Fluxing with CO₂ of rhyolitic magma at the beginning of an eruption—scholastic concept or reality?

The volumes of CO₂ released at some basaltic volcanoes (e.g., Etna, Stromboli) significantly exceed the limits of solubility in silicate melt, indicating flushing by deep, CO₂-rich fluids from crustal and mantle sources. A similar flushing mechanism in silicic magmatic systems has been less evident. Here, we propose that H₂O–CO₂ contents in quartz-hosted melt inclusions (MIs) can serve as an indicator of interaction between hydrous rhyolitic magma and carbonic fluid. Numerical modeling of this interaction process at the single-bubble scale, accounting for volatile diffusivity dependencies on water content, demonstrates that a uniformly dehydrated melt with highly variable CO₂ concentrations can form on relatively short timescales. In natural magma bodies where bubble coalescence and escape occur, this process generates melt compositions that form subvertical arrays on H₂O–CO₂ diagrams. Such compositions, preserved in melt inclusions, are commonly found in pyroclastic deposits from catastrophic intraplate rhyolitic eruptions, such as those of the Yellowstone caldera. Reinterpretation of extensive published MI data from the pre-Huckleberry Ridge Tuff-A ashfall deposits at Yellowstone reveals episodes of carbonic fluid-magma interaction that likely initiated explosive eruptions. The first eruptive cycle is associated with flushing at the basal boundary layer of the magma chamber by a fluid enriched in CO₂ and Li. The resulting bubbly layer would ascend rapidly, being captured in the earliest erupted magma that forms the base of the ash sequence. MIs from this level show a strong positive correlation between CO₂ and Li concentrations. In contrast, a significant negative correlation between CO₂ and Li is observed in MIs from near the top of the 2-meter-thick ash sequence. This pattern is attributed to a second, distinct flushing episode by a Li-poor fluid, which culminated in the massive early Huckleberry Ridge Tuff (HRT-A) ignimbrite eruption that overlies the ashfall deposits. Other samples exhibit a strong positive correlation between Li and H₂O, explainable by syneruptive diffusive loss of both components. Diffusive water loss from reentrants and MIs consistently indicates a syneruptive magma decompression rate of ~0.02–0.06 MPa/s. Finally, the unusual population of reentrants and MIs with uniformly low H₂O (1–1.5 wt.%) but variable-to-high CO₂ contents at the top of the ash section can be explained by dehydration during interaction with (or formation from) a water-poor melt. This melt was likely generated by remelting of largely solidified rhyolite, underplated by basalt, as an alternative or complementary process to CO₂ flushing.