The Bowen-Fenner Debate Revisited: A Review of Basalt Crystal Fractionation and the Generation of Andesite

The Bowen-Fenner debate over the nature of basalt crystal fractionation is still unresolved. Bowen's classic 1928 book detailed his basalt-andesite-rhyolite fractionation scheme, whereas Fenner emphasized the evidence for Fe-enrichment during most basalt crystallization, thus anticipating discovery of the Skaergaard. In the 1940's Bowen suggested only wet basalt generates common andesite. To address this debate MELTS simulations were performed with near-constant parents but varied water contents from 0-4 wt per cent. For comparison, data were compiled from mid-ocean ridge (MORB), back-arc basin (BABB), and arc basalts. There is a strong correlation between empirical fractionation patterns and the simulations. Dry basaltic magma exemplified by MORB shows fractionation patterns controlled by early plagioclase crystallization, whereas nearly all wet arc basaltic magmas have patterns controlled by crystallization of mafic minerals, with the later onset of highly calcic plagioclase. BABB have fractionation patterns between those two extremes. The fundamental antithetic behavior of Al and Fe during basalt crystal fractionation due to the sensitive control over plagioclase crystallization by water is completely obscured by the use of Harker and AFM diagrams. Mafic arc basalts which show evidence for significant crystallization of magnetite are largely lacking - most arc basalt fractionation patterns are consistent with 2-3 per cent H2O in the primitive melt and oxygen fugacity near the NNO buffer. A few mafic arc volcanoes have steep increases in Al content or slightly declining Fe content with only slight increases in FeO*/MgO suggestive of significant magnetite crystallization. Magma at these centers must be somewhat more hydrous and/or oxidized than is the normal case. Arc basalt clinopyroxene compositions support the inference that common high-alumina basalt (HAB) is a derivative magma, as empirical and experimental data, and MELTS simulations, all show coupled increases in Al and Fe/Mg in cpx, consistent with crystallization from increasingly aluminous wet magma. Further plagioclase-dominated fractionation of HAB in arcs in most cases generates Fe-rich basalt or basaltic andesite, consistent with lack of significant magnetite in HAB and mafic Fe-rich HAB groundmass compositions. Differentiation of HAB (by AFC?) may include Al-rich tholeiitic andesite, but not typical calc-alkaline andesite. The latter is likely made by magma mixing, as suggested by ubiquitous mineralogic disequilibrium, low Al content, and P systematics. P behaves incompatibly in virtually all common, non-alkaline, mafic arc magma, and apatite crystallization does not commence soon enough in mafic arc magma to buffer P concentrations in common andesite. Typical basalt to andesite mass balance fractionation models ignore P and fail P mass balance.