Due to its electrical properties graphene (G) has been successfully used as a sensing element for the detection of different gases reaching very high sensitivities (ppm or better), which are ascribed to the doping induced by adsorption. The sensitivity depends indeed critically on the chemical nature of the gas and is lower for CO than for other poisoning species. The nature of the active sites is, however, still unclear. If it were due to physisorption, the values of the adsorption energy cannot explain the need for high temperature re-generation of the sensing element. Chemisorption must thus be involved, either at defects or by doping, determining the magnitude of the heat of adsorption and consequently the sensitivity and the range of temperatures at which the sensor can operate. In order to clarify these issues we investigated experimentally adsorption of CO on G supported on polycrystalline Cu and Ni(111) by HREELS and XPS.
No signature of adsorbed CO was found after exposure at RT while just a few L at 100 K evidenced chemisorbed CO for G supported on Ni(111). On the other hand G on Cu is inert also at low temperature. This result indicates that the nature of the substrate plays an essential role in the adsorption process [1,2]. The heat of adsorption q is estimated to be about 0.58 eV/molecule at low coverage, so that an equilibrium coverage of 0.1 ML is expected at RT under a CO partial pressure of only 10 mbar. We identify top-bridge graphene as the most reactive configuration .
Doping G/Ni(111) by N2+ ion bombardment allows for the formation of a second, more strongly bound moiety, characterized by a CO stretch frequency of 236 meV and by an initial heat of adsorption (0.85 eV/molecule). The presence of N (in pyridinic or substitutional sites) enhances therefore significantly the chemical reactivity of G/Ni(111) towards CO.
Finally we investigated the role of isolated defects, which we created by low energy Ne+ ions bombardment on single layer graphene supported on different substrates (polycrystalline Cu and Ni(111)) . No CO adsorption occurs for defected G/Cu, while HREELS peaks form promptly for G/Ni(111). Two moieties, desorbing just above 350 K, are present under vacuum conditions after exposure at RT. The CO stretch frequencies and the ratio of their intensities indicate that they are due to chemisorbed CO at the G/Ni(111) interface close to the vacancies rather than at the defected G layer. The red-shift of the C1s binding energy indicates that in such regions detachment of the G layer from the substrate occurs.
Unfortunately defects of G/Ni(111) are not a good candidate as active sensing site since amending of vacancies occurs, as demonstrated by the reduction of the adsorbed coverage in subsequent CO doses followed by annealing at 380 K.
 M. Smerieri et al, ChemCatChem 7, 2328 (2015)
 E. Celasco et al, PCCP 18, 18692 (2016)
 E.Celasco et al, J. Chem. Phys. 146 104704 (2017)