Earth science

Magma chambers on a slow burner

An assessment of crystallization processes occurring in magma chambers in the ocean floor finds an unexpected enrichment in trace elements, reviving an old theory of the cycling of magma in these chambers. See Article p.698

The world's ocean basins are constantly being regenerated by an 80,000-kilometre-long volcanic system of mid-ocean ridges, where Earth's mantle melts to form magma that eventually produces the basaltic floor of the oceans. The composition of ocean-floor basalts is one of the main sources of information about Earth's deeper interior. On page 698 of this issue, O'Neill and Jenner1 re-examine the chemical compositions of basaltic lavas from this global magmatic system. They find new, and remarkably systematic, chemical relationships between the concentrations of 'incompatible' trace elements (so named because they are largely excluded from magmatic crystals) and that of magnesium oxide (MgO).

As expected, the content of incompatible elements increases in the basaltic-liquid component (the melt) of magmas, because MgO-bearing crystals precipitate in sub-oceanic magma chambers (reservoirs), causing the MgO content of the liquid to decrease. But O'Neill and Jenner show that the observed incompatible-element increase is much greater than conventional crystallization processes can explain. Their proposed solution to this dilemma would require a revision in the way geochemists calculate the composition of parental magmas entering magma chambers, and therefore also the composition of the mantle rocks from which these magmas are derived.

When basaltic lava comes into contact with cold sea water, it is chilled into glass. Geochemists like to analyse such glasses because they preserve the chemical composition of the lava particularly well, and they can thus tell the researchers much about the composition of the underlying mantle in which the melt forms. However, this view of the mantle is blurred because there are several intervening stages between melt formation and the eruption of lava. These are: partial melting of the mantle at depth (greater than about 30 km); extraction of the melt from the partially molten mush; its emplacement in shallow magma chambers; the formation and settling out of magmatic crystals in these chambers; and, finally, eruption of the remaining liquid on the ocean floor.

Two fundamentally opposing views of the mantle composition inferred from these glasses have prevailed over the past 40 years. One holds that the mantle has an essentially uniform composition, and that the compositional variability of the erupted basaltic lavas is produced primarily by processes occurring in the shallow magma chambers. The other view holds that magma-chamber processes have only minor effects on the erupted lavas that can be easily corrected for, and that the variations in lava composition mainly reflect differences in the composition of the mantle source and in the specifics of the melting process.

This latter view has gradually gained the upper hand, because much of the observed chemical variability of the lavas correlates with variations in the isotopic composition of the elements strontium, neodymium, hafnium and lead. These elements are the products of very slow radioactive decay, and therefore accumulate only during long residence times in the mantle. The observed differences in isotopic composition can therefore not be produced in short-lived magma chambers, but require long-term differences in parent–daughter ratios in the (mantle) source of the melts.

A crucial requirement when going backward from observed compositions of erupted basalts to their mantle sources is to evaluate the effects of partial crystallization and loss of the crystals in magma chambers. This is widely assumed to involve 'fractional crystallization', whereby newly formed crystals are immediately removed from chemical interaction with the liquid. Laboratory experiments have shown that the crystallization process in ocean-ridge magma chambers invariably involves the magnesium-bearing mineral olivine. The net effect of this is that the MgO content of the liquid progressively decreases as freshly crystallized olivine is removed from the liquid, whereas there is an increase in the contents of incompatible trace elements (such as barium, thorium and neodymium) because they are excluded from the crystals.

This was thought to be well understood — until O'Neill and Jenner plotted the incompatible-element contents against MgO for two recently assembled global data sets2,3. They found excellent linear correlations (with the expected negative slopes) between incompatible-element and MgO contents. However, they were startled to find that these slopes are consistently greater than the maximum allowed from fractional-crystallization theory.

If fractional crystallization does not explain this effect, what process does? One possibility is that lavas that have higher incompatible-element contents start out with systematically lower parental MgO contents. But that would mean that the sources of these magmas could not contain olivine, even though this is the most common of all upper-mantle minerals. Nevertheless, it has been proposed4 that some mid-ocean-ridge basalts are mixtures of liquids formed from peridotite, the 'standard' olivine-bearing mantle rock, and other liquids formed from eclogite or pyroxenite, which are olivine-free rocks that form from subducted, recycled oceanic basalts. Melts from such recycled basalts should also have a higher-than-normal content of incompatible elements and a lower-than-normal MgO content. Such recycled basalts should also have different isotopic compositions of neodymium, for example. However, the expected correlations between neodymium isotopes and MgO have not been documented for any global set of ocean-ridge basalts.

As a way out of the dilemma, O'Neill and Jenner revive and generalize a model that was originally proposed by O'Hara5 and later modified by Albarède6, but which has been mostly forgotten. This model envisages a magma chamber that is periodically refilled with fresh parental liquid from below. The fresh liquid mixes with the pre-existing liquid, and the mixture is tapped by a volcano, whereupon crystallization resumes. This 'trick' of replenishment with fresh parental magma keeps the MgO content of the liquid from falling too rapidly and allows a much greater build-up of incompatible-element concentrations in the residual liquid than would be possible with closed-system fractional crystallization.

At first sight, these effects might seem to be of interest mostly to aficionados of the details of magma-chamber processes. But they imply a much-reduced role for chemical heterogeneity of the mantle, as well as for the effects of partial melting, because most of the incompatible-element variability is now ascribed to processes occurring in the magma chambers.

How plausible is this model? O'Neill and Jenner propose that a global assemblage of magma chambers exists, in which crystallization processes vary locally, but which as an ensemble conform to the O'Hara–Albarède model. Although this model is apparently quite successful in describing the global observations, the question of why these locally variable crystallization processes should average out to this idealized model remains a mystery.

The authors have tested their model by comparing predicted and measured partition coefficients of incompatible trace elements for crystals forming in magma chambers (the partition coefficient is the concentration of an element in a crystal divided by its concentration in the liquid). For the most part, the agreement is impressive, but barium and potassium are significant exceptions. These elements behave like highly incompatible elements in the basalts. In other words, their partition coefficients should be close to zero, which is actually the case in mantle minerals. But their experimentally determined partition coefficients in plagioclase (one of the main minerals that form in shallow magma chambers) are high enough to raise questions about the model.

Clearly, further experimental work is needed to resolve these issues. In the meantime, O'Neill and Jenner's paper indicates the need for a re-examination of the nature of a magmatic process that is volumetrically by far the most significant on Earth. An accurate assessment of the crystallization process is needed to infer the composition of the mantle from which the ocean-floor basalts are derived.


  1. 1

    O'Neill, H. St C. & Jenner, F. E. Nature 491, 698–704 (2012).

  2. 2

    Jenner, F. E. & O'Neill, H. St C. Geochem. Geophys. Geosyst. 13, Q02005 (2012).

  3. 3

    Arevalo, R. Jr & McDonough, W. F. Chem. Geol. 271, 70–85 (2010).

  4. 4

    Sobolev, A. V. et al. Science 316, 412–417 (2007).

  5. 5

    O'Hara, M. J. Nature 266, 503–507 (1977).

  6. 6

    Albarède, F. Nature 318, 356–358 (1985).

Download references

Author information

Correspondence to Albrecht W. Hofmann.

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Hofmann, A. Magma chambers on a slow burner. Nature 491, 677–678 (2012).

Download citation

Further reading


By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.