Upper mantle structure beneath the Fiji Plateau: seismic observations of second P arrivals from the olivine--spinel phase transition zone

Large and impulsive second P arrivals (4–12 s after the initial P arrivals) from intermediate depth earthquakes located in the New Hebrides island arc are observed at distances of 7–14° on the records produced by short-period seismographs in Fiji. These second P arrivals are caused by refraction from a zone of rapidly increasing P velocity at about 370 km depth beneath the Fiji plateau, presumably caused by the olivine-spinel phase transition in the upper mantle. This conclusion is consistent with: (1) the observed onset distance of the travel-time cusp; (2) the change with distance of Δt (the time difference between first and second P arrivals); (3) the change of Δt with the depth of the events; (4) the slowness of the first and second P arrivals; and finally, (5) comparisons with the predicted arrival times frop other published velocity models which contain a rapid P velocity increase in the upper mantle. The range of velocity models which fit our data is shown by giving examples of models with the maximum and minimum values for the depth, thickness and amount of velocity increase in the transition zone. These include: (1) a model with a discontinuous increase in velocity of 0.81 km/s occurring at 340 km depth (shallowest zone, thinnest zone, and smallest velocity increase); (2) a model with an increase of 1.55 km/s occurring over a thickness of 47 km and centred at 346 km depth (thickest zone and largest velocity increase); and (3) a model with an increase of 1.13 km/s occurring over a thickness of 21 km and centred at 393 km depth (deepest zone). However, a consideration of the effects of mantle self-compression, temperature, and Birch's law restricts further the family of acceptable velocity models that fit our data. Models that meet these restrictions have an increase in velocity greater than 1.0 km/s occurring over a thickness less than 30 km and centred between 360 km and 390 km depth. For sources in the New Hebrides, where the inclined seismic zone has a dip of 70–80°, the descending lithospheric slab does not seriously affect the pattern of second P arrivals. Applying Snell's law to a simple model of the slab shows that the slab decreases Δt by less than one second, and decreases the onset distance of the travel-time cusp by less than 0.5 degrees. However, no second p arrivals at comparable distances are observed at Fiji from intermediate depth earthquakes occurring in the Tonga-Kermadec island arc. The simple model of the slab shows that the dip of the Tonga-Kermadec slab is too shallow to allow rays to leave the slab at the proper angle to become second P arrivals. The observed predominant frequency of the second P arrivals is only about 25 per cent higher than that of the first arrivals. This suggests that the average Qp for the first P arrival rays, which traverse only the upper 200 km of the mantle beneath the Fiji plateau, cannot differ greatly from the average Qp for the second P arrival rays, which penetrate to the phase transition zone. A survey of other studies of upper mantle velocity structures confirms that differences in the thickness and in the depth of the phase transition zone do exist throughout the world, although the comparison is complicated by differences in the type and uniformity of the data sets and in methods of analysis. Our results from the Fiji plateau (a recently generated oceanic lithosphere) differ significantly from the results obtained in northeastern Australia, but are remarkably similar to the results obtained in the western United States.