Reconstruction efficiency and discovery potential of a very large volume Mediterranean neutrino telescope

Transcription

1 Reconstruction efficiency and discovery potential of a very large volume Mediterranean neutrino telescope Apostolos G. Tsirigotis, Antonis Leisos, Spyros Tzamarias Physics Laboratory Hellenic Open University KM3NeT detector design The HOU Reconstruction & Simulation (HOURS) software package New tool developments Advances in track reconstruction and error estimation combining Kalman Filter with a Likelihood Neutrino Detector performance Sensitivity and discovery potential for point sources Point sources discovery improvements Signal from Gamma Ray Bursts & Diffuse flux sensitivity

2 KM3NeT lay-out KM3NeT in numbers ~ OMs ~300 (600) DU 20 storey/du ~40m storeyspacing ~1 km DU height ~ m DU distance ~3-6 km3 volume ~220 MEuro cost Detection Unit (DU): mechanical structure holding OMs, enviromental sensors, electronics,... DU is the building block of the telescope Optical Module (OM): pressure resistant sphere cointaining photo-multpliers

3 HOURS - Simulation software chain MuPage ~one hour life time for this study

4 GEANT4 Simulation in HOURS Any detector geometry can be described in a very effective way All the relevant physics processes are included in the simulation Full GEANT4 simulation SLOW Fast Simulation 100 to several thousand times faster than full Simulation (depended on muon energy) Parametrizations for: Angular Distribution of Cherenkov Photons - EM showers (from e-, e+, γ) - HA showers (from long lived hadrons) - Low energy electrons (from ionization) - Direct Cherenkov photons (from muon) Each parametrization describes the number and time profile of photons arriving on a PMT in bins of: Shower energy (E) (EM and HA showers) PMT position (D,θ) relative to shower vertex/muon position, PMT orientation (θpmt,φpmt) θpmt D Shower vertex/ muon position θ φpmt pmt axis Shower/muon direction

5 HOURS Optical noise filtering, prefit and muon reconstruction Prefit and Filtering based on: Optical Module Hit clustering (causality) filter & prefit using the clustering of candidate track segments (no apriori knowledge of the neutrino source) Causality filters and prefit using the apriori known direction of the neutrino source Muon reconstruction algorithms Combination of χ2 fit and Kalman Filter is used to produce many candidate tracks The best candidate is chosen using the Multi-PMT Direction and arrival time Likelihood (track quality criterion) d (x,y,z) L-dm θc dγ Track Parameters Muon energy reconstruction using the Charge Likelihood θ: zenith angle φ: azimuth angle (Vx,Vy,Vz): pseudo-vertex coordinates dm (Vx,Vy,Vz) pseudo-vertex

6 Kalman Filter Muon Track Reconstruction Algorithm State vector Initial estimation x0, C0 exp Update Equations t H k = x x=x k 1 timing uncertainty Kalman Gain Matrix Updated residual and chi-square contribution (rejection criterion for hit) Many (40-200) candidate tracks are estimated starting from different initial conditions ( x0, C0 ). The best candidate is chosen using the MultiPMT Direction and arrival time Likelihood (track quality criterion) A.G.Tsirigotis et al, Nucl.Instrum.Meth.A 602 (2009) 91

7 A new tracking criterion based on the directional information offered by the Multi-PMT and the time residuals distributions

8 Multi-PMT direction Likelihood PDFs of the angle, θ, between the Ch wavefront direction and the active direction of the Multi-PMT PDF d, s θ ; n Separate parametrization for n=1,2,...18 active small pmts. Muon Track 31 3 PMTs inside a 17 glass sphere Ch er en ko v θ For the parametrization only the angular acceptance and the directions of the small PMTs in the OM are used. ln(p(θ;n)) wa ve f ro nt D n=1 n=2 n=5 Charge weighted averaged direction of active PMTs cos(θ) N q i d i D= i=1

9 The directionality criterion is used for the selection of the best track candidate. Signal PDF d, s,i θ i ; ni Noise PDF d,n =constant i=1,2,...n the active Multi-PMTs ni=the number of active elements in the i th Multi-PMT th θ =the angle between the weighted average direction of the i active i Multi-PMT with the reconstructed Cherenkov wavefront For the selection of the best candidate track also the timing likelihood is used ti: hit arrival time, PDF t, s, i t i t exp ; q i,d i Signal Noise PDF t, n,i t i t exp ; d i texp:expected arrival time of direct photon qi: hit charge, di : Hit distance from track 1 p.e. charge OM distance from track : 50m, 100m, 140m Time delay (ns) The timing PDFs depend on the filtering and prefit stage They are created for the hits that pass these stages 3 p.e. charge OM distance from track : 50m, 100m, 140m Time delay (ns)

10 For all the candidate tracks form the direction*timing likelihood for all hits that pass the final filtering stage (common for all candidate tracks) Ltotal = [ pn, i N hit, qi PDF t, n,i PDF d,n 1 pn N hit,q i PDF t, s, i PDF d, s,i ] pn, i N hit, qi Probability of a hit to be noise Timing PDFs Direction PDFs Signal PDF t, s, i t i t exp ; q i,d i Noise PDF t, n,i t i t exp ; d i Signal PDF d, s,i θ i ; ni Noise PDF d,n =constant The candidate track with the largest Likelihood is chosen Maximize further the Likelihood for the chosen candidate track

11 Muon energy estimation N hit L E =ln i=1 N nohit P Qi, data ; E, D, θ P 0 ; E, D, θ i=1 Qi, data Hit charge (assumedly known exactly) normalized to the charge of a single p.e. pulse Probability depends on muon energy, E, distance from track, D, and PMT orientation with respect to the Cherenkov wavefront, θ: Convolution with the P Q i,data ; E, D, θ = F n ; E, D, θ G Q i, data ; n, n σ PMTresolution n=1 F(n; E, D, θ) PMT charge response function (simplified model with Gaussian) Not a poisson distribution, due to discrete radiation processes Muon energy estimation resolution L(E) Log(E/GeV)

12 Charge estimation resolution (talk by G. Bourlis) OM total charge = sum of the PMT charges OM total tot = sum of the tot values of the PMTs ET Enterprise Ltd. D783FL 3-inch diameter (tested in NIKHEF) ~22% charge estimation resolution OM Total charge error (pe) 31 3 PMTs inside a 17 glass sphere

13 Discovery potential for various KM3NeT layouts

14 Studied Detector Geometrical Layouts-Towers 2 x 154 Towers with a mean distance between them: 180m -> Instrumented volume 5.8 km (308towers_180m) 3 150m -> Instrumented volume 4.0 km3 (308towers_150m) Each Tower consists of 20 bars, 6m in length and 40m apart. One DOM at each end of the bar. 130m -> Instrumented volume 3.0 km3 (308towers_130m) 2x 154 Towers detector footprint

15 Studied Detector Geometrical Layouts-Strings 2 x 308 Strings with a mean distance between them: 130m -> Instrumented volume 6.1 km3 (616strings_130m) 100m -> Instrumented volume 3.6 km3 (616strings_100m) 2x Each strings consists of 20 DOMs 40m apart m... HEP2012 Ioannina, Greece, Apr il 5 8

16 5σ-50% discovery potential point source at -60 declination E-2 neutrino energy spectrum, binned technique 9 2 x10 E GeV cm s 308towers_180m 308towers_150m 308towers_130m 616strings_130m 616strings_100m

17 Galactic Candidate ν Sources SNRs Origin of Cosmic Rays => SNR paradigm Assuming that VHE γ emmiters are CR accelerators. As an example, assuming that the RXJ (the most luminous γ ray source) is a hadron accelerator then the ν spectrum can be calculated from the γ spectrum: (S.R. Kelner, F.A. Aharonian and V.V. Bugayov Phys. Rev D 74, (2006))

18 5σ-50% discovery potential point source at -60 declination RXJ1713 spectrum, binned technique) x Φ(Ε) 308towers_180m 308towers_150m 308towers_130m 616strings_130m 616strings_100m

19 RXJ1713 5σ-50% discovery potential (0.6o radius, uniformly emitting disk, binned technique) x Φ(Ε) 308towers_180m 13 years for discovery 308towers_150m 9.5 years for discovery 308towers_130m 9 years for discovery 616strings_130m 7.5 years for discovery 616strings_100m 7 years for discovery Hess RXJ

20 308Towers_180m detector performance Angular resolution Triggered events Background rate 3500 m depth Reconstructed events Effective area 170 atmospheric neutrinos per day 20 million atmospheric muon events per day

21 Galactic Candidate ν Sources Discovery Improvements - I To appear in the Proceedings of VLVnT2011

22 Background filtering technique using the apriori known neutrino point source Causality criterion H Optical Module (OM) position V d pseudo-vertex d Muon momentum direction (generated by a neutrino from a hypothetical source) μ source a b H V νμ Expected arrival time to OM of a photon emitted by the muon with the Cherenkov angle, θc (direct photon): ct expected =a b tanθ c H V a= d V a d b= H The vertical distance of OM to the muon track

23 Background filtering technique using the apriori known neutrino point source Causality criterion 1, H 2 should satisfy: Two direct photons with arrival times t1, t2 on the OMs with positions H H cδt d Δ = Δb tanθ c Project the hits position and vertex on a plane perpendicular to the known direction. x d Δt=t 1 t 2 = H 1 H 2 ΔH Δb=b 1 b 2 Then from simple geometry: V b1 b2 H d Δ d Δb = cδt d Δ H Δ H tanθ c 2 H d Δ d H H 1 Δ H

24 Background filtering technique using the apriori known neutrino point source Causality criterion 1, H 2 should satisfy: Two direct photons with arrival times t1, t2 on the OMs with positions H Δt=t 1 t 2 = H 1 H 2 ΔH Δb=b 1 b 2 H cδt d Δ = Δb tanθ c Project the hits position and vertex on a plane perpendicular to the known direction. x d Then from simple geometry: V b1 b2 H d Δ d Δb = cδt d Δ H Δ H tanθ c 2 H d Δ d H H 1 Δ H Causality criterion between two hits using the known direction of the source H H tanθ c Δ H d Δ d ct s cδt d Δ d 800m ΔH t s =10ns Relax the criterion (light dispersion, time jitter) Longitudinal distance between the two OMs to the direction of the muon track H d Δ d 67.5m one absorption length Δ H Lateral distance

25 Prefit and reconstruction technique using the known neutrino direction Causality criterion is used as background filtering <0.3% of noise hits survive >90% of signal hits survive For every three OMhits (on different OMs) that satisfy the causality criterion a pseudo-vertex can be found analytically. Cumulative distribution of the distance between the estimated pseudo-vertex and the MC-true pseudo-vertex This technique Many candidate pseudo-vertexes are found using different triplets of hits For signal events (Eν>100GeV) the clustering in space of all the candidate pseudo-vertexes can estimate the MC-true pseudo-vertex with accuracy ~ 2m Direct Walk technique (clustering of candidate track segments) The estimated pseudo-vertex and the known direction is used to further reduce the number of noise hits ~0.03% of noise hits survive ~90% of signal hits survive Combination of χ2 minimization and Kalman Filter is used to produce many candidate tracks The best candidate is chosen using the timing and Multi-PMT direction Likelihood

26 Prefit and reconstruction technique using the known neutrino direction Reconstruction efficiency and angular resolution E-2 neutrino generated spectrum (15GeV 100PeV) 2.9km3 neutrino detector with 6160 DOMs (arranged in 154 Detection Units (Towers)) Reconstructed tracks with at least 8 hits on different DOMs Reconstruction efficiency vs neutrino energy for events with at least 3 L1 signal Hits Point spread function for reconstructed events This technique This technique Direct Walk technique (clustering of candidate track segments) log E ν /GeV Direct Walk technique (clustering of candidate track segments) log E ν /GeV

27 Reconstruction technique using the known neutrino direction Estimation of fake signal For each atmospheric neutrino/shower event: Assume a candidate neutrino direction pointing to a hypothetical astrophysical source Apply filtering and prefit using the assumed direction Track reconstruction Accept the event if the angular difference between the assumed direction and the reconstructed muon direction < 1o (For point source searches this angular difference is further optimized) With the application of this technique the required observation time for 5σ discovery of the RXJ1713 galactic source is reduced from 13 years to 9 years (Towers 180m and binned technique) Fake signal can be further reduced by applying tracking quality criteria using the estimated tracking error. Fake tracks carry a very small weight in the unbinned method

28 Galactic Candidate ν Sources Discovery Improvements - II To appear in the Proceedings of VLVnT2011

29 Unbinned technique Use of the full experimental information on a track by track basis: reconstructed muon energy, and track resolution (muon reconstruction parameter errors) Energy distribution of signal (RXJ1713) and background (atmospheric neutrinos) Histogram: True Line: Predicted from reconstruction errors Signal Background Angle between reconstructed muon track and parent neutrino (Degrees) Reconstructed Energy (log(e/gev))

30 Comparison of the Discovery Potentials (RXJ ) Discovery Potential in units of the reference flux Number of events (N) Angular profile (S) Energy (E) RXJ1713 1,00 Number of observation years

31 Discovery Potential in units of the reference flux 5-σ DISCOVERY POTENTIAL (RXJ1713) 1,00 Number of observation years

32 Required years of observation time for 5-σ DISCOVERY (RXJ1713) 8.4y 7.6y 6.4y 1,00 6.0y 5.6y Number of observation years The combination of the known source direction technique with the unbinned method reduces the number of observation years to 4 years (PRELIMINARY)

33 Extragalactic ν Sources GRBs dnneutrino/dtdω (y-1 sr-1) Alert from satellite detectors (known time & direction) Short time window (<2h) for GRB prompt neutrino emission High energy neutrinos (> 100TeV) Application of energy cut on the reconstructed muon energy 4500m detector depth 8.4ev/y Detection of down-going GRB neutrinos events is feasible 1000 GRBs/year Half KM3NeT 3500m detector depth 7.5ev/y background Expected Neutrino fluence from GRBs R. Abbasi, et al., IceCube Collaboration, Astrophys. J. 710 (2010) 346 GRB zenith angle (degrees) The expected number of detected neutrino events from GRBs per year and steradian. 300 GRBs/year (4500m depth) Half KM3NeT 2.5 signal events/year 0.45 background events/year

34 Diffuse Flux ν Ultra high energy neutrinos from A multitude of objects such as Active Galactic Nuclei or GRBs The interaction of cosmic rays with intergalactic matter, radiation, cosmic microwave background No tight angular cut for reducing the background of atmospheric neutrinos isotropic diffuse flux Rely on a cut on the reconstructed muon energy, Eμreco. KM3NeT (E 2) diffuse ν flux sensitivity (energy cut Eμreco > 500 TeV) (GeV 1 cm 2 s 1 sr 1 ) Diffuse flux sensitivity of the KM3NeT neutrino telescope for one year of observation time.

35 Conclusions We improved the track reconstruction and the tracking error estimation by employing a Likelihood criterion and a fit at the last stage of the tracking chain KM3NeT will cover most of the ν sky with unprecedented sensitivity Promising Galactic Candidate ν Sources Discovery potential for Galactic Candidate ν Sources is further improved using the reconstructed energy estimation, the angular resolution on a track by track basis and the application of advanced filters using the known source's direction The most louminous galactic source (RXJ1713) can be discovered in less than 5 years of observation time The detector optimization is ongoing (for different depths, water properties,...) This research has been co-financed by the European Union (European Social Fund ESF) and Greek national funds through the Operational Program "Education and Lifelong Learning" of the National Strategic Reference Framework (NSRF) - Research Funding Program: THALIS HOU - Development and Applications of Novel Instrumentation and Experimental Methods in Astroparticle Physics The KM3NeT project is supported by the EU in FP6 under Contract no and in FP7 under Grant no

36 Backup slides

37 Muon energy reconstruction and pulse arrival time corrections depend on pulse amplitude Examples Data from HELYCON scintillator counters (¾ inch fast pmts, used also in H.E.S.S.) using atmospheric muons Bias (slewing) in evaluating the pulse arrival time using the pulse inflection point (black dots) or threshold crossing (red dots) -Time correction (ns) Data from 2003 NESTOR run (15 inch pmts) with calibration LED in deep sea Pulse amplitude (mv) How well can we estimate the pmt charge using 1,2,3 thresholds per 3-inch pmt in KM3NeT? How the charge estimation resolution affects the muon direction and energy reconstruction resolution?

38 Physics Case & Main objectives Main physics goals Origin of Cosmic Rays and Astrophysicalν sources Galactic Candidate ν Sources (SNRs, Fermi Bubbles, microquasar, ) Extragalactic Candidate ν Sources (AGN, GRB, ) Diffuse Fluxes Implementation requirements Construction time 5 years Operation over at least 10 years without major maintenance Cabled platform for deep-sea research Environmental sciences Geology and geophysics Marine biology and oceanography

39 Sky view of the KM3NeT FOV for up-going neutrinos shown 24h per day visibility up to declination δ~-50 Visibility > 25% of time Visibility > 75% of time KM3NeT covers most of the sky (87%) including the Galactic Centre

40 Front End Electronics Read-out SCOTT ASIC Time over threshold with adjustable thresholds Digitised output Zero suppression System on chip FPGA for data buffering and formatting

41 Optical Module - Multi-PMT 31 3 PMTs (~30% max QE) inside a 17 glass sphere with 31 bases (total ~6.5W) Cooling shield and stem First full prototype under test Single vs multi-photon hit separation Large (1260 cm2) photocade area per OM