Preprint typeset in JINST style - HYPER VERSION An extreme high resolution Timing Counter for the MEG experiment Upgrade M. De Gerone a, F. Gatti a,b, W. Ootani c, Y. Uchiyama c, M. Nishimura c, S. Shirabe d, P.W. Cattaneo e, M. Rossella e a Istituto Nazionale di Fisica Nucleare, Sezione di Genova, Via Dodecaneso 33, 16146, Genova (GE), Italy b Universitá degli Studi di Genova, Via Dodecaneso 33, 16146, Genova, Italy, c International Center for Elementary Particle Physics, University of Tokyo 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-0033, Japan d Department of physics, Kyushu University 6-10-1 Hakozaki, Higashi-ku, Fukuoka 812-8581, Japan e Istituto Nazionale di Fisica Nucleare, Sezione di Pavia, Via Agostino Bassi, 6, 27100, Pavia (PV), Italy E-mail: degerone@ge.infn.it ABSTRACT: The design and test of Timing Counter elements with resolution below 30 ps are presented. The detector is being designed for the upgrade of the MEG experiment, MEG II, looking for the µ eγ decay with an improved sensitivity of about one order of magnitude. It is based on several small plates of scintillator with a Silicon PhotoMultipliers dual-side readout. In this paper, the optimization of the single counter elements (SiPMs, scintillators, geometry) is described. Moreover, the results obtained with a first prototype tested at the Beam Test Facility (BTF) of the INFN Laboratori Nazionali di Frascati (LNF) are presented. KEYWORDS: Photon detectors for UV, visible and IR photons (solid-state) (PIN diodes, APDs, Si-PMTs, CCDs, EBCCDs etc), Scintillators, scintillation and light emission processes (solid, gas and liquid scintillators), Timing detectors. Corresponding author.
Contents 1. Introduction: the MEG experiment 1 2. The Timing counter Upgrade 1 3. Single counter optimization 3 3.1 SiPM comparison 3 3.2 Scintillator comparison 6 3.3 Geometry optimization 6 4. Beam test 7 4.1 Setup 7 4.2 Data analysis 8 4.3 Results 9 5. Conclusions 10 1. Introduction: the MEG experiment The MEG experiment has been running since 2008 at the Paul Scherrer Institute (Villigen, CH), looking for the µ eγ decay. The MEG collaboration recently published its last results [2] based on the analysis of data collected in the years 2009-2011: BR(µ eγ) 5.7 10 13 @90% C.L. While the analysis of the 2012-2013 data is still ongoing, an upgrade program of the MEG experiment (MEG II) started since 2012 [3], aiming to improve the experiment sensitivity by an order of magnitude, down to 5 10 14. In order to reach such a sensitivity, most of the current detector will be re-designed or modified. In this paper, the development of an extremely high resolution detector for the measurement of the positron timing will be described in details. 2. The Timing counter Upgrade The MEG detector [4] is designed to measure with the highest possible resolution the kinematic variables that define the signature of the decay µ eγ. Photons are detected by a Liquid Xenon detector placed outside the magnetic spectrometer where positrons are reconstructed (see figure 1). The spectrometer is made of a superconductive magnet, a set of Drift CHambers (DCH) and the Timing Counter (TC). The DCH system, together with the specially designed field provided by the COBRA magnet, measures positron energy and emission angle, while the purpose of the TC is to measure the positron time of impact. The current Timing Counter [5] is made of two identical arrays (placed inside the magnet up- 1
Figure 1. Schematic view of the positron spectrometer; upgraded Drift Chamber and Timing Counter designs are showed. Figure 2. Picture of the current Timing Counter. PMTs and scintillating bars lodged in a black plastic socket are visible. and down-stream the target position) of 15 scintillating bars (Bicron BC404), with 80 4 4cm 3 size. They are arranged in a barrel-like shape in order to fit the mechanical constraints of the spectrometer. Each bar is read-out on both sides by a fine mesh PhotoMultiplier Tube (PMT, Hamamatsu R5924). Signals from PMTs are properly processed in order to be fed into the trigger and DAQ system. A picture of the current Timing Counter is showed in figure 2. The Timing Counter has been running since 2008, showing very good and stable performances: a mean time resolution of 65 ps was stably reached over 5 years of run [6]. Despite the good results obtained, some issues suggest that the baseline design of the detector has to be changed to increase the resolution: the PMTs operation in high magnetic field and helium environment implies a deterioration of the PMTs transit time spread and gain, also using a fine mesh PMT; large size scintillating bars generate uncertainties on impact point reconstruction and spread of the trajectories of the optical photons inside the scintillator itself; the large amount of material crossed by the positron in the TC bar doesn t allow an effective usage of those events with more than one bar hit. All these problems originate from the usage of PMTs and large size scintillator bars. Thus, the natural choice is to modify the design of the Timing Counter basic element, reducing the size of the scintillator elementents and increasing the detector granularity. Concerning the read-out, the recent development of fast high gain solid state detector like Silicon PhotoMultipliers (SiPMs) permits to have compact devices with excellent timing performances. A detector made of several scintillator plates (from now on: pixel) with SiPMs based read-out allows to overcome the limitations of the current TC: magnetic field has no influence on SiPM operation; higher granularity of the detector results in smaller uncertainties from impact position measurement; thanks to the smaller amount of material along the positron trajectory, it will be possible to take advantage of the information coming from all the pixels crossed by the particle. 2
Figure 3. Schematic view of the new timing counter design. On the left: overview of the detector. On the right: detail of the single counter configuration. The last point is quite remarkable, because of the time resolution is expected to improve as 1/N hit, where N hit is the number of pixels crossed by the positron. Moreover, the small size of the single element results in a more flexible configuration of the detector, allowing the possibility to tailor the position and the density of the pixels along the detector. Also, very short rise time scintillator (like Bicron BC422, see section 3.2) can be used even in presence of a short attenuation length, which is no more an issue. Single counter good performances has already been proved [7], [10]. In the following, the research and development work on several prototype in order to choose the best material and devices will be presented. 3. Single counter optimization The optimization of the single counter configuration started from a systematic study among the SiPMs and the scintillating materials currently available to compare the properties relevant for our application. Then, an optimization of the pixel size has been performed, taking in mind that the final configuration has to be a trade-off between good timing resolution, high efficiency and a reasonable number of channels. 3.1 SiPM comparison Silicon Photomultipliers are considered to be a good solution for the new TC, thanks to their characteristics: good time resolution, quite high gain, compactness. In the last years, some manufacturers started to produce SiPMs, with different characteristics. We tested different models from Hamamatsu Photonics, Advansid, Ketek and SensL. All these devices share some features: they have size 3 3 mm 3, in order to be easily coupled to few mm thick scintillator pixels; moreover, they are sensitive in the near ultraviolet range (in order to match common plastic scintillators emission spectra); The summary of the SiPMs tested is showed in table 1. For each model, the noise level (dark count rate and cross-talk) together with the PDE has been evaluated. Moreover, also the breakdown dependence on temperature has been evaluated. Finally, the timing resolution has been measured on a prototype pixel with fixed sizes. 3
Table 1. Summary of SiPMs model tested. Manufacturer Model Type Note S10362-33-050C Conventional (Old) MPPC Ceramic package S10931-050P surface mount Hamamatus Photonics S12572-050C(X) New (standard type) MPPC Metal quench resistor S12572-020C(X) 25 µm pitch S12652-050C(X) Trench-type MPPC Metal quench resistor 3X3MM50UMLCT-B Improved fill factor Advansid NUV type Ketek PM3350 prototype-a Trench Type SensL MicroFB-30050-SMT B-Type Fast output Setup SiPMs are put in a thermal chamber, which allow to keep Device Under Test (DUT) at fixed temperature (23 C in the following measurements). Signals are transmitted on a coaxial cable to a voltage amplifier (developed at PSI, based on MAR-6SM amplifier [7]), then they are sampled at 5 Gs/s by a waveform digitizer (DRS4 evaluation board [8], [9] also developed at PSI). Dark noise and cross-talk The noise level of the device is evaluated by looking at the waveforms acquired by a random trigger. The dark count rate is calculated from the probability of observing zero photo-electron P(Np.e.= 0) in a fixed time window. The result is shown in figure 4(a) as a function of the applied over-voltage. The cross-talk probability is calculated from the P(Np.e. 2)/P(Np.e. 1) ratio including a correction for the accidental coincidence of the dark pulses. The result is shown in figure 4(b). The cross-talk probability almost linearly increases with the over voltage. The standard-type SiPMs, namely SiPMs without a trench structure, turned out to have worse performance with respect to the trench type whose improved structure strongly reduces the noise level. Anyway, the typical energy release in a pixel should guarantee an adequate signal-to-noise ratio also for those SiPMs with higher dark count and cross-talk rates. PDE The PDE for NUV light is measured with a LED whose wavelength (370âĂŞ410 nm) approximately matches the scintillator emission peak. The LED intensity is adjusted in such a way that the average number of observed photo-electrons ranges between 0.5 to 1.0. The relative PDE is then calculated from P(Np.e.= 0) in accordance with Poisson statistics, and thus the measured PDE value does not include the effect of cross-talk nor after-pulsing. The result is shown in figure 4(c). The highest PDE is obtained with Hamamatsu S12572 model, with 50 µm pitch cell. A more detailed description of noise and PDE studies can be found in [12]. Breakdown versus temperature dependence The BreakDown voltage (BD) versus temperature dependence has been measured by plotting the I-V characteristic of each SiPM at different temperatures (see figure 5(a)), in the range 20 45. Then, BD voltage has been plotted versus temperature and data are fitted linearly, as showed in figure 5(b), where the behavior for Advansid NUV SiPMs is displayed. The are shown in table 2. Time resolution The basic setup for the timing resolution measurement is the same as the one already described above. A scintillator pixel with size 60 30 5mm 3 is read-out on both sides by 4
) 3 Dark count rate (Mcps/mm 1.2 1 0.8 Old MPPC Old MPPC SMD New MPPC (50µm) New MPPC (25µm) New MPPC (Trench) KETEK Prototype-A AdvanSiD NUV SensL B-series 0.6 0.4 0.2 0 0 2 4 6 8 10 Over voltage (V) (a) Cross-talk probability 1 0.9 0.8 0.7 0.6 0.5 Old MPPC Old MPPC SMD New MPPC (50µm) New MPPC (25µm) New MPPC (Trench) KETEK Prototype-A AdvanSiD NUV SensL B-series 0.4 0.3 0.2 0.1 0 0 2 4 6 8 10 Over voltage (V) (b) Relative PDE 1.4 1.2 1 Resolution (ps) 100 90 80 Old MPPC New MPPC (50µ m) New MPPC (25µm) KETEK Prototype-A SensL B-series New MPPC (Trench) AdvanSiD NUV 0.8 70 0.6 60 0.4 Old MPPC Old MPPC SMD 50 0.2 New MPPC (50µm) New MPPC (25µ m) New MPPC (Trench) KETEK Prototype-A 40 AdvanSiD NUV SensL B-series 0 0 2 4 6 8 10 Over voltage (V) (c) 30 2 4 6 8 Over voltage/sipm (V) (d) Figure 4. Summary of the SiPMs comparison. a) Dark Count. b) Cross Talk Probability c) PDE d) Time Resolution. All results are given as a function of the applied over voltage. I-V curve VS Temperature: Advansid BD vs T - Advansid Current [µa] 10 1 ADV 15 ADV 20 ADV 25 ADV 30 ADV 35 ADV 40 ADV 45 BD Voltage [V] 26.4 26.3 26.2 26.1 26 p0 25.24 ± 0.05983 p1 0.02393 ± 0.001892-1 10 25.9 25.8 25.7-2 10 24 26 28 30 Voltage [V] 25.6 25.5 20 30 40 Temperature [ ] (a) (b) Figure 5. a) I-V curves for different temperatures acquired with Advansid NUV SiPM. b) Linear fit to the experimental data. an array of 3 SiPMs connected in series. SiPMs are coupled to the pixel with optical grease. A 35 ns coaxial cable ( 7.5 m) is used to transport signals to amplifier, simulating the final experimental conditions. Counters are excited by using a 90 Sr β-source, which provides electrons with 2.2 MeV endpoint energy. An external reference counter (RC) made by a small piece of scintillator (BC422, size: 5 5 5mm 3 ) coupled to a Hamamatsu S10362-33-050C SiPM is used for triggering 5
Table 2. Summary of BD versus temperature coefficients. Manufacturer BD vs T [mv/ C] Hamamatsu 49 Advansid 24 Ketek 16 purposes. The timing is extracted by means of a software constant fraction discrimination method applied on the recorded waveform. Discriminating fraction lies in the range 5 10% depending on SiPM model. The time resolution of the system is evaluated as the width of the T distribution, where T = T re f (T 0 + T 1 )/2, being T re f and T i the time measured by the reference counter and each SiPM array respectively. The summary of the results is showed in figure 4(d) as a function of the applied overvoltage. The best results are obtained with Hamamatsu model S12572, which can reach a resolution of σ( T ) 43 ps. More generally, Hamamatsu SiPMs show resolution in the range 43 50 ps, while other models show 15 ps worst resolution. 3.2 Scintillator comparison Three types of ultra fast plastic scintillator from Saint-Gobain Crystals, BC418, BC420 and BC422, were tested. The main characteristics of each scintillator are summarized in table 3, where also the characteristic of the BC404 are listed. The test was performed using 60 30 5mm 3 pixels. The best resolution is obtained with BC422, the one with the fastest risetime. The optimization of scintillator wrapping is discussed in a previous work [10] and a mirror-type reflector (3M radiant mirror film) is found to give the best resolution. However, in these measurements no wrapping is adopted for a comparative study. Table 3. Summary of the properties of plastic scintillators tested. Obtained time resolution on a 60 30 5mm 3 sample is also listed. Properties BC404 BC418 BC420 BC422 Light Yield (% Anthracene) 68 67 64 55 Rise Time (ns) 0.7 0.5 0.5 0.35 Decay time (ns) 1.8 1.4 1.5 1.6 Wavelength peak (nm) 408 391 391 370 Attenuation length (cm) 140 100 110 8 Obtained resolution (ps) - 48 51 43 3.3 Geometry optimization Pixels with different height and length were also tested. We used Bicron BC422 for these tests. The results are showed in figure 6. As expected, the time resolution worsens increasing the pixel sizes. On the other hand, a Geant4 based Monte Carlo simulation shows that efficiency increases by increasing pixel sizes. Thus the goal is to find the best compromise between efficiency and time resolution. The possibility to use different pixel sizes in different regions of the detector in order to increase the efficiency is under study. 6
Figure 6. Timing resolution as a function of pixel length for different pixel height. 4. Beam test In order to test the detector in experimental conditions similar to the final one and check the multiple hit scheme, a small prototype was built and tested at the Beam Test Facility (BTF) at the INFN Laboratori Nazionali di Frascati [11]. The BTF beam can be tuned in such a way to provide electrons with energy similar to the MEG signal (48 MeV in our test) with electrons average bunch multiplicity lower than 1. We decide to test both Hamamatsu and Advansid counters, the ones with the best tradeoff between time resolution and temperature dependence. Figure 7. Picture of the prototype setup tested at the BTF. The reference counter is illuminated by the spot of the laser tracker. 4.1 Setup We prepared two sets of pixel prototypes with BC418 scintillator, with 90 40 5 mm 3 sizes, equipped with Hamamatsu S12572-050C(X) (8 counters) and Advansid NUV (6 counters) SiPMs. Pixels are wrapped with 3M Radiant Mirror Film. Pixels are mounted on a moving stage with a stepping-motor remotely controlled, which allows to move the detector in the x y plane (z points along the beam axis). The whole system is mounted on 7
an optical bench which is enclosed in a shielded black box. The same reference counter described in section 3.1 is placed along the beam trajectory in front of the pixels. A lead glass calorimeter is placed behind the pixels and is used as beam monitor during data acquisition. The whole system is aligned to the beam line by using a laser tracker. Signals from SiPMs are fed into six DRS4 evaluation boards, synchronized by using a 25 MHz sine wave which is distributed to each board on a dedicated channel. Acquisition frequency is 2.5 Gs/s. A picture of the setup is showed in figure 7. 4.2 Data analysis Charge analysis Events are selected by cutting on the charge distribution of the first couple of pixels. An example of distribution is showed in figure 8, where the bunch multiplicity is clearly visible. This can be seen in figure 9, where the distribution of the first counter charge before (blue line) and after (red hatched area) the cut is showed. Moreover, we applied also a cut on the 3000 H Entries 92427 Mean 1.098 RMS 0.517 2500 2000 1500 1000 500 0 0 0.5 1 1.5 2 2.5 3 Charge counter 1 [a.u.] Figure 8. Charge distribution of the first couple of pixels. The selected events (single electron bunch) are marked in red. Figure 9. Charge distribution of the first pixel. Red dotted area shows same distribution after having applied charge cuts. reference counter charge spectrum, by selecting the events which lies around the Landau peak of the charge distribution. The timing resolution is then evaluated in two different ways: by taking the width of the T distribution, having defined T as: T = T ref 1 N T = 1 N N i=1 N j=1t j 1 N T i, (4.1) N i=1 T i, (4.2) where T ref and T i, j is the time measured by the reference counter and by the pixels respectively. In formula 4.2 the sum is made over two different subgroups of pixels. In both cases, we can evaluate the timing resolution as a function of the number of hits used in the time averaging. DRS calibrations As described in section 4.1, we used six DRS boards to register all signals coming from RC and pixels. Each DRS evaluation board has four channels available; as mentioned in 4.1, the 4th channel of each board is dedicated to acquire a common sine wave which is used as synchronization. Dedicated runs were taken in order to evaluate the contribution coming from 8
electronic jitter. This was found to be 18.7 ps and 16.2 ps for pixels whose arrays are read-out by the same or different boards, respectively. 4.3 Results We checked the multiple hit scheme, with the same method described in section 3.1. The measured resolution for both SiPM types is shown in figure 10, where the contribution from the electronic, described in section 4.2 is taken into account. The RC resolution, which was checked with dedicated runs and found to be 27ps is also subtracted. We didn t see any clear dependence of the time resolution from pixel number, which is an indication that the multiple scattering contribution os not significant due to the small scintillator thickness crossed by positrons. Finally we checked Figure 10. Summary of the resolution obtained with single pixel, both for Hamamatsu and Advansid sample. Figure 11. Time difference distribution between 8 pixels (Hamamatsu). The gaussian fit gives a resolution of 27 ps. the time resolution as a function of the number of hits. We studied the time distribution defined in equations 4.1 and 4.2 for different groups of SiPMs. As expected, the best result is obtained with the maximum number of hits. The corresponding plot of T distribution is showed in figure 11. The gaussian fit gives 27 ps as σ of the distribution. The summary of the results are showed in Figure 12. Summary of the obtained resolution as a function of the number of hit. The expected resolution and the estimated 1/N hit behavior are also showed. 12, where the expected behavior 1/N hit is also showed. The resolution scales as expected nicely following the 1/N hit behavior. 9
5. Conclusions We presented the R&D work on the upgrade of the Timing Counter for the MEG II experiment. The basic ideas of the new design, namely the good time resolution achievable with small scintillator counters read-out by SiPMs and the improvement of the overall time resolution by averaging the time measurement over multiple hits were fully exploited and tested. Optimizing the choice among different type of SiPM and scintillators lead us to obtain extremely good time resolution with a single counter. The best results obtained was σ( T ) 43 ps with a 60 30 5 mm 3 BC422 with double side readout based on Hamamatsu SiPMs scintillator pixel. A beam test performed at the Beam Test Facility in Frascati proved experimentally the multiple hit scheme. A resolution σ( T ) < 30 ps with eight counters was measured. Acknowledgments The authors would like to thank the Beam Test Facility crew, the mechanical and electronics workshops at INFN Section of Genova and Paul Scherrer Institute for their valuable help. References [1] G. Blankenburg et al., Neutrino masses and LFV from minimal breaking of U(3) 5 and U(2) 5 flavor symmetries, [arxiv:1204.0688v2] [hep-ph]. [2] J. Adam et al., [MEG Collaboration], New Constraint on the Existence of the µ + e + γ Decay, Phys.Rev.Lett. 110 (2013) 201801. [3] A.M. Baldini et al., [MEG Collaboration], MEG Upgrade Proposal, [arxiv:1301.7225] [physics.ins-det]. [4] J. Adam et al., [MEG Collaboration], The MEG detector for µ + e + γ decay search, Eur. Phys. J. C 73 (2013) 2365. [5] M. De Gerone et al., Development and Commissioning of the Timing Counter for the MEG Experiment, IEEE Trans. Nucl. Sci. 59 (2012) 379. [6] M. De Gerone et al., The MEG timing counter calibration and performance, Nucl. Inst. Meth. A 638 (2011) 41. [7] A. Stoykov et al., A time resolution study with a plastic scintillator read out by a Geiger-mode Avalanche Photodiode, Nucl. Inst. Meth. A, 695 (2012) 202. [8] http://www.psi.ch/drs/evaluation-board, Nucl. Inst. Meth. A, 623 (2010) [9] S. Ritt et al., Application of the DRS Chip for Fast Waveform Digitizing, Nucl. Inst. Meth. A, 623 (2010) 486. [10] W. Ootani, Development of Pixelated Scintillation Detector for Highly Precise Time Measurement in MEG Upgrade, Nucl. Inst. Meth. A, 732 (2013) 146. [11] G. Mazzitelli et al., Commissioning of the DAΦNE beam test facility, Nucl. Inst. Meth. A, 515 (2003) 524. [12] Y. Uchiyama [MEG collaboration], Nuclear Science Symposium Conference Record, IEEE, Seoul, Korea, 2013., in press 10