The Fallen Angel Particle

At long last, the ‘Chiral Majorana’ paper is retracted from Science. It took 4 years. I was not on the team led by Laurens Molenkamp that identified the multitude of problems with it. But I have seen the evidence uncovered by the original investigators, and I repeated and confirmed  many of the steps. I will not share at this time what the evidence is, hoping that the investigators will do so themselves. It is important for everyone to know what they found. Because it is brilliant work. And because we can all learn from them how to look at data with our eyes open, and how not to give into the lull of fairy tales. 

_{‘Chiral Particle Fallen Angel’ courtesy of Dall-E Mini}

I can attest that the evidence is overwhelming and goes to the basic level of whether the data ever existed? It didn’t. The entire paper is made up. The most positive thing I can say is that some of the figures were inspired by real data. But this is actually in itself a horrible thing – by imitating the overall signal shape from previous works, they made it more believable to some.

The data in the paper look remarkably similar to the theory predictions of the late Shoucheng Zhang. In fact, Zhang himself joined the paper as a co-author, which indicates he was impressed. He took on the task of promoting this work, referring to it as ‘The Angel Particle’, in analogy to the Higgs boson, known as ‘The God Particle’. Because we know that the similarity of the data to his theory was a result of data fabrication, we can say for sure that his predictions have not, in fact, been realized.

Many others were not impressed from the start. When the paper first appeared on arXiv, my own reaction, after looking at it for ten minutes, was ‘No Way.’. Others, such as Jay Sau, Xiao-Gang Wen, and their collaborators put forward simple Kirchhoff-level arguments why the claim cannot be right. This is one common folly of people who try to fabricate or falsify results – if they imitate physics that is itself wrong, they are likely to be scrutinized and found out. This was the same for the Delft ‘Quantized Majorana Conductance’ where my friend and collaborator Vincent Mourik and I precipitated the retraction. Despite that paper being accompanied by its own theory, from the University of Maryland, which we also could not verify, the physics behind it was simply wrong. Majorana is not expected to be quantized in nanowire devices, surely not in the short and imperfect ones that the Delft group had (See this paper, Figure 5 and discussion).

Nevermind this pushback, Kang Wang, the last author of the freshly retracted paper, received praise and accolades for his work. For example, he won the Neel Medal literally for the discovery of Chiral Majorana, a discovery that was entirely fabricated. Interestingly, despite Science knowing all the facts of the case, its publisher AAAS kept Kang Wang’s appointment as an editor at Science Advances, a sister journal to Science.  I was shocked when I received a referee request signed by him. How can AAAS allow a person who had published a fabricated paper in Science make decisions on what to publish, for the same organization?

Setting aside Kang Wang, it is disappointing that so many other co-authors also refused to retract this obviously fabricated paper. The three co-authors that agreed with the retraction demonstrated integrity and bravery.  Of the authors who refused to sign, I am particularly perplexed by Jing Xia of UC Irvine. His involvement in the paper was actually fairly peripheral. His group offered Kang Wang the use of a dilution fridge and helped set up the measurement. As I understand it, the UCLA authors quickly submitted  results for publication, and at the time other authors trusted them implicitly. 

But given that the raw data supporting the claim of chiral Majorana were supposedly taken in Xia’s lab, it should have only taken him a few hours of looking into the matter, to figure out that the claims he had signed-off on, were fabricated. Four years later, he still does not have enough courage or conviction to remove his name. Extremely disappointing. 

All that said, the behavior of individual authors pales by comparison with the absolutely shameful handling of the matter by UCLA and UC Irvine. When first approached, and shown damning evidence of data fabrication, the deans at both schools, both physicists, quickly proclaimed that they had investigated, and found no problems. Editors at Science then decided that their hands were tied. They could not fathom taking the initiative to perform an editorial retraction, i.e., one made without the approval of all authors or the institution. Four years later, an editorial retraction is now exactly what they have done.

The editorial decision, and the lack of mention of any institutional investigation in the retraction notice,  suggest that UCLA has not made progress with its investigation. It is appalling that the University of California, the largest state university, has utterly failed to assure the accuracy and integrity of their research record. They should apologize to the public, and compensate the government for funds and efforts that were wasted because of their failure.

To Science’s credit, during the time that they were dragging their heels on editorial retraction,  they published an attempt to reproduce the chiral Majorana result. No Majorana plateaus were found in data that an independent group acquired on similar samples. This work was led by the very talented Cui-Zu Chang and his collaborators at Penn State, and Wurzburg. It is fairly extraordinary for Science, to publish a straightforward repudiation of their own paper, as a separate paper. Of course, Science has a ridiculous policy that it will only consider technical comments responding to prior papers for 3 months after publication – making it impossible to publish experimental reproduction studies in our field as comments. They take a lot longer. 

There is some degree of justice in Science’s decision. These researchers spent time and money trying to repeat Kang Wang’s claims. Perhaps they believed in them and wanted to have their own Chiral Majoranas. Or they may have felt a duty to clear up this topic. 

I am hopeful that editors at Science are moving in a positive direction, in trying to improve quality control. They also recently published a paper led by Georgios Katsaros of IST Austria that demonstrated with exceptional clarity the absence of any Majorana in full-shell nanowires. This was a repudiation of an earlier claim, also in Science, that was published in 2020 by Charlie Marcus and others at Microsoft and several universities. Vincent Mourik and I found that their own additional data disproved the claims in their paper. Marcus’s paper is under the Editorial Expression of Concern by Science since July 2021. The Chiral Majorana was under Expression of Concern between December 2021 and today, November 17 2022, the day it was retracted.

On reproduction, Nature is not taking as fair an approach as Science. The editors only permit reproduction studies in lesser journals, a step or two down their vast pyramidal scheme. For example, Quantized Majorana Conductance was published, and then retracted, from Nature – but our reproduction study of it was only offered a spot in Nature Physics. The first missing Shapiro step paper was published in Nature Physics by a group from Purdue, but a negative reproduction study from NYU was published in Nature Communications. There are other examples. This is like a newspaper publishing corrections to a front page article on the back page. On the other hand, Nature does better than Science on their retraction process, having already retracted two Majorana papers (1 and 2), and one on room-temperature superconductivity. All of them still took too long, with most of the time taken by the editors just sitting on it.

Back to Kang Wang. He has written a blog in Chinese, in which  he basically argues that his samples are ‘better’ or at least ‘different’ to those at PSU. This is one of the commonly used arguments that appears sensible to people accused of unreliable research. They either try to argue that they or their experiments are ‘better’, known as the ‘virtuoso defense’ or they claim that their experiments were done differently – usually they amplify minor differences that play no role in conclusions. It is often easy to see through this.

Another step that the accused take is they often claim that they have since followed-up and confirmed their own findings that were found unreliable. Kang Wang did this – he put a paper on arXiv containing ‘even better’ conductance quantization than what was fabricated for Science on Chiral Majorana. As far as I know that paper has not been published.

The Delft first author Hao Zhang had more success with this method. He submitted to Physical Review Letters a paper with ‘even better’ quantization of his ‘Majorana’ ‘plateau’. Editors at PRL received critical feedback, but chose to publish the paper, knowing fully well that the physics behind it is not valid and that the same author had to retract the same claim from Nature.

Nature played a role in making it possible. They published a misleading retraction notice on ‘Quantized Majorana Conductance’. In it, they allowed the authors to say that the problems were limited to ‘signal miscalibration’ and they omitted acknowledgement of inappropriate data selection. (Another problem with retractions initiated by  authors, is that authors can choose the wording to their advantage).  Of course, if the only problem were calibration, an inadvertent error, then another paper with ‘calibration done properly’ could be valid. Still, the PRL  editors actually knew that the real problem was non-representative data selection. The value of conductance was not quantized. The authors simply selected from values both greater and smaller than the ‘special’ value.

Since Physical Review Letters has agreed to publish their ‘quantized within 5%’ paper, I showed the PRL editors data from my group containing the same features as in Hao Zhang’s PRL, but where the quantization is half of what naive theory says it should be. These are half-quantized plateaus like those in Kang Wang’s paper, but in a Majorana nanowire. I find it plausible that I could spin a story that this is a physics breakthrough and persuade editors at PRL to send it for review. Perhaps this is the real chiral Majorana? In our experiments we can find any value of conductance, and fit it to any theory – this fact is well-established by now.

The UCLA Science paper is not the only paper by the same key authors that should have been investigated. For example, editors at Physical Review Letters have also seen evidence that another paper by Qing Lin He and others from Kang Wang’s group contains manipulated data. They have shrugged it off. I am a lifetime member of APS and I am extremely dismayed by this. I did renew my AAAS membership yesterday, because with this editorial retraction, it is clear that at Science, published by AAAS, some editors are trying to do the right thing, even if they act with delays that cannot be justified.

The Editor-in-Chief of Science recently published an editorial in which he proposed changes to the process for handling challenged results. He wants to separate the question of scientific validity from the question of research integrity. Oftentimes, the journal simply wants to know if the paper is accurate, but the university investigates whether misconduct has occurred which takes a long time and holds back retractions of unreliable results. I agree with the editors’ proposal, because it can allow them to move ahead quickly with retractions as soon as they establish results are invalid. Any evidence of misconduct should still be pursued, and retraction should not be the only goal.

Editors at journals can and should still do a lot more. Demand full data before acceptance, retract papers where the authors refuse to provide data, send paper for independent post publication review if concerns are raised, the journal should not depend on universities which have an obvious conflict of interest. The editors have a system for reviewing and publishing. They also need a system for investigating. 

Some of these measures could help catch unreliable work before it is officially published, while others will aid in correcting the record more efficiently. In the meantime, we should brace ourselves for more revelations and retractions, on this topic and many others, across all disciplines and fields of science.