Where do cosmic rays come from?

In 1991 a detector in Utah recorded a single particle with an energy of about 3 times 10 to the 20 electronvolts, or 51 joules.

It was not a fluke of one instrument. In 2021 the Telescope Array recorded another, of 244 exa-electronvolts, and reported it in 2023. A 2022 search of the Pierre Auger Observatory’s data, in Argentina, found nothing above 166, though the two experiments’ energy scales differ. Arrivals at these energies are extraordinarily rare, and the exposures say how rare: Auger collected thirty thousand particles above 8 exa-electronvolts from 76,800 square kilometre steradian years of watching, and above 100 exa-electronvolts the flux is under one particle per square kilometre per century. That is why the detectors cover hundreds or thousands of square kilometres.

Above about 8 times 10 to the 18 electronvolts, the arrival directions are not uniform. The Pierre Auger Observatory, reporting on thirty thousand events in 2017, found a dipole of about 6.5 per cent: a slight excess from one half of the sky, its axis some 125 degrees from the centre of our own galaxy. With more data it now stands at about 7.4 per cent and 6.8 standard deviations, about 113 degrees from the galactic centre.

That is the cleanest evidence that the particles above this energy are made outside our galaxy. A galactic population would be expected to be brightest towards the galactic centre, and it is not.

A proton above a certain energy collides with photons of the microwave background and loses energy in the process. The effect puts a horizon on the highest-energy particles: above the threshold the sources have to lie within about a hundred megaparsecs, and for a particle as energetic as the 1991 one within about thirty. In cosmological terms that is the neighbourhood.

So the source is outside the galaxy and inside the local universe, which is a small enough volume to contain a catalogue of candidates and a large enough one to contain a great many of them.

Most astrophysical objects cannot hold a particle long enough to accelerate it this far. What survives that test is a short list: neutron stars, the jets of active galaxies, gamma-ray bursts, and accretion shocks in the intergalactic medium and in galaxy clusters.

Each can be made to work on paper. None has been tied to a specific event at the highest energies, and the absence of clear sources in those arrival directions is itself one of the puzzles. The strongest hints so far are statistical: in Auger’s data, an excess of arrivals above 38 exa-electronvolts across a wide region around the radio galaxy Centaurus A, at four standard deviations.

A charged particle is deflected by every magnetic field it passes through, including our own galaxy’s, and those fields are only roughly mapped. Auger puts typical deflections across the galaxy at a few tens of degrees for a particle of 10 exa-electronvolts per unit of charge. A cosmic ray’s arrival direction therefore carries a smeared memory of its origin rather than an address.

One way around this is to look for neutral messengers produced alongside them. In 2017 a neutrino of about 290 tera-electronvolts, detected under the Antarctic ice, arrived from the direction of a blazar that was flaring in gamma rays, and telescopes from radio to gamma rays followed it up. The IceCube team disfavoured a chance coincidence at the level of three standard deviations. Archival data later showed an earlier excess of neutrinos from the same blazar, and a second galaxy, NGC 1068, has since shown an excess of neutrinos at 4.2 standard deviations. That is the closest anyone has come to catching a source in the act, and it does not reach the extreme end: the 2017 neutrino points to protons or nuclei of at least tens of peta-electronvolts.

The 2021 event reported by the Telescope Array arrived from a direction pointing back at a void in the large-scale structure of the universe. The 1991 event had the same problem: Elbert and Sommers searched its arrival direction and found no likely source within 50 megaparsecs.

Either the particle was bent far more than models of the galaxy’s magnetic field allow, because it was a heavy nucleus or the fields are stronger than modelled, or a nearby source has gone unidentified, or the physics of particles at these energies is not fully known. The Telescope Array paper listed these rather than choosing. A later analysis by Unger and Farrar took it for an iron nucleus, found that it then fits the usual picture, and proposed a brief outburst in an otherwise ordinary galaxy no more than 50 megaparsecs away.