Where cosmic neutrinos come from: blazars, NGC 1068 and the Milky Way

A flaring black hole 4 billion light-years away, a single 6-PeV event that confirmed a 1960 prediction, a dusty nearby galaxy and the glow of our own Milky Way: the evidence behind the 2026 Physics Nobel, and how to check it yourself.

By 2014 IceCube had proved that neutrinos arrive from beyond the solar system (part 2). But they seemed to come evenly from the whole sky, and a glow from everywhere does not name a source. This last part follows the decade-long hunt for the sources: a real-time alert that sent telescopes around the world to one flaring galaxy, a single extraordinary event that confirmed a 1960 prediction, a nearby galaxy that shines in neutrinos but hardly in gamma rays, and our own Milky Way. You can redo the key numbers with the relativistic kinematics calculator, the cosmology calculator and the Counting statistics helper.

In short

  • 2017: one neutrino pointed to TXS 0506+056, a “blazar” (a giant black hole firing a jet towards us) about 4 billion light-years away, which was flaring in gamma rays at the same time.
  • 2016 event, published 2021: a 6.3 PeV antineutrino hit an electron and made a W boson, the Glashow resonance predicted in 1960. It also showed that cosmic neutrinos include antineutrinos.
  • 2022: the active galaxy NGC 1068 (about 47 million light-years away) is a steady neutrino source. 2023: neutrinos from the plane of our own galaxy.
  • Neutrinos are the only messengers that are both made only by cosmic-ray factories and travel straight through everything.

Why neutrinos are special messengers

Astronomers have three ways to study the most violent places in the universe: light (especially gamma rays), cosmic rays (fast protons and nuclei), and neutrinos. Watch them leave the same source:

Animation: gamma rays, cosmic rays and neutrinos leaving a source (needs JavaScript).

  • Gamma rays travel straight, but can be absorbed by gas and dust near the source, or on the way by collisions with the faint light filling space. They can also be made by fast electrons, so a gamma-ray source is not necessarily a proton accelerator.
  • Cosmic rays are charged, so magnetic fields bend their paths. By the time they reach Earth they come from everywhere, except at the very highest energies.
  • Neutrinos are made only when fast protons or nuclei hit gas or light and make pions. They are neutral, so they fly straight, and they pass through almost anything. One neutrino from a source is a direct sign that the source accelerates protons.

22 September 2017: an alert, a flare, a galaxy

From 2016 IceCube began sending real-time alerts: whenever a single high-energy track likely to be cosmic arrived, its direction went out to telescopes within a minute. On 22 September 2017 an event of roughly 290 TeV, called IceCube-170922A, arrived from the constellation Orion. The Fermi gamma-ray satellite saw that a known blazar, TXS 0506+056, sat right in the error circle and was in a bright flare. The MAGIC telescopes in the Canary Islands then caught very-high-energy gamma rays from it. The chance that this was a coincidence was about 3σ.

The team then went back through nine and a half years of archived data from that direction. They found more: between late 2014 and early 2015, an excess of about 13 ± 5 neutrinos over about five months, at 3.5σ, with no gamma-ray flare at the time. Two independent signs pointing to one object made TXS 0506+056 the first credible source of high-energy neutrinos (published in Science in July 2018).

A blazar is the core of a galaxy with a supermassive black hole that launches a jet of plasma at close to the speed of light, and we happen to look straight down the jet. TXS 0506+056 has a redshift of z = 0.3365, so its light, and its neutrinos, set out about 3.9 billion years ago.

A 1960 prediction caught at 6.3 PeV

In 1960 Sheldon Glashow pointed out that an electron antineutrino hitting an electron at rest could make a W boson, the carrier of the weak force, if the collision had exactly the W’s mass-energy. Because the electron is so light, that needs an antineutrino of about 6.3 PeV, far beyond any accelerator. At that energy the chance of interaction jumps up sharply: a resonance.

On 8 December 2016 IceCube recorded a cascade depositing about 6.05 PeV, just what a W decaying in the ice should leave once the energy carried off by unseen particles is added. The analysis, published in Nature in 2021, also found early “muon echoes” that suggested the shower contained hadrons, as a W decay would. With one event the significance is modest (2.3σ), but it confirmed the resonance and showed that cosmic sources make antineutrinos, which narrows down how the sources work. The resonance energy is a nice exercise you can do with a calculator (below).

NGC 1068: a galaxy that hides its gamma rays

A single source like TXS 0506+056 does not explain the whole diffuse flux; blazars can only provide a part. The next breakthrough came from a decade of patience. With about ten years of improved, carefully recalibrated track data (2011–2020), IceCube found an excess of about 79 neutrinos at TeV energies from the direction of NGC 1068, also called Messier 77, a well-known “Seyfert” galaxy in Cetus with a dust-shrouded supermassive black hole. The significance was 4.2σ (Science, November 2022).

The surprise: NGC 1068 is far brighter in neutrinos than in gamma rays. If the gamma rays made with the neutrinos had escaped, they would outshine what gamma-ray telescopes see. So they must be absorbed close to the black hole, in the hot, dense corona around it, where X-rays and gas are thick enough to stop gamma rays but not neutrinos. That fits the hint from the diffuse flux in part 2: many neutrino sources may be gamma-ray dark. Since then a few similar Seyfert galaxies (such as NGC 4151) have shown weaker hints.

Our own galaxy shines in neutrinos

Cosmic rays inside the Milky Way hit gas in the galactic disc and must make neutrinos. Gamma-ray satellites see the matching gamma-ray glow along the plane of the galaxy. But the plane runs mostly across the southern sky, where IceCube’s tracks are swamped by muons. The trick that worked used cascades, which can be seen from every direction, and a deep-learning classifier that kept many more good cascade events than earlier methods. In 2023 IceCube reported neutrino emission from the galactic plane at 4.5σ. For the first time we saw our own galaxy in something other than light.

What comes next

  • KM3NeT, a cubic-kilometre detector being built on the floor of the Mediterranean Sea, recorded on 13 February 2023 a muon so energetic that the neutrino behind it probably carried around 220 PeV, the most energetic neutrino ever seen (Nature, 2025). Its origin is still unknown.
  • Baikal-GVD in Lake Baikal is now about half a cubic kilometre and growing. P-ONE is planned in the Pacific off Canada’s coast, and TRIDENT in the South China Sea.
  • IceCube Upgrade adds seven densely instrumented strings in the deep ice to calibrate the ice better and study neutrino oscillations. IceCube-Gen2 would reach about eight cubic kilometres with radio antennas for the highest energies.

Checked with the tools

1. The Glashow resonance energy. For ν̄ₑ + e⁻ → W⁻ with the electron at rest, the collision energy is √s = √(2 E mₑ). Setting √s = m_W = 80.37 GeV gives E = m_W² / (2 mₑ) ≈ 6.32 PeV, the energy where IceCube’s event sits.

Collision energy of an antineutrino hitting an electron at restCollision energy of an antineutrino hitting an electron at rest110100 GeV1 TeV10 TeV100 TeV1 PeVAntineutrino energy (log scale)√s (GeV, log scale)W boson mass, 80.4 GeV6.3 PeVfibub.com
Relativistic kinematics tool's fixed-target formula (electron at rest, W mass 80.37 GeV). The resonance needs a 6.3 PeV electron antineutrino: IceCube saw one event there, from December 2016, published in Nature in 2021.

2. How long the neutrinos travelled. With the cosmology calculator and the Planck 2018 universe, TXS 0506+056 at z = 0.3365 gives a light-travel time of 3.9 billion years: the “about 4 billion light-years” quoted in the news. For NGC 1068 its redshift alone (0.0038) gives about 55 million years, while distance methods that do not depend on the galaxy’s own motion give about 47 million light-years (14 Mpc). For such a close galaxy, its own motion through space muddles the redshift.

How long the neutrinos travelledHow long the neutrinos travelled024600.20.40.60.8Redshift zLight-travel time (billion years)TXS 0506+056: 3.9 billion yearsNGC 1068: 55 million yearsfibub.com
Cosmology calculator maths (Planck 2018: H₀ = 67.66, Ωm = 0.311). Redshifts: TXS 0506+056 z = 0.3365 (Paiano et al. 2018); NGC 1068 z ≈ 0.0038. For a galaxy this close its own motion matters, so its distance is better taken from other methods (about 14 Mpc, or 47 million light-years).

3. Why sources took a decade. A steady source adds neutrinos in proportion to time, while the random scatter of the background grows only as its square root, so the significance grows like √time. With round numbers near the NGC 1068 scale (8 extra neutrinos a year over 60 background in the search spot), the expected significance passes 4σ only after about a decade. Halve the background (better direction measurement) and the wait drops to a few years: this is why IceCube’s 2022 result came with recalibrated, sharper track directions.

Years of data needed to see a steady sourceYears of data needed to see a steady source02468102.557.51012.515Years of dataExpected significance (σ)5σ4.2σ (NGC 1068, 2022)8 per year over 60 background4 per year over 608 per year over 15fibub.com
Our own illustration with the Counting statistics tool's Poisson maths: counts set to their expected values, one search spot, no trials penalty. Rates are round numbers, chosen near the scale of the NGC 1068 search (about 79 extra events over roughly a decade).
Published Our check
Glashow resonance energy 6.3 PeV (event deposited 6.05 PeV visible) 6.32 PeV
TXS 0506+056 light-travel time about 4 billion years 3.9 billion years (z = 0.3365, Planck 2018)
NGC 1068 4.2σ from about 79 events over about a decade about 4σ after 10 years for 8 per year over 60 (illustration)

How we made the graphs

The same script as in parts 1 and 2, scripts/blog-figures/nobel-physics-2026.mjs, makes these charts with threshold() and sqrtSFixed() from src/lib/sci/phrel.ts (Relativistic kinematics), cosmology() with the Planck 2018 preset from src/lib/sci/cosmo.ts (Cosmology calculator) and gammaP() from src/lib/sci/phcount.ts (Counting statistics). Inputs: the W mass (80.37 GeV) and electron mass from the Particle Data Group, the blazar’s redshift from Paiano and colleagues (2018). The source rates in chart 3 are our own round numbers, not IceCube’s.

How we test the maths

tests/sci-tools/relativistic-kinematics.mjs checks the threshold and √s formulas against textbook cases (the antiproton threshold of 6 m_p kinetic energy, π⁰ photoproduction at 144.69 MeV, √s for a beam on a fixed target). The cosmology code is checked in tests/sci.mjs against universes with exact answers: the Einstein–de Sitter universe, an empty universe (where the luminosity distance has a closed form) and the closed-form age of a flat universe with matter and a cosmological constant. The Poisson code is checked against SciPy. All of these run on every change.

Check it yourself

  1. Open the relativistic kinematics calculator and choose the GeV unit.
  2. In 4. Reaction threshold, set the Beam particle to nu, the Target particle (at rest) to e and the Products to w. The threshold beam energy should read about 6.3 × 10⁶ GeV, that is 6.3 PeV.
  3. Change the target to p (a proton at rest). The threshold drops to about 3.4 TeV: a W from a proton target needs far less energy, because the proton is 1,836 times heavier. (In reality it is a quark inside the proton that is hit, so there is no sharp resonance there.)
  4. Open the cosmology calculator, keep Planck 2018 (flat), set Start from to Redshift z and type 0.3365. Read the Lookback time: about 3.9 billion years.
  5. Open the Counting statistics helper and in 4. Poisson probabilities enter μ = 600 and k = 680 (ten years of 60 background plus 8 signal a year). The chance of 680 or more is about 0.07 %: roughly 3.2σ.

What you can use it for

Sources

  • Nobel Prize in Physics 2026, announced 6 October 2026: Francis Halzen, 'for decisive contributions to the IceCube Neutrino Observatory and the discovery of high-energy neutrinos of astrophysical origin'. NobelPrize.org and the Royal Swedish Academy of Sciences (kva.se).
  • IceCube, Fermi-LAT, MAGIC and other collaborations (2018), 'Multimessenger observations of a flaring blazar coincident with high-energy neutrino IceCube-170922A', Science 361: eaat1378.
  • IceCube Collaboration (2018), 'Neutrino emission from the direction of the blazar TXS 0506+056 prior to the IceCube-170922A alert', Science 361: 147-151.
  • Paiano S, Falomo R, Treves A, Scarpa R (2018), 'The redshift of the BL Lac object TXS 0506+056', ApJ Letters 854: L32 (z = 0.3365).
  • IceCube Collaboration (2021), 'Detection of a particle shower at the Glashow resonance with IceCube', Nature 591: 220-224.
  • Glashow SL (1960), 'Resonant scattering of antineutrinos', Physical Review 118: 316.
  • IceCube Collaboration (2022), 'Evidence for neutrino emission from the nearby active galaxy NGC 1068', Science 378: 538-543 (about 79 events, 4.2σ).
  • IceCube Collaboration (2023), 'Observation of high-energy neutrinos from the Galactic plane', Science 380: 1338-1343 (4.5σ).
  • KM3NeT Collaboration (2025), 'Observation of an ultra-high-energy cosmic neutrino with KM3NeT', Nature 638: 376-382 (event KM3-230213A).
  • Cosmological parameters: Planck Collaboration (2020), A&A 641: A6. W boson mass: Particle Data Group (2024), rounded.

If you are an author or publisher and want a credit or a passage changed or removed, tell us and we will act quickly.

Tools in this article

Read next