Signatures of quasi-Dirac neutrinos in diffuse high-energy astrophysical neutrino data

TL;DR

This paper tests quasi-Dirac neutrino models by searching for ultra-long-baseline oscillation signatures in IceCube's diffuse high-energy neutrino flux. By combining Cascade and ESTES track samples and modeling the flux with a broken power-law spectrum, the authors fit for a common mass-squared splitting $\delta m^2$ across generations, incorporating cosmological source distributions via $\rho(z)$ and a variety of spectral shapes. They find that the absence of disappearance features around 50–100 TeV disfavors $\delta m^2$ in the range $(5-7.5)\times10^{-19}\,\mathrm{eV}^2$ (3σ) for SFRD-like redshift evolution, with robust results across alternative redshift and spectral assumptions; extensions to two distinct $\delta m^2$ values yield only modest improvements in fit, highlighting the need for more data. Overall, the work demonstrates that diffuse all-sky neutrino measurements can probe new regions of massive-neutrino parameter space and guide future observational strategies, complementing point-source analyses for confirming ultra-long-baseline oscillation phenomena.

Abstract

Although the sources of astrophysical neutrinos are still unknown, they are believed to be produced by a population of sources in the distant universe. Measurements of the diffuse, all-sky astrophysical flux can thus be sensitive to flavor and energy-dependent propagation effects, such as very long baseline oscillations. These oscillations are present in certain neutrino mass models, such as when neutrinos are quasi-Dirac. Assuming generic models for the source flux, we find that these oscillations can still be resolved even when integrated over wide distributions in source redshift. We use two sets of IceCube all-sky flux measurements, made with muon and all-flavor neutrino samples, to set constraints at the $3σ$ level on quasi-Dirac mass-splittings between $(5 \times 10^{-19}, 8 \times 10^{-19})~\textrm{eV}^2$. We also consider systematic uncertainties on the source population and find that our results are robust under alternate spectral hypotheses or physical redshift distributions. Our analysis shows that spectral features in the all-sky neutrino measurements provide strong constraints on massive neutrino scenarios and are sensitive to uncharted parameter space.

Figures (10)

• Figure 1: Constraints on the Quasi-Dirac parameter space. The test statistic difference with respect to the broken power-law null hypothesis, as a function of the QD mass-squared difference $\delta m^2$, is plotted for analyses based on IceCube's 2025 CombinedFitIceCube:2025ewu (blue) and on a combination of Cascades (2020) IceCube:2020acn and ESTES (2024) IceCube:2024fxo results (purple). Brazil bands indicate the regions contained by 95% of realizations drawn from the best-fit BPL spectrum in IceCube:2025ewu. Previous constraints from Ref. Martinez-Soler:2021unz and sensitivities from Ref. Carloni:2022cqz are shown in grey.
• Figure 2: Source distribution and integrated oscillation probability.Left: Three choices for the source distribution in redshift: a flat distribution, the SFRD of Ref. Elias-Chavez:2018dru, and the distribution of BL Lacs from Ref. Groth:2025aan. Middle: Distribution of neutrinos with energy $E = 40TeV$ over the effective distance ($L_\textrm{eff}$), assuming a given redshift distribution $\rho(z)$ and power-law emission spectrum with index $\gamma$. Right: QD survival probability as a function of detected neutrino energy, integrated over the population of sources at all redshifts.
• Figure 3: Constraints on the QD mass-squared difference, under different choices for the source redshift distribution $\rho(z)$.Left: Redshift distributions of four source populations from Ref. Groth:2025aan, as well as the SFRD of Ref. Elias-Chavez:2018dru. Note that as discussed in the text, models in which neutrino sources are distributed uniformly over redshift, such as that of the LL AGNs in this plot, are disfavored by source studies Capel:2020txc. Right: The resulting $\textrm{TS}$ curves calculated based on the CombinedFit flux points, using a BPL emission spectrum.
• Figure 4: Flux predictions and IceCube CombinedFit measurements.Below: The best-fit QD flux hypotheses (grey lines) for three different values of the mass splitting $\delta m^2$, compared with the CombinedFit piecewise flux. The black points indicate the predicted piecewise flux in each bin according to the QD model. Above: The test statistic difference as a function of $\delta m^2$. The shaded bands indicate the 68% and 95% most probable regions, generated by sampling from the published CombinedFit best-fit BPL flux.
• Figure 5: Flux predictions and IceCube Cascades + ESTES measurements. The best-fit QD flux hypotheses (grey lines) for three different values of the mass splitting $\delta m^2$, compared with the Cascades and ESTES piecewise fluxes. Above: The test statistic difference as a function of $\delta m^2$. The shaded bands indicate the 68% and 95% most probable regions, generated by sampling from the published CombinedFit best-fit BPL flux.