Maarten van Kampen wrote a great article a few months ago ago, about a high-profile Nature retraction by the US-based businessmen and university professors Ranga Dias and Ashkan Salamat. In short, he wrote how sleuths exposed elaborate data forgery in their much celebrated groundbreaking discovery of a room temperature superconductor, Snider et al 2020 (and in another study by a common author). The Nature paper was retracted in September 2022, but the notice pretended it was due to some squabble about the details of a background subtraction method.

Now Maarten wrote a follow up. Because the same cheaters just published yet another Nature paper, about an even better room-temperature superconductor!

Salamat is still assistant professor at the University of Nevada, Dias is still assistant professor at the University of Rochester sitting on huge grants. Their company Unearthly Materials raised many millions in investments.

This was the Nature paper, retracted on 26 September 2022:

Elliot Snider , Nathan Dasenbrock-Gammon , Raymond McBride , Mathew Debessai , Hiranya Vindana , Kevin Vencatasamy , Keith V. Lawler , Ashkan Salamat, Ranga P. Dias Room-temperature superconductivity in a carbonaceous sulfur hydride Nature (2020) doi: 10.1038/s41586-020-2801-z 

This is its sequel, published in Nature on 8 March 2023:

Nathan Dasenbrock-Gammon , Elliot Snider , Raymond McBride , Hiranya Pasan , Dylan Durkee , Nugzari Khalvashi-Sutter , Sasanka Munasinghe , Sachith E. Dissanayake , Keith V. Lawler , Ashkan Salamat , Ranga P. Dias Evidence of near-ambient superconductivity in a N-doped lutetium hydride Nature (2023) doi: 10.1038/s41586-023-05742-0

Not all media fell for this alleged breakthrough farce, there was critical reporting. Like in Quanta magazine, and a detailed investigation by the the American Physical Society (APS).

And now Maarten joins in.

---

Superconductive fraud: the sequel

By Maarten van Kampen

A successful movie is bound to get a sequel. This is why we do not only have “Back to the Future”, but also “Back to the Future Part II” and naturally a “Back to the Future Part III”. The same seems to apply to high-impact papers.

After the huge box-office success of “Nature 2020: Room-temperature superconductivity in CSH” this March our Nature studios released a sequel with the same star-studded cast: “Nature 2023: Near-ambient superconductivity in N-doped LuH_x“. Also this publication landed as a bomb, receiving news coverage all around the globe. A day before the official release lead actor Ranga Dias gave an in-person sneak preview for a packed room, referencing “Back to the future” and stating that “those things can become possible” with high temperature superconductors. This, to be honest, not in front of the image of a hoverboard or a flying DeLorean, but of a MagLev train.

The first Nature 2020 release was followed by a large ‘controversy’, and at the end of last year that title received its deserved Retraction Note. The background of this retraction was covered by For Better Science in Anatomy of a retraction 2: superconductive fraud. It is clear from the title that we are also fond of sequels, but we really do not want to wait for the Nature 2023 Retraction Note. Instead, we will cover the huge mess surrounding the work of Dias and his upcoming retractions.

Nature 2020

Last year I got upset about the retraction of the Nature 2020 paper. The paper’s critics had to face an enormous amount of opposition, ranging from refusal of sharing data and methods to cease-and-desist letters. And this with second to none support from the prestigious Nature journal it appeared in. The end result of the critics’ two year journey was a retraction, opposed by all authors, with a meek Retraction Note. It contains a shrouded statement that mentions the use of an undisclosed background subtraction procedure of which the validity is “called into question”.

The full story is far worse and contains much more than legal threats. For example, that Nature paper was not the only one to be retracted: also a Physical Review Letters paper of a common author fell for obvious fraud:

M. Debessai , T. Matsuoka , J. J. Hamlin , J. S. Schilling , K. Shimizu Pressure-Induced Superconducting State of Europium Metal at Low Temperatures Physical Review Letters (2009) doi: 10.1103/physrevlett.102.197002 

And what the Retraction Note forgets to mention is that the Nature authors are to this date unable to explain how they obtained their published data from the purportedly measured data.

The figure below gives a flavour of the issues: the red-blue curve was published in the paper and should have been derived from (only) the measured data in black. However, none of the authors could explain how to get from the one to the other. And to make matters worse: the published data turned out to be utterly improbable, see the right panel. It consisted of a smooth mathematical function interrupted by sudden steps at every colour change. In their arXiv paper (and later in an unfortunate choice of a journal) van der Marel and Hirsch conclude that it is most likely that the measured data was constructed from the published curve. And, for what it is worth, I agree.

Nature 2023

The above is pretty much a run-of-the-mill case on For Better Science. In fact, so far it only stands out because it is a bit higher profile than most.

To make sure they would never be forgotten, Range Dias and the Nature Publishing Group pulled an enormous stunt: March 8^th 2023 the same group of authors published yet another Nature paper. And this one presents an even more stunning room-temperature superconductor: N-doped LuH_x. The material is again superconducting at room temperature, but now at a far lower pressures than their earlier find. And that obviously makes applications much more feasible.

To make things extra painful, the new paper was submitted in April 2022. This was at the height of the previous ‘controversy’ and at a time that Dias’ lawyers were threatening his critics with cease-and-desist letters. And whilst the new high-impact paper was busy being peer-reviewed, the earlier paper got retracted over the objections of each and every author for “its validity being called in question”. Thanks to this dialectic approach, Dias et al new Nature paper got published in March 2023, because: what can go wrong?

A high profile publication following a high profile retraction obviously requires a solid investigation to clear any suspicion. Dias’ current employer, the University of Rochester, did its due diligence by conducting not one but two internal inquiries. This very insightful APS article reported on the outcome of those:

“According to a spokesperson, both inquiries “determined that there was no evidence that supported the concerns.” But the university has not made the remit of the investigations public and has not provided any rationale for the investigations or details on how they reached their conclusions.”

APS, 9 March 2023

So this is what we know: “no evidence that supported the concerns” and no details shared. I asked the University of Rochester for the background subtraction method that was used, as the investigations surely must have unearthed it. No reply yet.

When looking at things from the very bright side, there is actually only good news. Two institutional investigations cleared the authors of all suspicions of research misconduct for the previously retracted paper, showing us that the ground-breaking discovery of a room-temperature superconductor in the Nature 2020 paper was actually true. And a new peer-reviewed paper, thoroughly vetted by Nature, presents a second and far more practical room-temperature superconductor. That paper comes with data, and post-publication review did thus far not turn up anything really unbecoming. No steps and mathematical functions this time! And, vindicated, the Nature authors are republishing an update of their 2020 results in a new preprint on arXiv:

Hiranya Pasan , Elliot Snider , Sasanka Munasinghe , Sachith E. Dissanayake , Nilesh P. Salke , Muhtar Ahart , Nugzari Khalvashi-Sutter , Nathan Dasenbrock-Gammon , Raymond McBride , G. Alexander Smith , Faraz Mostafaeipour , Dean Smith , Sergio Villa Cortés , Yuming Xiao , Curtis Kenney-Benson , Changyong Park , Vitali Prakapenka , Stella Chariton , Keith V. Lawler , Maddury Somayazulu , Zhenxian Liu, Russell J. Hemley, Ashkan Salamat, Ranga P. Dias Observation of Conventional Near Room Temperature Superconductivity in Carbonaceous Sulfur Hydride arXiv (2023) doi: 10.48550/arxiv.2302.08622 

But there is more…

Under normal circumstances this affair would be put on a long pause. Physics groups over the whole world would begin their attempts to reproduce the results, using nothing more than the ‘recipe’ in the Nature paper. And they have to start with the recipe because the authors and university have gone commercial, patenting the materials and starting a spin-off company. Unfortunately, due to intellectual property issues they cannot share their superconducting material. This is incidentally the same reason that was given for not sharing the raw data of the Nature 2020 paper.

There is a chance the results will be reproduced, but let’s quote Nature‘s editorial in saying that “doubts remain”. And assuming these doubts materialize, then it will take years of hard-to-publish failures (wrong recipe!) for that doubt to grow to the level of a certainty. And then, without any splash or noise, the story will fall into oblivion.

This could all happen, where it not that Dias has left a trail of seemingly low-level fraud.

Steps in Nature 2020 resistance data

Let’s first start with ‘oddities’ that by now have already shown not to stick. The finds were made by James Hamlin, co-author on the 2009 PRL paper that was retracted as a result of Jorge Hirsch’ investigation in the Nature 2020 paper. Since then, Hamlin actually became an active ‘science detective’.

The 2020 paper was brought down by its impossible steps in the magnetic susceptibility data, shown again in red in the left panel below. The published data turned out to be the sum of two parts: a smooth mathematical function in orange and a stepped function in blue. No resistance data was shared, but Hamlin extracted this data from the vector-based figures in the paper. And he noticed similarly weird steps in the resistance data, see the inset of the right panel below:

The issue looks like a carbon-copy of the earlier find: the published resistance data in red shows steps, and is in fact the sum of the smooth function in orange and the stepped function in blue. And while for the susceptibility data the authors could blame a mysterious background subtraction procedure, the resistance data should be directly measured. Measured data can show digitization steps, but these should not sit on top of a smoothly varying background.

Nature was informed but did not deem it necessary to investigate as the article was retracted:

“Thank you for your email and apologies for the slow reply.
We have retracted the paper because there were issues with data analysis and processing of the magnetic susceptibility data – as stated in the retraction note. Our involvement and investigation ends there.

Best regards
Tobias Roedel”

E-mail from  2022-11-07

This reply is from November 2022, with the 2023 Nature paper then being under review and more than two months before it got accepted. But we can rest assured that all is fine, as two institutional investigations could find nothing that corroborated the concerns.

[ Addition 2023-04-01 ] In a presentation earlier this month James Hamlin presents the below slide, showing that already 2022-02 he requested resistance data via the Nature editor. But did not receive it. He additionally seems a bit disappointed that in this post-retraction Science article senior author Salamat claims that the resistance data are not questioned! I personally like the ‘After doubts grew, blockbuster Nature Paper is withdrawn‘ subtitle of the article.

Plagiarism

Hamlin also found that his own 2007 PhD thesis was plagiarized by Dias’ 2013 PhD thesis, with an example shown below:

And, as the insightful APS article on the case tells us, this is basically fine:

“In response to the allegations that he plagiarized Hamlin’s thesis, Dias says he has done nothing wrong: “I have appropriate citations.” Washington State University, which awarded Dias his PhD declined to comment on whether they have carried out a misconduct investigation. A statement from the University of Rochester says, “Dr. Dias has taken responsibility for these errors and is working with his thesis advisor…to amend the thesis.”

I was also very amused to hear that Hamlin’s thesis again got plagiarized in this arXiv paper from Dias and Salamat. Jorge Hirsch, one of the Nature 2020 first critics, had been battling to receive the raw data from the authors with Dias being quoted saying

“Hirsch is a troll, we are not going to feed this troll”.

The raw data was finally released in the above arXiv paper, in 80 pages of images of data tables. And if that was not enough, the description of the experimental details of the susceptibility measurements were lifted near-verbatim from Hamlin’s thesis:

Is this plagiarism misconduct? Or should we consider it an inventive way of trolling critics?

Data fabrication in Clossal Density-Driven Resistance…

Things are getting more interesting with the next find. Hamlin went through Dias thesis (2013) and noticed a surprising similarity with this 2021 PRL paper:

Dylan Durkee , Nathan Dasenbrock-Gammon , G. Alexander Smith , Elliot Snider , Dean Smith, Christian Childs , Simon A. J. Kimber , Keith V. Lawler, Ranga P. Dias , Ashkan Salamat Colossal Density-Driven Resistance Response in the Negative Charge Transfer Insulator MnS2 Physical Review Letters (2021) doi: 10.1103/physrevlett.127.016401 

Figure 1b of the PRL paper studying MnS₂ contained curves that look quite a bit like those in a thesis figure on GeSe₄, both reproduced below:

The pair of curves highlighted in yellow appear visually similar in both figures and also some features in the 13 GPa curves marked blue seem to match.

In this PubPeer post the most fair comparison is made, showing that over the full temperature range the curves are at most similar whilst on a 0-120 K range they are near identical. The limited range is shown below:

In the figures the two materials share the same temperature axis, but for each figure a different vertical scale is used. Still, the different steps and kinks in the curves make this ‘coincidence’ so very, very unique that this data cannot correspond to measurements of two different materials. At best this is (yet another) unexplained ‘setup background’, more likely this is lazy fabrication. Note that 5 of the 9 authors on the retracted Nature 2020 paper also appear on this PRL.

The authors and the journal were informed on 27 October 2022. Simon Kimber, one of the authors, showed the expected shock after receiving the above analysis:

“My name is Simon Kimber, and I am a co-author on a PRL with Dias and Salamat. We received information on the 27/10/22 from Dr. Hamlin, that our high pressure resistance data on MnS2 were remarkably similar to that for GeSe4 in Dias’ thesis.
The agreement was so striking that the next day (28/10/22) I contacted all co-authors and PRL recommending retraction. I heard nothing in response.
After this matter became public in Feb (see PubPeer), I aggressively contacted divisional editors etc. We now have a response, which in my opinion only reinforces the case for retraction.
This is a huge mess, and in my opinion, will only get worse. As you might imagine, the stress, and the silence from co-authors and APS, have been utterly devastating.”

nanoscale.blogspot.com

It is jaw-dropping to read that of the ten PRL authors apparently only one saw the severity of the case and was basically ignored. And it may be coincidental, but Kimber happens to be the only author that is not affiliated with either the University of Rochester, home base of senior author Ranga Dias, or the University of Nevada Las Vegas, the affiliation of Ashkan Salamat. Also note that these concerns were raised end of October last year. The new Nature paper had yet to be accepted, Tobias Roedel had not yet dismissed the concerns on the stepped resistance data, and presumably one of the institutional investigations was still running. To paraphrase Kimber: “This is a huge mess”.

March 13^th 2023, some four months after the first report, PRL issued an Expression of Concern because

“Questions have arisen regarding the integrity of the data.”

I can only imagine this to be the first step to a retraction, with the editorial office now figuring out whether to call this “data not accurately reported” (PRL 2009) or maybe allude to some sort of fabrication.

Data fabrication in Pressure-induced Transformations of …

February this year the above find was published on PubPeer. Inspired but lazy I ran Dias’ thesis through ImageTwin and found a relatively minor issue in a 2016 Scientific Reports (SciRep) paper co-authored by Dias. This paper is basically a publication of the carbonyl sulfide (OCS) data in Dias’ 2013 thesis, plus a bit more.

Minseob Kim , Ranga Dias , Yasuo Ohishi , Takehiro Matsuoka , Jing-Yin Chen , Choong-Shik Yoo Pressure-induced Transformations of Dense Carbonyl Sulfide to Singly Bonded Amorphous Metallic Solid Scientific Reports (2016) doi: 10.1038/srep31594

In the collage below it can be seen that Figure 2 of the SciRep paper (left) is the merger of two figures from the thesis, respectively showing pressure-dependent Raman and pressure-dependent resistance measurements.

A somewhat innocent explanation for the above would be ‘best-looking data’ selection. Maybe the authors were pleased with the Raman result but decided to do extra resistance measurements. And managed to obtain two nicely matching runs that unfortunately looked very different from earlier results. And then decided to go for better publishable beauty, tossing out the odd resistance data and incidentally recycling a nice image from an earlier run.

PubPeer user Peziza proteana comes with a very different explanation. And it is truly bizarre, verifiable, and does check out. This commenter must be a true aficionado of the Dias genre, watching this 2017-02 feature length movie on metallic hydrogen to at least its 48^th minute. Dias had at that time just reported the break-through discovery of solid, metallic hydrogen:

Ranga P. Dias , Isaac F. Silvera Observation of the Wigner-Huntington transition to metallic hydrogen Science (2017) doi: 10.1126/science.aal1579 

This was done in association with Science studios and resulted in a jealous editorial from Nature with the title “Physicists doubt bold report of metallic hydrogen“. One simply cannot escape the similarity with their 2023 editorial: “Hopes raised for room-temperature superconductivity, but doubts remain“.

Criticisms included “the infrared spectra shown in the supplementary material did not match the data provided in tabulated format“, “their signal is spurious (not from the sample) and/or that they have not reached the claimed pressure” (referenced on PubPeer). Finally, in 2017 it was reported that this “only piece on Earth of a metal that could have revolutionised life as we know it” mysteriously disappeared “after catastrophic failure of diamond holding it under enormous pressure“. Or that was what Dias and Silvera claimed.

Back to the video. After 48 minutes the slide in panel (a) below is shown:

It shows the SciRep figure on carbonyl sulfide to the left and a systematic study of a range of other sulfides to the right. And that right figure is incidentally near-identical to Fig. 5.13 from Dias’ thesis, reproduced in panel (b) above. In the thesis the SiS₂ curve is missing. This is no coincidence, as the thesis correctly mentions that SiS₂ is difficult to handle due to its highly reactivity with moisture (forming H₂S).

Now comes the bizarre part. The black SiS₂ curve in the screenshot of the presentation looks somewhat like the OCS curve in the thesis, shown earlier in the right panel of Figure 5. I imagine Peziza tried to scale the image of the OCS curve on top of the screenshot of the SiS₂ curve. And then found that the curves themselves do not overlap, but that many of their points sit at near-identical (vertical) resistance values. In the figure below, reproduced from the PubPeer post, the digitized curves are compared to each other:

The left panel shows the original OCS and SiS₂ resistance versus pressure curves. There is not much commonality, except that both curves have the same number of points that sit at very similar ‘heights’. When plotting both curves using a single pressure series, the right panel is obtained. All but four of the resistance values sit at the same value. To be exact, the first 21 ‘matching’ points are identical to within ±1.5%, the last one is 13% off. The resistance itself has changed by more than a million (!) times.

I did some due diligence, repeating the digitization and coming to the exact same result. I also checked whether other curves on the slide maybe also shared many identical resistance values, for example due to some measurement artifact. This is not the case. And, like the PRL 2021 data discussed earlier, this simply cannot have happened by chance.

The same PubPeer thread has another comment comparing the SciRep OCS ‘Run 2′ data with the thesis’ SnS₂ measurement. Finding a more limited but similarly strange coincidence of 9 consecutive measurement points that span 5 orders of magnitude in resistance, yet happen to ‘hit’ near-identical resistance values.

Added 2023-03-31: visual comparison of the overlap between SciRep 2016 OCS and thesis SnS₂ data:

It seems that the most reasonable explanation for the ‘same resistance values’ issues is fabrication. In the PRL 2021 case the temperature values on the x-axis were left untouched and the data was only vertically scaled. Here the resistance values on the y-axis were (mostly) left as-is whilst the pressure values on the x-axis were newly fabricated.

It will be up to an internal investigation (or two) to find out what actually happened. In the mean time it is hard to tell whether any of the published datasets are valid or not.

Why and how?

No idea, obviously. But some speculation to the why: maybe a combination of Jan Hendrik Schön and the lure of big money?

Schön is a infamous physicist that during my PhD time was scoring breakthrough after breakthrough in the field of nanotechnology and condensed matter physics. To get a feeling: he published 7 Nature and 10 Science papers in 2000 and 2001, more than one every two months. This all came to an abrupt end when it was noticed that many of his published measurements were identical down to the noise. And once noticed, the fraud proved to be of childish quality, recycling curves with identical noise from paper to paper. Schön insisted on it all being mistakes, so this is in fact not much of an explanation. The assumed ‘craving for fame’ probably is.

The monetary side is very visible in the current case. As was noted in the previous For Better Science article, the authors have founded a start-up company Unearthly Materials, with its intellectual property licensable through the university’s URVentures (disclaimer: that site seems to have issues currently so I cannot check whether this still holds). This mix of research and money making has proven to be rather toxic, with IP and patent applications being used to deny access to raw data (Nature 2020) or material samples (Nature 2023). And with the university’s financial involvement one may start to doubt institutional impartiality.

In a video that was removed from YouTube, but still is available on Vimeo, a pitch is made to investors. With the promise of a nascent Superconducting Economy the authors show to have raised some $1 million in a seed round and $20 million in a Series A. And they tease investors with a Series B that aims to raise $250 million with a $1 billion cap. Of note: the YouTube video was likely removed because not all Series A investors were actually on board…

The right screenshot shows the seven person team of Unearthly Materials. The brains consist of Range Dias, Isaac Silvera (Dias’ former postdoctoral advisor at Harvard) and Askhan Salamat, discovers of metallic hydrogen and room-temperature superconductivity. The rest seems involved in commercialization and litigation.

Conclusion

I follow Simon Kimber here: “This is a big mess”. Two university investigations into the Nature 2020 debacle finding nothing amiss, but at the same time leaving open what happened between the published and potentially measured data. Nature retracting a paper over the objections of all its authors, and at the same time reviewing a new paper from those same authors. And ignoring additional concerns coming in they publish the miraculous find with a mix of fanfare and ‘doubts remain’.

There are two positive aspects here. For one, the card house seems to topple so rapidly that it may inflict some reputational damage to Nature and the University of Rochester. More institutional investigations will surely be needed. Secondly, Ranga Dias has discovered a third class of high temperature superconductors, again reaching room temperature at nearly 10x lower pressures than the Nature 2020 CSH. So if all goes well we will get our “Back to the Future Part III“!

---

Afterword by LS

(recycled from past Friday Shorts)

Now, you might wonder: why did Nature accept another fake garbage from the same crooked duo, after a retraction no less? Will they listen to the sleuths’ arguments about Dias’ history of cheating?

Well. Up there at editorial offices, they don’t believe in research fraud as a character trait. It’s the “each paper on its own merit” attitude which basically means they scrutinise all new submissions from exposed fraudsters for the same types of forgeries those fraudsters were originally caught on. If the editors find nothing, the new submission is deemed trustworthy.

Sure there will be many of those peers who refuse to give thumbs-up to a serial fraudster. But: as a peer reviewer you can’t reject a manuscript just because you mistrust the author, such are the rules. So Nature has been probably going through several potential reviewers until they found someone who will agree. In fact, it seems even those who waved the paper through were not sure, their reports aren’t published. This editorial by ChangQing Jin and David Ceperley, which Maarten referred to above, sceptically suggests: “Hopes raised for room-temperature superconductivity, but doubts remain“. Usually such accompanying editorials are written by the peer reviewers of the original paper.

In any case, I fully agree with this statement in Quanta:

““For me, it is difficult to imagine a second retraction,” said Mikhail Eremets, a physicist at the Max Planck Institute for Chemistry in Germany who led the discovery of hydride superconductors. “We should consider it seriously in spite of the prehistory.”

This neatly sums up the academic attitude. Simple mortals have unforgivable character flaws, but scientists don’t. You won’t go to a restaurant where the chef was caught urinating in the soup. But you will eagerly celebrate the next Nature paper for someone previously caught on fraud.

Nature never retracted a paper just because it was irreproducible. And even when new kinds of fraud were to be found: the elite journal won’t retract to save face.

I demand that Nature gives Haruko Obokata and Charles Vacanti a second chance.

---