Basaltic volcanism : deep mantle recycling, Plinian eruptions, and cooling-induced crystallization

Mafic magma is the most common magma erupted at the surface of the earth. It is generated from partial melting of the mantle, which has been subdivided into end-members based on unique geochemical signatures. One reason these end members, or heterogeneities, exist is subduction of lithospheric plates back into the mantle. The amount of elements, such as Cl and K, removed during subduction and recycled into the deep mantle, is poorly constrained. Additionally, the amount of volatiles, such as Cl, that are recycled into the deep mantle will strongly affect the behavior of the system. I have looked at Cl and K in HIMU source melts to see how it varies. Cl/Nb and K/Nb suggest that elevated Cl/K ratios are the result of depletion of K rather than increased Cl recycled into the deep mantle.
After the mantle has partially melted and mafic melt has migrated to the surface, it usually erupts effusively or with low explosivity because of its low viscosity, but it is possible for larger eruptions to occur. These larger, Plinian eruptions, are not well understood in mafic systems. It is generally thought that basalt has a viscosity that is too low to allow for such an eruption to occur. Plinian eruptions require fragmentation to occur, which means the melt must undergo brittle failure. This may occur if the melt ascends rapidly enough to allow pressure to build in bubbles without the bubbles expanding. To test this, I have done decompression experiments to try to bracket the ascent rate for two Plinian eruptions. One eruption has a fast ascent, faster than those seen in more silicic melts, whereas the other eruption is unable to be reproduced in the lab, however it began with a increased viscosity in the partly crystallized magma. After fragmentation and eruption, it is generally thought that tephra do not continue to crystallize. We have found that crystallinity increases from rim to core in two basaltic pumice. Textural data along with a cooling model has allowed us to estimate growth rates in a natural system, which are similar to experimental data.