A study that was once trumpeted as evidence for the existence of an exotic quantum state that could revolutionize computing has turned out to be anything but. A 2018 Nature paper1, based on work led by researchers at a Microsoft laboratory in the Netherlands, has now been officially retracted2 owing to what the authors call “insufficient scientific rigour” in the original data analysis.
The retraction is a setback for this approach to quantum computing, but scientists say it should still be possible to create and study the exotic states, known as Majorana fermions, that were the subject of the research. And researchers at Microsoft and elsewhere are still optimistic about the company’s plans to employ the phenomenon in a future quantum computer.
“Even if this is a step back, this is how we make progress,” says Ali Yazdani, a physicist at Princeton University in New Jersey.
Searching for Majoranas
Topological states have become one of the hottest areas of basic research in physics and materials science in the past decade. Theoretical physicists have predicted that in certain materials, the collective behaviour of electrons can display the properties of Majorana fermions, hypothetical elementary particles that could simultaneously be matter and antimatter.
The theory says that these collective quantum states would be topological, meaning that they would ‘remember’ how they moved around with respect to one another, in the same way that strings in a braid ‘remember’ how they were intertwined (topology is a branch of mathematics that studies knots and braids, among other things). This should make the Majorana states robust carriers of information, suitable for building a quantum computer that can do certain calculations faster than any ordinary classical computer can.
In practise, it has been devilishly difficult to build actual devices that can enable Majorana fermions to form. Researchers have tried various approaches, and Microsoft decided to bet heavily on one that had been pioneered by Leo Kouwenhoven, a physicist at Delft University of Technology in the Netherlands. In 2012, he had reported tantalizing hints of the existence of Majoranas to a standing-room-only crowd at a physics conference3. In 2016, the company hired Kouwenhoven — among other stars in the field — and tasked him with founding a Microsoft lab on the Delft campus.
Kouwenhoven’s approach involves trying to create Majorana fermions inside ‘hybrid’ nanowires, which combine a layer of a semiconductor material with a superconducting one laid on top of it. Theory predicted that at very low temperatures, thermal vibrations in the superconductor — which enable electrons to travel through such materials without any heat losses — also ‘leaked’ an effect on the semiconductor underneath. Under certain conditions, such as the application of a magnetic field, a pair of Majorana fermions, one at each end of the nanowire, should spontaneously form in the semiconductor.
No one yet knows how to detect Majorana fermions directly, but to at least prove they exist, Kouwenhoven and his collaborators looked for an effect that was considered a ‘smoking gun’: the electrical conductance of the nanowires’ tips should have a sudden peak as researchers apply a voltage and then dial it down to 0. This was the signature that the team reported in Nature. (Nature’s news team is editorially independent of its journal team.)
“This paper was considered a big deal,” says Sergey Frolov, a physicist at the University of Pittsburgh in Pennsylvania, whose subsequent work helped to trigger the retraction. “It was taken widely as, if not the ultimate proof, the strongest proof to date” for the existence of Majorana fermions, he says.
Based in part on his own attempt to measure the effect in his lab, Frolov suspected that the evidence presented in the 2018 paper did not tell the whole story. He points in particular to a study4 led in 2014 by physicist Silvano De Franceschi, another former member of Kouwenhoven’s team. De Franceschi and his collaborators had found that a separate, non-topological quantum phenomenon called Andreev states can, under certain conditions, mimic the experimental signatures that are expected in Majorana states. “Andreev states can explain a number of grand claims” made by several research groups over the years, says De Franceschi, who is now at the University of Grenoble Alpes in France.
To be sure that they have created a Majorana state, experimenters should repeat their measurement many times while slightly changing the experimental conditions, such as the strength of the magnetic field, Frolov says. In a paper published in Nature Physics in January5, Frolov and his collaborators describe an experiment that produced what looks like the coveted Majorana signature, but contradicts it when the conditions change slightly. “We could artificially create a combination of parameters to produce similar-looking data” to the plots in Kouwenoven’s Nature paper, Frolov says.
Frolov adds that, in November 2019, he obtained an extended set of data from one of the Nature paper’s authors. When he plotted these data, he found evidence that directly contradicted the central claim of the paper. Frolov and Vincent Mourik, another former Delft colleague who is now at the University of New South Wales in Sydney, Australia, then wrote to Nature to express their concerns about the paper, and in April 2020 the journal issued an editorial expression of concern.
In a preprint posted in January6, and in their 8 March retraction note, the authors say that, when the inconsistencies were pointed out to them, they re-analysed all the existing raw data from their original measurements and repeated the experiments with updated settings. When they did this, they found that there wasn’t the evidence to support their previous conclusion. “When the data are replotted over the full parameter range, including ranges that were not made available earlier, points are outside the 2-sigma [95%] error bars. We can therefore no longer claim the observation of a quantized Majorana conductance,” they wrote in the retraction notice. “We apologize to the community for insufficient scientific rigour in our original manuscript.”
Kouwenoven and Hao Zhang, the other corresponding author of the retracted paper who is now at Tsinghua University in Beijing, did not respond to requests for comment on Frolov’s analysis of their results.
How the problems with the original paper came about is still not fully understood. In May 2020, Delft University of Technology announced that their research-integrity committee was investigating “whether the research, data analysis and writing of the publication were executed in accordance with the applicable guidelines”. The committee appointed a panel of four external experts to review the experimental data and the paper. Their report, released on 8 March, said that the researchers had interpreted their data over-optimistically. “We found no evidence of fabrication: all data in the publication seem to be genuine results of measurements,” the report says. “However, the research program the authors set out on is particularly vulnerable to self-deception, and the authors did not guard against this.”
“If this were done intentionally, it clearly would have been a serious offense. However, based on the material made available to them and after discussions with the authors, the experts did not find evidence for intent,” the report says.
The broader investigation is ongoing, says Lieven Vandersypen, the director of research at the university’s quantum-technology institute. He says the staff has had a broader discussion on the lessons to learn from this incident. “Science always means being critical, doubting your results and discussing them.”
Microsoft is still committed to the topological approach to quantum computing. It remains to be seen whether Majorana states exist, and whether they can eventually beat other approaches that are much further advanced, researchers say. “This approach is particularly beautiful and elegant from the theoretical physics and mathematical point of view,” says Yasser Omar, a physicist at the University of Lisbon. “And it may still succeed.”
Nature 591, 354-355 (2021)