The authors of a Nature paper that could have meant a great leap forward for Microsoft’s computing power are retracting it today after other researchers flagged serious problems in the work.
The researchers, led by Leo Kouwenhoven, a physicist at Delft Technical University in the Netherlands who is also employed by Microsoft, published “Quantized Majorana conductance” on March 28, 2018. Along with work at other labs, the paper, which claimed to have found evidence for a long-elusive particle known as a Majorana fermion, prompted this quotation in a BBC story:
“We will have a commercially relevant quantum computer – one that’s solving real problems – within five years,” says Dr Julie Love, Microsoft’s director of quantum computing business development.
But researchers who asked for data related to the study quickly identified potential issues. The paper received an expression of concern last April, and as part of a preprint acknowledging their errors, the authors announced they would retract the paper on January 27 of this year, as WIRED reported in February. Per WIRED:
Two physicists in the field say extra data Kouwenhoven’s group provided them after they questioned the 2018 results shows the team had originally excluded data points that undermined its news-making claims. “I don’t know for sure what was in their heads,” says Sergey Frolov, a professor at the University of Pittsburgh, “but they skipped some data that contradicts directly what was in the paper. From the fuller data, there’s no doubt that there’s no Majorana.”
Here’s the retraction notice, which credits Frolov and others:
In this Letter, we reported electrical measurements and numerical simulations of hybrid superconducting–semiconducting nanow-ires in a magnetic field. We reported plateaus in the conductance at 2e2/h, which we interpreted as evidence for the presence of Majorana zero-modes. However, several inconsistencies were pointed out by Sergey Frolov and Vincent Mourik between the raw measurement data that was made available to them and the figures that were published in the paper. We therefore re-analysed all the existing raw data for our original measurements and rebuilt the original experimental set-up for a re-calibration of the conductance values. We established that the data in two of the figures (Fig. 2a and Extended Data Fig. 4b) had been unnecessarily corrected for charge jumps (corrections that were not mentioned explicitly in the paper), and that one of the figure axes was mislabelled (Fig. 4b). The new conductance calibration shifted the plateau values by 8 per cent, above 2e2/h, which affects all the figures1. When the data are replotted over the full parameter range, including ranges that were not made available earlier, points are outside the 2-sigma error bars. We can therefore no longer claim the observation of a quantized Majorana conductance, and wish to retract this Letter. After informing Nature of this decision, Nature issued an Editorial Expression of Concern2 and initiated the retraction process. In ref. 1 we provide all the raw data underlying the published figures as well as the unpublished datasets. Ref. 1 also contains the analysis methods and a side-by-side comparison between the original and the corrected figures. In ref. 3 we provide a new manuscript with corrected and extended datasets, discussed in the context of new insights on zero-energy states in systems with inhomogeneous potentials and disorder. We thank Piet Brouwer, Klaus Ensslin, David Goldhaber-Gordon and Patrick Lee for the expert evaluation report available via ref. 1. We also thank Michael Wimmer and Bernard van Heck for their help with the analyses. We apologize to the community for insufficient scientific rigour in our original manuscript.
The paper has been cited more than 300 times, according to Clarivate Analytics’ Web of Science.
Update, 2300 UTC, 3/8/21: Frolov and Mourik, who posted their analysis here, issued this statement:
The retraction note published today in Nature misrepresents the sequence of events. It omits that the two of us have contacted Nature, not the authors. We took this step after our communication with the authors exhausted itself. Only after Nature asked the authors for explanation have they made their decision to retract.
We are pleased that our findings are supported by the experts recruited by Delft University of Technology. The expert report puts “plateau” in quotes. The original claim of a “robust” conductance plateau is a result of selecting a dataset with irreproducible artefacts. The retraction note does not directly acknowledge data selection issues.
We praise the authors for making additional data public on Zenodo. This allows us to share what we found a year ago in early 2020. We will be explaining our findings to the public in the coming days.
We request that Delft University of Technology provides access to data from other related experiments. We wish to independently re-analyze those results and verify conclusions. Access to data is guaranteed by the Netherlands Code of Conduct for Research Integrity.
Like Retraction Watch? You can make a tax-deductible contribution to support our work, follow us on Twitter, like us on Facebook, add us to your RSS reader, or subscribe to our daily digest. If you find a retraction that’s not in our database, you can let us know here. For comments or feedback, email us at [email protected].