The presence of melts in the first part of upper mantle is image by many geophysical studies. The nature of these melts is referring by different petrological and experimental studies that show a melt rich in volatiles, principally in CO2. To apprehend the deep carbon cycle, the first necessity is to know the mobility of these melts, like the viscosity. As a function of geodynamical context, the temperature and the pressure change, therefore different carbonated melts are produce. In this study, we investigate different SiO2/CO2/H2O content in melt to determine their influences on the viscosity. To investigate the viscosity, the experiment is making in ESRF in Grenoble. We produce in-situ measurements in Paris Edinburg cell with falling sphere viscometry experiments. Experiments are produce between 1 and 3 GPa and liquid temperature of sample. One result show a viscosity of 0.15 Pa.s for 20wt% of CO2. It is one order of magnitude more viscous than dolomite (Kono et al., 2013) and average 5 times less viscous than basalt (Sakamaki et al., 2013). Primarily results obtain in ESRF show a no linear effect of SiO2/CO2 content. Volatiles have an effect on viscosity and this has implication on melt migration in the mantle, so on the deep carbon cycle.