News & Views | Published:


The lost continents

Nature volume 448, pages 655656 (09 August 2007) | Download Citation

Once subducted into the mantle, material from Earth's continental crust seems to disappear. But its distinctive isotopic signature has been found back at the surface — in volcanic rocks on a Pacific island.

Plate tectonic theory requires that much of Earth's crust returns to the underlying mantle through the process known as subduction. Most of this subducted material is solidified volcanic lava — basalts — from oceanic crust (see, for example, ref. 1). But continents do not completely escape this fate; material eroded from the continental crust is deposited on the ocean floor as a veneer of sediments up to a few hundred metres thick. This layer is subducted, when the time comes, along with the basaltic crust.

The Samoan islands (here, Ta'ū in American Samoa) are the tips of a volcanic plume rising from deep in Earth's mantle. Image: S. R. HART

But what happens to the continental material after that? Does convection in the mantle stir it around so that it loses its distinctive identity? Or does at least some of it reappear in volcanic rocks formed at ocean ridges or at isolated 'hotspots' of volcanic extrusion under the sea? Jackson et al.2 (page 684 of this issue) have found what might be a 'smoking gun' for the second option. In unprecedented studies of the isotopic and trace-element compositions of basalts dredged from the flanks of a Samoan island, they find almost unequivocal evidence for traces of recycled ancient continental crust in extruded volcanic rocks.

Samoa, a chain of islands in the South Pacific just east of the International Date Line, is thought to owe its existence to a plume of volcanic material rising from the deep mantle. It has long been singled out as a possible harbour of traces of recycled continental material because its rocks contain more radiogenic strontium (87Sr, produced by β-decay of the very long-lived rubidium isotope 87Rb) relative to normal strontium (86Sr), than is usual3.

Moderately increased 87Sr/86Sr ratios indicate that a rock has spent a long time in an environment with a higher Rb/Sr ratio. High Rb/Sr ratios are characteristic of the continental crust, but can also come about through the melting of mantle rock. Because of its large ionic radius, rubidium fits poorly into the crystal structures of mantle minerals, and so becomes concentrated in melted rock fractions much more effectively than strontium does. This effect is strongest at low melt fractions, so high Rb/Sr ratios could also be the signature of small amounts of melt that formed in one region of the uppermost part of the mantle (the asthenosphere) and migrated into another part (the lithosphere). This process, known as mantle metasomatism, has become increasingly favoured by mantle geochemists4. As a result, the burden of coming up with incontrovertible proof has shifted onto those who argue that some volcanic mantle upwellings contain recycled continental material.

Previous assessments have had to rely on extrapolations of isotope-ratio and trace-element trends in various extruded rocks to an endpoint thought to correspond to values found in the continental crust. As the range of ratios covered by the data was small, and their distance from the inferred continental ratio large, this procedure was akin to wagging a large dog with a small tail. One could not know if the ratio of the distant inferred end-point of the trend really did represent the continental crust, or something well short of it — a metasomatized mantle region, for example.

The breakthrough presented by Jackson et al.2 is an extension of the range of observed 87Sr/86Sr ratios (and the correlated neodymium isotope ratio 143Nd/144Nd) in volcanic extrusions well into the region of typical old continental rocks (Fig. 1). Even so, these isotope ratios might, in principle, be explained by rocks processed through an extreme version of the metasomatic model. But the authors also ascertain values of various other tell-tale trace-element ratios that are a factor of four lower than those found in metasomatized rocks, and are entirely characteristic of an origin in the continental crust5.

Figure 1: Adopting continental values.
Figure 1

The strontium and neodymium ratios 87Sr/86Sr versus 143Nd/144Nd in volcanic rocks from several mantle 'hotspots' or plumes. High strontium and low neodymium ratios are both characteristic of ancient continental rocks, but possibly also of mantle regions enriched by infiltration of small amounts of mantle melt. Data for Hawaiian islands are representative of most ocean island volcanoes, which show isotopic evidence for normal mantle enrichment and depletion processes. Published data from Samoa, the Marquesas and the Society Islands, all island groups in the South Pacific, deviate from the normal mantle trend (which also contains the composition of the total 'bulk' Earth) in the direction of high 87Sr/86Sr values, indicating that they might contain recycled continental material. Jackson and colleagues' data2 (red) greatly extend this trend. Together with other 'continental' geochemical indicators, these data strongly support the continental recycling hypothesis for this (relatively rare) class of volcano. (Isotope data assembled from ref. 7.)

One possible loophole in this interpretation is that the continental isotopic and trace-element signature might have been picked up by the basalts in transit from their melting site in the mantle, at depths of 100 km or so, through the original Pacific Ocean floor — which is itself covered with predominantly continent-derived sediment. Jackson et al. effectively counter the spectre of such 'crustal contamination' through non-recycled material by analysing actual Pacific sediments near Samoa, finding that the lead isotope ratios of these samples are incompatible with those of Samoan lavas.

If the origin of the Samoan magma source is now settled, one may reasonably infer that other nearby ocean islands of similar isotopic composition (known as enriched-mantle 2, or EM2)6, such as the Society Islands (including Tahiti) and the Marquesas, have similar sources. This would be a mixture of mantle rocks and recycled oceanic crust, with a sprinkling of sediment derived from ancient upper continental crust and subducted along with the rest of the package. Similarly, a few other ocean islands, such as Pitcairn, might contain small amounts of recycled pelagic sediment.

The troubling point remains, however, that even if all EM-type volcanoes contain recycled sediments, instances in which this can be recognized are rare. After all, wherever subduction occurs, sediments are likely to enter the mantle. Jackson and colleagues2 point out that if the current rate of sediment subduction (0.5–0.7 km3 per year) is representative of the past 4 billion years, the total amount of subducted sediment will make up only about 0.15% of the mass of the mantle. That might not seem much, but it is about a third of the present-day mass of the continents.

These sediments are also rather highly enriched in incompatible elements (those most likely to be removed by mantle melting), up to a factor of about 100 compared with primitive mantle compositions. They are even more enriched when compared with present-day mantle, which has been depleted in these same elements by the removal of continental crust. Thus, in the present-day mantle, possibly 10–20% of the highly incompatible elements rubidium, uranium and thorium might have been reinjected into the mantle from crustal sources by sediment subduction. And the very fact that the mantle is isotopically remarkably heterogeneous demonstrates that convective stirring is not particularly efficient.

Given all these factors, I find it remarkable that so little of the subducted sediments can be recognized in recycled form in both mid-ocean ridges and hotspot volcanoes. Much of the budget of subducted trace elements must therefore either be short-circuited back to the surface during subduction-related volcanism, or be hidden in a relatively stable place such as the subcontinental lithosphere. That is a question for another day. For now at least, such material has at last been identified in one place where it does appear.


  1. 1.

    Nature 443, 516–517 (2006).

  2. 2.

    et al. Nature 448, 684–687 (2007).

  3. 3.

    & Nature 296, 821–825 (1982).

  4. 4.

    et al. Geochem. Geophys. Geosyst. 5, doi:10.1029/2003GC000623 (2004).

  5. 5.

    , , & Earth Planet. Sci. Lett. 79, 33–45 (1986).

  6. 6.

    & Annu. Rev. Earth Planet. Sci. 14, 493–571 (1986).

  7. 7.

Download references

Author information


  1. Albrecht W. Hofmann is at the Max Planck Institute for Chemistry, Postfach 3060, 55020 Mainz, Germany.

    • Albrecht W. Hofmann


  1. Search for Albrecht W. Hofmann in:

About this article

Publication history



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.

Newsletter Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing