of veins/dikes. All melts produced during the smelting stage in each source region have gas contents sufficient to cause them to 0016-7037/ $- see front matter 2007 Elsevier Ltd. All rights reserved. * Corresponding author.

www.elsevier.com/locate/gca Geochimica et Cosmochimica Acta 71 (2007) 2876–2895erupt explosively and be lost. However, since in all but the shallowest source regions part of the melting sequence occurs with-out smelting, fractional melting implies that a significant fraction of UPB melts may have erupted more placidly to form a thin crust (3.3 km thick for a 100 km radius body). Our calculations suggest that melt extraction was so rapid that equilibrium trace element partitioning may not have been attained. We present a model for disequilibrium fractional melting (in which REE partitioning is limited by diffusion) on the UPB, and demonstrate that it produces a good match to the ureilite data. The disequilibrium model may also apply to trace siderophile elements, and might help explain the ‘‘overabundance’ ’ of these elements in ureilites relative to predictions from the smelting model. Our results suggest that melt extraction on the UPB was a rapid, fractional process, which can explain the preservation of a primitive oxygen isotopic signature on the UPB. 2007 Elsevier Ltd. All rights reserved.Fractional melting and smelting on the ureilite parent bod