Possible role of amphibole in the origin of andesite: some experimental and natural evidence

Experiments in the system high-A1 basalt (HAB)-water have been conducted in the melting range at pressures between 1 atm. and 10 kbar, defining the amphibole stability field and the composition of liquids which coexist with this amphibole. Plagioclase is the anhydrous liquidus phase between 1 atm. and 10 kbar but in the hydrous runs this role is taken by olivine at <7 kbar and then by clinopyroxene at higher pressures. Because amphibole is never on the high-A1 basalt liquidus it is not likely that andesite is derived from primary basalt by pure fractional crystallisation, although as we discuss, other mechanisms including equilibrium crystallisation might implicate amphibole. If primary basaltic magma undergoes closed-system equilibrium crystallisation, then the amphibole field will be intersected at between 50 and 100°C below the liquidus. The compositions of melts coexisting with amphibole alone do not match those of any of the natural andesite or dacitic lavas associated with the particular high-A1 basalt investigated. Like natural andesites, they become rapidly silica enriched, but they also become far more depleted in TiO₂ and MgO. However, the compositions of liquids lying directly on the divariant amphibole-out reaction zone, where amphibole +liquid coexist with clinopyroxene or olivine (±plagioclase), do resemble those of naturally occurring low-silica andesites. With increasing temperature pargasitic amphibole breaks down via incongruent melting reactions over a narrow temperature range to form a large volume of relatively low-silica basaltic andesite liquid and a crystalline assemblage dominated by either clinopyroxene or olivine. Our important conclusion is that basaltic andesite liquid will be the product of reaction between cooling, hydrous mafic liquid and anhydrous ferromagnesian phases. The solid reactants could represent earlier cumulates from the same or different magma batches, or they could be peridotite wall-rock material. Because the amphibole-out boundary coexisting with liquid is one of reaction, it will not be traversed so long as the phases on the high temperature side remain. Thus, the assemblage amphibole+clinopyroxene±olivine±plagioclase+liquid is one in which the liquid is buffered (within limits), and results reported here indicate that this buffering generates melts of low-silica andesite composition. When tapped to lower pressures these liquids will rise, eventually to fractionate plagioclase-rich assemblages yielding silicarich andesite and dacite melts. Conversely, the partial melting of hornblende pyroxenite, hornblende peridotite or hornblende gabbro can also yield basaltic andesite liquids. The phase relationships suggested by these experiments are discussed in the light of naturally occurring phenocryst and xenolith assemblages from the east Sunda Arc. Primary magmatic additions to the lithosphere of volcanic arcs are basaltic and voluminous upper crustal andesite in these terranes, complemented by mafic and ultramafic crystalline deposits emplaced in the lower crust or close to the Moho. Together these components constitute total arc growth with a basaltic composition and represent the net accreted contribution to continental growth.