Shaken, Not Stirred: Lack of Magma-chamber Overturn in a Caldera Setting Recorded by Whole-pumice, Mineral, and Melt Inclusion Chemistry. The Tshirege Member of the Bandelier Tuff, New Mexico, USA

Understanding the physical and chemical dynamics of caldera magma chambers prior to VEI 7+ eruptions is important for eruption forecasting and societal preparedness. The 1.26 Ma, 400 km3 Tshirege Member of the Bandelier Tuff is the later of two zoned rhyolitic ignimbrites erupted from the Valles caldera in the Jemez Mountains volcanic field, New Mexico, USA. To avoid effects of crystal-glass sorting, lithic contamination, and post-emplacement thermal alteration of glasses and mineral phases, we sampled whole pumices from non-welded tuff, including zones of otherwise welded tuff quenched against cold paleovalley walls. The tuff is broadly systematically chemically zoned from early-erupted high-silica rhyolite enriched in incompatible trace elements to late-erupted low-silica rhyolite. Whole-pumices indicate the upper portions of the Tshirege magma system were unaffected by convective stirring prior to the eruption, while some mixing and overturn is reflected by pumices from deeper in the system (i.e. stratigraphically higher in tuff sheet), likely closer to recharge sources. Analyses of well-preserved minerals, glasses and melt inclusions, and application of mineral thermometers and barometers, show a vertical stratification of temperature (710-840 °C), pressures of ~0.16 GPa, and variable [H2O] (~2.0-3.5 wt. %). Whole-pumice and melt inclusion chemistry supports crystal accumulation and subsequent melting and remobilization of the cumulate pile as a major contributor to the overall compositional zoning. 2-D thermal models indicate that the creation of the temperature gradient over ~10ka from mush development to mobilization would require volumetrically unreasonable influx of recharge magma (flux rate >30 km3/yr), suggesting that recharge heating is likely only affecting the lowest portions of the chamber. As a result, the thermal gradient likely existed prior to mobilization of the mush. Heating alone cannot affect thermal loosening of a mechanically locked mush to an eruptible crystallinity in reasonable time frames, and so input of volatiles such as H2O or CO2 from a second boiling of recharge magmas is required to melt and remobilize the crystal mush pile. The 1.26 Ma eruption was thus triggered by a final recharge event recorded as dacite pumice clasts distributed throughout the tuff.