Evaluation Tools for the Performance of a NESTOR Test Detector

Evaluation Tools for the Performance of a NESTOR Test Detector G. Bourlis, A. Leisos, S. E. Tzamarias and A. Tsirigotis Particle and Astroparticle Physics Group School of Science and Technology Hellenic Open University A. Staveris NESTOR Institute for Deep Sea Research, Technology and Neutrino Astroparticle Physics Abstract In March 2003 a NESTOR floor fully equipped with electronics and associated environmental sensors was deployed in the Ionian Sea at a depth of about 4000m. The deployed detector was operated continuously for more than a month and more than five million events were accumulated. In this note we describe the methods and the tools which have been developed in order to evaluate the performance of this NESTOR test detector. 1. Introduction NESTOR (Neutrino Extended Submarine Telescope with Oceanographic Research) is a deep underwater neutrino detector, located in the Mediterranean sea, approximately 15 nm south-east of the coast of Pylos (Greece), at a depth of 4000 meters [1,2].The basic detector unit is a rigid hexagon, made out of titanium with a diagonal of 32 m. 1

Figure 1 (Left) An artist view of a full Nestor tower consisted of twelve floors separated by 30 meter. (Right) A titanium floor during deployment. At the tip of each arm of the hexagonal floor there is a pair of two 15 inch photomultiplier tubes (PMTs) inside benthos glass housings [3], one looking upward and the other down wards. The electronics which are responsible for signal sensing, triggering, digitisation and data transmission to the shore are housed inside a large titanium sphere (1m in diameter) located at the center of the hexagonal floor. The electrical pulses of the PMTs are digitised by the Analog Transient Waveform Digitizers (ATWDs) of the floor electronics board (developed at Lawrence Berkeley National Laboratory). The digitised waveforms are transmitted to shore, where the raw data are recorded [4]. By stacking 12 of these floors in the vertical, with a distance between them 30m, they create a tower shown in Figure 1, which is connected to the shore by an electro-optical cable (18 fibers plus 1 conductor). In March 2003, using the cable ship RAYMOND CROZE of France Telecom, a hexagonal floor was deployed fully equipped with electronics and associated environmental sensors to a depth of 4000m. The deployed detector was operated continuously for more than a month and over five million events were accumulated. The data have been collected by using several event selection trigger modes, PMT coincidence levels and PMT thresholds. In addition, plethora of calibration events was taken using the LED flash units. Of this data, some two million events were accumulated under constant running conditions 2

with a 4-fold or higher coincidence level trigger and 30mV threshold for each PMT: this event sample has been used in the following analysis of the detector performance and for track reconstruction. In the next Sections we present our study results by comparing several distributions of experimental parameters with the corresponding distributions obtained by Monte Carlo simulation. 2. Performance of the Data Acquisition System The Data Acquisition System [5,6,7] proved to be very robust without any single failure during the run. The Data Acquisition cluster [5] which was responsible for the Data Storage, the Online Monitor and the Data Quality Monitor operated continuously without any crashes or unexpected hiccups. Data were collected in runs under constant experimental conditions. However in order to evaluate the detector performance several sets of experimental parameters were used to collect data. Totally 141 run files were saved in storage media corresponding to 120 GBytes of accumulated data. The readout and DAQ chain was operated with practically no dead time and the monitored experimental parameters (environmental and operational) remained stable within accepted tolerances. The PMT counting rates remained stable during the whole running period at a level of around 50 khz per PMT, due principally to Cherenkov light emitted by electrons from K 40 decays [8]. It was checked that the PMT counting rate measured by the readout electronics was not affected by the trigger conditions. As it is shown in Figure 2 the PMT counting rates evaluated from data collected with different trigger coincidence levels remained constant indicating thus that the formation and the dead time of the trigger does not affect the readout performance. 3

Figure 2 The counting rates of the up-looking PMTs with respect to the coincidence level. The stability of the rates assures that there is not any bias to the trigger selection. The averaged trigger rate, corresponding to the existence of four or more PMT pulses above 30mV amplitude inside the trigger window (four fold coincidence), was 3.76 Hz. This is in a very good agreement to the estimated value of 3.79 Hz derived from the Monte Carlo simulation [8]. According to the Monte Carlo estimation [9], only a small fraction (5.5%, 0.21Hz) of this rate is due to the atmospheric muons passing around of the detector. When the PMT thresholds were increased to 120mV, the measured trigger rate was found to be 0.29Hz, in agreement also with the Monte Carlo estimation of 0.30Hz. Table 1 represents the measured and the Monte Carlo estimated trigger rates for two different PMT thresholds. 4

Thresholds at 30 mv Thresholds at 120 mv Measured Total Trigger Rates (greater or equal to 4fold) M.C. Prediction (atmospheric muons +K 40 ) 3.76± 0.08 0.29± 0.01 3.79 ± 0.08 0.30± 0.01 M.C. Prediction (atmospheric muons only) 0.216 ± 0.01 0.12± 0.01 Table 1 Measured and expected trigger rates for different threshold values. Furthermore, the measured trigger rates corresponding to different PMT coincidence levels found also to be in a very good agreement with the Monte Carlo estimations, for different PMT thresholds. In Figure 3, we present the Monte Carlo estimated rates in comparison with the measured values for 9 selection trigger coincidence levels and two different PMT threshold settings. 5

Figure 3 Trigger rates as a function of the coincidence level, for two threshold settings. The points represent the data, the solid line the Monte Carlo estimation including background and the dashed line the Monte Carlo estimation for the contribution of the atmospheric muons. In addition, several studies have been made to ensure that all the collected light on the PMTs can be attributed to known sources. Since higher coincidence levels reject the combinatorial background and more photons are collected by the PMTs in events corresponding to atmospheric muons than in background events, the dependence of the total number of collected photons per event on the coincidence level has been studied. The total number of accumulated photoelectrons, inside the coincidence window, has been used as a measure of the total number of collected photons. In Figure 4 they are compared a) the mean value of the number of accumulated photoelectrons inside the coincidence window, as a function of the coincidence level b) to the Monte Carlo prediction. There is a faster than a linear increase with the coincidence level indicating thus that with higher coincidence levels we are selecting 6

atmospheric muon events. The description of this effect by the Monte Carlo is in a very good agreement with our measurements. This agreement indicates that the Monte Carlo contains all the major sources of the experimental signal. Figure 4 Total number of photoelectrons inside the coincidence window as a function of the coincidence level for two threshold settings. The points represent the data and the histogram gives the Monte Carlo estimation. 7

3. Performance of the Calibration System Many calibration runs were performed in order to check the synchronization and the stability of the gains [10]. The LED flushers were positioned about 20 meters above and below the NESTOR floor calibrating the up-looking and the down-looking PMTs respectively. Self generated electronic pulses were used in special runs to calibrate the gain of the ATWD [11] channels while the K 40 background has been used as a stable standard candle in order to monitor the gain stability of the PMTs. The PMT pulse height distributions from each data file selected under normal conditions were compared to a standard shape defined at the beginning of the run and found to be extremely stable for all of the PMTs of the detector during the whole running period (Figure 5). Figure 5 The pulse height distribution for the up-looking PMTs established at the beginning of the run (solid histograms). The crosses represent the pulse height distribution measured a few days later under the same experimental conditions. 8

Another sensitive test is found to be the examination of the pulse height distributions of each individual PMT that contribute to events selected with a high multiplicity coincidences: these are typically pulses produced by atmospheric muons. The pulse height distribution of a typical PMT (in units of the mean value of the pulse height distribution corresponding to the emission of one only photoelectron), when participating in a 6-fold or higher coincidence, is shown in Figure 6 and it is compared to a Monte Carlo estimation. The agreement between the measured and predicted spectra, which has been verified for all PMTs of the detector, indicates that the collected light is produced by the sources that are used in the detector simulation.. Figure 6 The pulse height distribution of a typical PMT, in units of the mean value of the one photoelectron distribution, participating in a high level coincidence. The crosses represent the data whilst the histograms show the corresponding Monte-Carlo prediction. 9

The LED calibration system has been used extensively [10] to study the dependence of the arrival time definition of each PMT pulse on its amplitude. The results of this study (slewing and statistical error) have been parameterized as functions of the pulse amplitude and had been used in the track reconstruction analysis. The global arrival time distribution of the accumulated photoelectrons during data collection was studied using events selected by triggers corresponding to at least six PMT pulses inside the trigger window. We have used the distribution of the arrival time of any digitized PMT pulse, weighted by the pulse amplitude (in units of the mean value of the one photoelectron pulse height distribution) and normalized to the total number of selected events. This distribution expresses the correlation of the Cherenkov light intensity and the arrival time of the corresponding PMT pulse. In Figure 7 it is shown the comparison of this arrival time distribution between Monte Carlo produced events and real data for three different cases. In the first case all the PMT pulses for all the events that have at least six PMT pulses inside the trigger window are considered (Figure 7a). In the second case we have used all the PMT pulses for the events that have survived the track reconstruction algorithm (Figure 7b) and in the final case we have used only the PMT pulses which participate in the track fitting procedure (Figure 7c). In all cases the Monte Carlo simulation agrees 1, within statistical errors, with the experimental global arrival time distribution of the accumulated photoelectrons indicating that there are not out of time light producing sources which have not been considered in the simulation. 1 The small peak around 400 ns which appears in the real data distributions is caused by a known malfunction of the ATWD digitization. Because it lies outside the trigger window does not affect the track reconstruction procedures. 10

Figure 7 Global arrival time distribution of the accumulated photoelectrons (red points) compared with the Monte Carlo expectation (black points) for events with six or more PMT pulses inside the trigger window. a) All PMT pulses included, b) All PMT pulses for reconstructed events, c) PMT pulses included in the fitting procedure. 11

4. Tracking Performance The raw data collected by the detector are analysed taking into account the information from the calibration database and the database containing the operational parameters of the run. The processing of the digitised waveforms [12] restores the original shape of the PMT pulses and extracts the arrival time as well as the amplitude of the pulses in units of the mean of the single photoelectron distribution. These two information along with the position of the PMT define a hit for the detector. The hits are used to the standard tracking algorithm [13] in order to estimate the five track parameters (zenith angle, azimuthal angle and the pseudo vertex position). Furthermore quality cuts are applied in order to improve the reconstruction efficiency and resolution [13]. The data analysis flow described above is presented in the Figure 8. Figure 8 The data analysis flow From the total data sample collected with a 4-fold or higher coincidence trigger and 30mV ATWD threshold, a subset containing 45800 events has been selected that have six or more PMT pulses (hits) within the 60 ns time window. These events have been analysed in order to reconstruct muon tracks. The arrival time of the digitized PMT pulses was used to estimate the muon track parameters by means of a χ 2 fit whilst the 12

PMT pulse heights were used to reject ghost solutions and poorly reconstructed tracks. The details of the reconstruction strategy and the relevant studies are reported elsewhere [14]. Figure 9 shows the distribution of the impact parameter of the reconstructed tracks in comparison with the Monte Carlo prediction. This small (six meters in radius) detector can resolve and reconstruct muons which are passing close to the periphery of the star. Muons passing through the detector produce pulses with similar arrival times and cannot be unambiguously reconstructed while muon tracks passing far away from the detector also are difficult to be reconstructed. Figure 9 The estimated impact parameter of the reconstructed tracks (crosses) compared with the Monte Carlo predictions (solid histogram). Figure 10 shows the distribution of the azimuth angle of the reconstructed tracks. As expected, the distribution in azimuth of the muon tracks at the detector depth is not affected by the detector response or the reconstruction efficiency. 13

Figure 10 The experimental distribution of the reconstructed azimuth angles (solid points) compared with the Monte Carlo prediction (histogram). Notice that the distributions shown in Figure 9 and 10 are in a very good agreement with the Monte Carlo predictions proving that the detector characteristics and performance have been described honestly in the simulation. Finally, the zenith angular distribution of the reconstructed tracks is compared to the Monte Carlo prediction in Figure 11. Although the model that describes the atmospheric muon flux at the detector depths (Okada model [15]) does not contain any muons coming below the horizon, due to resolution effects the distribution exceeds the 90 degrees. In order to quantify the level of agreement between the measured data and the predictions of the Okada model, the χ 2 probability (statistical similarity) of the experimental points to the Monte Carlo prediction was calculated. This was found to be 52%, demonstrating a very good agreement. 14

Figure 11 Zenith angular distribution (θ) of reconstructed tracks for the data (triangles) and Monte Carlo (solid points) event sample. The insert plot shows the same distributions on a linear scale. 5. Measurement of the Atmospheric Muon Flux at the Detector Depth The number of atmospheric muons (N) arriving at the detector depth per unit solid angle (Ω), per unit time (t) and per unit area (S), dn, is usually parameterized dω dt ds as [16,17]: dn =Io cos α ( θ ) dω dt ds (1) where I o is the vertical intensity. The index α has been found to be equal to 4.5±0.8 in previous measurements at 3697m water depth at the NESTOR site [18]. The vertical intensity was evaluated by integrating equation (1) and setting the total number of muons equal to the total 15

number of the reconstructed data tracks (D=745) divided by the total efficiency (ε) in reconstructing atmospheric muon tracks. I = o D ( α +1) < ε > 2 πτs (2) where T stands for the total experimental time of 609580 s during which this data subset was accumulated. The total efficiency (ε) has been estimated from the Monte Carlo simulated data as the ratio of the number of reconstructed tracks to the corresponding number of atmospheric muons generated with energies greater than 1 GeV at the detector depth (in a circle of 100 m 2 radius at 100m above the detector). The total efficiency, corresponding 3 to a Monte Carlo production model following the angular distribution of equation (1), with α=4.5, found to be: ε = 3.89 10-4 ±0.04 10-4. The vertical atmospheric muon intensity, found using the formula (2) gives: Ι = ± ο 9 9-2 1 1 8.8 10 1.3 10 cm s sr (3) where the estimated error is calculated from statistical uncertainties in the data and Monte Carlo simulation and the measurement error on the index σ. This is in good agreement with predictions of the vertical intensity of the 9-2 1 atmospheric muons at a depth of 3800m.w.e, by Okada ( Ι = 8.8 10 cm s sr 9 1 and Bugaev et al ( Ι = 9.0 10 cm s sr NESTOR measurements [18] of ο -2 1 ο ) [19, 20] as well as with the previous I 9.8 10 4.0 10 cm s sr 9 9-2 -1-1 o = ± at depths between 3700 and 3900m. It is also consistent with the DUMAND measurement [17] 8 8-2 -1-1 of I 1.31 10 = ± 0.4 10 cm s sr at a depth of 3707m. o A more accurate analysis of the data with a simultaneous estimation of the index α and the vertical muon intensity I 0 is published elsewhere [21,22]. 1 ) 2 Approximately twice the light transmission length in the water at the experimental site. 3 The re-weighting of the Monte Carlo events, produced with the Okada model, to follow the differential flux of equation (1), is described in [21]. 16

6. Conclusions In this note we described the tools and methods that we have developed in order to evaluate the performance of the NESTOR detector deployed in March 2003. The Data Acquisition system proved to be very reliable and the measured trigger rates consistent with the prediction of the Monte Carlo simulations. The performance of the Calibration system was demonstrated with the pulse height distributions and the time profile of the recorded pulses for different experimental conditions. In all cases the distributions were in a very good agreement with the Monte Carlo expectations. Finally the tracking performance of the detector was examined and the measurement of the atmospheric muon flux at the detector depth found to be consistent with the OKADA model. Acknowledgments The authors of this note wish to thank the members of the NESTOR collaboration, the stuffs of the NESTOR Institute as well as the academic and technical personnel of the School of Science and Technology of the Hellenic Open University for their help, scientific, technical and financial support. 17

References [1] NESTOR: Proceedings of the 2nd NESTOR International Workshop, L. K. Resvanis editor (1992); Proceedings of the 3nd NESTOR International Workshop, L. K. Resvanis editor (1993); website http://www.nestor.org.gr. [2] L. K. Resvanis et al. High Energy Neutrino Astrophysics (1992), V. J. Stenger, J. G. Learned, S. Pakvasa and X. Tata editor. [3] E. G. Anassontzis, et al, Nuclear Instruments and Methods A479, pp 439-455 ( 2002). [4] S.E. Tzamarias, ``NESTOR first results. Electronics-DAQ-Data Analysis'', Published in Amsterdam 2003, Technical aspects of a very large volume neutrino telescope in the Mediterranean Sea. [5] The architecture of the Data Acquisition System at the shore Laboratory of the NESTOR Experiment, HOU-NS-TR-2004-05-EN. [6] A Data Quality Monitor System For The NESTOR Experiment, HOU-NS-TR-2004-06-EN [7] An Online Monitor System For The NESTOR Detector, HOU-NS-TR-2003-01- EN [8] NESTOR Data Analysis: Background Sources and Rejection Techniques, HOU-NS-TR-2004-04-EN [9] A. Tsirigotis, phd thesis in preparation, Hellenic Open University [10] Performance of the NESTOR Calibration System, HOU-NS-TR-2004-02-EN [11] Stuart Kleinfelder, Analog Trasient Waveform Digitizer, LBNL 1998. [12] NESTOR analysis tools : Signal processing, HOU-NS-TR-2004-08-EN [13] NESTOR Analysis Tools : Track fitting, HOU-NS-TR-2004-09-EN [14] A. Tsirigotis, phd thesis in preparation, Hellenic Open University. See also Operation and performance of the NESTOR test detector, Nuclear Instruments and Methods in Physics Research A 552(2005) 420-439 18

[15] A. Okada, Astroparticle Physics 2, 393 (1994). [16] Peter K. F. Grieder, "Cosmic Rays at Earth", Elsevier, Amsterdam (2001), Chapter 4. [17] J. Babson et al, Physical Review D42, 3613 (1990). [18 I. F. Barinov et al, Proceedings of the 2nd NESTOR International Workshop, page 340, L. K. Resvanis editor (1992). [19] Edgar V. Bugaev, et al, Proceedings of the 3nd NESTOR International Workshop, page 268, L. K. Resvanis editor (1993). [20] Edgar V. Bugaev, et al, Physics Review D58, 054001 (1998). [21] Measurement of the atmospheric muon flux at the NESTOR Site, HOU-NS-TR-2004-11-EN [22] Measurement of the cosmic muon flux with a module of the NESTOR Neutrino Telescope, Astroparticle Physics, Vol 23, 377-392, 2005 19