The Galactic Neutrino Sky: Predictions from Gamma-ray Source Populations

The paper demonstrates that the observed high-energy neutrino emission from the Galactic plane, detected at a significance of 5.7σ with a prominent excess toward the inner Galaxy, can be naturally explained by the spatial distribution of Galactic neutrino sources. The authors construct ReGal-γ, a template of resolved Galactic γ-ray sources that may also produce high-energy neutrinos, by combining γ-ray source catalogs spanning GeV–PeV energies (Fermi-LAT 4FGL-DR4, H.E.S.S. HGPS, 1LHAASO, and 4HWC) and selecting candidate hadronic emitters, excluding pulsars, pulsar wind nebulae, TeV halos, and extragalactic sources. They also use an independent γ-ray source template (the CTA template) and two unresolved source models (Model A from Abe et al. 2024 and Model B from Schwefer et al. 2023) to test robustness.

Compared with models of Galactic diffuse emission, ReGal-γ predicts a neutrino intensity that is more strongly concentrated toward the inner Galaxy. The authors combine the source templates with the CRINGE diffuse emission model and unresolved source populations, assuming a hadronic fraction χ (the fraction of γ-ray emission from hadronuclear interactions) and converting γ-ray fluxes to neutrino fluxes using the relation: (dNν/dEν) per-flavor ≈ (χ/2) × (dNγ/dEγ) at Eγ = 2Eν. They find that the combined model reproduces both the spectral energy distribution and the Galactic longitudinal count profile reported by IceCube without requiring additional renormalization of the emission models, with an average hadronic fraction of χ ∼ 0.5–1 for the inner Galaxy.

The longitudinal profiles at 1, 10, and 100 TeV show that diffuse emission dominates in the outer Galaxy, while the source contribution is more significant in the inner Galaxy. The ReGal-γ template has more sources in the northern sky due to the addition of sources from 1LHAASO and 4HWC, and at high energies it reflects the scarcity of southern-sky sources due to the absence of air-shower observatories in the Southern Hemisphere. The CTA template, in contrast, is dominated by spectral extrapolations from lower energies and is subject to large uncertainties.

The authors convolve the neutrino templates with IceCube's effective area (a function of declination and neutrino energy) and smear the cascade flux templates by 17.5° to account for angular uncertainties, while no smearing is applied to tracks. They find that both ReGal-γ and the CTA template, when combined with the diffuse background and unresolved source populations, can explain the excess observed in the cascade sample in the inner Galaxy region where the IceCube GP fit is most sensitive. The paper notes that most southern-sky sources are drawn from HGPS, which surveys only b ≤ 3°, substantially narrower than the b ≤ 15° region considered in the IceCube analysis, so bright undetected sources may lie just outside the Galactic plane and could further contribute to the neutrino flux in the central bin.

The authors conclude that the spectral and longitudinal distributions of high-energy neutrino emission from the Galactic plane can be simultaneously explained by a Galactic neutrino emission model based on γ-ray observations, combining an unrenormalized diffuse emission component with a population of γ-ray sources not powered by pulsars. The source distribution inferred from γ-ray observations is more concentrated toward the inner Galaxy compared to the diffuse emission, which may explain why the observed neutrino profile is similarly peaked in this region and why fits based solely on diffuse emission templates such as Fermi-π0 and KRAγ require additional renormalization. Despite differences between the ReGal-γ and CTA templates (older catalogs, different spectral models for some sources, different methods of combining catalogs and characterizing spectra), the conclusion that including source populations provides a better explanation of the neutrino spatial distribution remains robust.

The paper also discusses caveats: the γ-ray source templates are constructed from catalogs based on different observing facilities and strategies, resulting in nonuniform sensitivities across sky regions and introducing incompleteness and selection biases. Southern-sky air-shower observatories such as SWGO will be essential for constructing a full-sky very-high-energy γ-ray catalog with more uniform sensitivity. Future measurements of the energy-dependent neutrino morphology from IceCube and KM3NeT will provide tighter constraints on these models and help determine the spatial distribution of Galactic neutrino sources.

Improvements to AI Systems:

1. Astrophysical Source Classification and Selection: Train an AI model to automatically classify γ-ray sources as hadronic (neutrino-producing) versus leptonic (pulsar/PWN/TeV halo) using multi-wavelength spectral features, spatial morphology, and variability. This would replace manual catalog filtering, enabling faster and more consistent construction of neutrino-source templates from future surveys (e.g., SWGO, CTA).

2. Template Construction with Incomplete Sky Coverage: Develop an AI that learns to predict missing source distributions in under-surveyed sky regions (e.g., southern hemisphere) by extrapolating from known source populations, luminosity functions, and Galactic structure models. This would correct the selection bias noted in the paper (HGPS limited to b ≤ 3°) and produce more complete neutrino templates.

3. Hadronic Fraction (χ) Inference: Build a Bayesian neural network that jointly infers the per-source hadronic fraction χ from γ-ray spectra, cosmic-ray density maps, and interstellar medium column densities. This would replace the global χ assumption (0.5–1) with spatially resolved, energy-dependent values, improving neutrino flux predictions.

4. Multi-Template Robustness Analysis: Implement an AI-based ensemble method that automatically combines multiple source templates (ReGal-γ, CTA, unresolved models) with weights learned from observed IceCube data, quantifying systematic uncertainties from catalog differences (energy ranges, spectral models, source overlap) without manual tuning.

5. Energy-Dependent Morphology Prediction: Train a generative model (e.g., normalizing flow) on simulated neutrino maps from combined diffuse+source templates to predict the longitudinal and latitudinal profiles as a function of neutrino energy (1, 10, 100 TeV). This would allow direct, data-driven comparison with future IceCube and KM3NeT measurements, identifying energy regimes where source vs. diffuse contributions dominate.

6. Automated Renormalization-Free Fitting: Develop an AI that fits observed neutrino data to combined templates without requiring arbitrary renormalization, using physical priors (e.g., χ ∈ [0,1], known cosmic-ray flux) and hierarchical modeling to infer source population parameters directly from the data, as demonstrated by the paper’s success with unrenormalized models.

7. Caveat-Aware Uncertainty Quantification: Create an AI system that propagates catalog incompleteness, sensitivity limits, and angular smearing (17.5° for cascades) into final neutrino flux predictions, using probabilistic programming to output credible intervals on the Galactic plane excess, rather than point estimates.

What the Improved AI System Can Do:

• Automatically generate full-sky, bias-corrected neutrino source templates from any combination of γ-ray catalogs, with per-source hadronic fractions and uncertainties.

• Predict neutrino spectral energy distributions and longitudinal profiles at arbitrary energies, directly comparable to IceCube/KM3NeT data, without manual renormalization.

• Distinguish between competing Galactic emission models (diffuse-only vs. diffuse+source) with quantified confidence, using robust multi-template ensembles.

• Forecast the impact of future observatories (SWGO, CTA) on reducing southern-sky source incompleteness and improving neutrino source localization.

• Provide real-time, physically interpretable fits to new neutrino data, enabling rapid testing of hadronic emission hypotheses and identifying anomalous regions requiring deeper γ-ray follow-up.

High-energy neutrino emission from the Galactic plane has been detected at a significance of 5.7 sigma, with a prominent excess toward the inner Galaxy. We show that this excess can be naturally explained by the spatial distribution of Galactic neutrino sources. By combining gamma-ray source catalogs spanning GeV-PeV energies and selecting candidate hadronic emitters, we construct ReGal- gamma, a template of resolved Galactic gamma-ray sources that may also produce high-energy neutrinos. Compared with models of Galactic diffuse emission, ReGal- gamma predicts a neutrino intensity that is more strongly concentrated toward the inner Galaxy. Combined with models of diffuse cosmic-ray emission and unresolved gamma-ray sources, the template reproduces both the spectral energy distribution and the Galactic longitudinal count profile reported by IceCube without requiring additional renormalization of the emission models. We test this result using an independent gamma-ray source template constructed from different catalogs and find that our conclusion is robust against uncertainties in source modeling. Future measurements of the energy-dependent longitudinal and latitudinal neutrino distributions will provide tighter constraints on these models and help determine the spatial distribution of Galactic neutrino sources.