The activation of CO2 by interaction with Na atoms on tungsten was studied in a joint experimental/theoretical effort combining MIES, UPS (HeII) and first principles calculations. Experimentally, both the adsorption of Na on tungsten, followed by CO2 exposure to the Na-modified surface at 80 K, and the adsorption of CO2 on tungsten, followed by Na exposure to the CO2 covered substrate, were studied. Below about 120 K CO2 physisorbs on pure W(011), and the distance between the three main spectral features is as for gas phase CO2 (E(B) = 8.4, 12.1, 14.1 eV). When offered to a Na monolayer (ML) deposited onto W, CO2 is converted into a chemisorbed species. The spectral pattern is different from physisorbed CO2, and the three spectral features are shifted towards lower binding energies (E(B) = 6.3, 10.7, 13.9 eV). The chemisorption continues until all available Na species are converted into Na+ species. Additional CO2 offered to the system becomes physisorbed on top of the chemisorbed species. When a CO2 monolayer, physisorbed on tungsten at 80 K, is exposed to Na, the interaction leads initially to a decrease of the surface work function and to a rigid, global shift of all CO2 induced features towards larger binding energies by about 2 eV. Only beyond a minimum Na coverage of about 0.5 ML, chemisorbed species can be detected. We conclude that, initially, transfer of the Na(3s) electron to the tungsten substrate takes place. Above 0.5 ML Na coverage, back donation of charge to CO2 takes place whereby the physisorbed carbon dioxide species become converted into chemisorbed ones. The experimental results are interpreted with the help of first principle calculations carried out on suitable slab models. The structures and surface binding mode of the chemisorbed CO2 species are described. The calculated density of states for the most stable situations is in qualitative agreement with experimental data.