The Case for Reactive Crystallization at Mid-Ocean Ridges

Evidence from abyssal peridotites suggests that significant chemical reaction with peridotite can occur during the early stages of cooling and crystallization of mid-ocean ridge basalt (MORB) magmas. We evaluate the hypothesis that reactive crystallization (crystallization influenced by such melt^rock reaction) could cause magma compositions to evolve along a different chemical trajectory than expected for fractional crystallization, and that reactive crystallization might be common in MORB petrogenesis. If correct, this hypothesis implies that a component of major element variability in fractionation-corrected MORB, commonly interpreted solely in terms of mantle source composition and potential temperature, could reflect reactive crystallization. The compositional evolution of MORB magmas undergoing reactive crystallization is predicted using thermodynamic calculations.We find that the decreasing melt MgO content during reactive crystallization is accompanied by nearly constant Mg-number [defined as molar MgO/ (MgOþFeO)], whereas melt SiO2 and Na2O contents evolve to higher values than in fractional crystallization. However, the extent of crystallization as a function of temperature is essentially identical during the initial 30^40% of both fractional and reactive crystallization. Comparison of melt transport and reaction timescales in a 1-D, steady-state, porous flow column shows that melt migration via grain-scale porous flow at the transition from melting to crystallization beneath ocean ridges will most probably give rise to reactive crystallization, whereas melt transport through the thermal boundary layer in larger conduits (dikes) will lead to fractional crystallization. Scatter in fractionation-corrected major element compositions could therefore reflect sample-to-sample variations in melt transport dynamics. Using a global compilation of MORB glass compositions, we show that 40^70% of the variability in fractionation-corrected MgO contents observed worldwide is also typically present in groups of samples collected from within 30 km of each other. Such short length-scale variability in MgO, and hence in the temperature of primitive magmas, cannot be due to variability in mantle potential temperature. There is a negative correlation in the variability of fractionation-corrected MgO (and most other compositional variables) with spreading rate.We infer that this negative correlation reflects a greater role for reactive crystallization in the thicker thermal boundary layers present beneath slow-spreading ridges. We demonstrate the ability of combined reactive and fractional crystallization to account for major element variability at several case-study locations, and argue that reactive crystallization can explain many observations of 30 km scale variability. Interpreted in terms of reactive crystallization, fractionation-corrected MgO variability could potentially bound geodynamic parameters such as the depth of onset of diking and the fraction of gabbro emplaced into residual mantle peridotite beneath the igneous crust.