Earth science

Water may be a damp squib

Experiments on silicon diffusion in the mineral olivine cast doubt on the widely held belief that water has a significant effect on the rheological properties of Earth's upper mantle. See Letter p.213

In 1965, Griggs and Blacic demonstrated1 that wet quartz is much weaker than dry quartz, and suggested that this should also be the case for other silicates. Indeed, the authors concluded: “These observations raise the possibility of great weakness in the earth's deeper crust and outer mantle at temperatures far below the melting point.” In other words, water might control the viscosity of all rocks. Since then, experiments on other silicates have reinforced that view2, and small amounts of water bound into normally dry minerals such as olivine are now attributed with almost divine powers in shaping the way that Earth works. On page 213 of this issue, Fei and colleagues3 show that water may be much less important in controlling many large-scale processes occurring in the Earth than was previously thought.

The ability of small amounts of water (or, more strictly, the concentration of hydroxide, OH) to weaken minerals and rocks is known as hydrolytic weakening, and is implicated in a wide range of Earth and planetary processes. Indeed, it is argued that plate tectonics itself may owe its existence to hydrolytic weakening4. The tectonic difference between Venus and Earth may be because Earth has kept more of its water5. Moreover, Earth could be subducting more water into its mantle than it loses from volcanoes at mid-ocean ridges and hotspots, thereby further weakening the mantle and enhancing mantle convection rates, despite the overall background cooling6. The effect of water on the viscous strength of rocks is conveniently called upon when normal rock-deformation processes cannot be invoked as an explanation, and phrases such as 'water weakens rocks by orders of magnitude' are commonplace in Earth science. Fei and colleagues' results throw a spanner in the works by suggesting that hydrolytic weakening in olivine is a much smaller effect than generally thought.

It needs to be said upfront that Fei et al. did not directly measure the viscosity of wet versus dry rocks. They approached the problem from a different angle by measuring silicon diffusion. This is because the viscosity of high-temperature rocks is often controlled by the most slowly diffusing atom, which is silicon in the case of olivine. In dry olivine, for instance, the activation energies for creep and for silicon diffusion are almost the same, suggesting that creep in olivine is controlled by silicon diffusion2. Fei and colleagues found that, surprisingly, water increases silicon diffusion by less than a factor of ten over three or more orders of magnitude in water content. If the viscosity of olivine is controlled by silicon diffusion, then this effect is much less than required for the several orders of magnitude of weakening commonly cited in the literature.

So which theory is right? First of all, it could simply be that the deformation mechanism in olivine is not controlled by silicon diffusion at all, and that the similar activation energy of creep and diffusion is just a coincidence. Another possibility is that the deformation experiments showing a strong hydrolytic weakening effect2 were performed under water-saturated conditions, possibly enhancing other deformation mechanisms (such as sliding on crystalline-grain boundaries) and thereby producing an artificially weakened rheology. However, the strain rates in the deformation experiments should then show a dependence on grain size — something that the authors of the studies took pains to point out is not observed. And finally, Fei and colleagues' diffusion experiments were performed on iron-free olivines; ferric iron and other ionic species may affect both diffusion and deformation.

However, support for a small hydrolytic-weakening effect was published earlier this year7. These authors used a newer deformation apparatus to measure the rheology of wet olivine up to a pressure of 7 gigapascals (equivalent to a depth of about 200 kilometres). First, they found that wet olivines were only about 1.5 times weaker than dry olivines, and second, they saw no measurable dependence of viscosity on water after the first few parts per million or so of water. In other words, a small amount of weakening occurred from a small concentration of water, after which the strength remained the same regardless of the water content. Certainly, the authors did not see the large dependence of viscosity on water content seen in the earlier deformation experiments2.

So is it possible that the large effect of water was never really there in the deformation experiments2 to begin with? Wet olivine is certainly weaker than dry olivine, but the deformation experiments on which marked hydrolytic weakening is based were performed at relatively low pressures (less than 0.5 GPa). At these pressures, the solubility of water in olivine is low, restricting the range of water contents in individual studies and perhaps making it difficult to determine the exact dependence of viscosity on water concentration. However, it is worth noting that the amount of water in the latest deformation experiments7 is also restricted, and so the uncertainty in determining the dependence of viscosity on water content could be aimed at their — opposite — results too.

It is early days. Fei and colleagues' diffusion experiments need to be repeated and extended to other compositions — particularly iron-bearing olivines. Deformation experiments need to be performed on olivines containing more water, and on polycrystalline material as well as single crystals. Deformation and diffusion mechanisms appropriate to Earth conditions and compositions need to be worked out. And, of course, Earth is not only olivine, so what about the other minerals of which it consists, such as pyroxenes, garnet, wadsleyite, ringwoodite and perovskite? How does their strength depend on water?

Finally, what about all those Earth processes that seem to require hydrolytic weakening? It is worth pointing out that there are other ways of softening minerals and rocks. Melts, strain localization, grain-size reduction and changes in deformation mechanism can all produce interesting dynamical behaviour8. So, regardless of whether hydrolytic weakening is or is not a strong effect, plate tectonics exists, and Venus is definitely different from Earth.


  1. 1

    Griggs, D. T. & Blacic, J. D. Science 147, 292–295 (1965).

    ADS  CAS  Article  Google Scholar 

  2. 2

    Kohlstedt, D. L. Rev. Miner. Geochem. 62, 377–396 (2006).

    CAS  Article  Google Scholar 

  3. 3

    Fei, H., Wiedenbeck, M., Yamazaki, D. & Katsura, T. Nature 498, 213–215 (2013).

    ADS  CAS  Article  Google Scholar 

  4. 4

    Regenauer-Lieb, K., Yuen, D. A. & Branlund, J. Science 294, 578–580 (2001).

    ADS  CAS  Article  Google Scholar 

  5. 5

    Moresi, L. & Solomatov, V. Geophys. J. Int. 133, 669–682 (1998).

    ADS  Article  Google Scholar 

  6. 6

    Korenaga, J. J. Geophys. Res. 116, B12403 (2011).

    ADS  Article  Google Scholar 

  7. 7

    Girard, J., Chen, J., Raterron, P. & Holyoke, C. W. Phys. Earth Planet. Inter. 216, 12–20 (2013).

    ADS  CAS  Article  Google Scholar 

  8. 8

    Bercovici, D. & Ricard, Y. Phys. Earth Planet. Inter. 202–203, 27–55 (2012).

    ADS  Article  Google Scholar 

Download references

Author information



Corresponding author

Correspondence to John Brodholt.

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Brodholt, J. Water may be a damp squib. Nature 498, 181–182 (2013).

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

Sign up for the Nature Briefing newsletter for a daily update on COVID-19 science.
Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing