Upper Limit on the Diffuse Flux of Cosmic $ν_μ$ with the ANTARES Neutrino Telescope

0.65 MB·English

by  Simone Biagi

1
1
Upper Limit on the Diffuse Flux of Cosmic ν with the ANTARES
0 µ
2
Neutrino Telescope

n
a Simone Biagia∗ for the ANTARES collaboration
J

9 aDipartimento di Fisica dell’Universit`a and INFN – Sezione di Bologna,
1 Viale Berti Pichat 6/2, 40127 Bologna, Italy

]
E A search for a diffuse flux of astrophysical muon neutrinos, using data collected by the ANTARES neutrino
H telescope from December 2007 to December 2009 is presented. A (0.83×2π) sr sky was monitored for a total of
334daysofequivalentlivetime. Thesearchedsignalcorrespondstoanexcessofevents,producedbyastrophysical
.
h sources, over the expected atmospheric neutrino background without any particular assumption on the source
p
direction. Since the number of detected events is compatible with the numberof expected background events, a
o- 90%c.l. upperlimitonthediffuseνµfluxwithaE−2spectrumissetatE2Φ90% =5.3×10−8 GeV cm−2 s−1 sr−1
r intheenergyrange20TeV–2.5PeV.Othersignalmodelswithdifferentenergyshapewerealsotestedandsome
st rejected.
a
[

1 1. Introduction trumofcosmicneutrinosisexpectedtobeharder
v thanthatofatmosphericneutrinos,thesignalwe
0 TheANTAREShigh-energyneutrinotelescope are looking for corresponds to an excess of high
7 is a three-dimensional array of photomultiplier
energy events in the measured energy spectrum
6 tubes (PMT) distributed over 12 lines installed
without any particular assumption on the source
3
deep in the Mediterranean Sea, each line includ-
. direction.
1 ing 75 PMTs [1]. A neutrino telescope in the
Electrons (in the so-called “leptonic models”)
0
Northern hemisphere includes the Galactic Cen-
1 orprotonsandnuclei(“hadronicmodels”)canbe
tre in its field of view and is complementary to
1 accelerated in astrophysical processes. Hadronic
: the IceCube Antarctic telescope [2]. models[3]predictthattheenergyproducedinthe
v
The main goal of the experiment is the
i sourcesiscarriedawaybycosmicrays,γ-raysand
X search for high-energy neutrinos from astrophys-
neutrinos. A benchmark flux for the measure-
r ical sources. If the sensitivity of point source ment of diffuse neutrinos is the Waxman-Bahcall
a search techniques is too small to detect neutrino
(W&B) upper bound [4]. Using the CR observa-
fluxes from individual sources, it is possible that tions at E ∼ 1019 eV (E2 Φ ∼ 10−8 GeV
many sources could produce an excess of events cm−2s−1srC−R1) the diffuse fluCxRofCmRuon neutrinos
over the expected atmospheric neutrino back-
is constrained at the value:
ground. In this proceeding the search for very-
highenergyextraterrestrialmuonneutrinos from E2Φ <4.5/2×10−8 GeV cm−2 sr−1 s−1 (1)
ν ν
unresolved sources is presented using data col-
(the factor1/2isaddedtotakeintoaccountneu-
lected by the ANTARES telescope from Decem-
trino oscillations).
ber 2007 to December 2009.
Atmosphericmuonsandneutrinosarethemain
2. Neutrino tracking and energy recon-
sources of background in a neutrino telescope.
struction
The former can be suppressed by applying re-
quirements on the direction of the events, the
Muon neutrinos are detected via charged cur-
latter is an irreducible background. As the spec-
rent interactions: ν +N → µ+X. The arrival
µ
times and the amplitudes of the Cherenkov light
∗[email protected] signals detected by the PMTs [5] are used to re-
1
2
construct the trajectory of muon neutrinos and
to estimate their energy.
The track reconstruction algorithm defined in
[6]isbasedonalikelihoodfitthatusesadetailed
parametrization of the probability density func-
tionforthe photonarrivaltimes. Thetrackposi-
tionanddirection,theinformationonthenumber
of hits (N ) used for the reconstruction and a
hit
quality parameter Λ are the main outputs of the
algorithm. Λ is determined from the likelihood
andthe numberofcompatiblesolutionsfoundby Figure 1. Definition of the variable Ri on the i-th
thealgorithmitself. Λcanbeusedtorejectbadly PMT. In this example, both ARS0 and ARS1 are
reconstructed events. For Eν > 10 TeV, an an- fired;after theintegration dead-time,both chipscol-
gular resolution for muon neutrinos better than lected light again. In this example, Ri =4.
◦
0.3 is accomplished by the ANTARES detector.
2.1. Monte Carlo simulations short-lived charmed particles D → K + µ +
ν becomes a significant source of atmospheric
The Monte Carlo (MC) simulation tools [7,8] µ
“prompt leptons”. The Recombination Quark
include the production of Cherenkov light, the
Parton Model (RQPM) is used in this simula-
generation of the optical background caused by
tion, since it gives the largest prompt contribu-
radioactive isotopes and bioluminescence present
tion among the models considered in [12].
inseawater,andthedigitizationofthePMTsig-
Atmospheric muons. The ANTARES trig-
nals. In particular, the PMT simulation also in-
ger rate [13] is dominated by atmospheric muons
cludestheprobabilityofadetectedhitgivingrise
that represent the main background for a neu-
to an afterpulse. The simulation of afterpulses is
trino telescope. A small fraction (approxi-
criticalwhentheenergyestimatordefinedin§2.2
mately5%)oftriggereddowngoingmuonsismis-
isappliedtoMCevents. Theafterpulseprobabil-
reconstructedasupgoing;their rejectionisa cru-
ity was measuredin laboratoryusing ANTARES
cial point in this analysis.
[9] and NEMO [10] PMTs and it was confirmed
The MUPAGE package [14] was used to sim-
withdeep-seadata. Upgoingmuonneutrinosand
ulate atmospheric muon samples. One year of
downgoing atmospheric muons have been simu-
equivalent live time with a total energy E ≥
lated and stored in the same format used for the T
1 TeV and bundle multiplicity m = 1 ÷ 1000
data.
Signal and atmospheric neutrinos. MC was generated. The total energy ET is the sum
of the energy of the individual muons in an at-
muonneutrino eventshavebeengeneratedinthe
energy range 10 ≤ E ≤ 108 GeV and zenith mospheric muon bundle. Triggered ANTARES
ν
angle between 0◦ ≤ θ ≤ 90◦ (upgoing events). eventsmainly consistofmultiple muonsoriginat-
ing in the same primary CR interaction [15].
ThesameMCsamplecanbedifferentlyweighted
toreproducethe“conventional”atmosphericneu-
2.2. Energy dependent variable
trinospectrum(Bartol),∝E−3.7 athighenergies
ν The only way to separate atmospheric and as-
[11], and the expected astrophysical signal spec-
trophysical neutrinos is through a discrimination
trum, ∝ E−2. The normalization of the signal
ν basedontheenergy. Anoriginalenergyestimator
flux is irrelevant when defining cuts, optimizing
isdefined,whichisbasedonhitrepetitionsinthe
procedures and calculating the sensitivity. Here
PMTs due to the different arrival time of direct
a diffuse flux test signal is defined equal to:
and delayed photons (Fig. 1). Direct photons
E2 Φ =1.0×10−7 GeV cm−2 s−1 sr−1. (2) are emitted at the Cherenkov angle and arrive
ν ν
at the PMTs without being scattered. Radiative
Above 10 TeV, the semi-leptonic decay of processescontributetoenergylosseslinearlywith
3
Table 1
Numberof expected eventsfor MC and data.
µatm νatm νsig Data
Reco 2.2·108 7.1·103 106 2.5·108
Upgoing 4.8·106 5.5·103 80 5.2·106
1st-level 9.1·103 142 24 1.0·104
2nd-level 0 116 20 –
Expected events in 334 days of equivalent live time
for the three MC samples: atmospheric muons,
atmospheric neutrinos (Bartol+RQPM), astrophys-
ical signal from eq. 2, and data. Reco: at the
reconstruction level; Upgoing: reconstructed as
upgoing; 1st-level: after thefirst-level cuts; 2nd-level:
after thesecond-level cut.
Figure 2. Reconstructed muon energy using the R
energy estimator as a function of the true muon en- are present. For a muon neutrino sample, R is
ergy. Thedisplayedeventsareselectedafterthequal-
linearly correlated with the log of the true muon
itycutsdefinedin§3.1. Onlyneutrino-inducedmuons
energyE intherangefrom10TeV(R≃1.26)
are selected. true
to 1 PeV (R ≃ 1.73). R can be used to estimate
the muon energy E , see Fig. 2. The distribu-
reco
tion of log(E /E ) has a FWHM=0.8. The
the muon energy for E > 1 TeV. The resulting reco true
µ
resolutioniscomparableorbetterwithrespectto
electromagneticshowersproduceadditionallight.
other energy reconstruction algorithm [17].
Photons originating from secondary electromag-
Thisenergyestimatorisrobustbecauseitdoes
netic showers or scattered Cherenkov radiation
not depend on the number of active PMTs and
arrive on the PMTs delayed with respect to the
on non-linear effects on charge integration.
direct photons,witharrivaltimedifferencesupto
hundreds of ns [3]. The fraction of delayed pho-
tons increases with the muon energy. 3. Cosmic neutrino signal selection
Thesignalproducedbythe PMTsisprocessed
The data collectedfromDecember 2007to De-
bytwoAnalogue RingSampler (ARS) [16]which
cember2009areanalyzed. Inthis period,thede-
digitize the time and the amplitude of the sig-
tectorconfigurationchangedseveraltimeswith9,
nal (the hit). They are operated in a token ring
10 and 12 active lines. For this reason, three dif-
scheme. If the signal crosses a preset threshold,
ferentdetectorconfigurations,basedonthenum-
typically 0.3 photo-electrons, the first ARS inte-
ber of active lines and optical modules, were re-
grates the pulse within a window of 40 ns. If
produced in MC simulations. Data runs are se-
triggered, the second chip provides a second hit
lectedaccordingtothedata-qualityrequirements
withafurther integrationwindowof40ns. After
explained in [8]. The total live time is 334 days:
digitization, each chip has a dead time of typi-
70 days with 12 lines, 128 days with 10 lines and
cally 250 ns. After this dead time, a third and
136 days with 9 lines.
fourth hit can also be present.
ThenumberofrepetitionsR forthei-thPMT
i 3.1. Rejection of atmospheric muons
isdefinedasthenumberofhitsinthesamePMT
As described in §2.1, the rejection of atmo-
within500nsfromtheearliesthitselectedbythe
spheric muons is a crucial point in the search for
reconstructionalgorithm(Fig. 1). In mostcases,
acosmicneutrinosignal. Thiscontaminationcan
R =1or2,butitcouldbealso3or4. Themean
i be strongly suppressed by applying requirements
number of repetitions in the event is defined as
onthegeometryoftheeventsandonthetrackre-
R= Ri ,whereN isthenumberofPMTs construction quality parameter Λ. Two different
NPPMT PMT
in which hits selected by the tracking algorithm levels of cuts are defined in order to remove the
4
Λ -4 120 Λ -4 0.02
-4.2 -4.2
0.018
-4.4 100 -4.4
0.016
-4.6 -4.6
0.014
80
-4.8 -4.8 0.012
-5 60 -5 0.01
-5.2 -5.2 0.008
-5.4 40 -5.4 0.006
-5.6 -5.6 0.004
20
-5.8 -5.8 0.002
-6 100 150 200 250 300 350 400 0 -6 100 150 200 250 300 350 400 0
N N
hit hit
Figure 3. ScatterplotsofthereconstructionqualityparameterΛversusthenumberofhitsN foratmospheric
hit
muon (left) and signal neutrino (right) MC simulations after the first-level cuts. The pink line represents the
second-level cut described by eq. 3. The color code is in unitsyear−1.
contamination of mis-reconstructed atmospheric 3.2. Discriminationfrom atmosphericneu-
muons from the final sample. trinos
First-level cuts. Events are selected according A cut on the energy dependent variable R, de-
tothesecriteria: (i)upgoingparticleswithrecon- fined in §2.2, is used to separate the diffuse flux
◦
structed zenith angle θ < 80 (corresponding signal from the atmospheric ν background. The
rec µ
to 0.83×2π sr); (ii) Λ>−6; (iii) N >60; (iv) optimal cut value is obtained through a blinding
hit
reconstruction with at least two lines. The first- procedure on MC events, without using informa-
level cuts reduce the rate of mis-reconstructed tions from the data. The numbers of expected
events by almost 3 orders of magnitude, as in- signal (n ) and background (n ) events are com-
s b
dicated in Table 1. puted as a function of R. Then, calculating the
Second-level cut. The remaining atmospheric so-called Model Rejection Factor (MRF) defined
muonshavethequalityparameterΛwhichonav- in [18], the best cut is obtained and used as the
eragedecreaseswithincreasingN . Fig. 3(left) discriminator between low energy events, domi-
hit
shows the correlationbetween Λ and N for at- natedbytheatmosphericneutrinos,andhigh en-
hit
mospheric muons. In orderto completely remove ergy events, where the signal could exceed the
the expected rate of mis-reconstructed events in background. After the optimization of all the
∗
the MC sample, a cut value Λ is defined as a parameters, the observed data events (n ) are
obs
function of N : revealed (un-blinding procedure) and compared
hit
withtheexpectedbackgroundfortheselectedre-
Λ∗ = −4.59−5.88·10−3Nhit Nhit ≤172 (3) gion of R. If data are compatible with the back-
(cid:26) −5.60 Nhit >172 ground, the upper limit for the signal flux is cal-
culated using the Feldman-Cousins method [19]
∗
Removing all events with Λ < Λ , the atmo-
at a 90% confidence level (c.l.).
spheric muons are completely suppressed. Inde-
The cumulative distributions of the R vari-
pendent MC atmospheric muon simulations us-
able for atmospheric neutrino background (Bar-
ing CORSIKA (see details in [8]) confirm that
tol+RQPM) and diffuse flux signal (eq. 2) are
the maximum contamination in the final sam-
computed for the three configurations of the
ple is less than 1 event/year. As can be seen
ANTARES detector and the corresponding live
in Fig. 3 (right), the signal is highly preserved
times. The MRF is calculated as a function of R
fromthesecond-levelcut. Theeffectsofthefirst-
using these cumulative distributions. The mini-
and second-level cuts on signal and atmospheric
mumfoundforR=1.31determinesthecutvalue
neutrinos are also given in Table 1.
5
s flux at these energies (25-30%) [11].
nt Data
ve Atms ν The number of expected background events
e 10 Signal ν with R ≥ 1.31 is 8.7 for Bartol MC only. Most
prompt models described in [12] give negligi-
ble contributions; the RQPM model predicts the
largest contribution of 2.0 additional events with
1 respect to the conventionalBartol flux. An aver-
age over all the consideredmodels gives a contri-
butionof0.3events. AcombinedmodelofBartol
flux plus the average contribution from prompt
models is adjusted with the data/MC normal-
10-1
1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 ization factor obtained in the R < 1.31 region.
R
Hencethenumberofexpectedbackgroundevents
for R≥1.31 is 10.7.
Figure 4. Distribution of the R parameter for the
134neutrinocandidatesinthe334daysofequivalent 4.2. High energy events and upper limit
live time. Points represent data, the filled histogram
TheMCsimulationshavebeentestedandcom-
istheatmosphericneutrinoMC(Bartolmodelonly).
pared with data. In particular, the R distribu-
The dashed line represents the maximum contribu-
tionsshowareasonableagreementbothforatmo-
tion (RQPM) of “prompt” neutrinos. The MC pre-
sphericmuons[20]andforatmosphericneutrinos
dictions are not normalized to the data. The signal
in the low energy region R <1.31 (c.f. Fig. 4).
normalized at the upper limit (eq. 4) is shown as a
fullline. ThecutatR=1.31isindicatedasavertical Asaconsequence,thesignalregionwithR≥1.31
line. was un-blinded and 9 high-energy neutrino can-
didates are found.
Systematic uncertaintiesonthe expectednum-
for the energy dependent variable [20]. Assum-
ber of background events in the R ≥ 1.31 region
ing the Bartol (Bartol+RQPM) atmospheric ν
µ are evaluated considering: (i) the contribution
flux, 8.7 (10.7) background events and 10.8 sig- of prompt neutrinos, estimated as +1.7 events;
naleventsareexpectedforR≥1.31. Thecentral −0.3
(ii) the uncertainties from the conventional neu-
90% of the signal is found in the neutrino energy
trinoflux,thatdependmainlyontheuncertainty
range 20 TeV<E <2.5 PeV.
ν on the absolute flux as a function of the energy
and on the spectral index, evaluated to be ±1.1
4. Upgoing neutrino candidates
events. The uncertainties on the detector effi-
ciency(angularacceptanceoftheopticalmodule,
4.1. Low energy events R<1.31
water absorption, scattering length, trigger sim-
Events surviving the second-level cut are up-
ulation and the effect of afterpulses) amount to
going neutrino candidates. The first step of the
5%: they affect the detection both of signal and
un-blindingistorevealtheeventswithR<1.31.
background neutrinos in the high energy region.
In this region, 125 events are found. A compar-
The number of observed events is compatible
ison with MC predictions is shown in Fig. 4 as
with the number of expected background events.
a function of R. The events with R ≥ 1.31 in
The 90% c.l. upper limit on the number of sig-
Fig. 4 are uncovered after the final un-blinding
nal events µ (n ,n ) for n = 9 observed
of the data sample. MC predictions are lower 90% obs b obs
events and n = 10.7±2 background events in-
by ∼ 20% with respect to the detected events. b
cluding the systematic uncertainties is computed
Bartol atmospheric neutrino MC predicts 104.0
withthemethodof[21]: thevalueµ (n )=5.7
events with R < 1.31, and Bartol + RQPM pre- 90% b
is obtained. The upper limit on the diffuse flux
dicts 105.2 events. The discrepancies between
is given by Φ =Φ ·µ /n :
predictedandmeasuredeventsarewellwithinthe 90% ν 90% s
systematic uncertainties of the absolute neutrino E2Φ =5.3×10−8 GeV cm−2 s−1 sr−1. (4)
90%
6
Table 2
Tested flux models.
-1 r] Frejus MACRO Model R∗ Nmod ∆E90% µ90%/Nmod
s
-1 s (PeV)
-2 m10-6 MPR 1.43 3.0 0.1÷10 0.4
c
V P96pγ 1.43 6.0 0.2÷10 0.2
e (MPR)/2
G S05 1.45 1.3 0.3÷ 5 1.2
E [ SeSi 1.48 2.7 0.3÷20 0.6
N/d10-7 Baikal NT-200 ( e+ )/3 Amanda-II UHE Mpp+pγ 1.48 0.24 0.8÷50 6.8
2 Ed Amanda-II ( e+ )/3 Astrophysical fluxmodels, thevalueof theR∗ which
ANTARES- 2007-09 minimizes the MRF, the expected number of events
N , the energy range ∆E in which the 90% of
(W&B)/2 mod 90%
-8 events are expected, and the ratio µ90%/Nmod. See
10103 104 105 106 107 108 [20] and references therein.
E(GeV)

Figure5. TheANTARES90%c.l. upperlimitfora
E−2 diffuseνµ+νµ flux obtained with this analysis, REFERENCES
compared with the results obtained by other experi-
ments and theoretical predictions. See [3] and refer- 1. P. Coyle, these proceedings.
ences therein. 2. T. Montaruli, these proceedings.
3. T. Chiarusi and M. Spurio, Eur. Phys. J. C65
(2010) 649.
This limit holds for the energy range between 20
4. E. Waxman and J. Bahcall, Phys. Rev. D59
TeVto2.5PeV.Theresultiscomparedwithother (1998) 023002.
measuredfluxupperlimitsandtheoreticalpredic- 5. J.A. Aguilar et al., Nucl. Instrum. Meth. A570
tions in Fig. 5. (2007) 107.
Sometheoreticalpredictionsofcosmicneutrino 6. A. Heijboer, arXiv:0908.0816 (2009).
fluxes with a spectral shape different from E−2 7. J. Brunner, 1st VLVnT Workshop, Amsterdam.
are also tested. For each model a cut value R∗ is http://www.vlvnt.nl/proceedings/
8. J.A. Aguilar et al., Astropart. Phys. 34 (2010)
optimized following the procedure in §3.2. Table
179.
2 gives the results for the models tested: the val-
∗ 9. J.A. Aguilar et al., Nucl. Instrum. Meth. A555
ues of R , the numbers N of ν signal events
mod µ (2005) 132.
for R ≥ R∗, the energy intervals where 90% of
10. S.Aielloetal.,Nucl.Instrum.Meth.A614(2010)
the signal is expected, the ratios between µ
90% 206.
(computed according to [19]) and Nmod. A value 11. G.D. Barr et al., Phys. Rev.D70 (2004) 023006.
of µ90%/Nmod < 1 indicates that the theoretical 12. C.G.S. Costa, Astropart. Phys. 16 (2001) 193.
modelisinconsistentwiththeexperimentalresult 13. M.Ageronetal.,Astropart.Phys.31(2009)277.
at the 90% c.l. [20]. 14. Y.Becherinietal., Astropart.Phys.25(2006) 1;
G. Carminati et al., Comp. Phys. Comm. 179
(2008) 915.
5. Conclusions
15. J.A.Aguilaretal.,Astropart.Phys.33(2010)86.
Using data from 334 days of equivalent live 16. J.A. Aguilar et al., Nucl. Instrum. Meth. A622
(2010) 59.
time collected with the ANTARES telescope, a
17. A. Romeyeret al., hep-ex/0308074 (2003).
search for a diffuse flux of high energy cosmic
18. G.C. Hill and K. Rawlins, Astropart. Phys. 19
muon neutrinos was made. The 90% c.l. upper
(2003) 393.
limitonthediffuseν fluxwithaE−2spectrumis
µ 19. G.J.FeldmanandR.D.Cousins, Phys.Rev.D57
set at E2Φ = 5.3×10−8 GeV cm−2 s−1 sr−1
90% (1998) 3873.
in the energyrange20TeV – 2.5PeV.Other sig- 20. J.A. Aguilar et al., arXiv:1011.3772 (2010). Sub-
nal models with different energy shape are also mitted to Phys. Lett. B.
testedandsomeofthemexcludedatthe90%c.l.. 21. J. Conrad et al., Phys. Rev.D67 (2003) 012002.