Data from a different technique for probing events within the Earth, interpreted in terms of a new hypothesis about the effects of water at depth, raise tantalizing questions about recycling of a tectonic plate.
In places such as the Pacific Rim, the sea floor collides with and is pushed below the land masses that border the ocean. This process of subduction, and its consequences in the form of earthquakes and volcanism, are the result of plate tectonics, in which the formation of sea floor at mid-ocean ridges is accommodated by the recycling of tectonic plates in the Earth's interior. But the recycling is a complex process: the interaction of the subducting plate, or slab, with the surrounding material itself produces volcanism and, in some settings, new sea floor. On page 399 of this issue, Booker et al.1 present dramatic models of the subduction system beneath the Andes that suggest that the recycling processes run deeper than previously thought. Their interpretation depends on new data, which in themselves are striking but which also allow the authors to make some inferences about the influence of water at depth.
Water plays an important role in subduction recycling: its release from the oceanic plate can affect the rheology of the rock as it moves into the mantle, below the Earth's crust, and also increases rock melting2. Yet the amount of water released, the depth of release and the pathways it takes are poorly defined, as are the depths of melting and the route that the melt takes as it moves from close to the subducting plate to its eruption in the overlying volcanic arc.
Seismic techniques for looking into the Earth will be familiar to most people. But there are other approaches, one being the magnetotelluric (MT) method, which offers promise for locating melt and identifying the distribution of water in the mantle. This technique uses naturally occurring electric currents in the ionosphere, created by the capture of charged particles by the planet's magnetic field, to measure Earth's electrical conductivity. That conductivity depends partly on composition and temperature. But it can be dramatically increased by small amounts of partial melt, provided that the melt forms an interconnected network3. And it can also be affected by water in the mantle, in the form of dissolved hydrogen, both above and below the transition that occurs globally at around 410 km, where the mantle composition undergoes a change from one mineral phase (olivine4) to another (wadsleyite5). MT has been used only infrequently to address subduction-zone processes, mostly because these systems are close to the ocean and require complex onshore–offshore investigations6,7. Not only are the results of Booker et al.1 dramatic in their own right, they also represent a significant advance in MT imaging of subduction systems.
Where Booker et al. have made their measurements, the subducting slab flattens beneath South America before plunging down into the mantle. This makes the area somewhat anomalous — an inactive volcanic arc sits about 200 km above the slab, instead of the more normal 100 km. But it does mean that the key subduction processes occur farther inland beneath the continent, and can be imaged using terrestrial MT (see Fig. 1 of the paper on page 400).
The most significant part of Booker and colleagues' model is the conductive anomaly that appears to coincide with the top of the subducting slab. The introduction of fluids from the slab, resulting in melting, seems one obvious explanation for the anomaly. Yet what is surprising is the great depth — at least 250 km but perhaps as much as 400 km — to which this melting seems to be occurring. Earthquake activity from the adjacent slab (there is very little activity in the region sampled by Booker et al.) shows intermediate-depth earthquakes ending around 200 km, and it has been suggested that this marks the depth limit of water-bearing mineral phases8,9. In fact, the MT data indicate that there is electrical conductivity along or just above the flat part of the slab (above 200 km), suggesting some water loss at these depths.
There are few other clues that can help to account for this apparent deeper melting. Analyses of samples from an overlying inactive volcano, Pocho, are equivocal, because much of the chemistry of the samples reflects processes occurring in the crust10; in any event, their age means that they might easily reflect a different melting geometry. The prolonged passage of an old slab under the continent at relatively shallow depths is unusual, and might provide an explanation for why the slab has been able to hold on to its water to greater depths than usual. Perhaps it results in the breakdown of mineral phases that do not normally contribute to the water budget.
Booker et al.1 explain the deep melting by invoking the ideas of Bercovici and Karato11. This new hypothesis suggests that a layer of melt exists just above the 410-km boundary as slowly upwelling mantle undergoes a change from water-rich wadsleyite to olivine, liberating water and inducing melting. In Booker and colleagues' models, the melt column seems to originate as much from the 410-km boundary as it does from the subducting slab, and so the suggestion of a link between the two is not surprising. The authors propose that the liberation of additional water from the slab induces further production of buoyant melt, a necessary condition if the melt is to rise as is seen.
This link to Bercovici and Karato's model is intriguing, but I'm not completely sold on it. Undoubtedly, the argument would be strengthened by better data on the structure across the 410-km boundary which, as the authors point out, is not the best-resolved feature of the model. Booker et al. suggest that the mantle to the west of (or beneath) the slab is dehydrated as a result of melt extraction at the East Pacific Rise, part of the mid-ocean ridge system in the Pacific Ocean. Although this might be true for the upper 60–80 km or so of oceanic plate from which melt and water have been extracted, other data suggest that there is plenty of remaining water that can increase electrical conductivity in the mantle below about 80 km depth12. This means that only the region adjacent to the subducting slab would be expected to be dry. If there is water around, we would expect a more or less uniform increase in conductivity at 410 km across the region. Although their models show a stepwise increase across the 410-km boundary, Booker et al. point out that the data are consistent with a flat 410-km transition throughout the region. Why is all this important? Well, it speaks to outstanding issues in terms of resolution of this critical part of the system.
Regardless of the details of the 410-km boundary, the authors' primary observation — an electrical conductor that must surely represent a subduction-related melt column rising from depth — is striking. And, as they point out, issues pertaining to the 410-km boundary and the link to the Bercovici–Karato hypothesis can be addressed with measurements made with a longer chain of MT stations. If the interaction with a melt layer at 410 km is the explanation, it should be seen in other subduction systems. It hasn't been seen elsewhere yet, but maybe we just need to look more carefully.
Booker, J. R., Favetto, A. & Pomposiello, M. C. Nature 429, 399–403 (2004).
Hirth, G. & Kohlstedt, D. L. Earth Planet. Sci. Lett. 144, 93–108 (1996).
Roberts, J. J. & Tyburczy, J. A. J. Geophys. Res. 104, 7055–7066 (1999).
Karato, S. Nature 347, 272–273 (1990).
Xu, Y. et al. Science 280, 1415–1418 (1998).
Wannamaker, P. E. et al. J. Geophys. Res. 94, 14127–14144 (1989).
Echternacht, F. et al. Phys. Earth Planet. Inter. 102, 69–87 (1997).
Kirby, S. et al. AGU Geophys. Monogr. 96, 195–214 (1996).
Hacker, B. R. et al. J. Geophys. Res. 108, doi:10.1029/2001JB001129 (2003).
Kay, S. M. & Gordillo, C. E. Contrib. Mineral. Petrol. 117, 25–44 (1994).
Bercovici, D. & Karato, S. Nature 425, 39–44 (2003).
Lizarralde, D. et al. J. Geophys. Res. 100, 17837–17854 (1995).