[Paper Review] Recent neutrino oscillation result with the IceCube experiment

Shiqi Yu, Jessie Micallef|arXiv (Cornell University)|Jul 29, 2023

Astrophysics and Cosmic PhenomenaPhysics and Astronomy3 citations

TL;DR

This paper presents a new measurement of atmospheric muon neutrino oscillations using IceCube's DeepCore detector, employing Convolutional Neural Networks (CNNs) for improved neutrino interaction reconstruction. The analysis yields precise constraints on the neutrino mixing angle $ heta_{23}$ and mass splitting $ riangle m_{32}^2$, with results consistent with global world measurements and enhanced sensitivity due to advanced machine learning techniques.

ABSTRACT

The IceCube South Pole Neutrino Observatory is a Cherenkov detector instrumented in a cubic kilometer of ice at the South Pole. IceCube's primary scientific goal is the detection of TeV neutrino emissions from astrophysical sources. At the lower center of the IceCube array, there is a subdetector called DeepCore, which has a denser configuration that makes it possible to lower the energy threshold of IceCube and observe GeV-scale neutrinos, opening the window to atmospheric neutrino oscillations studies. Advances in physics sensitivity have recently been achieved by employing Convolutional Neural Networks to reconstruct neutrino interactions in the DeepCore detector. In this contribution, the recent IceCube result from the atmospheric muon neutrino disappearance analysis using the CNN-reconstructed neutrino sample is presented and compared to the existing worldwide measurements.

Motivation & Objective

• To measure the atmospheric neutrino mixing angle $ heta_{23}$ and mass splitting $ riangle m_{32}^2$ using neutrinos from cosmic ray interactions in Earth's atmosphere.
• To improve sensitivity to neutrino oscillation parameters by leveraging deep learning techniques in neutrino reconstruction.
• To utilize the DeepCore sub-detector’s enhanced low-energy sensitivity to access GeV-scale neutrinos and study oscillation patterns over long baselines.
• To reduce systematic uncertainties by incorporating background-like neutrino candidates into the calibration and simulation framework.
• To compare the new CNN-reconstructed results with existing worldwide measurements, validating the method’s precision and robustness.

Proposed method

• Employed Convolutional Neural Networks (CNNs) to reconstruct neutrino interactions using digitized waveforms from photomultiplier tubes in the IceCube detector.
• Used a dense configuration of digital optical modules (DOMs) in the DeepCore sub-detector to detect Cherenkov light from relativistic charged particles produced in neutrino interactions.
• Reconstructed neutrino energy $E$ and interaction vertex using CNNs trained on simulated and real data, improving energy and angular resolution below 100 GeV.
• Applied the two-flavor oscillation formula $P( u_ u o u_ u) riangleq 1 - ext{sin}^2(2 heta_{23}) ext{sin}^2(1.27 riangle m_{32}^2 L / E)$ to model $ u_ u$ disappearance probability.
• Utilized neutrino events with $ ext{cos}( heta_{ ext{zenith}}) \lesssim 0$ (through-going events) to maximize sensitivity to oscillation parameters.
• Calibrated and simulated the detector using the same framework as previous results, with added machine learning-based reconstruction to refine event selection and background modeling.

Figure 1: Distribution of $\nu_{\mu}$ survival probability with color representing the value of probability at given values of $\cos(\theta_{\rm{zenith}})$ and E with oscillation parameters from the previous result [ 9 ] .

Experimental results

Research questions

• RQ1 What are the improved constraints on $ heta_{23}$ and $ riangle m_{32}^2$ using CNN-reconstructed neutrino events in DeepCore?
• RQ2 How does the application of Convolutional Neural Networks enhance the reconstruction of low-energy neutrino interactions compared to traditional methods?
• RQ3 To what extent do the new results agree with the global world average measurements of atmospheric neutrino oscillation parameters?
• RQ4 What is the sensitivity gain from using deep learning in the reconstruction of neutrino events in the 5–100 GeV energy range?
• RQ5 How do background-like neutrino candidates contribute to reducing systematic uncertainties in the oscillation parameter extraction?

Key findings

• The analysis achieves improved reconstruction accuracy for GeV-scale neutrinos using Convolutional Neural Networks, enabling better energy and angular resolution.
• The measured value of $ riangle m_{32}^2$ is consistent with the global world average, with reduced uncertainties due to enhanced event selection and calibration.
• The mixing angle $ heta_{23}$ is measured with high precision, showing no significant deviation from maximal mixing, consistent with previous global fits.
• The sensitivity to oscillation parameters is strongest for neutrinos traversing the Earth (long baseline, $L \sim 1.3 \times 10^4$ km) with energies between 5 and 100 GeV.
• The CNN-based reconstruction reduces systematic uncertainties by better modeling the detector response and distinguishing signal from background-like events.
• The results are in good agreement with existing worldwide measurements, validating the use of deep learning in high-energy neutrino physics.

Figure 2: IceCube neutrino observatory’s main in-ice array and DeepCore sub-detector (left) and surface layout of strings (right) where red-filled DeepCore strings and orange-circled IceCube main strings are used in the CNN reconstruction.