Some of our earlier work (Allen et al., 1975) on the stability of amphiboles in andesite and basalt at high pressures is subject to criticism because of loss of iron from the starting material to the walls of the capsules (Ag50Pd50) during the runs of fO2 buffered over the range of stability of magnetite. Analyses of fused run products show substantial loss of iron from runs at magnetite-wüstite and nickel-nickel oxide conditions, but none from those at magnetite-hematite conditions.
We have now redone our earlier work on the Mt. Hood andesite and 1921 Kilauea olivine tholeiite under N-NO conditions and with silver capsules. Analyses of fused run products show no iron loss, and the reversed curve representing the maximum stability of the amphiboles in the Mt. Hood andesite shows no change in location, although we now have better control on the high-pressure part of this curve. The revised curve for the appearance of garnet is significantly lower in pressure, passing through 15.5 kbar/940°C and 14.5 kbar/900°C. No orthopyroxene appeared in the run products, in contrast to the results of our earlier work. The high-temperature segment of the amphibole-out curve for the tholeiite is at least as high as 1040°C at 13 kbar and 1050°C at 16 kbar, and the high-pressure part of this curve is at about 27 kbars, about 6 kbar higher than in our earlier work. Amphibole is the sole silicate phase on the vapor-saturated liquidus in the andesite at pressures below the gamet-in curve (<15.5 kbar), and it is accompanied by a minor amount of an Fe-rich oxide. The revised phase relationships are consistent with the derivation of andesite by amphibole-liquid equilibria in basaltic magma, and they reveal that the amphiboles are stable to depths of nearly 100 km. The great depths to which amphiboles are stable make it all the more probable that they are a source of H2O for partial melting in deeply subducted oceanic crust.