Mantle Melting, Melt Extraction and Source Heterogeneity: Issues and Constraints

Charles H. Langmuir Lamont-Doherty Earth Observatory, Palisades, NY 10964, USA

There has been active debate about fundamental aspects of magma generation in the mantle: the magnitude and distribution of temperature variations, whether melting is fractional or equilibrium, the extent of melt reaction during ascent, and variations in the composition of the mantle source. These issues can be addressed with greater rigor today because of the growing data base on young volcanic rocks, new experimental data, improvements in modeling, and U-series disequilibria. It is useful conceptually to divide mantle processes into global and local aspects. Global variations refer to average chemistry over hundreds of km, and reflect long wavelength gradients and processes-- such as mantle temperature, thickness of the crust and lithosphere, and spreading or convergence rate. Local variations are pertinent to a single ridge segment or hot spot, or between adjacent volcanoes at a convergent margin. Chemical variations, and their associated controlling processes, can differ fundamentally between global and local scales.

Our understanding of mid-ocean ridge basalts (MORB) plays a key role in addressing these issues. MORB are the rosetta stone of igneous petrology because of their unique advantages for the understanding of igneous processes: fresh glass compositions, low volatile contents. reduced oxidation states, simple mineralogy, lack of pre-existing crust, global coverage, tectonic constraints
(e.g. crustal thickness, spreading rate), constrained age and samples from all levels of the system -- volcanics, crustal cumulates, and residual mantle. Comparison with MORB allows constraints for other tectonic settings to be far better than if these other settings are considered in isolation.

Even for MORB the situation is not simple. The critical constraint from basalt petrology is not widely recognized, and comes from the positive correlation between pressure of melting (from Fe) and extent of melting (from Na), which requires ascent of magma without re-equilibration at shallow levels, and temperature variations (Klein and Langmuir, 1987). Another important aspect is that fractional melting leads to higher mean pressures, and hence higher Fe and lower Si contents, than equilibrium melting (Langmuir et al., 1992). Data from abyssal peridotites is consistent, and shows higher extents of melting for samples from shallow regions, and fractional melting (i.e. no re-equilibration) (Johnson and Dick, 1992). Recent work questions Fe as a pressure indicator, and points out that this model does not account for variations in the Sm/Yb ratio (Shen and Forsyth, 1994). Sm/Yb variations imply a large garnet influence for the deepest ridges, where the major element models imply the least garnet influence. This has led to a revised model which calls on mantle heterogeneity and surface cooling with little temperature variation (Shen and Forsyth, 1994) Th-Pr-Ra studies have also been interpreted to require porous flow with re-equilibration at shallow mantle levels (e.g. Lundstrom et al., 1995), rather than fractional melting.

The new experimental results for Si conform with the predictions from Fe, and give the same results for pressure variations. In addition, recent data on the global systematics of Th-U disequilibria are consistent with the major element models (Bourdon et al., subm.). The conflict with Sm/Yb can be resolved if volatiles produce a region of deep, low degree melts below the dry solidus. For cold mantle, the dry solidus is shallower than garnet stability, but the wet solidus is in the garnet field. The deep, garnet-
influenced melts may make up 20-30% of melt production. The deep melts are enriched in trace elements and dominate the trace element budget, producing high Sm/Yb, but the major elements are little affected. For hot mantle, the melt volume from deep melts is swamped by the far larger volume of melt production above the dry solidus, such that the contribution is only 2-3% of total melt. In addition, two dimensional effects and the inability to extract the melt from the margins of the melting regime lead to more depleted compositions. These results reconcile the conflict between Sm/Yb and major elements, and support the existence of mantle temperature variations beneath ridges.

Examination of MORB chemistry along gradients in depth, spreading rate and hot spot influence provide additional constraints on source and process. For example, strong evidence for mantle heterogeneity in major elements (as well as trace elements and isotopes) comes from transects along ocean ridges crossing the Azores and Iceland hot spots. While both hot spot sources appear to be enriched in Na, the Na enrichment in the Azores is greater, and Iceland is enriched in Fe and Ti relative to the Azores. The Iceland and Azores type of plume represents two great plume classes-- Azores, Canaries, Fernando, Cape Verde, Tristan etc. on the one hand, and Iceland, Hawaii, Marquesas, Samoa, Cook-Australs etc. on the other. Local major element variations within hot spots are remarkably systematic, and are similar in form to local variations observed at slow spreading ridges.

The systematics and understanding from MORB are very useful in approaching the more complex environment of convergent margins. At convergent margins as at ridges, chemical variations separate into global and local aspects. While there is inevitable uncertainty concerning the effects of the continental crustal filter, the major element systematics of least differentiated arc basalts are remarkably consistent with the melting systematics derived from experiments and MORB (Plank and Langmuir, 1988). Once this context is understood, effects such as depletion of the mantle wedge and the variable influence of sediment and fluid on local variations become far more systematic.


Bourdon, B., Zindler, A., Elliott, T. & Langmuir, C. Nature (submitted).

Johnson K. & Dick, H., J. Geophys. Res. 97, 9219 (1992).

Klein, E. & Langmuir, C., J. Geophys. Res. (1987).

Langmuir, C., Klein, E. & Plank, T., In Mantle Flow and Melt Generation at Ocean Ridges, AGU Monograph 183-280 (1992).

Lundstrom, C., Gill, J., Williams, Q. & Perfit, M., Science 270, 1958-1961 (1995).

Plank, T. & Langmuir, C., Earth Planet. Sci. Lett. 90, 349-370 (1988).

Shen, Y. & Forsyth, D., J. Geophys. Res. 100, 2211-2237 (1994).