Diffusion and melt inclusions

I haven’t written much about actual lab science recently, and Dr. Carl Spandler, a former postdoc of ours at the ANU, recently published a funky nature paper using data from our laser lab, so here’s a summary of what they did. I should point out that I didn’t do any of the gruntwork for this research. I did, however, provide some minor assistance for some of the followup studies.

A summary of the paper, and the abstract are free. The full paper requires a subscription.

But first, a description of melt inclusions.

As a magma (or silicate melt) cools, it starts to crystallize. In the case of basalt, the first crystals to form are generally olivine. Sometimes, as these crystals are growing, they grow in a geometry that allows a blob of the melt from which they grow to become trapped as an inclusion inside of the crystal. If the crystal is then transported somewhere else, it keeps this entombed blob of melt.

Sometimes, melt inclusions are anomalous. An anomalous melt inclusion in any melt inclusion with a trace element composition different to what one would expect to find. One trace elemental component that is often anomalous is the relative abundances of the lanthanides, known to geologists as the rare earth elements. For those who can’t remember the names of these elements, a mnemonic is available here.

In the Earth’s mantle, rare earth elements all generally occur as trivalent oxides in melts, silicates, or (rarely) phosphates. Because they have the same charge, similar oxygen affinity, and gradually decreasing ionic radius, their behavior in geochemical systems relative to each other can be predicted fairly well. Thus, deviations in the relative abundances of REE’s compared to a particular reference value are used to infer all sorts of geological stories.

Many anomalous melt inclusions have anomalous REE patterns, and these have been used to constrain the origin of these melt inclusions. For example, the heavy REE have a small enough ionic radius to easily fit into the crystal structure of garnet, while the light REE do not. So if a rock containing garnet partially melts, and that melt is in equilibrium with some residual garnet, then the melt will have a very high La/Lu ratio, because some of the rock’s Lu will stay behind in the garnet, while none of the La will.

Because garnet is only stable in the mantle at high pressures, a “garnet signature” REE pattern can be used to infer a deep source of melting- and most basalts do not show significant garnet signatures. A melt inclusion with a garnet REE signature in a rock with no bulk garnet signature would be said to be anomalous.

Of course, in order to be geologically meaningful, the REE in a melt inclusion have to be effectively trapped by the crystal. If the temperatures and residence times are too large, then solid state diffusion might allow the melt inclusion to equilibrate with the melt outside of the crystal. While most melt inclusion people have previously assumed that the existence of melt inclusions requires them to not re-equilibrate, the purpose of the experiments presented in the Spandler et al. paper was to determine whether or not REE diffusion can occur in typical magmatic systems.

So, this is what they did:
1. Get a population of normal MORB melt inclusions that were unlikely to have any anomalous inclusions in them.
2. Determine the temperature at which these inclusions were trapped in the host olivine.
3. Determine the composition of the olivine that traps the inclusions.
4. Calculate what the composition of a basalt should be, in order to be in equilibrium with the olivine at the trapping temperature of the inclusions, under a fixed fO2 and atmospheric pressure.
5. synthesize a basalt of that composition. A synthetic basalt made from lab reagents will have no REE in it.
6. Dope the synthetic basalt with several hundred ppm of the following REE: Pr, Eu, Tb, Ho, Lu
Presumably these were chosen for the following reasons: AS odd-numbered elements, they have lower abundance, so the ration of synthetic to natural is greater for a fixed concentration. Also, the detection limits and counting stats for the mass spec are better, because all but Eu are monoisotopic clear mass numbers, so you can count all the ions, not just those from a minor isotope.
7. Heat the synthetic basalt up to the trapping temperature, toss in the intact olivines containing natural melt inclusions, and let them sit for varying time periods.
8. Quench, extract the olivines, polish them down to expose the melt inclusions, and see if any of the doped elements diffused into the melt inclusions.

Not only did they find that diffusion occurred, but they were able to determine what the diffusion coefficients were. And applying those coefficients to magmatic systems showed that REE will diffusively re-equilibrate on a timescale of years. Short-lived nuclide and geophysical constraints suggest it takes thousands to tens of thousands of years for melts to migrate from their mantle sources to the surface. Thus, anomalous melt inclusions must be trapped in a late stage of magma migration, as any melt inclusion captured early on would re-equilibrate long before it was erupted the surface.

While the paper was languishing in review, Carl described the results in a talk at Goldschmidt. It was that talk that caused Al Hoffman to blow his top, which was highly entertaining for us pudknockers in the back row. But melt inclusion research is an incredibly finicky and laborious line of study, so I can see how being shown that it can’t possibly mean what you think it means could be upsetting.

Anyway, that’s the lab denizen’s view of the study. It would be interesting to see what a skeptical petrologist makes of it.

C. Spandler, H. St C. O'Neill & V. S. Kamenetsky 2007. Survival times of anomalous melt inclusions from element diffusion in olivine and chromite. Nature 447, 303-306