S-Pconversion from the transition zone beneath Tonga and the nature of the 670 km discontinuity

Using more than 300 recordings of deep Tonga earthquakes from the Warramunga Seismic Array (central Australia), we have searched for short-period S-P conversions from the 670 km discontinuity arriving in the P-wave codas. The array seismograms were stacked and plotted in sections to optimize observation of S670P yielding the following results: For the northern part of Tonga (17.3°–18.4°S), S670P-phases are observed with apparent conversion depths between about 660 and 680 km. For the central part (20.2°–22.3°S), S670P is not clearly observed. For the southern part (23.1°–25.7°S), S670P is observed for conversion depths between about 660 and 700 km. Intermittent occurrence of S-P conversion may result from different characteristic source radiation patterns among the three sections. Overall uncertainties in conversion depths (resulting from unmodelled lateral velocity variations and hypocentral location errors) are perhaps a few tens of kilometres. The S-P rays must travel either within or very close to the subducted slab because of the source/receiver/slab-dip geometry, so we conclude that large (>50 km) deformations of the 670 km discontinuity do not occur beneath Tonga. Our observations appear to be inconsistent with a chemical discontinuity at 670 km, but are easily explained if the 670 km discontinuity is a strongly pressure-dependent and, perhaps, moderately endothermic phase change in both the slab and surrounding mantle.