Fermi Data Revise the Cosmic-Ray Picture in Starburst Galaxies

Cosmic rays are energetic particles whose interactions with gas, light and magnetic fields illuminate galaxies across the electromagnetic spectrum. A new study argues that more careful accounting of gamma rays changes what astronomers can infer about these particles in galaxies undergoing intense star formation.

In an author manuscript posted to arXiv on November 4, 2026, and accepted by PASJ following peer review, Hinako Ochi and colleagues analyze 15 years of observations from the Fermi Large Area Telescope. Their seven targets include the bright starburst galaxies M82 and NGC 253, the nearby galaxies M31 and M33, and three fainter starbursts. These are previously detected sources, not new galaxy discoveries.

The central result concerns a downturn in the gamma-ray energy output of M82 and NGC 253 below a few hundred million electronvolts. According to the authors, this feature removes the need for the unusually large cosmic-ray electron contribution inferred in some earlier work.

Why the spectrum matters

Cosmic-ray protons can collide with interstellar gas, producing neutral pions that decay into gamma rays. Electrons also generate gamma rays through interactions with gas and by boosting lower-energy photons. These processes leave different spectral signatures, making the distribution of gamma-ray energies a diagnostic of the underlying particles.

But Fermi’s ability to pinpoint a photon’s origin worsens at lower energies. Emission from a target galaxy can become entangled with nearby sources and diffuse foreground radiation.

For M82 and NGC 253, the researchers compared a simple power-law spectrum with a curved alternative. They report that the power-law model artificially raised the galaxies’ inferred low-energy emission by assigning less of it to surrounding components. The curved model better matched the measured spectra. Their analysis also used stricter selections for photons with better directional information at low energies.

Combining the revised Fermi measurements with published radio and higher-energy gamma-ray observations, the team modeled the galaxies’ proton and electron populations. The inferred electron-to-proton ratios were roughly one electron per 25–30 protons at 1 GeV, falling to about one per 120–150 at 10 GeV. The authors describe these values as comparable to those in the Milky Way.

Magnetic fields hold more energy

The resemblance does not extend to the balance between particles and magnetic fields. Under the study’s adopted environmental parameters, the inferred cosmic-ray proton energy density is about eight times below the magnetic-field energy density in M82 and about 2.5 times below it in NGC 253.

That challenges the assumption that cosmic rays and magnetic fields hold approximately equal energy, an assumption often used to estimate galactic magnetic fields from radio emission.

These are model-dependent conclusions, not independent measurements of every ingredient. The fits treat each starburst as a single emission region and adopt gas densities, radiation fields and magnetic-field strengths from literature-based estimates. The authors acknowledge substantial magnetic-field uncertainty. However, halving M82’s adopted field made the electrons needed to explain its radio emission overproduce low-energy gamma rays, preventing a satisfactory fit.

How effectively do galaxies trap particles?

The study also revises the gamma-ray brightness of Arp 220 and NGC 2146 downward by approximately a factor of two. The authors attribute those changes to improved background models and additional photon statistics.

The revised measurements bring the sampled starbursts close to or below the expected calorimetric limit: the benchmark gamma-ray output when cosmic-ray protons lose their energy through interactions rather than escaping. Ratios of measured luminosity to that benchmark range from 0.33 to 1.20. The upper value exceeds unity numerically, so the result is agreement near the benchmark, not a strict demonstration that every estimate lies below it.

Ordinary star-forming galaxies in the comparison generally show ratios around 10%. The Small Magellanic Cloud, included through published measurements, is an exception at 28%; the authors suggest individual sources may contribute significantly there.

Andromeda’s unresolved center

For M31, or Andromeda, the analysis confirms earlier evidence for two gamma-ray components. Its central component has a soft, curved spectrum and a 95% upper limit on its radius of 0.33 degrees—about 4.5 kiloparsecs, smaller than the galaxy’s prominent 10-kiloparsec gas ring.

The authors argue that unresolved sources must contribute substantially. Millisecond pulsars remain a possibility, but a population resembling typical Milky Way examples would not match the measured spectrum. An older, spectrally softer population is proposed as a possibility requiring further modeling.

The study thus offers revised constraints rather than a universal explanation: starbursts appear efficient at converting cosmic-ray energy into radiation, while identifying the emitters in quieter galaxies remains more complicated.