Melt generation beneath ocean islands: A U-Th-Ra isotope study from Lanzarote in the Canary Islands

New major and trace element data, and Sr, Nd, Ph, U, Th and Ra isotopes are presented for prehistoric and historic lavas from Lanzarote in the Canary Islands. These rocks are amongst the most primitive found on intraplate oce an islands ranging in composition from basanite to alkali basalt, with MgO contents >9.3% and Mg numbers >67. The youngest are from two of the three 1 824 vents, the largest group of samples is from the best known eruption epi sode, the 1730-36 Timanfaya eruptions, a smaller group of samples are from the northeast Corona region (similar to 50 ky) and the oldest samples are f rom the Famara complex and basement massif. The rocks have some of the char acteristics of HIMU GIB, including high Ce/Pb, Nb/Ce and low Nb/U and restr icted Sr-87/Sr-86 (0.70209-0.70332). There is significant Th-230/U-238 dise quilibrium (Th-230 excesses range from 6-76%) with some of the intermediate silica composition Corona samples showing the greatest disequilibrium. The historic samples exhibit Ra-226 excess. The major and trace element data h ave undergone fractionation corrections to Mg numbers of 70, requiring <5% olivine fractionation, and these inferred primary compositions ale used to evaluate a number of melt generation and mixing models. The fractionation-corrected compositions for the 1824 and the 1730-36 have been modelled as 1-4% melts of a source similar to primitive mantle. Howeve r, Yb is incompatible, and so the amounts of residual garnet were < similar to 5%, suggesting that there was no significant contribution from garnet p yroxenite source rocks. Rather the REE and the FeO contents are both consis tent with melting in the garnet-spinel transition, at depths of 60-70 km. ( Th-230/U-238) increases slightly with increasing La/Yb, in:the younger rock s, and they require Some form of dynamic melting model. In the preferred mo del the upwelling rate is kept constant, and the differences in the degrees of melting are attributed to different lengths of the melt column, with th e smaller degree melts being extracted from greater depths. Strikingly, (Ra -226/Th-230) increases with increasing degrees of melting, and so it reflec ts the time since extraction from the-melt column rather than variations in the melting processes. Intra-suite minor and trace element variations are due to magma mixing, and not to progressive changes in the degrees of parti al melting, and it is envisaged that such magma mixing occurred during the dynamic melting processes. Dynamic melting at depths of 60-70 km suggests t hat the regional uplift around the Canary Islands is at least in part due t o thermal erosion of the underlying lithosphere. Variations of average Ce/Yb, Tb/Yb, (Th-230/U-238), SiO2 and lithospheric a ge for different OIB highlight how the smaller degree melts tend to be gene rated at greater depths, and the mean pressure of melting increases with th e thickness of the lithospheric lid. However, there is no general link betw een (230Th/238U) and the:lithospheric age or thickness, and hence the integ rated degrees of melting. High buoyancy fluxes result in higher degrees of melting and low (Th-230/U-238) (Chabaux and Allegre, 1994), but for OIB gen erated within low buoyancy plumes, (230Th/238U) and the degrees of melting primarily depend on the depths of melt extraction. Differences in the avera ge composition of low buoyancy OIB depend on the thickness of the overlying lithosphere (Ellam, 1992; Haase, 1996), and the differences within an OIB suite, such as between the 1730-36 and 1824 lavas on Lanzarote, depend on t he depth of extraction from the melt column. Copyright (C) 1999 Elsevier Sc ience Ltd.