Dottorato di Ricerca in Fisica - XXVIII ciclo Search for high energy neutrino astrophysical sources with the ANTARES Cherenkov telescope Chiara Perrina Supervisor: Prof. Antonio Capone 25 th February 2014
Four questions (outline) Where do ultra high energy neutrinos (UHE n) come from? UHE neutrino astrophysical sources Why do we detect these particles? Motivations of the neutrino Astronomy How do we detect these particles? The ANTARES detector and its detection principle How do we search for down-going neutrinos? -My research: - event selection (cut-based / multivariate) - MRF analysis - preliminary results + future developments 2
UHE neutrinos Cosmic ray (CR) models provide UHE neutrino fluxes from the decay of p ± produced in interactions of accelerated protons (or nuclei) with radiation/matter fields in or near accelerating astrophysical objects. p e CR p 0 p n / Inverse Compton (+Bremsstr.) 3
Candidate n sources Galactic sources: Supernova Remnants (SNRs), Pulsars, Nebulae, binary systems, stellar black holes (mq), the Galactic Centre,... Extragalactic sources : Active Galactic Nuclei (AGNs), Gamma-ray Bursts (GRBs), 4
Why do we detect these particles? To identify high energy sources we need a particle neutral stable ν weakly interacting The only known particle that has these properties is the neutrino. Protons are deflected and/or absorbed Photons are absorbed Neutrinos Photons: abundant and easy to detect, but high energy photons (> 10 15 ev) interact with the CMB radiation producing e + e - pairs. Protons: at high energy (> 10 19 ev) interact with the CMBR, at lower energy are deviated by galactic magnetic fields. 5
Come si rivelano i neutrini? How to detect neutrinos? To be detected, neutrinos have to interact with matter by charged or neutral current interactions. Since neutrinos are weakly interacting and their fluxes from astrophysical sources are expected small (~ 10 GeV -1 sr -1 km -2 year -1 ), a large target/detection volume (~ km 3 ) is required. The best channel for neutrino Astronomy is n m m, since m can travel large distances in water (~ km). Cherenkov submarine telescopes are the best experimental solution. 6
Detection principle M. Markov: We propose to install detectors deep in a lake or in the sea and to determine the direction of the charged particles with the help of Cherenkov radiation (1960, Rochester Conference). m (~ n) path reconstruction by times and positions of the hits on PMTs PMTs matrix Cherenkov photons ( C ) (q 42, n H2 O=1.335) n m W m C q muon N X neutrino 7
The ANTARES detector Located at a depth of ~ 2.5 km in the Mediterranean Sea, ~ 40 km South-East off the coast from Toulon, France (42 50 N, 6 10 E); annual sky coverage of 3.5π sr; annually integrated 1.5π sr common view with the IceCube telescope at South Pole; Galactic Centre coverage. Buoy 12 lines, 25 floors/line, 3 PMTs/floor ( 15 PMTs) => 885 PMTs floor 14.5 m 350 m 2.5 km depth F. Montanet ~ 60 m 100 m Anchor Cables Junction Box 8
Signal and background Two types of physics background: atmospheric muons induced by CR interactions in atmosphere; muons by atmospheric neutrinos interactions. p, N m sig m atmo m back n astro n atmo m atmo p, N Up-going Down-going To reject atmospheric muons, considered only up-going events; the m back by atmospheric neutrinos is irreducible: to be subtracted on a statistical basis 9
Reconstruction of the muon track The muon travels at the light speed in vacuum (c); The Cherenkov photons travel at the light speed in the medium (c/n). We measure the arrival times of the photon on the PMT (t exp ). Known the positions of the PMTs, assuming the track parameters (x 0, y 0, z 0, q, f), we can calculate the theoretical arrival time of the photons on the PMTs (t teo ). By applying a recursive fit procedure varying the 5 parameters of the track, we maximize the likelihood of residual time between t exp e t teo. The parameter that provides an estimate of the quality of the reconstruction is N DOF is the number of degrees of freedom of the track (N hit 5), N comp the number of compatible solutions and L is the likelihood. 10
Angular resolution for the neutrino direction reconstruction Kinematics m RECO n TRUE angle m RECO m TRUE angle Experimental resolution 11
Event display Example of a reconstructed down-going muon, detected in all 12 lines: Example of a reconstructed up-going muon (i.e. a candidate neutrino) detected in 6 out of 12 lines: D O W N U P 12
Data set and selection criteria Data taking period: - May 2008 - December 2011, for the cut-based analysis; - the 50%, until October 2010, for the multivariate one. Run selection Runs with the detector in good DAQ conditions Events selection Selected only down-going events Selected only multi-line events Signal/background discrimination cos q Energy (number of hits and charge in the cut-based analysis) L 13
Cut-based selection 1 2 n ASTRO & n ASTRO MC m - ATMO & m + ATMO MC n ATMO & n ATMO MC 3 4 14
Energy estimators It is possible to estimate the energy of the event by the amount of photons recorded by PMTs: the higher the energy of the m, the greater the length of its path and the number of charged tracks originated by the electromagnetic interactions subsequent to radiation phenomena. When the muon energy increases, the number of PMTs in which a signal is recorded grows, then the NUMBER OF HITS and the number of Cherenkov photons that reach each PMT increase, i.e., increases the TOTAL CHARGE COLLECTED. NUMBER OF HITS TOTAL CHARGE COLLECTED log10(energy/gev) log10(energy/gev) 15
Cuts effect We accept the 15.37% of the signal. We reject the 99.64% of the background (background efficiency = 3.6x10-3 ). We can do better! 16
Selection variables used with BDT method 17
BDT optimization I optimized the performances of the BDT looking for the best set of parameters that I had to choose a priori: 1. number of trees; 2. maximum trees depth; 3. minimum number of events per leaf; 4. number of cuts. I considered the values in the table below, the best values - those that maximize s/sqrt(s+b) - are in bold type: Number of trees 850 (850, 800, 700, 550, 350, 100 ) Depth single tree 3 (3, 4, 5) Minimum events in leafs 100 (50, 100, 150, 200, 250) Maximum number of cuts on single variable 20 (5, 10, 20, 30) 18
Signal and background efficiencies 50% 20% 3 x 10-3 4 x 10-5 Cut-based selection: signal eff. = 15.37% background eff. = 3.6 x 10-3 19
Extraction of signal from background To simulate the signal I used the MC events of astrophysical n and anti-n (isotropically generated), and I selected the events placed in a region of 5 in r.a. and 3 in d around a source. I considered 56 hypothetical sources located at 0 of r.a. and between -40 and 80 of declination. To evaluate the background I used scrambled data. 20
Extraction of signal from background Around the MC track of each signal event I opened cones of different sizes (a, half cone angle). For each size, I evaluated the probability that the reconstructed track was contained in the cone of observation. Assuming a neutrino flux from the source (prop. to E n -2 ), I estimate the number of signal events expected from it (n s ) inside the cone. For the same cone I evaluated the number of background events (n b ) by real scrambled data. The goal was to define the aperture of the cone which optimizes the signal / background ratio and allows to determine the minimum value of the signal neutrino flux observable from the source, given the number of background events. The method used for the optimization is the Model Rejection Factor (MRF). n MC BDT > 0.2 reconstructed track a 21
The MRF method sensibility Assumed theoretical neutrino signal flux MRF in correspondence with the cone whose amplitude maximizes the signal / background ratio Minimum value of the n s which excludes at 90% c. l. the only-background hypothesis in agreement with n b (Feldmann and Cousins sensitivity) Number of signal events expected in agreement with 22
The result: the sensitivity This analysis (56 candidate sources with r.a. = 0 ) 23
Next steps Analysis of the whole data set (ongoing). Generation of a new MonteCarlo for the simulation of neutrino fluxes from point-like sources (ongoing). Soon the request of unblinding will be discussed within the ANTARES Collaboration. An important development of this analysis will be the search for neutrinos transient sources, such as GRBs or highly variable sources. The search for neutrinos from these sources restricts the time of observation to the flaring time big reduction of the background. 24