We have implemented full CO 2 ice cloud microphysics into the LMD Mars Global Climate31 Model (MGCM) and we have conducted the first global simulations. The microphysical model32 implementation follows the modal scheme used for water ice cloud microphysics in the MGCM,33 but includes specific aspects that need to be accounted for when dealing with CO2 ice clouds.34 These include nucleation of CO 2 on water ice crystals and CO 2 condensation theory adapted for35 the Martian conditions. The model results are compared to available observations globally, and36 separately for polar regions and equatorial mesosphere. The observed seasonal and latitudinal37 variability of the CO2 ice clouds is in general reproduced. The polar regions are covered by38 CO 2 ice clouds during the winter as observed. Instead of forming only in the lowest 10-15 km39 of the atmosphere, they extend up to several tens of kilometers above the surface in the model,40 dictated by the modeled temperature structure. We have also quantified the contribution of the41 cloud microphysics to the surface CO2 ice deposits. Snowfall from these clouds contributes up42 to 10% of the atmosphere-surface ice flux in the polar regions in our simulations, in the range43 that has been indirectly deduced from observations. In the mesosphere, notable amounts of CO 244 ice clouds form only when water ice crystals are used as condensation nuclei in addition to45 dust particles, and their spatial distribution is in agreement with observations. The mesospheric46 temperature structure, dominated by tides, dictates the longitudinal and seasonal distribution of47 these clouds. The seasonal and local time variations of the clouds are not fully reproduced by the48 model. There is a long pause in CO2 ice cloud formation in the model around the aphelion season,49 but clouds have been observed during this period, although with a lower apparition frequency.50 Modeled mesospheric clouds form mainly during the night and in the morning, whereas during51 the daytime, when most of the cloud observations have been made, the model rarely predicts52 clouds. These discrepancies could be explained by the strong dependence of the cloud formation53 process on mesospheric temperatures that are themselves challenging to reproduce and sensitive54 to the MGCM processes and parameters. The rare possibilities for nighttime observations might55 also bias the observational climatologies towards daytime detections. Future developments of56 the model consist in the inclusion of a possible exogenous condensation nucleus source in the57 mesosphere and the radiative eect of CO 2 ice clouds