Study of a scintillation counter consisting of a pure CsI crystal and APD Yifan JIN, Denis Epifanov The University of Tokyo Oct 20th, 2015 1
Outline Belle II calorimeter upgrade Electronics noise in the scheme with APD CsI(pure)+(1-4)APDs light output and equivalent noise energy Improvement of the light output Novel wavelength shifters with nanostructured organosilicon luminophores Summary and Plans 2
Photopentode VS APD There are two options for photo sensor, photopentode and APD. Photopentode Uoperating(V) > 800 APD S8664-55 400 APD S8664-1010 400 Gain 20 50 50 Sensitive area 20 0.25 1 (cm2) QE(%) 15 30 ± 10 30 ± 10 Capacitance ~ 10 85 270 (pf) (1/G)*(dG/dT) << 1% 3% 3% @G=50 The low capacitance and large sensitive area provide photopentode low noise. However, at bias voltage and QE, APD's performance is better. In the case of photopentode, the stochastic noise is the dominant term. However, for APD, both of them, stochastic noise and electronic noise components are important. 3
End cap ECL upgrade Hamamatsu APD S8664-1010 APD Hamamatsu S8664-1010 Ubias = 394 V: Gain = 50, Idark = 8 na @ T=25 oc (1/G)(dG/dT)[%/oC] In the CsI(pure) + Si APD option, APDs from Hamamatsu Photonics are investigated. The advantages of APD are: 1, compactness; 2, insensitivity to magnetic field; 3, bias voltage of about 400 V, and low dark current of the order of na. The main problem with APD is to reach desirable level of electronic noise. With the actual size crystal and 1 APD (1 x 1 cm2) Hamamatsu S8664-1010, we obtained ENE 2 MeV, while the required ENE 0.5 MeV. 4
Scheme of the counter Pure CsI crystal with 6x6x30 cm3 is used. Hamamatsu S8664 series have been used as photo sensors. The electronic noise is measured by RMS of the peak in the amplitude spectrum of the signal from calibration pulses from 5 the wave generator.
Study of electronic noise ENC2 = A2 τ + (B2/τ + E2)C2 + D2 Shot noise Thermal noise 1/f noise Shot noise coefficient, A = 2 Idark K1 g F/e Thermal noise coefficient, B = 4 k B T R S K 2 e 1/f noise coefficient, E= K3 Af Af is a noise coefficient of order 10-10 -10-12 V2 Additional noise e positron charge; Id dark current; F excess noise factor; C APD junction capacitance; τ shaping time; g APD gain; K1 K2 K3 shaper factor; τbest = BC/A ENC2best = 2ABC + E2C2 + D2 RS equivalent serial resistance of preamp D additional noise. 6
Measurement of total ENC and addition noise (D) Cfb U bias Rfb Rb τ=20 500 ns Rs Shaper Ccal ADC PC test channel 10ms 1850 50 us u0 After preamp After shaper 1050 650 At the shaping times from 20 ns to 500 ns, D is not constant. It varies strongly, which is explained by the relatively large additional parallel (ina) and serial (ena) noises. Fast shaper of better quality (like ORTEC 474, 579) might be helpful to decrease D 7
Measurement of thermal noise B, 1/f noise E Two well known capacitors C1 and C2 were used to measure B and E. B2/τ + E2 = (Q12 Q22)/(C12 - C22) Cfb B = (26.2 ± 0.8 ± 4.8) ns/pf U bias Rfb Rb E = (6.1 ± 0.1 ± 0.4) 1/pF τ=20 500 ns Rs Shaper Ccal test channel ADC PC u0 B2/τ = (4kBK2TRS)/τe2 2 BF862 FETs RS, equivalent serial resistance, it is dominated by reversal transconductance of the CAEN preamp's FET (BF862) RS 50 Ω was also measured with additional serial resistance at the CAEN preamp input. Therefore, we tried FET 2SK932-23, one of the best FET at short shaping times, with intention to get lower RS. However, no obvious improvement is obtained. There is no potential to reduce B and E! 8
Shot noise, excess noise factor F Cfb Ubias Q2no Iphoto=2 e Id τ g F K+(B2/τ+E) Cd2+D2 Q2with Iphoto=2 e (Id+Iphoto) τ g F K+(B2/τ+E) Cd2+D2 F = (Q 2 -Q with Iphoto 2 no Iphoto )/(2 e Iphoto τ g K) Rfb Rb Rs Shaper Iphoto APD Ccal test channel S8664-55: g = 50, F = 5.5 ± 0.5 ADC PC u0 S8664-1010: g = 50, F = 4.7 ± 0.5 Excess noise refers to the additional noise due to avalanche fluctuation, and can be expressed as: F =h*g+(2-1/g)(1-h) where h is the ratio of the hole impact ionization rate to that of electrons. 9
ENC vs. shaping time ENC2 - D2 = A2 τ + (B2/τ + E2)C2 A = (2 Idark K1 g F/e) B = (4 kb T RS K2) /e E= (K3 Af C2) Discrepancy between calculated and measured ENC2-D2 is due to the uncertainty in CAPD The agreement between the measured noise and noise calculated by the formula indicates good suppression of the correlated noises 10
Light output (LO) and ENE Cosmic muons are used to calibrate ADC channels in units of energy (MeV) CsI 6x6x30 cm3 MC Acosmic EXP ConversionADC(MeV/ch) = EMC /A peak peak Epeak(cosmic) 33 MeV ENE = σcal ConversionADC Cfb Ubias The light output is measured by comparison of the signal from cosmic muons (A ) with calibration Rfb Rb cosmic signal (A ). Acal Rs cal Iphoto Ncosm(ph.e.) = (Ccal U0 / e) (Acosm / Acal) 2 Sensitivity = Ncosm/Epeak/(APD gain = 50)/(SAPD [cm ]) MC Sensitivity = (26 ± 2) ph.e. / MeV / cm2 Shaper APD Ccal ADC PC test channel u0 11
ENE, several APDs per crystal S8664-55 S8664-1010 Total noise=apd noise + D 1 APD S8664-1010 has essentially larger dark current (26 na) in comparison with the average one (8 na), we introduce correction to ENE. ENE Measured Estimated 2 S8664-1010 (same Idark) (1.10 ± 0.11) MeV (0.97 ± 0.09) MeV 2 S8664-55 (1.71 ± 0.17) MeV (2.05 ± 0.18) MeV 4 S8664-55 (1.20 ± 0.12) MeV (1.36 ± 0.12) MeV The ENEs of the counter are far away from our goal. Further studies are needed. 12
To increase the light output The light collection coefficient strongly depends on the quality of APD coupling to crystal and reflectivity of the wrapping material. 1, Three types of optical grease were tested, OKEN-6262A, BC-630 and TSF451-50M. However, we didn't find anything better than OKEN-6262A. 2, White porous Gore-Tex teflon was confirmed as the best reflector at UV range [1]. The thickness of the white teflon was studied. It is shown that 200 um is sufficient, further increase on thickness provides no more than 5% improvement on signal. 3, Novel wavelength shifting (WLS) plates containing nanostructured organosilicon luminophores [2] provides essential increase of light output. [1] M.Janecek, IEEE Trans. Nucl. Sci. 59.3 (2012) 490. [2] S.A. Ponomarenko, et al., Scientific Reports 4 (2014) 6549. 13
Wavelength shifters with organosilicon luminophores Based on the nanostructured organosilicon luminophores (NOL-9,10,14) from LumInnoTech Co., the WLS plates were developed ((60 x 60 x 2) mm3). 302 nm 502 nm PLQY=90% 327 nm 588 nm PLQY=95% 337 nm 655 nm PLQY=78% The absorption and emission spectra of these NOL's match our needs very well (λcsi = 320 nm). The improvement of the APD QE is by a factor of 2 3. 14
Results with WLS plates With optical grease CsI(pure) WLS APD NOL-9 turns out to be the best WLS that provides an enhancement on signal by a factor of about 3. 15
Problems in previous measurement I 2 S8664-1010 APD's Previously, the gaps between using optical grease and not using optical grease for S8664-1010 and S8664-55 are dramatically different. 4 S8664-55 APD's Meanwhile, the signal from 4 small APDs is larger than that of 2 large APDs. 16
Signal degradation Every time we remove the optical grease on WLS, some luminophores dissolved in the optical grease is removed at the same time. Therefore, the signal degradation is caused. 17
Problems in previous measurement II Previously, among the four small APD's we used, one small APD, its operating point is much lower, and it is working at a much higher gain. So the signal is larger compared with that of 2 S8664-1010. We purchased four new S8664-55 APD's, and measured their operating bias voltage for G=50. S8664-55 A S8664-55 B S8664-55 C S8664-55 D 406 V 406 V 405 V 405 V Be aware of grease/erasing, equality of operating voltage for APD's. 18
4 new small APD's + new WLS (3)/(4) Error originates from the relative temperature gain variation ((1/G)*(dG/dT)), accuracy of simulation of cosmic peak position and statistical accuracy of the data. Between the crystal and APD's, we applied WLS. The APD's are coupled to the backside of WLS by OKEN optical grease. However, no optical grease is used between crystal and WLS. A factor of about 3 is earned on signal intensity by the use of WLS. No obvious difference between WLS (3) and (4) has been observed. 19
Attaching APD's on edge side In this measurement, the APD's are attached on one side of the edge due to the limited space of our shield box. A factor of 1.3 is earned by this configuration. In total, in comparison with the coupling of APD's to the crystal, we earned a factor of 4. 20
Shaper-FADC board All the previous measurements are implemented by the Clear Pulse Shaper with shaping time range 20-500 ns. Now we turn to use the shaper-fadc board with shaping time 30 ns. This shaper-fadc board was used in R&D with the matrix of 20 CsI (pure) + photopentode at BINP. Shaper FADC (BINP board): 1 differentiation + 4 integrations with τ = 30 ns FADC: 256 words buffer, ν = 40 MHz Fit functions: e t t 0 f calib t =A 0 e P0 4 5 t t 5 t t e 0 f cosmic t = A 0 e P0 5 Parameters of the fit: A0, t0, P0 4 4 t t 0 0 21
The result with Shaper-FADC The rising time (Trising 160 ns) of the signal from Clear Pulse ( CR-3RC) with τ = 50 ns is approximately the same as that of Shaper-FADC (CR-4RC) with τ 30 ns. 4 S8664-55 (edge side) 1 S8664-1010 (back side) 2 S8664-1010 (back side) 2 S8664-1010 (back side) (360 ± 30) kev (540 ± 50) kev (440 ± 40) kev expected to be (380 ± 40) kev One Large size APD S8664-1010 is old. It has triple dark current, which is the reason why the ENE with 2 S8664-1010 is not suppressed by a factor of 2. 22
Statistical noise Due to the fluctuation during avalaunche development (F), there is a statistical noise. Therefore: F, excess noise factor, F=4.7 for operating point (G=50) Nph.e, number of photoelectrons initially produced in APD (before amplification) δstat = FNph.e 2 δ total =δ 2 stat +δ 2 elec F=4.7 @g=50; F=5.6 @g=100 F + δ2elec Nph.e A2 To optimize the total noise, maybe we should lower the gain in order to decrease F and hence the statistical noise. δtotal /A = 23
Summary Hamamatsu APDs of S8664 series provide a promising option for Belle II end cap ECL upgrade. Several APDs per crystal allow us to decrease further ENE and provide readout redundancy. Essential increase of the light output of the CsI(pure)+APD(s) counter was achieved with WLS plates based on the nanostructured organosilicon luminophores (NOL-9). The ENE of the counter with 2 S8664-1010 APDs (back side) was measured to be ENE = (0.44 ± 0.04) MeV, expected to be (0.38 ± 0.04) MeV. The ENE of the counter with 4 S8664-55 APDs (edge side) was measured to be ENE = (0.36 ± 0.03) MeV. 24
Future plan Several configurations and geometries will be tested. Different optical greases between WLS plate and APD. Find optimal optical glue. Beveled WLS. 25
Thank you for your time! 26
Back up 27
Improvement of the LO Refraction index Transparency @315 nm Light collection efficiency from the producer OKEN-6262A 1.453 (@ 590 nm) 85% TSF451-50M 1.404 (@ 590 nm) 98% 1.465 95% BC-630 1.00 0.85 0.95 Three types of optical grease were tested ( = 100 μm), OKEN-6262A provide the largest light output Effect of the thickness of white porous Gore-Tex teflon was studied, thickness of 200 μm was found to be optimal. signal amplitude 600 M.Janecek, IEEE Trans. Nucl. Sci. 59.3 (2012) 490. 500 451.4 476.9 491.2 498.3 122 um 185 um 2*122 um 500 um 400 300 200 100 0 thickness of teflon 28
Characteristics of APD APD is a dominant source of the signal temperature variations, which have to be compensated (1/G)(dG/dT)[%/oC] To compensate temperature variations of APD gain, we can organize temperature sensor - bias voltage feedback. 29