Searches for clustering in the time integrated skymap of the ANTARES neutrino telescope

Prepared for submission to JCAP Searches for clustering in the time integrated skymap of the ANTARES neutrino telescope S. Adrián-MartínezaA. AlbertbM. AndrécG. AntoneM. Ardida J.-J. AubertfB. BaretgJ. Barrios-MartíhS. BasaiV. Bertinf S. Biagij,k C. BogazzilR. Bormuthl,m M. Bou-Caboa M.C. BouwhuislR. BruijnlJ. BrunnerfJ. BustofA. Caponen,o L. CarametepJ. CarrfS. CecchinijT. ChiarusijM. Circellaq R. ConiglionevL. CorefH. CostantinifP. CoylefA. Creusotg C. CurtilfG. De Rosas,t I. DekeyserrA. Deschampsu G. De Bonisn,o C. Donzaudg,w D. DornicfQ. Dorostix D. DrouhinbA. DumasyT. EberleD. ElsässerzA. Enzenhöfere S. EscoffierfK. FehneI. FelisaP. Fermanin,o F. Folgere L.A. Fuscoj,k S. GalatàgP. GayyS. GeißelsödereK. Geyere V. Giordanoaa A. GleixnereJ.P. Gómez-GonzálezhK. Grafe G. GuillardyH. van Harenab A.J. HeijboerlY. HellouJ.J. Hernández-ReyhB. HeroldeA. HerreroaJ. HößleJ. Hofestädte C. HugondC.W JameseM. de Jongl,m M. KadlerzO. Kalekine U. KatzeD. KießlingeP. Kooijmanl,ac,ad A. Kouchnerg I. Kreykenbohmae V. Kulikovskiyaf,d R. LahmanneE. Lambardf G. LambardhD. LattuadavD. LefèvrerE. Leonoraaa,ag H. LoehnerxS. Loucatosah S. ManganohM. Marcelini A. Margiottaj,k J.A. Martínez-MoraaS. MartinirA. Mathieuf T. MichaellP. MigliozzisC. Muellerae M. NeffeE. Nezrii D. PalioselitislG.E. PăvălaşpC. Perrinan,o P. PiattellivV. Popap T. Pradierai C. RaccabG. RiccobenevR. RichtereK. Roensche A. Rostovtsevaj M. SaldañaaD.F.E. Samtlebenl,m A. Sánchez-LosahM. Sanguinetid,ak J. SchmideJ. Schnabele S. SchultelF. Schüssler1,ah T. SeitzeC. SiegereA. Spiese M. Spurioj,k J.J.M. SteijgerlTh. Stolarczykah M. Taiutid,ak C. TamburinirY. Tayalatial A. TrovatovB. Vallageah C. Valléef V. Van ElewyckgE. VisserlD. Vivolos,t S. WagnereJ. Wilmsae E. de Wolfl,ad K. YatkinfH. YepeshJ.D. ZornozahJ. Zúñigah 1corresponding author arXiv:1402.2809v2 [astro-ph.HE] 7 May 2014 aInstitut d’Investigació per a la Gestió Integrada de les Zones Costaneres (IGIC) - Universitat Politècnica de València. C/ Paranimf 1 , 46730 Gandia, Spain. bGRPHE - Institut universitaire de technologie de Colmar, 34 rue du Grillenbreit BP 50568 - 68008 Colmar, France cTechnical University of Catalonia, Laboratory of Applied Bioacoustics, Rambla Exposició,08800 Vilanova i la Geltrú,Barcelona, Spain dINFN - Sezione di Genova, Via Dodecaneso 33, 16146 Genova, Italy eFriedrich-Alexander-Universität Erlangen-Nürnberg, Erlangen Centre for Astroparticle Physics, Erwin-Rommel-Str. 1, 91058 Erlangen, Germany fCPPM, Aix-Marseille Université, CNRS/IN2P3, Marseille, France gAPC, Université Paris Diderot, CNRS/IN2P3, CEA/IRFU, Observatoire de Paris, Sorbonne Paris Cité, 75205 Paris, France hIFIC - Instituto de Física Corpuscular, Edificios Investigación de Paterna, CSIC - Universitat de València, Apdo. de Correos 22085, 46071 Valencia, Spain iLAM - Laboratoire d’Astrophysique de Marseille, Pôle de l’Étoile Site de Château-Gombert, rue Frédéric Joliot-Curie 38, 13388 Marseille Cedex 13, France jINFN - Sezione di Bologna, Viale Berti-Pichat 6/2, 40127 Bologna, Italy kDipartimento di Fisica dell’Università, Viale Berti Pichat 6/2, 40127 Bologna, Italy lNikhef, Science Park, Amsterdam, The Netherlands mHuygens-Kamerlingh Onnes Laboratorium, Universiteit Leiden, The Netherlands nINFN -Sezione di Roma, P.le Aldo Moro 2, 00185 Roma, Italy oDipartimento di Fisica dell’Università La Sapienza, P.le Aldo Moro 2, 00185 Roma, Italy pInstitute for Space Sciences, R-77125 Bucharest, Măgurele, Romania qINFN - Sezione di Bari, Via E. Orabona 4, 70126 Bari, Italy rMediterranean Institute of Oceanography (MIO), Aix-Marseille University, 13288, Marseille, Cedex 9, France; Université du Sud Toulon-Var, 83957, La Garde Cedex, France CNRS-INSU/IRD UM 110 sINFN -Sezione di Napoli, Via Cintia 80126 Napoli, Italy tDipartimento di Fisica dell’Università Federico II di Napoli, Via Cintia 80126, Napoli, Italy uGéoazur, Université Nice Sophia-Antipolis, CNRS, IRD, Observatoire de la Côte d’Azur, Sophia Antipolis, France vINFN - Laboratori Nazionali del Sud (LNS), Via S. Sofia 62, 95123 Catania, Italy wUniv. Paris-Sud , 91405 Orsay Cedex, France xKernfysisch Versneller Instituut (KVI), University of Groningen, Zernikelaan 25, 9747 AA Groningen, The Netherlands yLaboratoire de Physique Corpusculaire, Clermont Université, Université Blaise Pascal, CNRS/IN2P3, BP 10448, F-63000 Clermont-Ferrand, France zInstitut für Theoretische Physik und Astrophysik, Universität Würzburg, Emil-Fischer Str. 31, 97074 Würzburg, Germany aaINFN - Sezione di Catania, Viale Andrea Doria 6, 95125 Catania, Italy abRoyal Netherlands Institute for Sea Research (NIOZ), Landsdiep 4,1797 SZ ’t Horntje (Texel), The Netherlands acUniversiteit Utrecht, Faculteit Betawetenschappen, Princetonplein 5, 3584 CC Utrecht, The Netherlands adUniversiteit van Amsterdam, Instituut voor Hoge-Energie Fysica, Science Park 105, 1098 XG Amsterdam, The Netherlands aeDr. Remeis-Sternwarte and ECAP, Universität Erlangen-Nürnberg, Sternwartstr. 7, 96049 Bamberg, Germany af Moscow State University, Skobeltsyn Institute of Nuclear Physics,Leninskie gory, 119991 Moscow, Russia agDipartimento di Fisica ed Astronomia dell’Università, Viale Andrea Doria 6, 95125 Catania, Italy ahDirection des Sciences de la Matière - Institut de recherche sur les lois fondamentales de l’Univers - Service de Physique des Particules, CEA Saclay, 91191 Gif-sur-Yvette Cedex, France aiIPHC-Institut Pluridisciplinaire Hubert Curien - Université de Strasbourg et CNRS/IN2P3 23 rue du Loess, BP 28, 67037 Strasbourg Cedex 2, France ajITEP - Institute for Theoretical and Experimental Physics, B. Cheremushkinskaya 25, 117218 Moscow, Russia akDipartimento di Fisica dell’Università, Via Dodecaneso 33, 16146 Genova, Italy alUniversity Mohammed I, Laboratory of Physics of Matter and Radiations, B.P.717, Oujda 6000, Morocco E-mail: fabian.sc[email protected] Abstract. This paper reports a search for spatial clustering of the arrival directions of high energy muon neutrinos detected by the ANTARES neutrino telescope. An improved twopoint correlation method is used to study the autocorrelation of 3058 neutrino candidate events as well as cross-correlations with other classes of astrophysical objects: sources of high energy gamma rays, massive black holes and nearby galaxies. No significant deviations from the isotropic distribution of arrival directions expected from atmospheric backgrounds are observed. Contents 1 Introduction 1 1.1 The ANTARES neutrino telescope 2 1.2 Dataset 3 2 Autocorrelation analysis 4 2.1 Method 4 2.2 Reference autocorrelation distribution and comparison with data 5 2.3 Performance and sensitivity 6 2.4 Autocorrelation results and discussion 7 3 Two-point cross-correlation with external catalogues 8 3.1 High-energy gamma rays 8 3.2 The local universe 9 3.3 Massive black holes 9 4 Summary 10 1 Introduction The key question to resolve the long standing mystery of the origin of cosmic rays is to locate their sources and understand the mechanisms that accelerate these particles up to energies orders of magnitude above the energies reached by man-made accelerators. Over the last years it has become obvious that multiple messengers will be needed to achieve this task. Fundamental particle physics processes like the production and subsequent decay of pions in interactions of high energy particles predict that the acceleration sites of cosmic rays should also be sources of high energy gamma rays and neutrinos. The detection of astrophysical neutrinos and the identification of their sources is one of the main goals of neutrino telescopes operated at the South Pole (IceCube [1]), in Lake Baikal [2] and in the Mediterranean Sea (ANTARES [3]). Despite significant effort, no clear signature for point-like sources of astrophysical neutrinos has been found so far [4–9]. Currently, both the spatial distribution as well as the morphologies of sources potentially emitting neutrinos in the TeV energy range are unknown. Similar to the distribution of observed sources emitting high energy gamma rays, they are supposed to be distributed very inhomogeneously throughout our cosmic neighbourhood. A significant fraction of them should be located in the Galactic disk and be spatially extended (e.g. shell-type supernova remnants). It is therefore interesting to study the intrinsic clustering of the arrival directions of neutrino candidates. In this analysis an improved autocorrelation method is used to this end. As no prior information about the potential sources is required biases are naturally reduced. Since it covers a large angular range, i.e. neutrino emission regions of very different sizes, this study complements the searches for point-like sources and could provide hints for underlying, yet unresolved, source morphologies and source distributions. Exploiting the expected multimessenger signatures of potential sources the introduced method is extended to searches for correlations between the arrival directions of neutrino candidates and other classes of – 1 – ! storey Figure 1. Schematic view of the ANTARES telescope. The inset shows a photograph of an optical storey. astrophysical objects: sources of high energy gamma rays, massive black holes and nearby galaxies. 1.1 The ANTARES neutrino telescope The ANTARES telescope [3] became fully operational in 2008. The detector comprises twelve detection lines anchored at a depth of 2475 m and 40 km off the French coast near Toulon. The detector lines are about 450 m long and host a total of 885 optical modules (OMs), each comprising a 17” glass sphere which houses a 10” photomultiplier tube. The OMs look downward at 45◦in order to optimise the detection of upgoing, i.e. neutrino induced, tracks. The geometry and size of the detector make it sensitive to extraterrestrial neutrinos in the TeV-PeV energy range. A schematic layout of the telescope is shown in Figure 1. The neutrino detection is based on the induced emission of Cherenkov light by high energy muons originating from charged current neutrino interactions inside or near the instrumented volume. All detected signals (hits) are transmitted via an optical cable to a shore station, where a computer farm filter the data for coincident signals in several adjacent OMs. The muon direction is then determined by maximising a likelihood which compares the time of the hits with the expectation from the Cherenkov signal of a muon track. Details on the event reconstruction are given in Ref. [7,10]. Two main backgrounds for the search for astrophysical neutrinos can be identified: downgoing atmospheric muons which have been mis-reconstructed as upgoing and atmospheric neutrinos originating in cosmic ray induced air showers at the opposite side of the Earth. Depending on the requirements of the analysis both backgrounds can, at least partially, be discriminated using various parameters such as the quality of the event reconstruction or – 2 – number of hits 0 50 100 150 200 250 probability -5 10 -4 10 -3 10 -2 10 -1 10 selected data events atmospheric neutrinos (Bartol) ) -2 astrophysical neutrinos (E Figure 2. Normalised distribution of the number of hits used in the event reconstruction for data (black markers) and Monte Carlo simulations (atmospheric neutrinos following the parametrisation from Ref. [11]: black histogram; astrophysical neutrinos: red, dotted histogram). an estimator of the energy of the muon, e.g. the number of hits used in the track reconstruction. The latter, being strongly correlated with the energy of the original neutrino, helps to discriminate events of atmospheric origin from neutrinos produced in astrophysical sources. Atmospheric neutrinos have a much softer energy spectrum (∝E−3.7) compared to the generic E−2spectrum expected from Fermi acceleration in astrophysical sources. As illustrated in Figure 2this difference affects the distribution of energy dependent parameters (or energy estimators) and can therefore be used to enhance the background discrimination. In addition, analysing the reconstructed arrival directions of the events allows to search for an excess over the isotropic atmospheric backgrounds. Both features will be exploited in the analysis described in this paper. 1.2 Dataset The data analysed here has been recorded by the ANTARES neutrino telescope between 2007 and 2010. During the beginning of this period (2007-2008) the detector was in its commissioning phase, increasing from five active detection lines to the full detector with twelve lines by mid 2008. After imposing basic data quality requirements, the event selection criteria have been optimised by means of Monte Carlo simulations to yield the best average upper limit on the neutrino flux in a search for point-like sources [7]. These criteria are mainly a cut on the reconstructed zenith angle, θ > 90◦, a requirement on the reconstruction quality parameter, Λ, as well as a cut on the estimated angular uncertainty of the track reconstruction, β < 1◦. A total of 3058 neutrino candidates are found in 813 days of effective lifetime. Monte Carlo simulations show that the contribution from misreconstructed atmospheric muons is about 15 % in this dataset. A skymap in galactic coordinates of these events is shown in the upper left plot of Figure 7. – 3 – number of events from a point-like source 2 4 6 8 10 12 14 16 18 20 effectσprobability for a 3 0 0.2 0.4 0.6 0.8 1 nHit estimator estimatorX/dEd no energy estimator Figure 3. Probability to detect with a 3σsigma significance a single point-like source following an E−2energy spectrum as a function of its neutrino luminosity, i.e. the number of detected neutrinos. The blue triangles denote the standard autocorrelation method without an energy estimator. The black squares show the performance including the dE/dXenergy estimator [13] and the red circles denote the finally used method using the nHit estimator. 2 Autocorrelation analysis 2.1 Method The most commonly used method to detect intrinsic clusters within a set of Nevents is the standard two-point autocorrelation distribution. It is defined as the differential distribution of the number of observed event pairs, Np, in the dataset as a function of their mutual angular distance, ∆Ω. This technique has already been applied to the first two years of ANTARES data. No significant clustering has been detected [12]. Here, an improvement of this method by using an estimator of the neutrino energy is presented. To suppress statistical fluctuations that would reduce the sensitivity of the method, the cumulative autocorrelation distribution is used. It is defined as Nˆ E(∆Ω) = N X i=1 N X j=i+1 wij ·[1 −H(∆Ωij −∆Ω)] ,(2.1) where His the Heaviside step function. The weights wij =wi·wjare calculated using the individual event weights wi=Rˆ Ei 0f(ˆ E) d ˆ E, where f(ˆ E)is the cumulative distribution of the energy estimator ˆ Efor the background. Astrophysical neutrinos are more likely to produce events with a higher value of the energy estimator ˆ Eithan atmospheric neutrinos. This is represented by a higher event weight wi. The used distribution is built from large statistics Monte Carlo simulations reproducing the actual data taking conditions, including, for example, the time dependent background fluctuations induced by bioluminescence. These simulations have been validated by extensive comparisons with data. An example is shown in Figure 2, where the number of hits used in the event reconstruction is depicted (see [7,10] – 4 – test statistic 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 probability -4 10 -3 10 -2 10 -1 10 1 background 5 signal events 10 signal events Figure 4. Normalized distribution of the maximum value of the test statistic derived from pseudoexperiments with scrambled data (black solid line). The red dashed (blue dotted) line corresponds to pseudo-datasets including a point-like source of 5 (10) events. for further details). The simulated events used to build the f(ˆ E)distribution follow an energy spectrum as expected for atmospheric neutrinos [11] (black histogram in Figure 2). Modifying the standard autocorrelation by these weights leads to a significant increase of the sensitivity to detect clustering of (astrophysical) events following a harder energy spectrum. This improvement is illustrated in Figure 3. Various possibilities exist for the estimation of the energy and the definition of the weights. As crosscheck of the stability and performance of the method, the full analysis has been performed using two different energy estimators: the number of hits used during the final step of the event reconstruction, nHit, as in the search for point-like sources [7], as well as a recently developed estimator exploiting the correlation between the energy deposit, dE/dX, and the primary energy [10,13]. Both provide very similar results. As shown in Figure 3, the nHit energy estimator shows a slightly better performance for weak sources and is therefore retained for the final analysis. Pseudo-experiments as described below are used to determine the optimal size of the angular steps ∆Ω. Increasing the number of angular steps enhances the angular resolution of the method but degrades the sensitivity due to the increasing number of trials (look-elsewhereeffect [14]). Taking into account the median angular resolution of 0.5◦[7], an optimum has been found for angular steps of about 0.1◦. 2.2 Reference autocorrelation distribution and comparison with data To detect structures in the sky distribution of the selected events, a reference autocorrelation distribution to compare with is needed. This reference is determined by scrambling the data themselves, a method which allows the reduction of systematic uncertainties potentially introduced by the use of Monte Carlo simulations. The scrambling is performed keeping the pairs of local coordinates (zenith, azimuth) in order to avoid losing information about possible correlations between them. The detection time is drawn randomly from another event within – 5 – events per source 2 4 6 8 10 12 14 16 18 20 effectσprobability for a 3 0 0.2 0.4 0.6 0.8 1 1 source 2 sources 3 sources events per source 2 4 6 8 10 12 14 16 18 20 effectσprobability for a 3 0 0.2 0.4 0.6 0.8 1 =0.5 degσ =1 degσ =2 degσ Figure 5. Probability for a 3σeffect using the autocorrelation method exploiting the nHit energy estimator. Left plot: Dependence on the number of injected point-like sources in the visible sky. Right plot: Dependence on the extension of a single source modelled by a 2-dimensional Gaussian. the same detector configuration to keep track of the changing layout of the detector due to its construction and maintenance. This method is applied to all selected events and a randomised sky map naturally reproducing the coverage of the unscrambled ANTARES data is constructed. This randomised sky is then analysed in exactly the same way as the data to derive the autocorrelation function. The randomisation process is performed about 106times and the derived autocorrelation distributions are averaged in order to reduce statistical fluctuations. Structures in the sky distribution of the data will show up as differences between the autocorrelation distribution of the data and the reference distribution. The comparison is performed by using the formalism introduced by Li and Ma [15]. This formalism provides a raw test statistic, t, as a function of the cumulative angular scale. As the comparison is performed bin-by-bin and as the scan is made over different angular scales, this result has to be corrected for the corresponding trial factor. To limit the number of trials the scan is performed only up to 25◦, a scale which includes most known extended sources and emission regions. Finally, the method proposed by Finley and Westerhoff [16] is applied by performing about 106pseudo experiments in which the autocorrelation distributions of randomised sky maps are compared with the reference distribution. For each simulated map the maximum value of the test statistic is calculated. The obtained distribution is shown as black solid line in Figure 4. To reach high significances the tail of the obtained distribution is fitted and extrapolated with an exponential function. The final p-value of the analysis is then calculated as the probability to obtain the same or a higher value of tfrom these background-only pseudo-experiments. 2.3 Performance and sensitivity The performance of the algorithm is determined using mock datasets built by scrambling the selected data events as described above. While keeping the total number of events constant, predefined source structures with various sizes and source luminosities are added. The angular resolution of the detector is taken into account by convolving the intrinsic source size with a two dimensional Gaussian with a width of σ= 0.5◦. The energy estimater for the injected – 6 –