Reconstructing CO2 concentrations in basaltic melt inclusions using Raman analysis of vapor bubbles

• A comparison between Raman-restored CO2 and model-restored CO2 is given.
• A melt inclusion H2O and CO2 diffusion model is proposed.
• Pressures increase by 600–4000 bars for melt inclusions corrected for CO2 loss.
• If possible, both CO2 restoration models should be used.

Melt inclusions record valuable information about pre-eruptive volatile concentrations of melts. However, a vapor bubble commonly forms in inclusions after trapping, and this decreases the dissolved CO2 concentration in the melt (glass) phase in the inclusion. To quantify CO2 loss to vapor bubbles, Raman spectroscopic analysis was used to determine the density of CO2 in bubbles in melt inclusions from two Cascade cinder cones near Mt. Lassen and two Mexican cinder cones (Jorullo, Parícutin). Using analyses of dissolved CO2 and H2O in the glass in the inclusions, the measured CO2 vapor densities were used to reconstruct the original dissolved CO2 contents of the melt inclusions at the time of trapping. Our results show that 30–90% of the CO2 in a melt inclusion is contained in the vapor bubble, values similar to those found in other recent studies. We developed a model for vapor bubble growth to show how post-entrapment bubbles form in melt inclusions as a result of cooling, crystallization, and eruptive quenching. The model allows us to predict the bubble volume fraction as a function of ΔT (the difference between the trapping temperature and eruptive temperature) and the amount of CO2 lost to a bubble. Comparison of the Raman and modeling methods shows highly variable agreement. For 10 of 17 inclusions, the two methods are within ± 550 ppm CO2 (avg. difference 290 ppm), equivalent to ±~300 bars uncertainty in estimated trapping pressure for restored inclusions. Discrepancies between the two methods occur for inclusions that have been strongly affected by post-entrapment diffusive H+ loss, because this process enhances bubble formation. For our dataset, restoring the CO2 lost to vapor bubbles increases inferred trapping pressures of the inclusions by 600 to as much as 4000 bars, highlighting the importance of accounting for vapor bubble formation in melt inclusion studies.