The Assembly and Emplacement of the Mushy Magma Model: A Historical Perspective

The "mush model" for magmatic systems has emerged as an alternative to the classic notion of a silicate liquid dominated reservoir, the so-called big tank model. The mush model is motivated by a concurrence of geochemical, geophysical and geological observations and new ideas on multiphase fluid dynamics. This presentation will review the historical development and remaining open questions about the mush model as it pertains to silicic systems. The observation that rhyolites have extreme depletions in Sr, Ba and Eu, as well as depletions in Zr and LREE, precluded an origin by direct crustal melting, instead requiring crystallization differentiation. This initially motivated the sidewall crystallization model, where less dense, evolved liquid originated and percolated upwards through a crystal-rich boundary zone, adjacent to a liquid dominated reservoir. The 'defrosting' or remobilization of this sidewall was proposed as a mechanism for producing complex temporal signatures in erupted suites, and this notion later found additional support in the recognition of so-called 'antecrysts.' Lab bench scale tank models of crystallizing salt solutions were offered as analogs for these boundary layer driven magmatic systems. However, it was recognized that features of this boundary layer model that did not agree with seismic, gravity and magnetotelluric, and geological observations. Seismic studies of silicic systems typically indicate P-wave velocity anomalies of 15% or greater for both shallow and deeper systems. But they do not show velocity anomalies that would indicate substantial regions of pure liquid in the core. Rather, the geophysical anomalies, are consistent with a with a spatially extensive crystal mush with an overlying thin melt lens. In addition the observation that erupted crystal poor liquids abruptly transition into crystal rich magmas with interstitial liquid compositions that are nearly identical to the crystal poor ones, provides evidence of a geometrical and source relationship between crystal poor and subjacent crystal rich mush. Lastly, it was appreciated that the boundary layer fluid dynamic models lacked geological verisimilitude, and invoked assumptions on heat transfer rates that were not in accord with geological conditions. Taken together this required a new conceptual model that could honor a broader range of constraints, and led to the 'full chamber' mush model as described by Hildreth (2001, 2004, 2007) and subsequently Bachmann and Bergantz (2004, 2008). However there are many open questions about this model, particularly how they are assembled, the physics of melt movement and mixing, and the way they respond to open system events. For example it is now recognized that crystal mushes can be remobilized rapidly and mineral isotopic and trace element zoning requires that the mush can go through some re-melting, consistent with the unzipping model of Burgisser and Bergantz (2011).