Impact of multi-messenger spectral modelling on blazar-neutrino associations

TL;DR

Blazars are candidate sources for astrophysical neutrinos, but the inferred connections depend sensitively on the assumed neutrino energy spectrum. The authors implement a Bayesian hierarchical framework (hnu) that integrates physically motivated $p\gamma$ spectra from lepto-hadronic models with traditional power-law spectra, applied to IceCube muon-track data from the Northern sky. They find that peaked $p\gamma$ spectra reduce low-energy associations yet can maintain or enhance links to high-energy events, and that informative priors on $E_{\mathrm{peak}}$ dramatically constrain fluxes and $\bar{n}$. Overall, physics-informed spectral modelling improves interpretability of blazar–neutrino connections and highlights new promising candidates for follow-up, while also pointing to the need for time-dependent analyses and improved detector systematics.

Abstract

Blazars are interesting source candidates for astrophysical neutrino emission. Multi-messenger lepto-hadronic models based on proton-photon (p-gamma) interactions result in predictions for the neutrino spectra (''p-gamma spectra'') which are typically strongly peaked at PeV energies. In contrast, statistical analyses looking to associate blazars and high-energy neutrinos often assume a power-law spectral shape, putting the emphasis at lower energies. We aim to examine the impact of such spectral modelling assumptions on the associations of neutrinos with blazars. We use hierarchical_nu, a Bayesian framework for point source searches, and incorporate the theoretical predictions for neutrino spectra through a dedicated spectral model and priors on the relevant parameters. Our spectral model is based on recent predictions for a selection of intermediate and high synchrotron peaked blazars that have been found to be spatially close to high-energy events detected by IceCube. We apply our model to the 10 years of publicly available muon track IceCube data aimed at point source searches, focusing on the Northern hemisphere. Out of 29 source candidates, we find five sources, including TXS 0506+056, that have an association probability $P_\mathrm{assoc} > 0.5$ to at least one event. The p-gamma spectra typically lead to a lower overall number of associated events compared to the power-law case, but retain or even enhance strong associations to high-energy events. Our results demonstrate that including more information from theoretical predictions can allow for more interpretable source-neutrino connections.

Figures (8)

• Figure 1: Top panel: Averaged $p\gamma$ spectrum (blue) and approximation used in this work (orange). Bottom panel: Ratio of $p\gamma$ and approximation.
• Figure 2: Joint $\bar{n}$ and $\gamma$ posterior. Filled contours show 68%, 90% and 95% credible regions from darkest to lightest. Black contours show 68%, 90% and 95% confidence levels found with SkyLLH. The MLE is marked by a black cross. Top and right panel show marginalised posterior densities of the same credibilities. Black errorbars show MLE and $1\sigma$ uncertainties found with SkyLLH.
• Figure 3: Analysis of TXS 0506+056 using a power-law. The colour of dots and lines reflects the posterior-averaged event association probability to TXS 0506+056. Left panel: Scatter plots of analysed events, projected onto the sky. The dots' size is not connected to energy or angular uncertainty. ROI is centered on TXS, marked by a cross. The alert event IC170922A is marked by a red circle. Right panel: Marginalised neutrino energy posteriors of all events, transformed to $\log_{10}(E_\nu/\GeV)$. The upper axis shows the reconstructed muon energy of events, $\hat{E}$. The reconstructed energy and energy posterior of IC170922A are linked by a dashed line.
• Figure 4: Analysis of TXS 0506+056 using the $p\gamma$ spectrum. Coloured bands are 50%, 68% and 95% credible regions of fluxes. Dark blue, green and yellow closed contours are 50%, 68% and 95% credible regions of joint posterior density of $E_\mathrm{peak}$ and peak energy flux. Grey line is best fitting neutrino flux found using SkyLLH, dashed contours are 68% confidence level on $E_\mathrm{peak}$ and peak energy flux. Red line is neutrino flux prediction. Left panel: Uninformative priors. Right panel: Informative $E_\mathrm{peak}$ prior. All energies are defined in the detector frame.
• Figure 5: Comparison of fits to TXS 0506+056 with different models and priors. Top panel: Posterior of expected number of events. Dashed lines show the posterior means. Bottom panel: Energy posterior of IC170922A. Dashed lines show the geometric means of the posteriors.