40 joules. That is the kinetic energy packed into a single subatomic particle that punched into the atmosphere above Millard County, Utah, on the night of 27 May 2021. By the time the surface detectors of the Telescope Array (TA) experiment finished triangulating the air shower, the reconstructed energy had settled at 244 ± 29 (statistical) +51/-76 (systematic) exa-electron volts, the most energetic cosmic ray the TA collaboration has ever recorded and the second-highest-energy cosmic ray ever measured, trailing only the 1991 "Oh-My-God" event detected by the Fly's Eye experiment at roughly 320 EeV. The collaboration named the particle Amaterasu, after the Japanese sun goddess [1].
That single number, 244 EeV, is the entry point into a 2026 debate that is quietly rewriting a core assumption of ultrahigh-energy cosmic ray (UHECR) physics: that the most energetic particles in the universe are protons or light-to-intermediate nuclei. A paper published in Physical Review Letters in May 2026 by B. Theodore Zhang, Kohta Murase, Nick Ekanger, Mukul Bhattacharya and Shunsaku Horiuchi argues that Amaterasu, and a population of similar events, are ultraheavy nuclei, with atomic numbers well above iron, in the range of selenium, tellurium, or platinum. If correct, the claim changes what cosmic ray observatories should be looking for, where these particles could come from, and how they get accelerated in the first place [2].
The Detection and the Void It Came From
The Telescope Array's surface array, 507 scintillator stations spread across roughly 700 square kilometres of west-desert Utah, registered Amaterasu on 27 May 2021. There was no coincident fluorescence detector trigger for this event, which matters: fluorescence telescopes measure the longitudinal development of the air shower and, from it, can estimate the depth of shower maximum (Xmax), the standard proxy for primary mass. Without that, the TA collaboration could reconstruct the particle's energy and arrival direction but not its identity. They published what they had in Science in November 2023, led by R. U. Abbasi and the rest of the TA collaboration [1].
The arrival direction was (RA, Dec.) = (255.9° ± 0.6°, 16.1° ± 0.5°). Trace that line back through the solar neighbourhood, accounting for the Galactic magnetic field, and the source region lands inside the Local Void, a roughly 60 megaparsec-scale underdensity in the local large-scale structure containing no known galaxies, no clusters, no obvious accelerator. The three explanations offered in the 2023 paper are still the field's working set: a large magnetic deflection from a recognisable source, an unidentified source in the local extragalactic neighbourhood, or incomplete knowledge of particle physics at the highest energies [1].
For most of the past three years, the debate has been framed as a source problem: where, exactly, did this thing come from? The 2026 papers, including the Zhang et al. result, are reframing it as a composition problem first, and a source problem second.
The Ultraheavy Hypothesis (Zhang et al., 2026)
The paper that put the ultraheavy-nuclei claim on the table is "Ultraheavy Ultrahigh-Energy Cosmic Rays," published in Physical Review Letters 136, 181002, on 7 May 2026, by B. Theodore Zhang (Yukawa Institute for Theoretical Physics, Kyoto), Kohta Murase (Penn State and Yukawa), Nick Ekanger, Mukul Bhattacharya, and Shunsaku Horiuchi [2]. The arXiv preprint dates to 27 May 2024, the same day the Amaterasu energy was first publicly highlighted.
The core argument is structural. Photodisintegration, the process that strips nucleons from a nucleus when it collides with a cosmic microwave background (CMB) or extragalactic background light (EBL) photon, has a threshold that scales with the nuclear binding energy. Iron-56, the heaviest stable nucleus produced in quantity by stellar nucleosynthesis, has a binding energy per nucleon around 8.8 MeV, and the dominant CMB/EBL interactions efficiently chop it up over cosmic distances. Push above iron, to elements like selenium-82, tellurium-128, or platinum-195, and the binding energy per nucleon climbs back toward the iron peak (or beyond it for the most stable heavy isotopes). Zhang and colleagues show that the energy loss lengths of these ultraheavy (UH) nuclei above 100 EeV are "significantly longer than those of protons and intermediate-mass nuclei" [2].
That longer loss length does two things. It explains how an Amaterasu-class event could plausibly arrive from a source inside or near the Local Void without being shredded en route. It also explains a long-standing puzzle in the field: the spectral tension between the Telescope Array and the Pierre Auger Observatory at the highest energies. The two experiments have, for nearly a decade, reported subtly different flux normalisations and shapes above about 50 EeV. The standard excuses have ranged from energy-scale calibration differences to declination-dependent exposure. Zhang et al. propose that an enhanced contribution of UH nuclei from a nearby transient source, one whose light has not yet reached us, could reconcile the two datasets [2].
They also make a testable prediction. The mean depth of shower maximum, ⟨Xmax⟩, above 100 EeV should be lower (shallower showers) for ultraheavy primaries than for pure iron primaries. AugerPrime, the upgraded Pierre Auger Observatory, and the planned Global Cosmic Ray Observatory (GCOS) will be able to measure Xmax with the precision needed to confirm or rule out the claim. The current UHECR dataset, the authors write, is "consistent with energy generation rate densities of UH nuclei from collapsars and neutron-star mergers," which gives the hypothesis concrete candidate accelerators [2].
The qualifier "may" matters. The paper is explicit that this is a possibility to be tested, not a confirmed identification. The arrival direction and energy are consistent with an ultraheavy interpretation. The composition cannot be measured directly from the surface array data that captured the event.
A Bayesian Search for Origins (Bourriche & Capel, 2026)
A separate 2026 study, published in The Astrophysical Journal on 28 January 2026 by Nadine Bourriche and Francesca Capel, both at the Max Planck Institute for Physics and the Technical University of Munich, attacks the problem from a different angle. Rather than asking what Amaterasu is, they ask where it came from [3].
Their tool is a simulation-based inference (SBI) framework that combines CRPropa 3D propagation simulations, including all relevant interactions and magnetic deflections in both Galactic and extragalactic fields, with Approximate Bayesian Computation. The output is a posterior distribution over source position, distance, energy, and magnetic-field properties for a single UHECR event, given its measured arrival direction and reconstructed energy.
Applied to Amaterasu, the framework "reveals a broader set of nearby source candidates than found in previous analyses" [3]. The catch is in the prior. Bourriche and Capel's fiducial model assumes the particle arrives as an iron nucleus (Z = 26), with comparison runs for protons, nitrogen, and silicon. They explicitly do not claim the particle is ultraheavy. Their work is complementary: it maps what the magnetic deflection field looks like under different composition assumptions, so that when (and if) the composition is pinned down by Xmax measurements, the posterior over source regions can be sharpened.
The conference-proceedings version of the same study, presented at the UHECR 2024 symposium in Malacca, Malaysia, is openly accessible and lays out the same CRPropa 3 + ABC methodology in slightly more introductory form [4].
A Competing Proton Hypothesis (Das, Hazra & Gupta, 2025)
A third paper, published in The Astrophysical Journal Letters in August 2025 by Saikat Das, Srijita Hazra, and Nayantara Gupta, takes the opposite end of the composition spectrum. It argues that Amaterasu arrived as a proton, accelerated by a specific, identifiable source: the blazar PKS 1717+177, at redshift z = 0.137, sitting 2.5° from the reconstructed arrival direction [5].
The proton scenario runs into an energy-loss problem of its own. A proton at 244 EeV cannot propagate unattenuated across hundreds of megaparsecs through the CMB; it would photopion-produce and lose energy long before reaching us. Das and colleagues invoke Lorentz invariance violation (LIV) at the relevant energy scale to keep the proton alive in transit. That is an exotic-physics assumption, and it puts the model in a different epistemic category from the Zhang et al. result, which uses only standard nuclear physics and a well-known interaction suppression.
The two interpretations, ultraheavy nuclei from a transient source on the Zhang side, proton from a blazar with exotic propagation physics on the Das side, are mutually exclusive. They are also, in a sense, two ways of answering the same question: what do you have to assume is different about the particle, or about the laws of physics, to make the Local Void arrival direction plausible?
What It Means for the Field
The Zhang et al. result, if borne out, has three downstream consequences. First, the candidate-source list shifts away from steady-state accelerators like active galactic nuclei and toward transients, collapsars (the collapsing cores of massive stars that produce long-duration gamma-ray bursts) and neutron-star mergers, both of which can plausibly accelerate nuclei as heavy as platinum. Second, the experimental priority list shifts: composition measurements, specifically Xmax and its fluctuations, become the discriminating observable, not just energy spectra and arrival directions. Third, the energy-loss budget of UHECRs gets a major revision, which in turn affects predictions for cosmogenic neutrino and photon fluxes at the highest energies.
Bourriche and Capel's Bayesian framework fills in the middle of the chain: once you have a composition hypothesis, you can ask where, in the local universe, a particle with that composition and that energy, deflected through the known magnetic field structure, could plausibly have come from [3]. The two analyses, ultraheavy composition plus source-inference under different priors, are designed to be used together.
The energy-reconstruction uncertainty from the original TA measurement, +51/-76 EeV systematic on top of ±29 EeV statistical, is large enough that the central value could plausibly range from about 170 to 295 EeV depending on what composition the analysis assumes [1]. That is not a small spread. It is the difference between "extreme but plausible" and "right at the boundary of what current physics can accelerate."
What Comes Next
For the science-curious reader, the practical translation of the Zhang et al. result is this: the most energetic single subatomic event ever measured in the Northern Hemisphere may not be what the field has spent three decades assuming it was. The data already in hand cannot decide the question. The data that can, Xmax distributions from AugerPrime and the future Global Cosmic Ray Observatory, will take years to accumulate.
In the meantime, the work to follow is the work to read. Zhang et al. (PRL 136, 181002, 2026) lays out the ultraheavy hypothesis and its testable predictions [2]. Bourriche and Capel (ApJ 997, 264, 2026) show how the source-origin question changes once composition is held up as a variable rather than a fixed assumption [3]. Das, Hazra and Gupta (ApJL 988, L8, 2025) demonstrate that the proton interpretation is still alive, but only with the help of exotic physics [5]. The 2023 Abbasi et al. detection paper remains the foundational document for the event itself [1].
If you want one thing to take away: do not treat the next headline about Amaterasu as a settled fact in either direction. The particle arrived on 27 May 2021, and the argument about what it is has barely begun. Follow the Xmax measurements as AugerPrime publishes them, and read the Zhang et al. paper closely when you do. The data will tell us what Amaterasu is. Until then, treat every confident claim, on either side, as provisional.