atomic carbon on Co(0001), deposited by ethylene decomposition, forms islands with a (√3 × √3)R30°structure at low C coverage (∼0.2 ML), whereas a high C coverage (0.5 ML, saturation) induces a reconstruction of the cobalt surface. Atomic carbon weakens the adsorption of CO and H 2 , but even a saturated atomic carbon layer does not block the surface for adsorption. Carbon−carbon coupling, i.e., polymeric carbon formation, was not observed for temperatures ≤630 K on the close-packed cobalt surface. Polymeric carbon, in the form of small graphene islands, forms on the close-packed terraces after heating of an acetylene-saturated surface. Graphene also forms upon heating of an atomic carbon covered surface on which ethylene was dosed at low temperature. In this case, step edges act as a nucleation point of the graphene islands, while their growth proceeds via the addition of C 2 H x species. In both cases, hydrogenated forms of carbon rather than atomic carbon are key precursors for graphene growth. Graphene covers the cobalt surface, thereby inhibiting adsorption of CO and hydrogen completely. The described graphene formation mechanism is seen as a relevant, low temperature route to detrimental carbon that would deactivate a cobalt FT catalyst. Atomic carbon is more reactive than graphene, as it is oxidized at lower temperatures than graphene. The graphene islands formed at relatively low temperatures are of poor structural quality and contain (islands of) encapsulated cobalt atoms.

We have developed an efficient and accurate formalism which allows the simulation at the ab initio level of inelastic electron tunneling spectroscopy data under a scanning tunneling microscope setup. It exploits fully the tunneling regime by carrying out the structural optimization and vibrational mode calculations for surface and tip independently. The most relevant interactions in the inelastic current are identified as the inelastic tunneling terms, which are taken into account up to lowest order, while all other inelastic contributions are neglected. As long as the system is under tunneling regime conditions and there is no physisorbed molecule on the surface or tip apex, this lowest order in inelastic tunneling (LOIT) approach reduces drastically the computational cost compared to related approaches while maintaining a good accuracy. Adopting the wide-band limit for both tip and surface further reduces calculation times significantly, and is shown to give similar results to when the full energy dependence of the Green's functions is taken into account. The LOIT is applied to the Cu(111) + CO system probed by a clean and a CO contaminated tip to find good agreement with previous works. Different parameters involved in the calculations such as basis sets, k sampling, tip-sample distance, or temperature, among others, are discussed in detail.

No abstract