Determining neutrino oscillation parameters from atmospheric muon neutrino disappearance with three years of IceCube DeepCore data

We present a measurement of neutrino oscillations via atmospheric muon neutrino disappearance with three years of data of the completed IceCube neutrino detector. DeepCore, a region of denser IceCube instrumentation, enables the detection and reconstruction of atmospheric muon neutrinos between 10 and 100 GeV, where a strong disappearance signal is expected. The IceCube detector volume surrounding DeepCore is used as a veto region to suppress the atmospheric muon background. Neutrino events are selected where the detected Cherenkov photons of the secondary particles minimally scatter, and the neutrino energy and arrival direction are reconstructed. Both variables are used to obtain the neutrino oscillation parameters from the data, with the best fit given by $\mathrm{\Delta }{m}_{32}^2={2.72}_{-0.20}^{+0.19}\times {10}^{-3}{\mathrm{eV}}^2$ and ${\sin }^2{\theta }_{23}={0.53}_{-0.12}^{+0.09}$ (normal mass ordering assumed). The results are compatible, and comparable in precision, to those of dedicated oscillation experiments.

Figure 1

A simulated 12 GeV ${\nu }_{\mu }$ interacting in DeepCore and producing an 8 GeV muon (42-m range) and a 4 GeV hadronic shower. In (a), the dashed vertical lines are detector strings, the star marks the position of the interaction vertex and the solid line is the muon track. Twenty DOMs record photons; they have colors related to the photon arrival time (lighter is earlier) where their size is proportional to the charge observed. In (b), the DOM depth as a function of the arrival time of photons at the string with most light collected is shown. Marker sizes scale with charge. The expected hyperbolae from simulation, a track fit and the same fit altered by 25° are also shown.

Figure 2

Zenith angle distributions of neutrino simulation and atmospheric muons derived from data for three subsequent steps in the event selection with increasing veto cuts. To go from the first to the second panel the veto cut which uses a muon track hypothesis is applied. A cut on the charge observed above and prior to the trigger is used to go from the second to the third panel. A comparison is also made to a 10% control sample of the data. The small excess in the data around $\cos {\theta }_z\simeq 0.3$ in the first panel are atmospheric muons that could not be tagged. Note that the region $\cos {\theta }_z>0$ is not used in the final analysis of the data.

Figure 3

Difference in arrival time between direct photons and the expected hyperbola from the track fit, comparing simulation with 5% of the final data sample.

Figure 4

True energy distribution of simulated neutrino events in the final sample, with hatched areas representing each component. Only events used for the final result are considered [$E_{\text{reco}}=[6,56]\mathrm{GeV}$, and $\cos ({\theta }_{\text{reco}})<0$]. The missing ${\nu }_{\mu }$ component is also shown.

Figure 5

Comparison between data and expectations for the case of oscillations and no oscillations. In each figure the zenith distribution for an energy band is shown (top), and the ratio of the data and the best fit to no oscillations is shown (bottom). The binning corresponds to that used for obtaining the best-fit oscillation parameters. Bands indicate the impact of the assumed systematic uncertainties.

Figure 6

Distribution of events as a function of reconstructed $L/E$. Data are compared to the best fit and expectation with no oscillations (top), and the ratio of data and best fit to the expectation without oscillations is also shown (bottom). Bands indicate estimated systematic uncertainties.

Figure 7

90% confidence contours of the result in the ${\sin }^2{\theta }_{23}-\mathrm{\Delta }{m}_{32}^2$ plane in comparison with the ones of the most sensitive experiments [8, 9, 10]. The log-likelihood profiles for individual oscillation parameters are also shown (right and top). A normal mass ordering is assumed.