Did Microsoft Researchers FALSIFY the Data to Claim the Breakthrough In Building a Quantum Computer?
Retraction of the controversial Nature paper no one talks about outside of the scientific community.
When there is a breakthrough in the development of quantum computers, it is made sure the news is covered in the media and presented to the public. This is how it was also with Microsoft in 2018 when scientists after 6 years of hard work claimed to finally be able to see Majorana fermions, needed for a topological computer, pursued by Microsoft. It was all around in the news.
However, the paper became controversial not just because no other group in the world managed to reproduce their claims, but because their data analysis started being questioned. Three years later, the researchers finally changed their claims drastically, and this followed with the retraction of the paper from the journal.
Realistically, how far away is the future of quantum computers?
The idea of a quantum computer dates back to the early 1980s. The potential of quantum computers had become more apparent in 1994 when Shor’s algorithm was developed. This algorithm proves that on a quantum computer factoring into prime numbers can be done in polynomial time, which is almost exponentially faster than the most efficient known classical factoring algorithm on ordinary computers.
Because quantum circuits are much more susceptible to noise than classical ones, most experts until the late 1990s still believed the error correction in quantum circuits was practically unreachable.
There were some significant developments since, but the hype about quantum computers blew up in recent years with huge investments in quantum computing research, both in the public and private sectors, leading to the catchy term ‘quantum supremacy’ recently.
But to be true, while a quantum computer would be overpowering in solving some classes of problems, in solving many other types of problems it would not be any better than the computers we know.
Because of the infancy of quantum technology and no apparent advantage of one approach over another, it is interesting to follow the race of building hardware for a quantum computer, where the main competitors currently are focusing on implementing completely different physical ideas. This is in stark contrast with the computers of today, which are all built on the same semiconductor physics technology.
The frontrunner quantum technologies today are superconducting electronic circuits pursued by Google and ion traps technology spearheaded by IBM.
The current race is still not the race for a quantum computer, but just for the quantum bits, named qubits.
A commercial quantum computer would actually require tons of qubits working together. And the most challenging part is probably not the number of qubits, as much as it is wiring them in a logical unit.
Why Microsoft wants to build a Quantum Computer on still unproven theory?
While Google, IBM, IMEC, BBN Technologies, Rigetti, Intel, and other start-ups are pursuing qubits with more established technologies, Microsoft put its bets on the challenging idea of producing a topological qubit.
For a superconducting quantum computer to be built by Google, the building blocks for are Josephson junctions, theoretically predicted in 1962, experimentally seen even earlier but not correctly interpreted, and the discovery received the Nobel prize in physics in 1973. IBM's quantum computer is to be built using cold trapped ions. The methods to cool and trap atoms with laser light were developed in the late 1980s and mid-1990s, for which the Nobel prize was awarded in 1997 and the first ultracold atoms were trapped in their lowest possible energetic state in 1995, for which the Nobel prize in physics was given in 2001. Both of these technologies are quite established today. On another hand, Microsoft is developing an idea of having a topological quantum computer where building blocks are Majorana quasiparticles. While there were few milestones reached in this research important for the science in general, the Majorana quasiparticle is still not conclusively produced! As of today, there are about 50 superconducting qubits made by Google, about 50 ion-trapped qubits made by IBM, and 0 topological qubits.
The initial question could be rephrased, why would be a topological quantum computer that much better, that it’s worth a risk?
A topological quantum computer would be computationally no more powerful than any other quantum computer, but its huge advantage compared to other leading technologies, is its very low error rate. Quantum states are very delicate and much more prone to error than the classical states used in today’s electronics. They are easily destroyed or tampered with by temperature and electromagnetic noise.
Topological materials which could either be synthesized or engineered as a hybrid from different more available materials, have a special property that they contain non-trivial boundary states. Such topological boundary states are much more robust to the external noise and more stable.
Microsoft’s path to a topological quantum computer
The concept of a topological quantum computer, using so-called Majorana quasiparticles, was born in 1997 by Alexei Kitaev. He at the time worked at the Landau Institute near Moscow, and after his idea became popular, he joined Microsoft for 3 years. His publicly available draft, which was officially published full 7 years later in the journal Annals of Physics, very soon initiated experimental ideas for creating a set-up for producing these Majorana quasiparticles.
The first practical proposals came in 2010 from two different teams of scientists led by Sankar Das Sarma from the University of Maryland and Felix von Oppen from the Free University of Berlin. At about the same time, they suggested Majoranas will appear on the interface between a semiconducting wire and a superconductor.
Majorana as a concept was a long-time known in high-energy physics.
Hypothetical Majorana particles were theoretically predicted far back in 1937, but no such particle was experimentally found yet.
The new idea was that Majorana quasiparticles could be found in specially engineered materials, like when a superconductor is in a contact with a nanowire, a very thin wire in which electrons are stacked next to each other. Such a Majorana quasiparticle wouldn’t be an elementary particle, but in this topologically non-trivial set-up, on the contacts between the wire and a superconductor, the edge electrons will effectively behave like Majoranas.
The idea got a tremendous boost in 2012 when a group of scientists at the Delft University of Technology in the Netherlands led by Leo Kouwenhoven found the signatures of Majorana quasiparticles in a hybrid superconductor nanowire under magnetic field, at extremely low temperatures. In 2014, in collaboration with the Delft University of Technology, an additional research institute in Delft, named QuTech, was founded as a Dutch initiative to study quantum computing, quantum internet, and qubit research.
Microsoft got in contact with scientists in Delft way back in 2010 while they were already actively working in the field, but Microsoft officially joined QuTech collaboration in 2017.
In this set-up from 2012, the semiconducting nanowire in proximity to the superconductor, Majoranas which don’t have charge or spin, arise as zero-energy modes, one at each end of a wire, at the contact with the superconductor. Such modes are not easy to be measured and furthermore, there could be other simple explanations not involving the sought-after Majoranas.
One way of ascribing the origin of these zero-energy modes to Majoranas is to show that the conductivity is discretized, meaning it doesn’t have a continuum of values, but just some specific values. Establishing Majoranas compared to other more trivial explanations involves showing the robustness of the zero-energy modes to the temperature and electromagnetic fields.
In the article from March 2018 published by Delft group led by Kouwenhoven in the prestigious journal Nature, they claimed exactly that. They claimed they found definitive proof that the zero-energy modes originate from Majoranas and that what follows are braiding experiments that could lead to the topological quantum computer.
This paper was a big win for Microsoft with a press release named Majorana trilogy completed by TU Delft. In 2019 Kouwenhoven was named the Scientific Director of the newly opened Microsoft Quantum Lab in Delft, Netherlands.
The controversy around the 2018 paper
While many groups at other Universities reproduced the results published in a paper led by Kouwenhoven in 2012, a few groups tried to reproduce major claims in the paper from 2018, but without success.
The article underwent scrutiny as some groups started questioning the methodology of data analysis. A group from the University of Pittsburgh, doubting the claims from the paper, said that they asked for complete raw data obtained in the measurements for over a year now.
In April 2020, Nature editors published a warning, calling for caution against using the claimed results of the 2018 paper, and announced they actively work with the authors to resolve the issues.
Finally, on January 27th this year, a revised draft version of the now-controversial article by the original group became publicly available. They retracted their earlier claims of observing Majorana quasiparticles in a material due to the ‘technical errors’, as reported in the revised paper. The revision explains that the traces of Majoranas could be also interpreted as trivial zero-energy modes. This drastic change in claims will follow the retraction of the original paper published in Nature.
However, it looks like the problem is bigger than the ‘technical errors’. Almost 3 years after the initial article was published, the full data obtained in their experiment was published too. This initiated a heated discussion on Twitter about the manner in which the raw data was processed. Experimental physicist Sergey Frolov from the University of Pittsburgh, who was one of the collaborators on the breakthrough article from 2012, pointed out that it looks like some 2018 data was conveniently chopped off, in order for the main conclusions to hold.
Questioning research integrity
Papers are retracted from publications all the time, and often because of innocent mistakes with big consequences. Science is an always-evolving field with its steps forward and backward. However, this recent controversy has initiated debates and webinars among scientists on research integrity. It is suggested that it should become compulsory for the big discoveries appearing in prestige journals to publish all their data, as no such requirement exists today.
Lots of scientists feel the burden of responsibility for bad science is in the hands of editors who have the final word on publishing or not publishing an article after feedbacks from the scientific community. So far the editors who approve articles for publishing are not publicly named and it looks like they are not weighing on transparency or accountability. Additionally, there is also the role of private companies, which optimistically sponsor research and have ambitious plans to, in this case, produce a quantum computer in a timeline they came up with.
In the meantime, the TU Delft Integrity Committee has also started an investigation, with the goal of finding out whether ‘the research, data analysis, and the writing of the publication, were executed in accordance with the ethical guidelines’. Kouwenhoven didn’t give any statement about it.
Future of a topological quantum computer
Important to being said is that the retraction of the 2018 paper claiming the detection of Majorana zero-modes doesn’t belittle the idea of having Majoranas in the proposed system or reduces prospects of topological quantum computer built from Majorana quasiparticles. It just means no one has yet observed Majorana quasiparticles.
As an example, in the history of semiconductors, the agreement between theoretical predictions and experimental results was bad in the 1920s, until it was found that the properties change dramatically even with small amounts of impurities in samples. This resulted in efforts for producing cleaner samples ultimately leading to semiconductors being applied everywhere around us. Some scientists think sample cleanliness could be very easily the roadblock for detection of Majoranas at the moment.
Today the stakes are high, but to be true, we are still far away from having a commercial quantum computer. The hardware is still developed in baby steps, so in the race of having a functional quantum computer, no one is really that much behind others.
