The origin of island arc high-alumina basalts

Abstract

A detailed examination of the hypothesis that high-alumina basalts (HAB) in island arcs are primary magmas derived by 50–60% partial melting of subducted ocean crust eclogite shows that this model is unlikely to be viable. Evidence suggests that the overwhelming majority of arc HAB are porphyritic lavas, enriched in Al2O3 either by protracted prior crystallization of olivine and clinopyroxene, or by plagioclase phenocryst accumulation in magmas of basaltic andesite to dacite composition. Experimentally-determined phase relationships of such plagioclase-enriched (non-liquid) compositions have little bearing on the petrogenesis of arc magmas, and do not rule out the possibility that arc HAB can be derived by fractionation of more primitive arc lavas. Although models invoking eclogite-melting can match typical arc HAB REE patterns, calculations indicate that the Ni and Cr contents of proposed Aleutian primary HAB are many times lower than such models predict. In contrast, Ni vs Sc and Cr vs Sc trends for arc HAB are readily explained by olivine (+Cr-sp) and clinopyroxene-dominated fractionation from more primitive arc magmas. GENMIX major element modelling of several HAB compositions as partial melts of MORB eclogite, using appropriate experimentally (26–34 kb)-determined garnet and omphacite compositions yields exceptionally poor matches, especially for CaO, Na2O, MgO and Al2O3. These mismatches are easily explained if the HAB are plagioclase-accumulative. Groundmasses of arc HAB are shown to vary from basaltic andesite to dacite in composition. Crystal fractionation driving liquid compositions toward dacite involves important plagioclase separation, resulting in development of significant negative Eu anomalies in more evolved lavas. Plagioclase accumulation in such evolved liquids tends to diminish or eliminate negative Eu anomalies. Therefore, the absence of positive Eu anomaly in a plagioclase-phyric HAB does not indicate that plagioclase has not accumulated in that lava. In addition, we show that plagioclase phenocrysts in arc HAB are not in equilibrium with the liquids in which they were carried (groundmass). Exceptional volumes of picrite and olivine basalt occur in the Solomons and Vanuatu arcs; the presence in lavas from these and other arcs (Aleutian, Tonga) of olivine phenocrysts to Fo94, finds no ready explanation in the primary eclogite-derived HAB model. We suggest that most lavas in intra-oceanic arcs are derived from parental magmas with >10% MgO; fractionation of olivine (+Cr-sp) and clinopyroxene drives liquids to basalt compositions with <7% MgO, but plagioclase nucleation is delayed by their low but significant (<1%?) H2O contents. Thus evolved liquid compositions in the basaltic andesite—andesite range may achieve relatively high Al2O3 contents (<17.5%). The majority of arc basalts, however, have Al2O3 contents in excess of 18%, reflecting plagioclase accumulation. We give new experimental data to show that HAB liquids may be generated by anhydrous, low-degree (<10%) partial melting of peridotite at P<18 kb. Relative to arc HAB, these experimental melts have notably higher Mg#(69–72) and are in equilibrium with olivine Fo87–89. Only further detailed trace element modelling will show if they might be parental magmas for some arc HAB.