Retraction Watch

Tracking retractions as a window into the scientific process

Physicists retract Nature paper on Earth’s core after findings aren’t reproducible

with 5 comments

cover_naturePhysicists have retracted a highly cited paper from Nature on the behavior of electrons at the center of the Earth after other researchers could not reproduce their findings.

The 2015 paper earned coverage in Science News and Live Science, where co-author Ronald Cohen explained:

There was a big problem in how you generate a magnetic field, and now, because of our results, that problem has basically gone away.

Here are more details about what the original paper claimed, courtesy of a press release from The Carnegie Institution for Science, where co-authors Peng Zhang and Cohen work:

The center of the Earth is very hot, and the flow of heat from the planet’s center towards the surface is thought to drive most of the dynamics of the Earth, ranging from volcanoes to plate tectonics. It has long been thought that heat flow drives what is called thermal convection — the hottest liquid becomes less dense and rises, as the cooler, more-dense liquid sinks — in Earth’s liquid iron core and generates Earth’s magnetic field. But recent calculations called this theory into question, launching new quests for its explanation.

In their work, Cohen and Zhang, along with Kristjan Haule of Rutgers University, used a new computational physics method and found that the original thermal convection theory was right all along. Their conclusion hinges on discovering that the classic theory of metals developed in the 1930’s was incomplete.

The retraction note for “Effects of electron correlations on transport properties of iron at Earth’s core conditions” explains the error that felled that conclusion:

In this Letter we reported density functional theory plus dynamical mean-field theory (DFT + DMFT) computations of the resistivity from electron–electron scattering at the conditions of Earth’s core, and found that the electron–electron scattering was about the same magnitude as the conventional electron–phonon scattering, giving a total resistivity that was sufficient to allow a classical thermal-convection-driven dynamo. However, L. Pourovskii, J. Mravlje, S. Simak and I. Abrikosov could not reproduce our findings, which led us to re-examine our computations. We found an error of a factor of two that is due to our neglect of spin degeneracy (two electrons per band), which would halve the electron–electron resistivity and probably make the electron–electron scattering insignificant for the geodynamo, at least for pure iron. We therefore wish to retract this Letter.

The smaller electron–electron scattering supports the high conductivity of iron that was predicted1from electron–phonon density functional calculations. However, preliminary calculations show that using the exact double counting2 recently developed for the DFT + DMFT method increases the electron–electron scattering. It is also probable that the Wiedemann–Franz law, assumed in our previous work, is not followed or has a non-constant Lorenz number in liquid metals3 or correlated systems4. Whether the resulting conductivity is consistent with a geodynamo driven by thermal convection requires further detailed calculations; the results will be reported elsewhere. The results and conclusions in the Letter that refer to resistivity at low temperatures (in Fig. 2b), and scattering rate and electronic structure (in Fig. 3) remain valid.

The paper has been cited 17 times, according to Thomson Reuters Web of Science.

We always appreciate it when authors are transparent about what exactly went wrong and act promptly to fix the record, so we’re designating this an example of “doing the right thing.”

This isn’t the first time the authors have had to correct the record — we’ve found six corrections for Cohen, which note typographical errors and issues with calculations, for example:

Last author Kristjan Haule has two corrections:

We’ve reached out to Haule and Cohen, and will update this post with anything else we learn.

Like Retraction Watch? Consider making a tax-deductible contribution to support our growth. You can also follow us on Twitter, like us on Facebook, add us to your RSS reader, sign up on our homepage for an email every time there’s a new post, or subscribe to our new daily digest. Click here to review our Comments Policy. For a sneak peek at what we’re working on, click here.

  • tekija April 14, 2016 at 3:03 pm

    Do 17 citations a “highly cited paper” make?

    • genetics April 15, 2016 at 12:47 pm

      Yes, within a year that’s not bad. Plus that very much depends on the topic. They publish a lot in Physical review B, which according to the journal is “ranked number one in total citations in condensed matter physics”. That journal has an impact factor of 3.736. So in the highest ranking condensed matter physics journal, a paper is on average cited less than 4 times within the first 2 years after it’s year of appearance. So yes, in that field I would say 17 citations in one year make it a highly cited paper.

    • Anonymous April 15, 2016 at 2:50 pm

      Indeed, this study had a broad impact for geophysics, as it was contradicting the finding by Alfe and coworkers that the conductivity of iron in Earth core was higher than previously thought, which suggested that previous deep-earth model were inaccurate and that there was a missing energy source for the earth core. No too many citations, but quite important indeed. It is too bad that Zhang, Cohen & Haule and they did not check more carefully their calculations before publishing their “story”.

  • Narad April 14, 2016 at 8:12 pm

    Even though I’m many rungs down on ladder, it’s comforting to see that I’m not the only one with a penchant for revisiting the final versions of my own work.

  • Narad April 14, 2016 at 8:13 pm

    ^ “the” ladder

  • Post a comment

    Threaded commenting powered by interconnect/it code.