The exhaust particles that were in the blank

In April 2025 a Japanese and Swiss team reported something new in the exhaust of commercial jet engines: tiny onion-shaped particles, built from concentric graphene-like shells, plus faint "trace" particles nobody had described before. Press releases went out in May, and coverage ran with the health angle for people living near airports. On 20 July 2026 the same authors retracted the paper. The onion particles were also on the blanks.

This is a good retraction, and I mean that as a compliment. It is also a clean lesson in why a blank matters most when you think you have found something new.

What the paper claimed

Akihiro Fushimi and colleagues, from Japan's National Institute for Environmental Studies, Zurich University of Applied Sciences (ZHAW), and other institutes, published the work in ACS ES&T Air. They sampled PW4000 and CFM56 turbofan engines in a test cell at Zurich Airport during 2021, at the engine exit and 15 meters downstream, and looked at the particles with high-resolution transmission electron microscopy (HRTEM).

They sorted what they saw into four types: turbostratic particles (ordinary soot), onion-like particles, amorphous particles, and trace amorphous particles. Downstream, soot was under 1% of the particles counted. The rest were mostly single spheres of about 10 to 20 nm. The paper called the structures of the three non-soot types newly identified, and suggested they formed by nucleation and partial pyrolysis, with lubrication oil as the likely source. It raised the possibility that those structures could change how the particles interact with the human respiratory tract.

What the retraction says

The notice is short. After publication, someone at a scientific meeting told the authors that onion-like carbon particles might form during the production of the carbon films used in TEM work. The authors ran validation experiments, including reanalysis of the TEM samples from the study, and:

"identified a significant number of onion-like particles and trace particles in the blank samples."

So, they concluded, it was hard to say those particles came from the jet engine at all, and that was one of the study's main findings. The notice calls these critical errors in the experiments. ZHAW's statement goes further, saying the findings are no longer considered valid scientific knowledge, and apologizes for the wrong conclusions.

The warning was 45 years old

The notice cites one reference, and it is the part of this story I keep coming back to. In 1980, S. Iijima published a high-resolution electron microscopy study in the Journal of Crystal Growth of extremely small graphitized carbon particles. Where did he find them? In amorphous carbon films prepared by vacuum deposition. He was not sampling a jet engine. The curved, layered carbon was already sitting in the kind of film microscopists use.

It was flagged again in 2001. P. J. F. Harris wrote a short paper in Carbon on carbonaceous contaminants in evaporated carbon support films, covering fullerene-like nanotubes and nanoparticles among other things, seen on grids from a number of commercial suppliers. His abstract says there are many cases in the carbon literature of these contaminants being mistaken for the sample. I read both papers at the abstract level only, and both are about carbon films in general, not the grids in this study.

Why a blank would matter so much here

Most contamination problems hide in a big signal. This one did not have a big signal to hide in. The paper's main text describes the collection: exhaust drawn over 400-mesh, collodion membrane-coated copper grids, with collection efficiencies of 6% for 10 nm particles and 2% for 30 nm particles. Each sample was then read through roughly 50 to 100 images, usually one particle per image.

My reading, not the authors': when you are counting particles one at a time on a surface that catches only a few percent of them, the surface itself becomes part of the measurement. A handful of background particles per grid is not a rounding error. If the thing you are classifying can be manufactured by the support film, the only way to tell it from the sample is to count an unexposed grid the same way, with the same eyes and the same classification rules, and compare the rates.

"Trace" particles are worth a second look too. The paper describes them as images without discernible contours, thin and ambiguous enough that they were identified and measured by hand. That is a hard category to defend without a blank count next to it.

What the paper did check, and what I could not see

To be fair to the authors, they did consider artifacts. The main text says the onion-like particles were shown not to be artifacts observed during HRTEM, and points to a section of the supporting information titled "evaluation of the artifact during HRTEM observation." I could not open that supporting file, so I do not know what it tested. What I can say is narrower: the word "blank" does not appear anywhere in the main text of the paper as archived on PubMed Central.

That is not proof that no blanks existed. The notice mentions reanalyzing the samples used in the study and finding particles in "the blank samples," which leaves open whether blanks were collected at the time and not examined closely, or prepared later. It does not say, and neither does it give a count, so "a significant number" is the only size we have.

What the retraction does not touch

The notice names the onion-like and trace particles. It does not say jet engines are clean, and it does not address the wider concern about ultrafine particles around airports. The paper's own introduction cites earlier work, including a 2019 study by the same lead author, tying exhaust nanoparticles near a runway to lubrication oil. That line of research was not the subject of this notice, and I have not reviewed it here. The whole article is retracted, though, so none of its numbers should be cited as if they stand, including the type fractions.

Retractions also travel slower than discoveries. EurekAlert now flags the press release at the top, and ZHAW added a notice to its news page. When I checked on 31 October, the English release on the NIES website, ZHAW's own 2025 blog post, and the Nanowerk and StudyFinds write-ups I opened showed no retraction note. If you saw this story in 2025, the correction probably never reached you.

How to read a "newly identified" claim

Is there a blank, treated identically? Not a clean sample from a different day. The same grid, film, filter or vial lot, carried through the same handling, read by the same rules.

Is the blank reported as a number? "Blanks were clean" is weaker than a count per unit area next to the sample count.

Could the substrate make the thing being found? Ask this first when a sample and its support are the same element. Carbon particles on a carbon-related film, or a trace metal in a metal-rich vessel, deserve extra suspicion.

How much of the signal survives if the background is real? Sparse counts with low capture leave very little margin.

Did an expert in the method review it? The fix here came from someone at a meeting who knew TEM films, not from the aerosol result itself.

I do not run electron microscopes, so I will not pretend to second-guess the imaging. But blanks are universal. In any spectroscopy lab, the first question about a surprising peak is whether it is also there when the sample is not. Report 216 covers how much a blank can wander. This is the other half of that lesson: sometimes a blank does not just wander, it contains your discovery.

Credit where due

The authors got a comment at a meeting, took it seriously, tested their own samples, and retracted their own paper. The notice was received on 10 June 2026 and published on 20 July. Nobody forced that, and a less honest group could have ignored it. Scientists finding their own mistake and saying so in public is the system working, and it is the reason this story is useful to anyone else.

The honest limits on this

I read the retraction notice and the paper's main text in full on PubMed Central. I could not open the supporting information, and the ACS site and ScienceDirect blocked automated access. The notice does not report how many blank particles were found, what kind of blank was used, or whether the film named in it matches the collodion-coated grids described in the paper's methods. The link I draw between low collection efficiency and blank sensitivity is my own inference. The Iijima and Harris papers were read as abstracts, and their relevance to this particular grid is the retraction notice's implication for Iijima and only my analogy for Harris.