Potentially huge amounts of water could be carried deep within the Earth by subducting oceanic crust. But it seems that most of that water is released, fuelling volcanism above subduction zones.
The hydrological cycle usually refers to the process by which water cycles through the atmosphere, rivers and ocean, and round again. But is there another hydrological cycle, one that circulates water through the depths of the Earth? It has long been suspected1 that surface material can be carried to the Earth's deep interior by subduction of ocean crust and associated mantle, and then returned to the surface by plumes rising through the mantle. In particular, from isotope data on basalt rocks erupted on oceanic islands, it has seemed that mantle plumes can carry deeply subducted oceanic crust and sediment2, as well as the residue of oceanic crust creation (the oceanic lithosphere)3, from depth back to the surface.
Does this process include relatively volatile species, such as water and carbon dioxide? The results reported by Dixon et al.4 on page 385 of this issue suggest not, at least to any great extent. They find that water is cycled much less efficiently in this way than might be expected — indeed, plumes that apparently contain material recycled from the Earth's surface contain less, not more, water than plumes dominated by the so-called 'common component', such as that beneath Iceland.
Magma readily gives up water at low pressure, the water escaping to the atmosphere during, and even before, eruption. So determining the water content of magma can be difficult. At depths of 1,000 m or more in the ocean, however, the pressure is sufficient for magma to retain its water as it quenches to glass by contact with sea water. Much of the work on volatiles in magmas has therefore centred on quenched glass recovered from submarine eruptions. These studies have established that when the mantle melts, hydrogen behaves much like the rare-earth element cerium5,6. To get around the variations in water introduced by partial melting and fractional crystallization, the H2O/Ce ratio can be used to compare the relative amounts of water in different magmas. Dixon et al.4 determined the water content of submarine glasses erupted on regions of the Mid-Atlantic Ridge thought to be influenced by mantle plumes. The water content of many of these basalts is surprisingly low, the H2O/Ce ratios being lower than in many other submarine basalts at places such as Easter Island7.
Furthermore, samples from the Discovery plume in the South Atlantic and the Great Meteor plume in the North Atlantic display a negative correlation between the ratios of H2O/Ce and of 87Sr/86Sr. The significance of this relationship is that it shows that the water deficiency is present in the mantle that produced the basalts, and the strontium isotopes also provide insight into the origin of that mantle. The 87Sr/86Sr ratio varies in the Earth only because of the very slow decay of 87Rb (rubidium) to 87Sr. Because the continental crust is enriched in Rb, the crust has a higher 87Sr/86Sr ratio than does the mantle, so a high ratio in basalts can signal the presence of recycled crustal material in the mantle. The most water-deficient samples studied by Dixon et al. have the highest 87Sr/86Sr ratios, suggesting that they are derived from mantle containing recycled crustal material.
In this respect, Dixon and colleagues' results are unexpected. One would have thought that recycled material would contain comparatively large amounts of water, for two reasons. First, marine sediments are rich in clay minerals, which can contain 10% or more water bound in their lattices. Second, circulating sea water hydrates the basaltic oceanic crust, eventually raising its structurally bound water content to 2–5% by weight. Additional water is present in the pore space in both sediments and basalt. These water concentrations are far above the ambient concentrations in the mantle, of perhaps 0.03%. The amount of water carried into the mantle in subducted oceanic crust and sediment exceeds 1 × 1012 kg yr−1 (ref. 8), enough to drain the oceans in little more than a billion years if the water were not returned from the mantle.
Dixon et al. argue that the low water content of recycled material results from efficient dehydration of the oceanic crust and sediment during subduction. This idea is not new. Release of water from subducting oceanic crust has long been believed to cause the magma production fuelling volcanoes that ubiquitously sit atop subduction zones9,10. Krakatoa, Mount St Helens, Mount Pinatubo and Soufriere Hills all lie above subduction zones, and the notoriously explosive nature of these and other such volcanoes is largely due to the high water content of their magmas. It is, however, a little surprising that the dehydration process is so efficient. Dixon et al. calculate that 92% of the water is extracted from subducting sediment, and 97% from the subducting oceanic crust. The deep hydrological cycle thus appears to be short-circuited, with most subducted water quickly returning to the surface through volcanism rather than being carried into the deep mantle. In other words, the subducted material is effectively put through the wringer before it sinks very far into the mantle.
It would be interesting to compare the flux of water released by subduction-zone volcanoes with estimates of the subduction flux and Dixon and colleagues' calculated dehydration efficiency. It would also be interesting to know if CO2 is released from subducting oceanic crust and sediment as efficiently as water is. These are difficult tasks, however — determining the water content of magmas erupted on land is problematic, and CO2 is lost from magma even more readily than water.
Finally, the new results4 also point to the importance of subduction-zone processes in shaping the composition of both the Earth's surface and its interior. Water released by dehydration could carry away much of the soluble-element content of subducting crust and sediment. Any prediction of the composition of deeply recycled crustal material must take account of these losses.
Armstrong, R. L. Rev. Geophys. 6, 175–199 (1968).
Hofmann, A. W. & White, W. M. Earth Planet. Sci. Lett. 57, 421–436 (1982).
Schaefer, B. F., Turner, S., Parkinson, I., Rogers, N. & Hawkesworth, C. Nature 324, 304–307 (2002).
Dixon, J. E., Leist, L., Langmuir, C. & Schilling, J.-G Nature 420, 385–389 (2002).
Michael, P. Earth Planet. Sci. Lett. 131, 301–320 (1995).
Dixon, J. E. & Clague, D. A. J. Petrol. 42, 627–654 (2001).
Simons, K. & Dixon, J. E. Geochem. Geophys. Geosyst. 3, pap. no. 2001GC000173 (2002).
Peacock, S. Science 248, 329–337 (1990).
Green, T. H. & Ringwood, A. E. Contrib. Mineral. Petrol. 18, 105–162 (1968).
Gill, J. B. Orogenic Andesites and Plate Tectonics, 247 (Springer, Berlin, 1981).
About this article
Geophysical Research Letters (2006)
Mineralogical Magazine (2005)