Inferring the Flavor of High-Energy Astrophysical Neutrinos at Their Sources

The sources and production mechanisms of high-energy astrophysical neutrinos are largely unknown. A promising opportunity for progress lies in the study of neutrino flavor composition, i.e., the proportion of each flavor in the flux of neutrinos, which reflects the physical conditions at the sources. To seize it, we introduce a Bayesian method that infers the flavor composition at the neutrino sources based on the flavor composition measured at Earth. We find that the present data from the IceCube neutrino telescope favor neutrino production via the decay of high-energy pions and rule out production via the decay of neutrons. In the future, improved measurements of flavor composition and mixing parameters may single out the production mechanism with high significance.

Figure 1

Flavor composition of high-energy astrophysical neutrinos at their sources, inferred from present IceCube measurements [23] (bottom) and from the projected sensitivities of the near-future IceCube upgrade [28] (center) and planned IceCube-Gen2 [29] (top), assuming production by pion decay. Here we enforce a prior of no ${\nu }_{\tau }$ production, i.e., $f_{\tau ,S}=0$. We show the most probable values (white dotted lines) and credible intervals (blue shaded regions) of $f_{e,S}$ and mark physically motivated neutrino production scenarios: pion decay, muon-damped, and neutron decay.

Figure 2

Left: Flavor composition of high-energy astrophysical neutrinos at Earth, approximating current IceCube measurements [23], expressed in terms of variations in the likelihood, $-2\mathrm{\Delta }\ln {\mathcal{L}}_{\oplus }$. The contours show the 68% and 95% confidence regions; This triangle was produced by the IceCube Collaboration using a frequentist approach. We include expectations from three benchmark production scenarios, computed with mixing parameters fixed at their best-fit values—shown as symbols—and varied within their $3\sigma$ ranges [58, 59]—shown as bounded regions. Right: Flavor composition at the neutrino sources, inferred based on current measurements of flavor in IceCube and of mixing parameters in oscillation experiments [58, 59]. We assume no prior on the flavor composition at the sources. The contours show the 68%, 90%, 95%, and 99% credible regions; this triangle was produced by the procedure introduced here using a Bayesian approach.

Figure 3

Left: The same as Fig. 2, left, but showing the projected flavor sensitivity of the IceCube upgrade, approximated from Ref. [28]. Right: The same as Fig. 2, right, but showing the projected performance of the IceCube upgrade in inferring $f_{\alpha ,\oplus }$.