Skip to main content

Thank you for visiting You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

Mantle geochemistry

Big lessons from little droplets

How does Hawaii look deep below the surface? Like viewing an object at a different magnification, studies of minuscule inclusions in volcanic rocks on the surface provide a fresh perspective on the question.

Mantle plumes are thought to be roughly elongate cylinders of rock that buoyantly rise up from deep within the Earth, manifesting themselves at the surface in features such as the Hawaiian islands and Iceland. Much attention has centred on Hawaii, because it is constructed from Earth's largest volcanoes distributed along two geochemically distinct alignments, and there is considerable debate about what these distinctions reveal about the underlying plume that feeds them. On page 837 of this issue1, Ren et al. add to that debate: they show that interpretations depend on the size of the material that is subject to chemical analysis.

The Hawaiian plume is roughly 100 kilometres in diameter, and rises from the lower mantle below a depth of 660 kilometres2. It begins to partly melt in the final stages of its upward journey, and lava erupts at the surface to make the Hawaiian volcanoes. These are heterogeneous in that each volcano has its own geochemical identity, which can vary with time3,4,5. However, lavas from Kilauea, currently the most active volcano, have a geochemistry similar to that of older lavas that erupted along the ‘Kea’ alignment of volcanoes4,5, which include Mauna Kea, Kohala and Haleakala to the northwest. The geochemistry of Kea volcanoes differs from that of the ‘Loa’ volcanoes4,5, which include Loihi, Mauna Loa, Kahoolawe and Koolau, a parallel alignment displaced to the west (see Fig. 1 on page 837).

These differences, expressed in radiogenic isotopes, and trace and major elements, have been interpreted to reflect the three-dimensional spatial organization of the chemical constituents in the plume before melting took place. This is referred to as the geochemical structure of the Hawaiian plume1,6,7,8. But there is no consensus about the form of this structure; different models are summarized in Figure 1.

Figure 1: Models of the geochemical structure of the Hawaiian mantle plume.

The diagram depicts three possible plume structures below the two alignments (Kea and Loa) of Hawaiian volcanoes, which are surface expressions of the tectonic plate on which they sit. a, A concentrically zoned and vertically continuous plume6. b, A bilateral bundle of filaments vertically continuous on the 50–500-kilometre scale7, with undefined spaces between filaments. Mauna Loa and Kilauea are two volcanoes on the Loa and Kea trends, respectively. c, A partly ordered structure, with streaks of red in a matrix of blue (modified from ref. 8). Red, average geochemical properties of volcanoes along the Loa track3,4,5,6,7,8; blue, average geochemical properties of volcanoes along the Kea track3,4,5,6,7,8. Ren et al.1 favour the partly ordered structure (c), which they believe is defined by streaks of pyroxenite (red, pyroxene-rich rocks) in a heterogeneous peridotite matrix (blue, olivine-rich rocks).

One model has geochemical heterogeneities ordered in a concentric and vertically continuous structure6 (Fig. 1a); Loa volcanoes sample the centre, Kea volcanoes sample the periphery. Another model has geochemical heterogeneities ordered in filaments clustered together like spaghetti7 (Fig. 1b), which is consistent with the expectation of stretching and shearing of heterogeneities in a plume9; Loa and Kea volcanoes sample a bilateral distribution of filaments7. A third model has a more random but partly ordered distribution of geochemical heterogeneities both vertically and laterally8 (Fig. 1c). The origin of these heterogeneities is a separate issue. However, there is a consensus that it has something to do with near-surface magmatism, crust production and sediment accumulation in the early Earth, subduction of this outer unit into the deep mantle, and an upward return in a hot plume that melts to make Hawaii.

All three models are interpretations of the geochemistry of whole-rock samples — ‘hand specimens’ — of centimetre scale. The approach taken by Ren and co-workers1 differs: using microanalytical techniques, they have acquired geochemical data on melt inclusions, about 10–100 micrometres in size, contained in millimetre-sized crystals of olivine, a common mineral in some volcanic rocks. They show that major-element and trace-element compositions of olivine-hosted inclusions from a single rock extracted from a single volcano can have the properties of both Loa and Kea volcanoes. They suggest that either the Loa or the Kea component is represented by whole rocks, but both are represented in olivine-hosted inclusions regardless of the specific geographical location of the volcano. This study supports the more random model8, although some hybrid of the ordered models (Fig. 1a, b) is still required to yield the Loa and Kea trends (Fig. 1c).

How is it possible that whole-rock samples from an individual volcano support either a Loa or a Kea source, but the inclusions support both sources? Why does the scale of the geochemical observation yield different interpretations? The answers lie in an understanding of how rock melts. The Hawaiian plume melts by forming millimetre-sized liquid droplets that inherit the geochemical properties of their source rock. Each drop of melt usually mixes with other drops during transport to the surface. Although it is not clear exactly where mixing takes place, it is known to homogenize the geochemistry on the centimetre scale of the hand specimen and the kilometre scale of an individual volcano. A rock from Kilauea can be thought of as a blend of one melt drop from a Loa source and 100 melt drops from a Kea source. However, even a well-mixed rock can contain olivines that crystallized from melt droplets before mixing took place. In this way, crystal growth can entomb the drops as tiny inclusions before they become part of the blend.

The large geochemical heterogeneities reported by Ren and co-workers for Hawaii are similar to those of inclusion studies for lavas from Iceland10. However, the Iceland case has been interpreted somewhat differently, as being the result of the continuous removal of small melt fractions from a single source composition10, a process called fractional melting. The authors of that study10 acknowledge that some of their data might also be explained by variable source compositions, as do Ren et al. for Hawaii. And it is likely that some of the geochemical variability reported by Ren et al. can be explained by fractional melting.

Further studies are evidently called for. What is becoming clear is that complementary inclusion and whole-rock geochemical studies expand the scale of observation in a way that is comparable to viewing an object with variable magnification.


  1. 1

    Ren, Z. -Y., Ingle, S., Takahashi, E., Hirano, N. & Hirata, T. Nature 436, 837–840 (2005).

    ADS  CAS  Article  Google Scholar 

  2. 2

    Morgan, W. J. Nature 230, 42–43 (1971).

    ADS  Article  Google Scholar 

  3. 3

    Frey, F. A. & Rhodes, J. M. Phil. Trans. R. Soc. Lond. A 342, 121–136 (1993).

    ADS  CAS  Google Scholar 

  4. 4

    Lassiter, J. C., DePaolo, D. J. & Tatsumoto, M. J. Geophys. Res. 101, 11769–11780 (1996).

    ADS  Article  Google Scholar 

  5. 5

    Hauri, E. H. Nature 382, 415–419 (1996).

    ADS  CAS  Article  Google Scholar 

  6. 6

    DePaolo, D. J., Bryce, J. G., Dodson, A., Shuster, D. L. & Kennedy, B. M. Geochem. Geophys. Geosyst. 2, doi:10.1029/2000GC000139 (2001).

  7. 7

    Abouchami, W. et al. Nature 434, 851–856 (2005).

    ADS  CAS  Article  Google Scholar 

  8. 8

    Blichert-Toft, J., Weis, D., Maerschalk, C., Agranier, A. & Albarède, F. Geochem. Geophys. Geosyst. 4, doi:10.1029/2002GC000340 (2003).

  9. 9

    Farnetani, C. & Samuel, H. Geophys. Res. Lett. 32, doi:10.1029/2005GL022360 (2005).

  10. 10

    Maclennan, J. et al. Geochem. Geophys. Geosyst. 4, doi:10.1029/2003C000558 (2003).

Download references

Author information



Rights and permissions

Reprints and Permissions

About this article

Cite this article

Herzberg, C. Big lessons from little droplets. Nature 436, 789–790 (2005).

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.


Quick links

Nature Briefing

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

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