Communications of the ACM
Majorana Meltdown Jeopardizes Microsoft's Quantum Computer
The creation and detection of Majorana particles in 2018 was probably no more than overinterpreted noise. Lead scientist Leo Kouwenhoven's integrity is now under investigation.
Credit: Alexey Drjahlov
In the race to build a large-scale quantum computer, Microsoft invested heavily in a scientific breakthrough by Dutch physicist Leo Kouwenhoven, then at QuTech in Delft, the Netherlands. Kouwenhoven first reported evidence of Majorana particles in 2012, and transferred from QuTech to Microsoft in 2016 on that ticket. In 2018, his group claimed in Nature to have actually detected these elusive particles in a nano device they designed .
Kouwenhoven's mission for Microsoft is to build a large quantum computer based on Majorana qubits. Though other types of qubits already function in small quantum computers, for instance at IBM, Google, and at QuTech unit Quantum Inspire, Majorana qubits would be more stable and easier to manage in large numbers. Qubits are the building blocks of a quantum computer. They are nano-structures or single particles that can exist in a superposition of two states (0 and 1), and entangling many qubits enables the quantum magic of exponentially speeding up certain types of computations.
However, Microsoft's hope of leap-frogging the quantum competition at IBM and Google has received a potentially lethal blow. In the fall of 2019, Sergey Frolov and his group at the University of Pittsburgh tried, and failed, to replicate Kouwenhoven's Majorana results, and have been communicating with him about this for months. On April 29, 2020, Nature issued an 'editorial expression of concern' about the 2018 article. A month before, a Delft University commission on scientific integrity has started an investigation which is still ongoing. On February 9 of this year, Kouwenhoven's group announced in a preprint on ArXiv in which they would formally retract their 2018 Nature article (the retraction happened March 8).
Nobody denies Kouwenhoven is a brilliant scientist who has done important work in this field, but the question now looming is: did he turn a blind eye to nagging suspicions that his Majoranas were not real? The pressure must have been enormous, because Microsoft had invested tens of millions of dollars in his Station Q Delft laboratory dedicated to Majorana research.
Even if confirmed, these Majoranas are not particles that would ever show up independently in a detector. Formally named Majorana Zero-Modes (MZM), they are quasi-particles — a quantum phenomenon emerging from the collective behavior of many electrons in a partly superconducting nano-device similar to a transistor. Detection is indirect, and hinges on measuring the electrical conductance of the device while varying a magnetic field.
Very few laboratories in the world can do these experiments. When Frolov — who used to be part of Kouwenhoven's research group at Qutech — tried to replicate the experiment, he found the purported Majorana signals could just as well be the result of erratic, trivial behavior of the device. After much back and forth, Kouwenhoven's group finally did a re-analysis of its 2018 data, from which they concluded: "Our experiments cannot exclude the ugly regime of random disorder." Put bluntly: they admit these Majoranas were probably noise.
Frolov also noticed in the 2018 Nature paper that an important graph had been doctored. Kouwenhoven admitted first author Hao Zhang had done this, but only "for esthetic reasons." While not by itself proof of unacceptable data manipulation, it is suspicious that this was not mentioned in the paper.
Of course, this was not a one-off experiment; Majoranas have been Kouwenhoven's research mission for years. In 2012, his group reported in Science to have found 'signatures' of Majoranas for the first time. "It was the original paper that started this field," says Frolov. He does not think this paper raises any data manipulation concerns, "though it most likely did not find Majoranas."
Kouwenhoven's research publications after 2018 make no specific claims of detecting Majoranas. But it seems odd that his group at Station Q Delft, who must have been doing research all the time with this same kind of superconducting nano device, never noticed any problems with the Majorana signals.
Said Frolov, "We have asked for access to data from prior research, but so far we were denied access. I do not have evidence that any papers (other than Nature 2018) require action such as retraction or correction or comment. But the authors are mandated by the Netherlands Code of Conduct for Research Integrity to provide access to data, so they are in violation."
Lex Bouter, professor of research integrity and one of the authors of the Netherlands Code of Conduct for Research Integrity, agrees: "There seems to be no valid excuse in this case for not releasing these data." Supposing Kouwenhoven ignored warning signals in his own research after 2018; does failing to report this constitute research misconduct? According to Bouter, the Code has no explicit rule about this, but it states in general that scientists should be 'open about uncertainties', which implies such signals should be taken seriously and may necessitate 'honorable self-retraction'.
Reports from Dutch integrity commissions are not made public by default, though the conclusions are later published in anonymized form. Bouter, who is not involved in this investigation, said, "Sometimes full reports are published, either as a sanction or to clear the name of the accused."
If Kouwenhoven, Frolov, or the Delft university board disagree with the commission's conclusion, they can ask LOWI, the independent Netherlands Board on Research Integrity, for a second opinion. If that still does not resolve the issue, either party can sign up for a protracted legal fight in the regular courts.
Arnout Jaspers is a freelance science writer based in Leiden, the Netherlands.
No entries found