Post-publication review of 'Flux-tuned topological superconductivity in full-shell nanowires' Vaitiekenas et al. Science 2020

In March 2020 a research group at Microsoft / University of Copenhagen published a paper in Science titled “Flux-induced topological superconductivity in full-shell nanowires” https://science.sciencemag.org/content/367/6485/eaav3392 (in what follows Science-2020). The authors claimed (both through explicit statements in the text and through their choice of what data to present) that some nanowires they have studied exhibit robust zero-bias peaks (i.e. peaks in conductance at zero voltage) coinciding with the odd lobes of a multi-lobed phenomenon known as the Little-Parks Oscillation. They interpret this as evidence for topological superconductivity, i.e. Majorana particles.

In October 2020 the authors shared additional data from those experiments on Zenodo http://doi.org/10.5281/zenodo.4263106. Having carefully studied the paper, as well as additional data, we conclude that data in a key figure, Figure 2, are not representative of the full experimental data obtained. Contradictions are found not only in analogous data from other devices, but even over larger parameter ranges from the same device.

Contrary to statements in the text, we find evidence that the authors’ supposedly robust zero-bias peaks fail to coincide consistently with the odd lobes of the Little-Parks Oscillations. We also find evidence that some supposed zero-bias peaks are split peaks, and that zero-bias peaks are gate-dependent, and not robust throughout the tunneling regime. Again these findings are contrary to explicit statements in the paper.

Left: Paper figure in the Research Article Summary shows a coincidence of a zero-bias peak and Little-Parks oscillations, but additional data contain examples where this coincidence is not present. 

With respect to data as shown in a key figure, Figure 2, we found multiple statements throughout the paper to be inaccurate or misleading as applied to the full data available. These statements, identified below, concern both the description of the data, as well as the interpretation of the data as relevant to ‘flux-induced topological superconductivity’ - the topic of the paper. Thus the core conclusions of the paper are invalidated. We do not see how this paper can remain in the present form, nor how the problem can be addressed by publishing any form of clarification such as a correction. 

Selection of data for publication is an inevitable part of the research process. Hundreds and thousands of datasets are obtained for each study in our field. Yet only 5-10 become figures in a given paper. Here we identify a situation in which the selection of data for this paper is not representative of the total data obtained. Our field of study is vulnerable to this problem. Data chosen for the figures tell a compelling story, while full data present contradictions. Sometimes the contradictions are so serious that the paper is retracted. This was the case for ‘Quantized Majorana Conductance’ by Zhang et al., Nature 2018. 

On July 2 2021, Science published an experiment from a group at IST Austria, which uses the same nanowires and the same measurement technique used for Figure 2 of Science-2020, but arrives at opposite conclusions (Valentini et al., Science 2021 https://www.science.org/doi/10.1126/science.abf1513). The analysis by the IST group presents a fuller body of data. While Niels Bohr Institute has stated that they conducted an internal investigation into Science-2020, we believe external and independent expertise is required to further assess this issue. We continue to request that the corresponding authors share full data obtained.

The Editor-in-Chief of Science issued an Editorial Expression of Concern on Science-2020 on July 30, 2021.