Petrological systematics of mid-ocean ridge basalts: Constraints on melt generation beneath ocean ridges

Mid-ocean ridge basalts (MORB) are a consequence of pressure-release melting beneath ocean ridges, and contain much information concerning melt formation, melt migration and heterogeneity within the upper mantle. MORB major element chemical systematics can be divided into global and local aspects, once they have been corrected for low pressure fractionation and interlaboratory biases. Regional average compositions for ridges unaffected by hot spots ("normal" ridges) can be used to define the global correlations among normalized Na₂O, FeO, TiO₂ and SiO₂ contents, CaO/Al₂O₃ ratios, axial depth and crustal thickness. Back-arc basins show similar correlations, but are offset to lower FeO and TiO₂ contents. Some hot spots, such as the Azores and Galapagos, disrupt the systematics of nearby ridges and have the opposite relationships between FeO, Na₂O and depth over distances of 1000 km.

Local variations in basalt chemistry from slow- and fast-spreading ridges are distinct from one another. On slow-spreading ridges, correlations among the elements cross the global vector of variability at a high angle. On the fast-spreading East Pacific Rise (EPR), correlations among the elements are distinct from both global and slow-spreading compositional vectors, and involve two components of variation. Spreading rate does not control the global correlations, but influences the standard deviations of axial depth, crustal thickness, and MgO contents of basalts.

Global correlations are not found in very incompatible trace elements, even for samples far from hot spots. Moderately compatible trace elements for normal ridges, however, correlate with the major elements. Trace element systematics are significantly different for the EPR and the mid-Atlantic Ridge (MAR). Normal portions of the MAR are very depleted in REE, with little variability; hot spots cause large long wavelength variations in REE abundances. Normal EPR basalts are significantly more enriched than MAR basalts from normal ridges, and still more enriched basalts can erupt sporadically along the entire length of the EPR. This leads to very different histograms of distribution for the data sets as a whole, and a very different distribution of chemistry along strike for the two ridges. Despite these differences, the mean Ce/Sm ratios from the two ridges are identical.

Existing methods for calculating the major element compositions of mantle melts [Klein and Langmuir, 1987; McKenzie and Bickle, 1988; Niu and Batiza, 1991] are critically examined. New quantitative methods for mantle melting and high pressure fractionation are developed to evaluate the chemical consequences of melting and fractionation processes and mantle heterogeneity. The new methods rely on new equations for partition coefficients for the major elements between mantle minerals and melts. The melting calculations can be used to investigate the chemical compositions produced by small extents of melting or high pressures of melting that cannot yet be determined experimentally. Application of the new models to the observations described above leads to two major conclusions: (1) The global correlations for normal ridges are caused by variations in mantle temperature, as suggested by Klein and Langmuir [1987] and not by mantle heterogeneity. (2) Local variations are caused by melting processes, but are not yet quantitatively accounted for. On slower spreading ridges, local variations are controlled by the melting regime in the mantle. On the EPR, local variations are predominantly controlled by ubiquitous, small scale heterogeneites. Volatile content may be an important and as yet undetermined factor in affecting the observed variations in major elements.

We propose a hypothesis, similar to one proposed by Allegre et al [1984] for isotopic data, to explain the differences between the Atlantic and Pacific local trends, and the trace element systematics of the two ocean basins, as consequences of spreading rate and a different distribution of enriched components from hot spots in the two ocean basins. In the Atlantic, the hot spot influence is in discrete areas, and produces clear depth and chemical anomalies. Ridge segments far from hot spots do not contain enriched basalts. Melting processes associated with slow-spreading ridges vary substantially over short distances along strike and lead to the local trends discussed above, irrespective of hot spot influence. In the Pacific, enriched components appear to have been more thoroughly mixed into the mantle, leading to ubiquitous small scale heterogeneities. Melting processes do not vary appreciably along strike, so local chemical variations are dominated by the relative contribution of enriched component on short time and length scales. Thus the extent of mixing and distribution of enriched components influences strongly the contrasting local major element trends. Despite the difference in the distribution of enriched components, the mean compositions of each data set are equivalent. This suggests that the hot spot influence is similar in the two ocean basins, but its distribution in the upper mantle is different. These contrasting relationships between hot spots and ridges may result from differences in both spreading rate and tectonic history. Unrecognized hot spots may play an important role in diverse aspects of EPR volcanism, and in the chemical systematics of the erupted basalts.

The observations and successful models have consequences for melt formation and segregation. (1) The melting process must be closer to fractional melting than equilibrium melting. This result is in accord with inferences from abyssal peridotites [Johnson et al., 1990]. (2) Small melt fractions generated over a range of pressures must be extracted rapidly and efficiently from high pressures within the mantle without experiencing low pressure equilibration during ascent. This requires movement in large channels, and possibly more efficient extraction mechanisms than nonnally envisaged in porous flow models with small residual porosity. (3) Diverse melts from the melting regime produce variations in basalts that are observable at the surface. (4) Basalt data can be used to constrain the melting process (e.g. active vs. passive upwelling) and its relationship to segmentation. The data cannot be used to constrain the shape of the melting regime, however, for many shapes lead to similar chemical results. (5) Highly incompatible elements and U-series disequilibria results appear not yet to be explained by melting models, and may require additional processes not yet clearly envisaged.