How Microsoft's Quantum Boast Went Bust

Illustration for article titled How Microsoft's Quantum Boast Went Bust
Photo: Ray Massey (Getty Images)

Microsoft is an outlier among the companies investing in quantum computing research. Unlike Google, IBM, or the handful of startups that have built noisy experimental prototypes out of superconducting circuits, ions, or photons, the company is trying to build a quantum computer by using objects known as Majorana particles—distinctive patterns of electrons in a tiny wire that proponents claim have fundamental advantages over rival designs.

Advertisement

The catch? Nobody has ever been able to prod electrons into forming a Majorana particle. Now, the outlook seems even bleaker: Earlier this month, Microsoft-affiliated researchers retracted a heavily publicized 2018 journal article in Nature that claimed strong experimental evidence that they’d created the particle. The paper, which Gizmodo covered at the time, underwent Nature’s peer review process, in which two to three expert reviewers who are anonymous to the authors recommend a manuscript to be published, rejected, or revised. “We apologize to the community for insufficient scientific rigour in our original manuscript,” the authors wrote in the retraction. The signal they’d presented as the signature of a Majorana particle suffered from a measurement error, invalidating their results.

The University of Delft in the Netherlands, which is the home institution of Leo Kouwenhoven, the physicist and Microsoft employee who led the experiment, also conducted an independent investigation of the team’s work. In a report published on the day of the retraction, the investigation team found that Kouwenhoven’s group selected their data in a biased way, such that their measurements looked more convincing. (Kouwenhoven did not respond to a request for comment.)

The investigation found that the team didn’t intend to mislead. “They were kind of sloppy,” said physicist Patrick Lee of the Massachusetts Institute of Technology, who was part of the independent investigation. “I can’t find a better way to describe it.”

The authors had designed their experiment based on earlier theoretical papers. These papers predicted that, under the right conditions, two Majorana particles, each behaving like half an electron, should form on both ends of a semiconductor wire wrapped in a shell made of a superconductor. To make a qubit—the fundamental building block of a quantum computer—you could then encode information by swapping the positions of the two half-electrons on the wire, in a process likened to braiding hair. Swapping right over left could represent 1, and left over right could represent 0. A device made from Majoranas is known as a topological quantum computer. Because this information is encoded in the orientations of the two particles and not as inherent properties of the particles themselves, a topological quantum computer is supposed to be less prone to errors than existing qubit designs. However, no one has yet managed to create a topological qubit, let alone a computer.

Despite the involvement of big-name consumer tech companies, quantum computing is still largely a research field. While some companies have produced small prototype devices, these quantum computers cannot solve useful problems. One of their main limitations is that they cannot execute the most perfectly designed algorithm without committing errors, and experts do not know how to correct those errors.

Advertisement

The University of Delft report suggested that the authors were so motivated to find a Majorana particle that they deceived themselves into thinking they’d seen it. The investigators quoted the physicist Richard Feynman: “The first principle is that you must not fool yourself—and you are the easiest person to fool.”

Doubts about the work began as early as November 2019, when physicist Sergey Frolov of the University of Pittsburgh found he could not replicate the paper’s experimental results. Partnering with Vincent Mourik of the University of South Wales, Frolov requested Kouwenhoven’s team share their data, and they found that the original paper contained improperly cherry-picked data. “It became clear that there was no justification for their claims,” said Frolov. Frolov and Mourik alerted the authors and Nature, and their analysis spurred the independent investigation and ultimately the paper’s retraction on March 8, 2021.

Advertisement

A spokesperson for Nature said in a statement, “We are committed to updating the scientific record when appropriate, in order to provide clarity to our readers, and endeavour to do so as soon as we have enough information to determine the best course of action. However, these issues are often complex and as a result, it can take time for editors and authors to fully unravel them.”

The retraction is a “wake-up call” for the researchers and for the community to be more careful in publishing their experimental results, said Lee.

Advertisement

But this paper’s demise does not doom topological quantum computing, according to Lee. “If you read the popular press, you get the idea that this [retraction] was a showstopper, that Microsoft has fallen on its face, and the investment is a failure,” he said. “I think that is not correct.”

In the paper, the physicists conducted a much more difficult version of an experiment commonly performed in an introductory physics class: applying a voltage across a wire and measuring its electrical resistance. In their case, they used a nanowire, several hundred times finer than a human hair, made of indium antimonide wrapped in superconducting aluminum and kept extremely cold, near absolute zero. (Technically, the team measured the material’s conductance, which is just the number 1 divided by the resistance.)

Advertisement

According to some theoretical predictions, when the nanowire’s electrons form a Majorana particle, the nanowire’s conductance should plateau at a certain value as you lower the voltage across the device to zero. The 2018 paper claimed to observe this plateau.

Then, some members of the team told the public they’d made a Majorana particle. “Now, the scientists provide a definite proof for Majorana existence paving the way towards Majorana quantum bits,” read a press release that the University of Delft issued when the paper was published. “This experiment closes a chapter in the quest for Majorana particles.” Soon after the paper’s publication, Julie Love, Microsoft’s director of quantum computing business development, told the BBC that the company would have a commercial quantum computer “within five years.” Most physicists treated the paper’s results as “a smoking gun” for the Majorana particle, said physicist David Goldhaber-Gordon of Stanford University, who was a part of the University of Delft investigation team.

Advertisement

But the plateau wasn’t a definitive sign of the Majorana at all—and some physicists knew that. Electrons behaving in other ways could also exhibit this plateau. Some physicists had even proposed that the Majorana wouldn’t cause a plateau at all, said Frolov.

In other words, the team and the ensuing media coverage overhyped the result. “In my view, this was not a significant paper, even if it were correct,” said Frolov.

Advertisement

Frolov is concerned about what the negative publicity means for the rest of the field. “This kind of retraction can precipitate negative things for the entire field, like canceled grants,” he said. One of his grant proposals was denied this January because a reviewer said that the experimental technique he uses—the same one employed by the authors of the Majorana paper—has been discredited, he said. “Nothing is wrong with the technique,” said Frolov.

The retraction implicates the authors, not their underlying strategy. “I have basically no doubt, that when the right ingredients are put together, that the Majorana should exist,” said Goldhaber-Gordon.

Advertisement

The paper and subsequent retraction offer a case study of how the scientific process actually plays out in the real world. Arguably, in this instance, the process worked. The truth ultimately came to light: Kouwenhoven’s team retracted their paper and explained what went wrong. The episode has also sparked new science. This January, Frolov published a paper in Nature Physics detailing how his team could recreate the plateau via a different electron phenomenon. Physicist Sankar Das Sarma of the University of Maryland, one of the coauthors on the retracted paper, has recently released new theoretical work indicating that the experiment requires materials with far fewer impurities to create a Majorana.

“This is the best example of the scientific process that I have seen in my life,” said Das Sarma. (Das Sarma worked on the theory section of the paper, which Lee confirmed was not a focus of the University of Delft’s investigation.)

Advertisement

But the retraction also shows that the scientific process is “fragile,” said Goldhaber-Gordon. Very few people have the expertise to even catch the group’s errors. “A point of danger in our scientific system is that it’s very hard to evaluate other people’s claims,” he said.

Frolov and Mourik could evaluate the retracted experiment because they used to work with Kouwenhoven. But even with their expertise, the process was time-consuming and stressful. “We are trying to make the scientific process work, and it is very hard,” said Frolov.

Advertisement

In choosing to expose the group’s mistake, Frolov and Mourik—who are less established in their careers than Kouwenhoven—also had to lay their professional reputations on the line. Complicating matters further, Frolov said that Kouwenhoven helped him in his early career. “He played a huge role in my life,” said Frolov. “He boosted my career by letting me work in his group.”

Now, his relationship with his former mentor hangs in limbo. “In November 2019, we met at a conference. We laughed; we drank beer; it was all good,” said Frolov. “And now I can’t imagine this happening again.”

Advertisement

“It took courage and a lot of work for [Frolov and Mourik] to come forward and push this,” said Lee.

Frolov is planning to cross-check another experiment, with the hope that it will deter others in the field from further sloppiness.

Advertisement

Microsoft appears to be sticking the course. “We remain confident in our topological approach to scaled quantum computing,” wrote Zulfi Alam, the vice president of Microsoft Quantum, in a statement on LinkedIn.

Das Sarma compares the pursuit of the Majorana particle to other fundamental physics discoveries. It took physicists only 15 years to discover W and Z bosons and 100 years to measure a gravitational wave after theorists predicted each of them to exist. “How long will it take? Honestly, I don’t know,” said Das Sarma. “I don’t want to make up a number.”

Advertisement

Sophia Chen is a science journalist covering physics and data science.

DISCUSSION

One of their main limitations is that they cannot execute the most perfectly designed algorithm without committing errors, and experts do not know how to correct those errors.

I thought the errors were inherent to quantum computing, which is probabilistic vs the deterministic conventional computing. Not “1" or “0", but “Eh, it’s probably close to 1? Definitely at least 0.8".