WAITER, THERE'S AN ECLOGITE IN MY JELLY SANDWICH

The Himalayan orogen is a popular arena for debates on the distribution of strength in the continental lithosphere. Competing models are the 'jelly sandwich', in which a weak lower crust separates a strong upper crust and upper mantle, and 'creme brulee', with a strong crust and weak upper mantle. For seismically active areas, the latter model predicts a continuous depth distribution of seismicity, while in the jelly sandwich case a bimodal distribution of seismicity with depth may be observed. In practice, much of the debate has centered on the depth of the Moho relative to the depth of the seismicity. Determining the depth of earthquakes requires the use of a velocity model. The Moho is typically detected via P to S conversions in teleseismic arrivals, and mapping this delay time to depth also requires a velocity model. We circumvent biases introduced by the choice of velocity models by directly comparing S-P delay times of receiver function conversions with delay times between S and P picks for local earthquakes. The entire crust is seismogenic under the High Himalaya and south. Further north under the southern Tibetan plateau, earthquakes separate cleanly into a group below the Moho and a group above a lower crustal interface. Our data thus suggest a weak (non-seismogenic) lower crustal layer with high seismic velocity under southern Tibet. A layer with partial eclogitization of crustal material has the necessary characteristics. A 'double Moho' in receiver functions indicating a fast lower crustal layer has been found in several experiments along most of the Himalayan arc, all limited to a band within similar across-arc distance, suggesting a possible reversal of the process to the north. The eclogite transition is influenced by temperature and water availability, and the high density of eclogite affects surface elevations and crustal volume balance. We explore the dynamic implications of eclogitization within the time scales dictated by Himalayan convergence.