Wednesday, October 31, 2007

Experimentally determined D values for noble gasses

Blogging on Peer-Reviewed Research
Thermochronic recently blogged about a high profile paper that recently suggested that the partitioning coefficients for argon between mantle minerals and melts is greater than one (Watson et al. 2007). As it turns out, Heber et al. 2007, published earlier this year, suggests that the D values are actually between 10-3 and 10-4 for all noble gasses. The Heber et al. 2007 paper describes a very well-thought out and executed research program, reads like a detective novel, and is generally everything that a scientific paper should be.

They perform a standard partitioning coefficient experiment- slowly cool a melt to grow crystals, quench it, and measure the concentration of the trace element of choice in both melt and crystal to determine the ratio. The brilliant thing about this study is the depth of self-critizism and examination which they perform to identify and eliminate potential contaminants or non-equilibrium effects. This stands in stark contrast to the Watson et al. paper, where they pretty much assume that their results are right, and arrive at a value 10,000 to 10 million times higher than the Heber et al. paper.

I don’t want to ruin a gripping read, but the things that Heber et al. overcome include (but are not limited to): Room temperature diffusion of He out of the glass between experiment completion and analysis time, bubble formation, melt inclusions, very low concentrations. The interpretations of the paper are well supported by the data, but do include an explanation of how these numbers can explain “primordial” OIB noble gas ratios.

Although the noble gas partition coefficients are low (on the order of 10-4), they are still 2 orders of magnitude higher than the partitioning coefficients of K, Th, and U, which are on the order of 10-6 in a hartzburgitic system. Thus, although noble gasses are incompatable, they are not as incompatable as the radioactive sources for 4He and 40Ar. So old depleted mantle will have a high primordial/radiogenic gas signature.

The fact that the Heber et al. and Watson et al. partition coefficients for argon differ by up to 8 orders of magnitude is striking. But I find the Heber et al. result more compelling, because the authors present, and then test, a dizzying array of potential complications in a rigorous manner. In contrast, Watson et al. make statements like,

“More generally, we believe that the inconsistency of our results and those of Broadhurst et al. with other experimental studies is due to the different experimental protocols used.”

While I have great respect for Professor Watson and his impressive research record, I am not interested in his beliefs, unless they are backed up by experimental evidence. And he makes no attempt to reconcile his results with previous work, much less try to disprove his own experiments. So a paper that dissects the potential causes of research discrepancies and addresses them is more convincing to me than one that relies on untested dismissals. As a final note, Watson et al. do not reference Heber et al. Heber et al. 2007 was published on 15 February, and available online since December 2006, while Watson et al. was not submitted until the beginning of May 2007.

Veronika S. Heber, Richard A. Brooker, Simon P. Kelley, Bernard J. Wood. 2007 Crystal-melt partitioning of noble gases (helium, neon, argon, krypton, and xenon) for olivine and clinopyroxene. Geochimica et Cosmochimica Acta 71 1041-1061.

E. Bruce Watson, Jay B. Thomas, Daniele J. Cherniak 2007 40Ar retention in the terrestrial planets. Nature 449 299-304.

update: sociological speculations on these papers is here.


Yami McMoots said...

I tend to cut Nature papers some slack for yhandwavy language, since they're writing to such a stringent page limit. I also never read the supplemental info to find out if my slack is justified therein.

C W Magee said...

IT isn't the hand-wavy language that bothers me, it is the cavielier manner in which someone who disagrees with (and/or ignores) all previous research by 4-7 orders of magnitude simply declines to try proving himself wrong or explaining the discrepancy.

If I go back to teaching, maybe I should start telling my students to submit to nature if they get exam problem answers that are in the wrong ball park.

C W Magee said...

Note that Sean at cosmic variance has posted a much more eloquent explanation at why overturning established scientific models requires extra work. He talks about cosmology, not diffusion, but the issues are similar. His post is here.

Thermochronic said...

I am unfortunately unable to access the Heber et al paper (can tomorrow at work) but in the meantime...

I think characterizing the Watson paper as cavalier is a little misleading. I have no idea why they don't reference Heber, although I would not be surprised if it had more to do with Nature length requirements, and the focus being on the results at hand. They have a longer paper with much more data that is in press (that includes among it's reviewers at least one of the Heber co-authors), so we'll see then.

My take on why there is such a difference, and why the Watson study is believable is this. First, the results were consistent and reproducible, regardless of how they fit with results from other experiments, the data set itself is tough to argue with. Different experiments in different conditions gave consistent results, that alone is telling you something.

Secondly, I think one of the big differences is what was actually used to make the measurements. Watson et al make direct measurements of Ar that entered crystals via diffusion. Heber et al., from my understanding, grow crystals in gas rich melts. Do growing crystals incorporate noble gases in this manner? Is it the dominant mechanism in the mantle? If you are going for diffusion measurements, why not do them directly? I think the fact that two different experimental set ups give drastically different results is telling us not that the experiments are bad, but that there are different operational mechanisms at different stages of a crystal's life. The Watson study seems to address diffusion much more directly.

As I said, the data are tough to argue with. The crystals had no Ar, they were put in cold seal pressure vessels with varying Ar atmospheres at different temperatures for different amounts of time, and then diffusion profiles were directly measured using Rutherford backscattering. That appears to be a more direct method of measuring gas diffusion than the Heber method.

But, I'll look at the paper tomorrow and see what I think. Perhaps worthy of another blog post?

I doubt Nature lets you go into your experimental details too much (although I have yet to have that problem), and I'd hesitate at assigning that omission to Watson's group (based on their average), and anticipate the longer paper (Chem Geo I think) to cover those bases.

mbrennwa said...
This comment has been removed by the author.
mbrennwa said...

Will the detective novel by Heber et al become available as an audiobook? Maybe the GCA editors might consider releasing a selection of their novels in this format... :-)

Anonymous said...

The paper on Ar solubility and partitioning that was submitted to Chemical Geology that contained all the data from which the Nature paper is now available on line. One reviewer from Heber et al's group held up the paper for months and demanded lots more work. Anyways, give the new paper a read.

Here's the link:

If you read the paper you'll see that these authors have conducted the most comprehensive study of Ar solubility and diffusion out of all the studies thus far.

Anonymous said...

The link didn't work. So to find the paper, go to chemical geology and click the link to "Full Text in ScienceDirect". In ScienceDirect you'll find the paper in the Articles in Press link.

Have fun with this one.

Dr. Lemming said...

Unfortunately only the abstract is available to those of us in the private sector. But it looks interesting. Although it is a shame that they didn't also study any K minerals or melts to get direct D values or results testable through excess argon detemrination.