Microsoft’s Big Win in Quantum Computing Was an ‘Error’

In a 2018 paper, researchers said they found evidence of an elusive theorized particle. A closer look now suggests otherwise.

In March 2018, Dutch physicist and Microsoft employee Leo Kouwenhoven published headline-grabbing new evidence that he had observed an elusive particle called a Majorana fermion.

Microsoft hoped to harness Majorana particles to build a quantum computer, which promises unprecedented power by tapping quirky physics. Rivals IBM and Google had already built impressive prototypes using more established technology. Kouwenhoven’s discovery buoyed Microsoft’s chance to catch up. The company’s director of quantum computing business development, Julie Love, told the BBC that Microsoft would have a commercial quantum computer “within five years.”

Three years later, Microsoft’s 2018 physics fillip has fizzled. Late last month, Kouwenhoven and his 21 coauthors released a new paper including more data from their experiments. It concludes that they did not find the prized particle after all. An attached note from the authors said the original paper, in the prestigious journal Nature, would be retracted, citing “technical errors.”

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.”

The 2018 paper claimed to show firmer evidence for Majorana particles than a 2012 study with more ambiguous results that nevertheless won fame for Kouwenhoven and his lab at Delft Technical University. That project was partly funded by Microsoft, and the company hired Kouwenhoven to work on Majoranas in 2016.

The 2018 paper reported seeing telltale signatures of the Majorana particles, termed “zero-bias peaks,” in electric current passing through a tiny, supercold wire of semiconductor. One chart in the paper showed dots tracing a plateau at exactly the electrical conductance value that theory predicted.

Frolov says he saw multiple problems in the unpublished data, including data points that strayed from the line but were omitted from the published paper. If included, those data points suggested Majorana particles could not be present. Observations flagged by Frolov are visible in the charts in the new paper released last month, but the text does not explain why they were previously excluded. It acknowledges that trying to experimentally validate specific theoretical predictions “has the potential to lead to confirmation bias and effectively yield false-positive evidence.”

Tweets by Sergey Frolov, a physicist who questioned missing data in the 2018 paper.

Microsoft provided a statement attributed to Kouwenhoven saying he could not comment, because the new paper that reinterprets his group’s results is undergoing peer review. “We are confident that scaled quantum computing will help solve some of humanity’s greatest challenges, and we remain committed to our investments in quantum computing,” he said. Nature added an “editorial expression of concern” to the 2018 paper in April last year, and a spokesperson said this week that the journal is “working with the authors to resolve the matter.” A spokesperson for Delft Technical University said an investigation by its research integrity committee, started in May 2020, is not complete. A person familiar with the process says the final report will likely find that researchers at Delft made mistakes but did not intend to mislead.

Whatever happened, the Majorana drama is a setback for Microsoft’s ambitions to compete in quantum computing. Leading computing companies say the technology will define the future by enabling new breakthroughs in science and engineering.

Quantum computers are built from devices called qubits that encode 1s and 0s of data but can also use a quantum state called a superposition to perform math tricks not possible for the bits in a conventional computer. The main challenge to commercializing that idea is that quantum states are delicate and easily quashed by thermal or electromagnetic noise, making qubits error-prone.

Google, IBM, and Intel have all shown off prototype quantum processors with around 50 qubits, and companies including Goldman Sachs and Merck are testing the technology. But thousands or millions of qubits are likely required for useful work. Much of a quantum computer’s power would probably have to be dedicated to correcting its own glitches.

Microsoft has taken a different approach, claiming qubits based on Majorana particles will be more scalable, allowing it to leap ahead. But after more than a decade of work, it does not have a single qubit.

Majorana fermions are named after Italian physicist Ettore Majorana, who hypothesized in 1937 that particles should exist with the odd property of being their own antiparticles. Not long after, he boarded a ship and was never seen again. Physicists wouldn’t report a good glimpse of one of his eponymous particles until the next millennium, in Kouwenhoven’s lab.

Microsoft got interested in Majoranas after company researchers in 2004 approached tech strategy chief Craig Mundie and said they had a way to solve one problem holding back quantum computers—qubits’ flakiness.

The researchers seized on theoretical physics papers suggesting a way to build qubits that would make them more dependable. These so-called topological qubits would be built around unusual particles, of which Majorana particles are one example, that can pop into existence in clumps of electrons inside certain materials at very low temperatures.

Microsoft created a new team of physicists and mathematicians to flesh out the theory and practice of topological quantum computing, centered on an outpost in Santa Barbara, California, christened Station Q. They collaborated with and funded leading experimental physicists hunting for the particles needed to build this new form of qubit.

Kouwenhoven, in Delft, was one of the physicists who got Microsoft’s backing. His 2012 paper reporting “signatures” of Majorana particles inside nanowires started chatter about a future Nobel prize for proving the elusive particles’ existence. In 2016, Microsoft stepped up its investment—and the hype.

Kouwenhoven and another leading physicist, Charles Marcus, at the University of Copenhagen were hired as corporate Majorana hunters. The plan was to first detect the particles and then invent more complex devices that could control them and function as qubits. Todd Holmdahl, who previously led hardware for Microsoft’s lucrative Xbox games console, took over as leader of the topological quantum computing project. Early in 2018, he told Barron’s he would have a topological qubit by the end of the year. The now-disputed paper appeared a month later.

While Microsoft sought Majoranas, competitors working on established qubit technologies reported steady progress. In 2019, Google announced it had reached a milestone called quantum supremacy, showing that a chip with 53 qubits could perform a statistical calculation in minutes that would take a supercomputer millennia. Soon after, Microsoft appeared to hedge its quantum bet, announcing it would offer access to quantum hardware from other companies via its cloud service Azure. The Wall Street Journal reported that Holmdahl left the project that year after missing an internal deadline.

Microsoft has been quieter about its expected pace of progress on quantum hardware since Holmdahl’s departure. Competitors in quantum computing continue to tout hardware advances and urge software developers to access prototypes over the internet, but none appear close to creating a quantum computer ready for prime time.

Frolov, the University of Pittsburgh researcher, says the questions around Kouwenhoven’s 2018 paper leave the small field of physics dedicated to detecting Majoranas “wounded,” facing a potentially unpleasant comedown after a period of high expectations. “We have good science to do with reasonable expectations, not magical expectations,” he says. Frolov says the group should release the full raw data from its experiments for outside scrutiny.

Frolov worked through the extra data with Vincent Mourik from Australia’s University of New South Wales, who says he shares Frolov’s concerns. Both previously worked with Kouwenhoven at Delft, before he was hired by Microsoft, including on the 2012 paper on Majorana particles.

Sankar Das Sarma, a theoretical physicist at the University of Maryland who has collaborated with Microsoft researchers, believes the technology will eventually work, but it could take a while. He was a coauthor on both the disputed 2018 paper and the new version posted last month.

Das Sarma says new theories developed over the past few years show that the methodology used in 2018 could not conclusively establish the presence of Majorana particles anyway. Purer materials, more complex experiments, and a lot more scientific progress are all needed, he says.

How far off that puts a Microsoft qubit is unclear. Das Sarma says Majorana-based quantum computing may be at a stage comparable to 1926, when the first patent for a transistor was filed. It took until 1947 for researchers to create the first working transistor; the miniaturizable silicon versions that enabled the computing industry were not developed until the late 1950s. “I see no reason why a Majorana fermion cannot exist or that once it exists you cannot control it,” he says. “But it may be 30 years away.”

More Great WIRED Stories

Products You May Like

Articles You May Like

Meta’s New Llama 3.1 AI Model Is Free, Powerful, and Risky
Spotify, Stop Trying to Become a Social Media App
Huge Microsoft Outage Linked to CrowdStrike Takes Down Computers Around the World
Waymo Is Suing People Who Allegedly Smashed and Slashed Its Robotaxis
Donald Trump and Silicon Valley’s Billionaire Elegy

Leave a Reply