Maarten van Kampen was ordered by the MIT Vice President for Research not to tell anything to any living soul, or else he commits a federal felony. Because the MIT professor Ju Li keeps making groundbreaking green nanotechnology discoveries on daily basis. For each retraction, dozens more papers get published and celebrated with MIT press releases!

We are here to help.

---

MIT review closed and decision final

By Maarten van Kampen

In December 2021, “fake spectra Googler” Thallarcha lechrioleuca found a fake XRD pattern in a Nature Communications paper:

Sa Li, Junjie Niu, Yu Cheng Zhao, Kang Pyo So, Chao Wang, Chang An Wang , Ju Li High-rate aluminium yolk-shell nanoparticle anode for Li-ion battery with long cycle life and ultrahigh capacity Nature Communications (2015) doi: 10.1038/ncomms8872 

The ultimate author of the paper, Ju Li, is Professor of Materials Science and Engineering and Battelle Energy Alliance Professor in Nuclear Engineering at MIT.

Five of the seven authors sport an MIT affiliation (shown in bold), including the first two authors that made equal contributions. The paper received funding from the NSF:

“We acknowledge the support by NSF grant DMR-1120901 and National Natural Science Foundation of China (NSFC–No.51221291 and 51172119).”

In August 2015, an MIT press release celebrated the papers breakthrough find, headlined ““Yolks” and “shells” improve rechargeable batteries”:

“Now a team of researchers at MIT and Tsinghua University in China has found a novel way around that problem: creating an electrode made of nanoparticles with a solid shell, and a “yolk” inside that can change size again and again without affecting the shell. The innovation could drastically improve cycle life, the team says, and provide a dramatic boost in the battery’s capacity and power.

The new findings, which use aluminum as the key material for the lithium-ion battery’s negative electrode, or anode, are reported in the journal Nature Communications, in a paper by MIT professor Ju Li and six others. The use of nanoparticles with an aluminum yolk and a titanium dioxide shell has proven to be “the high-rate champion among high-capacity anodes,” the team reports.”

The press release quotes Li saying his team “came up with the method serendipitously” and that his new technology is “probably the best anode material available” and “quite close to being ready for real applications.“

The paper declares that “The authors declare no competing financial interests“, but one month after its publication, the authors and their universities applied in September 2015 for the patent “ALUMINUM BASED ELECTROACTIVE MATERIALS” which was published in March 2016.

I “adopted” the paper, found some additional concerns, and reported the issues to MIT in January 2022. The paper is by now retracted, with a Retraction Note that just stops short of calling the issues fabrication. I recently also received the outcome of the MIT investigation that took over 20 months to conclude (highlights theirs):

“Based on the faculty member’s Inquiry of the complaint and referenced publications, MIT has determined that an Investigation into the allegations is not warranted because the available evidence is not sufficient to support the allegations that research misconduct may have occurred as alleged. Accordingly, MIT’s review of these allegations is now closed and this decision is final.”

Letter from MIT, highlights theirs.

I want to take the opportunity to review the available evidence and come to the opposite conclusions: one or more authors have fabricated results.

XRD

Thallarcha lechrioleuca looked at Fig. 9 of the supplementary information and noticed that the XRD curves had different peaks, but shared their noise and baseline:

This is very much not expected, as the curves purport to show a pristine aluminum yolk-shell nanoparticle anode (black) and its state after up to 511 charge/discharge cycles. And hence not only the peaks, but also the noise should differ between the measurements.

At the above resolution one may deem the available evidence “not sufficient”. Fortunately, the authors published their figure in vector format. This allows one to redraw the curves with thinner lines and to move them closer together:

The above figures leave no doubt. All five curves are identical down to the noise, except for a number of narrow regions around the peaks. This type of coincidence is found more often in scientific literature, with the case of Magnus Willander and Omer Nur being a good example. It was covered on For Better Science and has the benefit of a public institutional investigation. Let me quote the expert Skoglundh:

“Since the X-ray diffractograms of the two samples ZnO NW on CNT-UCT and ZnO NW on CNT-CT in Figure 4 of Article [2] contain significant ranges that are  identical (in terms of position and relative intensity between the diffraction peaks of the two samples, absolute intensity for one and the same peak for the two samples and, above all, with respect to noise) they can not come from two different measurements. At least one of these X-ray diffractograms must therefore be fabricated.”

Skoglundh report August 2020

I think this conclusion can be easily adapted for current paper: since the five curves contain significant ranges that are identical (above all with respect to noise) they cannot come from different measurements, and at least four of these X-ray diffractograms must therefore be fabricated. Except that MIT apparently sees this differently?

The issues with the XRD data do not stop here. In Figure 5 of the supplementary data the authors want to show that their TiO₂ shells protect the Al ‘yolk’ from oxidation. When processing the powder into a battery anode they need to grind it with a binder. Being thorough, the authors check the oxidation resistance of both the powder as-is, red curve (a), as well as after grinding, black curve (b):

In the two rightmost insets it can be seen that the grinding action introduces many repeating stretches into the black XRD curve (red boxes). These repeating stretches can be found in other retracted papers, see e.g. this example from the fraudster Xiangke Wang. And they appear to be mostly fabrications, with their creator using them to replace unwanted features with a noise-like background. This approach looks much better than some alternatives shown in the top-left inset: simply removing a stretch of data (right yellow highlight) or replacing it with just a few points (left highlights).

The red curve labeled (a) in Fig. S5 should represent the XRD pattern of unprocessed yolk-shell powder. And that pattern should be markedly different from the XRD curves in Fig. S9 that were measured on an actual battery anode and thus represent powder ground with a binder and applied to an anode. But alas, the red unprocessed curve of Fig. S5 is identical to the ‘before cycling’ curve of Fig. S9. This can be seen in the leftmost inset where above 20^o the red curve perfectly overlaps with the purple XRD pattern measured on the battery anode. It is very hard not to see falsification or fabrication here.

TEM & TEM-EDX

Thallarcha lechrioleuca not only has a keen eye for XRD anomalies, but also for ‘mix-ups’ in TEM imagery. And this paper has an interesting number of those.

In Figure 2 the authors show SEM and TEM imagery of their Al-TiO₂ yolk-shell particles. Except that some two years after publication the authors suddenly realized that there was a mix-up: the image in Fig. 2(b) was not showing a TEM image of Al-TiO₂ yolk-shell particles, but instead of MnO₂ core-shell particles. It was previously published in a 2014 paper by co-authors Sa Li and Chang-An Wang:

Sa Li and Chang-An Wang “Design and Synthesis of Hierarchically Porous MnO₂/Carbon Hybrids for High Performance Electrochemical Capacitors” Journal of Colloid and Interface Science (2014) doi: 10.1016/j.jcis.2014.09.065

This little mix-up was fixed with a Correction on 29 November 2017:

“This Article contains an error in which the scanning electron microscope image shown in Fig. 2b was included incorrectly. The original Fig. 2b showed yolk-shell microstructures, as previously published in Li, S. and Wang, C.-A. Design and synthesis of hierarchically porous MnO₂/carbon hybrids for high performance electrochemical capacitors. Journal of Colloid and Interface Science 438, 61–67 (2015). (©2014, with permission from Elsevier). We and our colleague Dr Yuming Chen have repeated the high-rate battery cycling tests of the material, which is supplied in the Supplementary Information associated with this correction, and this error does not have consequence on the science, data or conclusions of the Article, but we apologize for any confusion caused. The correct version of Fig. 2, which shows the inner aluminum yolk encapsulated by 2–4 nm-thick TiO₂ shell in the revised panel b, appears below as Fig. 1.”

The figure above shows Fig. 2 of the 2014 MnO₂ paper (left) and the uncorrected Fig. 2 of the 2015 Nature Communications paper (right). The green rectangles show the overlap between the two, after 180^o rotation.

Thallarcha had a good look at the figure and noticed that the TEM image in panel (c) was a magnified version of that in panel (b), see the red rectangles above. This creates two issues. For one, even after correction panel (c) presents a C@MnO₂ core-shell particle as an Al@TiO₂ yolk-shell particle. And most importantly: the authors seem to have been able to detect Al and Ti in a C@MnO₂ particle… Panels (d-f) show EDX maps corresponding to the TEM image in panel (c). And the purple dots in panel (d) show that the Ti atoms indeed sit in their expected ‘shell’ and panel (f) that the Al atoms form a nice ‘yolk’ core. And this is obviously very much impossible for a MnO₂ particle.

I can only see two not-so-nice explanations:

1. The authors not only “mixed up” their images, but also “mislabelled” the EDX mappings. And failed to notice this upon correction.
2. Panels (b-f) are actually showing Al-TiO₂ yolk-shell particles and the image of the 2014 paper is wrong.

I would consider these options either falsification or recklessly negligent bookkeeping, especially considering that this survived a correction.

It is also noteworthy that at this point five of the six panels in the original Fig. 2 seem to depict C@MnO₂ particles. Only panel (a) shows a SEM image of Al@TiO₂ yolk-shell particles. And I cannot help noticing that that image looks very much like the SEM imagery of C@MnO₂:

The issues with the TEM images and EDX maps actually go a bit further. The author(s) appear to be very keen to use Photoshop on their images. Take the original Fig. 2(b): the version published in the 2015 Nature Communications paper is radically ‘cleaned up’ compared to its original version in the 2014 paper, see the animation below:

Not only the support grid in the background has been removed, but also some less perfect particles near the edge have been erased.

This image is not the only one that has been ‘cleaned up’. Like Figure 2, Figure 4 of the Nature Communications paper shows SEM, TEM, and EDX images of the particles, but now after many cycles of use as battery anode. The TEM zoom-in and EDX maps are shown below:

EDX elemental maps are inherently noisy. This can be seen in panels (d) to (f) above, with the blurred dots and islands representing smoothed out detections of characteristic x-ray photons. Except that the purple Ti ‘shell’ in panel (d) shows a perfectly curved edge at the lower-left side that cuts cleanly through these dots and islands (red arrow) . The same holds for the Al ‘yolk’ shown in panel (f): it not only has a curved edge, but also sharp and perfectly horizontal and vertical ‘cuts’ indicated by the red dotted lines. And this is not how EDX works: these mappings have been edited to create the impression of a perfect Ti shell with a matching Al yolk. And that is again falsification.

This paper, with its fabricated XRD data, misrepresented and falsified EDX data, and Photoshopped TEM imagery is the basis for MIT patent 20160079592. The patent was filed one month after publication of the paper and contains all of the paper’s figures, including the fabricated ones. Its inventors are grateful for the NSF funding they received:

Retraction notice

Springer-Nature has retracted the paper on 13 September 2023 upon request of the authors. The retraction notice acknowledges all of the issues discussed above:

“The authors have retracted this article because there are a number of flaws in the figures and data. Figure 2c and the EDS mappings in Figures 2d, 2e and 2f were re-used from the authors previous work¹ where a different material was used. Figures 4b and 4c contain sections with an unexplained blank section not matching the background or image. The EDX images in Figures 4d–4f were modified before publication. The background noise in the XRD spectra in Supplementary Figure 9 shows a much higher degree of overlap than would be expected for different cycles.”

Retraction note to High-rate aluminium yolk-shell nanoparticle anode for Li-ion battery with long cycle life and ultrahigh capacity.

A TEM image showed the wrong compound. The corresponding EDX mappings were mislabelled to match their new setting. Other TEM and EDX images were Photoshopped. XRD patterns were fabricated from a single curve. And all of this left so little of the data untouched that the ubiquitous phrase “conclusions unaffected” is refreshingly absent from the retraction notice.

MIT and University of Wisconsin investigation

Five of the paper’s seven authors have an MIT affiliation. This includes the first and second author who contributed equally to the work:

Second author Junjie Niu signed the paper both with a MIT and a University of Wisconsin affiliation. Niu currently holds a Richard and Joanne Grigg Professorship at University of Wisconsin and his university page appears proud of his Nature Communications papers:

“Niu has published more than 80 papers in prestigious journals including Nature sister journals–Nature Nanotechnology and Nature Communications and two book chapters.”

I contacted both MIT and the University of Wisconsin in January 2022. I never heard back from the latter, but MIT promptly replied and even came with follow-up questions a few months later. The paper was retracted in September 2023. I learned, however, that this was a decision by the authors that was completely separate from the MIT review.

MIT finished their review in October 2023, so more than 20 months after concerns were raised. And I learned that a review is not an investigation. It appears that after raising concerns a single senior MIT faculty member took over 20 months to conclude that an “investigation into the allegations is not warranted because the available evidence is not sufficient to support the allegations that research misconduct may have occurred as alleged“.

Maria Zuber, Vice President for Research at MIT and the author of the below letter, tries to make it very clear that this is the end of it: “MIT’s review of these allegations is now closed and this decision is final“.

The letter also makes clear what Zuber thinks led to the retraction:

“The policies also provide that research misconduct does not include honest error or differences of opinion.”

I cannot fathom how one could reach a conclusion like this: XRD data have been fabricated, images have been Photoshopped, and prior results have been presented as something else. And all of this was already published in a retraction notice that was available before the ‘review’ was finished.

Prior art

MIT professor Ju Li is not new to retractions. It turns out that he is also author on a Science Advances paper that got retracted in November 2021:

Bu Yuan Guan, Le Yu, Ju Li, and Xiong Wen (David) Lou A universal cooperative assembly-directed method for coating of mesoporous TiO2 nano shells with enhanced lithium storage properties Science Advances (2016) doi: 10.1126/sciadv.1501554

“Recently, we found two duplicated images (Fig. 3A, Fig.S17A) in two figures that were published previously in two papers by B. Y. Guan et al. (2, 3). Because of this, our confidence on the validity of the data in the paper is largely compromised. Therefore, we wish to retract this research article promptly. We sincerely apologize that these errors were not discovered before the manuscript was published. All of the authors have agreed the paper should be retracted, except Dr. B. Y. Guan who opposed this retraction.”

This Science Advances paper also deals with TiO₂ nano-particles and, just like the Nature Communications study, it finds that these TiO₂ nano-shell particles show an excellent performance as anode material in Li batteries. Until it was retracted, that is.

The retraction of the Science Advances paper was covered by Retraction Watch and both Ju Li and ultimate author David Lou are quoted there.

“we checked Dr. Guan’s PhD thesis and his published papers, and we were shocked to find these two duplicated images in his previous papers. I actually immediately contacted Dr. Guan for an explanation, but he refused to communicate with me. After discussion with Prof. Ju Li, we immediately informed the Science Advances Editorial office.“

David Lou, Retraction Watch

It turned out that the Science Advances paper contained amongst others TEM images that were used in two previous publications where they represented different materials. And doesn’t that sound very familiar?

• Buyuan Guan, Tao Wang, Shangjing Zeng, Xue Wang, Dong An, Dongmei Wang, Yu Cao, Dingxuan Ma, Yunling Liu & Qisheng Huo “A versatile cooperative template-directed coating method to synthesize hollow and yolk-shell mesoporous zirconium titanium oxide nanospheres as catalytic reactors” NanoResearch (2014) doi: 10.1007/s12274-013-0392-9
• Buyuan Guan, Xue Wang, Yu Xiao, Yunling Liua, and Qisheng Huo “A versatile cooperative template-directed coating method to construct uniform microporous carbon shells for multifunctional core-shell nanocomposites” Nanoscale (2013) doi: 10.1039/C3NR34041D

The 2016 Science Advances paper seems to be pretty much a re-publication of the 2014 paper that itself has issues. The papers share their “unique” versatile/universal synthesis method, many of the coated core materials, “mislabelled” TEM images, and even a mislabelled particle size histogram:

Buyuan Guan, professor at the papermill-owning Jilin University, is author on all three papers and is fingered as the culprit. And whilst this is likely true, it would still be interesting to know what substantial contribution the three other authors made to the retracted Science Advances paper to deserve an authorship seat on that fabrication.

I cannot help help noticing that none of the above three papers have authors or affiliations in common with the retracted Nature Communications paper we started with. Except for Ju Li, that is. He must be have been extremely unlucky: publishing two high-profile papers on the use of TiO₂ nano-shell particles for lithium-ion batteries in just two years, and then get them both retracted.

The retracted Nature Communications paper does have another interesting link to the retracted Science Advances paper. The press release by MIT on the Al-TiO₂ yolk-shell particles quotes Li’s co-author on the Science Advances paper, David Lou, as expert:

“These yolk-shell particles show very impressive performance in lab-scale testing,” says David Lou, an associate professor of chemical and biomolecular engineering at Nanyang Technological University in Singapore, who was not involved in this work. “To me, the most attractive point of this work is that the process appears simple and scalable.”

MIT News, August 2015

See, “not involved in this work“. One can hope he wasn’t incidentally the peer reviewer of the paper.

With the MIT review not finding evidence for research misconduct, Li can presumably continue doing whatever he has been doing to score multiple retractions. In fact, since the 2015 Nature Communications paper MIT issued (news tab) another forty press releases celebrating Li’s discoveries.

Conclusion

A MIT professor got two papers on TiO₂ nano-shell particles retracted. The latest Retraction Note mentions that TEM and EDX images were modified and that data was misrepresented. And the sentence “The background noise in the XRD spectra … shows a much higher degree of overlap than would be expected for different cycles” stops just short of calling the XRD data fabricated.

MIT launched a 20+ month review into the issues with the paper. And had an senior staff member come to the baffling conclusion that there is insufficient evidence to support allegations that misconduct occurred.

Case closed and decision final!

---