[Paper Review] Search for neutrino flares from point sources with IceCube

J. L. Bazo Alba, E. Bernardini+1 more|arXiv (Cornell University)|Aug 28, 2009

Astrophysics and Cosmic Phenomena3 references3 citations

TL;DR

This paper presents a time-clustering, unbinned likelihood analysis to search for neutrino flares from pre-defined astrophysical sources using 22-string IceCube data, improving sensitivity over time-integrated methods by optimizing for variable time-scales. No significant flare excess was found, leading to upper limits on neutrino fluence for 10 variable sources, with the highest fluctuation at 3C 454.3 (p-value = 4.7%).

ABSTRACT

A time-dependent search for neutrino flares from pre-defined directions in the whole sky is presented. The analysis uses a time clustering algorithm combined with an unbinned likelihood method. This algorithm provides a search for significant neutrino flares over time-scales that are not fixed a-priori and that are not triggered by multiwavelength observations. The event selection is optimized to maximize the discovery potential, taking into account different time-scales of source activity and background rates. The method is applied to a pre-defined list of bright and variable astrophysical sources using 22-string IceCube data. No significant excess is found.

Motivation & Objective

• To develop a time-dependent search method for neutrino flares that does not rely on multiwavelength trigger data.
• To improve sensitivity over time-integrated analyses by focusing on short-duration flares with variable time-scales.
• To extend flare search capability to the southern sky, where high-energy atmospheric muons dominate background.
• To apply the method to a list of bright, variable astrophysical sources using 22-string IceCube data (276 days, 88.9% uptime).
• To derive upper limits on neutrino fluence for sources where no significant flare was detected.

Proposed method

• Uses a time-clustering algorithm to identify the most significant temporal accumulations of neutrino events around pre-defined source directions.
• Applies an unbinned maximum likelihood method to compute significance, defined as λ = -2 log[L(ν_s=0)/L(ν_s, γ_s)] where L is the likelihood function.
• Models background probability density functions (pdfs) for space (P^space), energy (P^energy), and time (P^time), each corrected for detector geometry, zenith dependence, and uptime.
• Uses P^time with a constant fit for the northern sky (atmospheric neutrinos) and a sinusoidal fit for the southern sky (seasonal muon background).
• Applies zenith- and energy-dependent P^space and P^energy pdfs to account for detector response and Earth absorption effects.
• Calculates signal pdf S_i as a product of spatial Gaussian (around source), energy-dependent spectral model (E^{-2}), and time pdf.

Experimental results

Research questions

• RQ1 Can a time-clustering algorithm with unbinned likelihood improve sensitivity to short-duration neutrino flares compared to time-integrated methods?
• RQ2 How does the background modeling (spatial, energy, time) affect the significance calculation in time-dependent neutrino flare searches?
• RQ3 What is the performance gain of the unbinned method over binned methods in the southern sky, where high-energy muons dominate?
• RQ4 What are the upper limits on neutrino fluence for known variable sources like FSRQs and LBLs, given no significant flare was detected?
• RQ5 How does the method perform across different time-scales, especially for flares shorter than the total observation period?

Key findings

• The time-clustering algorithm with unbinned likelihood improves sensitivity over time-integrated searches, with a 1.3× higher detection probability in the northern sky and 2× higher in the southern sky for a 5-event signal in a 7-day window.
• No significant neutrino flare excess was observed above atmospheric background across the 10 pre-defined sources, with the highest fluctuation at 3C 454.3 (p-value = 4.7%, not including trial factors).
• The fluence upper limit for 3C 454.3 is 2.22 GeV/cm² over a 0.5-day flare duration, assuming an E^{-2} spectrum.
• The southern sky sample (ultra-high energy events) achieved better sensitivity than the northern sky sample due to optimized energy cuts and background modeling.
• The method successfully extended the search window to the southern sky, where high-energy atmospheric muons are dominant and time-modulated.
• The analysis confirmed that azimuthal and zenith-dependent background corrections are essential for time-dependent searches with time-scales below one day, reducing background misestimation by up to 40%.