High-energy neutrinos originate in our own galaxy

Cosmic rays are high-energy particles that move through space with an energy range that roughly spans from a few GeV to several PeV, although most have energies on the lower end of the range. They’re constantly bombarding Earth’s atmosphere and are responsible for some chip-level errors that can flip the value of memory cells in electronic devices. Outside the planet’s protective atmosphere and magnetic field, the highly ionizing radiation poses dangerous acute and chronic health risks for astronauts.

Even if space technology advances enough to quickly and reliably carry people to Mars and beyond, cosmic rays may limit how long and thus how far humans can travel through space, if the shuttles lack radiation-hardened shielding. (See the Quick Study by Larry Townsend, Physics Today, March 2020, page 66.)

Despite their discovery more than 100 years ago, cosmic rays remain mysterious. Astronomers do know that some GeV cosmic rays originate from the Sun’s coronal mass ejections and other solar eruptions. But the theoretical hypotheses for where the high-energy flux comes from—supermassive black holes, active galactic nuclei, supernovae, or something else—are difficult to evaluate with much statistical certainty because of what happens to the cosmic rays as they travel through space.

Along their path to Earth, high-energy cosmic rays often encounter strong galactic magnetic fields, which scramble their trajectories so much that it’s impossible to determine where they came from. Some of the cosmic rays interact with the gaseous interstellar medium and produce particles whose trajectories are easier to track. Those interactions produce both charged and neutral pions. Neutral pions quickly decay and most commonly produce gamma rays. Although observations of the gamma-ray emission hint at where cosmic rays come from, the evidence isn’t conclusive because gamma-ray photons are readily absorbed by other matter in the interstellar medium. One solution to the origin question comes from the other product of cosmic-ray interactions: charged pions, which produce neutrinos.² 

Most visible mass in the Milky Way lies in its galactic plane, and in the interstellar space, hydrogen nuclei are spread out at an average density of one per cubic centimeter. When cosmic rays interact with interstellar hydrogen, gamma rays are produced; and they’ve already been observed by the Large Area Telescope aboard NASA’s Fermi Gamma-Ray Space Telescope. The cosmic-ray interactions also make neutrinos, which are most visible in Earth’s southern sky if atmospheric noise is minimized, according to calculations of the estimated neutrino flux. In part for that reason, the Amundsen–Scott South Pole Station in Antarctica hosts the IceCube Neutrino Observatory.

The detector is a cubic kilometer of ice with more than 5000 spherical optical sensors, which hang on cables drilled as deep as 2500 m. Most neutrinos pass through the detector area without interacting, but when one slams into a water molecule, it produces muons and other charged particles. If any of them are energetic enough, they travel faster than light does in the ice, which results in the emission of blue-tinged Cherenkov radiation, as illustrated in figure 1. (The effect is similar to the sonic boom that’s heard when a supersonic jet travels faster than the speed of sound.)

By measuring the Cherenkov radiation picked up by the optical sensors and the blue light’s spatial pattern, researchers can infer the particle’s energy and from what direction it arrived. One of the biggest challenges, however, is the overwhelming number of neutrinos and muons produced in Earth’s atmosphere by cosmic rays. For every astrophysical neutrino that IceCube observes, the facility measures 100 million atmospheric muons.

To filter the immense background noise from the astrophysical neutrino signal, the IceCube Collaboration had developed an event-selection protocol. A neutrino signal measured by IceCube is categorized either as a track event, in which a neutrino creates a muon that then leaves an energetic trail as it races through the ice, or as a cascade event, in which the neutrino releases most or all of its energy at a single location. Many atmospheric neutrinos leave track events in the IceCube detector, so selecting for just cascades greatly reduces the background noise.

The approach is computationally expensive. Many neutrino events need to be excluded, including those found in certain detector regions that are known to witness many atmospheric muons and neutrinos with energies close to those of atmospheric neutrinos. The exclusions have left the IceCube Collaboration with a high-quality data set that it has used for previous analyses—for example, in 2013, the team discovered an extragalactic source of neutrinos.³ The data set, however, was too small to draw conclusions with high statistical confidence about galactic neutrino emission.

To better analyze the 10 years of data collected from IceCube so far, the collaboration developed a hybrid artificial-intelligence technique, spearheaded by Mirco Hünnefeld of the Technical University Dortmund and Steve Sclafani of Drexel University.⁴ The first part of the technique uses a deep-learning neural network to identify cascade neutrino events. The neural network runs through the data quickly, and the resulting time savings allows more events, including those at the lower end of the energy range, to be included in the data set. Compared with earlier analyses, the neural network also improves by a factor of two the angular resolution of the direction of incoming neutrinos.

The second part of the technique uses the light patterns collected by the detectors to reconstruct the neutrino’s direction and energy. The previous reconstruction method was based on a computationally expensive set of Monte Carlo simulations, but the neural network approximates those simulations more efficiently. The new data set of 60 000 neutrino events is 20 times as large as previous sets. Francis Halzen, a theoretical physicist who is a member of the IceCube Collaboration, says that “given the new machine-learning techniques, it almost looked easy in retrospect, especially given our unsuccessful efforts to see our galaxy over the last decade.”

With those improvements, the collaboration found a signal of diffuse high-energy neutrinos from the galactic plane. Figure 2 shows the Milky Way as it’s typically been seen through radio, optical, and gamma-ray emissions. Those three wavelength regimes show a clear, bright galactic center flanked on either side by a thin plane of more diffuse emission. The Milky Way’s neutrino signal is about 10% of the total flux at 30 TeV, and although it may not be immediately clear, the signal is consistent with the gamma-ray emission. In fact, the IceCube Collaboration found that the chance for the neutrino signal to arise randomly from background noise is less than 1 in 100 000, or a statistical confidence level of 4.5 standard deviations.

The neutrino picture of the Milky Way so far shows a diffuse pattern of emission that arises when cosmic rays interact with interstellar gas. But neutrinos could be coming from galactic point sources of cosmic rays. To figure that out, the team will need to review more data. Similar to how the atmospheric background overwhelms the astrophysical signal, the suspected point-source neutrino signal may be masked by the diffuse neutrino signal shown in figure 2.

Astronomers already have some ideas for where galactic cosmic rays, and by extension neutrinos, could come from. Supermassive black holes could produce them (see Physics Today, August 2022, page 14), but the one at the center of the Milky Way may not be active enough to generate such high-energy particles.

Neutrinos could also be sourced from active galactic nuclei (see “IceCube pinpoints an extragalactic neutrino source,” Physics Today online, 12 July 2018), and some have already been spotted in an extragalactic supernova (see “A supernova for the ages, 30 years later,” Physics Today online, 23 February 2017). In the Milky Way, says Halzen, “finding the sources are our next priority, and we are on it.”