Photodetector

The photodetector is used to detect scintillation photons and output electric signals accordingly. We imposed some conditions for the selection. First, the cross-section of the photodetector must not exceed that of the crystal. Second, it must tolerate or be nonsensitive against magnetic field around 1 T to ensure that that the gain remains stable. Third, it should work in vacuum. The first two criteria inevitably refused PMT (photomultiplier tube) because of its typically large size and sensitivity to the magnetic field. Semiconductor sensors were found to be appropriate candidates; nowadays, there are several variations and products to choose from. Semiconductor photosensors employ the following steps for detecting photons:

1. Photons that enter the depletion layer⁴generate electron–hole pairs. The ratio of the number of electrons to that of incident photons is referred to as quantum efficiency.

4Outside the depletion layer, the separation of the electron–hole pairs does not work.

4.2. Design

60 80 100 120 140

(MeV/c) Pbeam

4 4.5 5 5.5 6 6.5

Energy Resolution (%) _GSO

LYSO

Figure 4.3: Energy resolutions of an old ECAL prototype using GSO and LYSO as a function of the electron momentum [78]. The error bars contain both static and systematic uncertainties.

2. The HV applied to the sensor results in a strong electric field whose high gradient protects them from the recombination, accelerates the electrons, and moves them to the electrode.

3. The drifted electrons further ionize atoms and generate new pairs. Repeating this process leads to the amplification of the incident photons. The number of electrons arising from a single photon indicates the gain.

4. The amplification forms an electric pulse on the electrode that is read by the readout electronics.

There are three categories in the amplification process for a given electric field strength: ionization, proportional, and Geiger modes. Each of these categories is associated with different products.

The PD (photodiode) works in the ionization mode, wherein the gain is almost unity; the outputs are too weak to use for the ECAL. In the Geiger mode, the amplification process saturates immedi-ately, and the number of output electrons is no longer proportional to the initial electrons. This feature is not appropriate for energy measurement, whereas the SiPM overcomes it by combining many pixels of very tiny Geiger-mode silicon sensors⁵. Although many experiments have adopted SiPM for mod-ern detectors, it is too weak for neutrons, which causes lattice defects and new excitation bands that considerably increase dark-current noise. Consequently, the ECAL adopted the APD that operates in the proportional mode.

To this end, there were two candidate products of Hamamatsu Photonics: 55 and S8664-1010. Both have a rectangular shape but they differ in terms of their sensitive area, i.e., 5×5 and 10×10mm², respectively. Their nominal bias is around 400 V, and the gain is about 50, which requires amplifier electronics. Finally, we select S8664-1010 as follows.

Table 4.2 lists the experiments conducted to examine their radiation hardness. The first assessment used the tandem electrostatic accelerator at Kyushu University in Japan, which accelerates deuterons

5The magnitude of an output charge from SiPM is proportional to the number of pixels that the photons hit unless multiple photons enter the same pixel simultaneously. It can maintain the countability of the incoming photons.

41

4.2. Design

Table 4.2: Radiation-tolerance tests for the APDs. (*) S8664-1010 was exposed by this flux; however, its tolerance was evaluated for up to 2.4×10¹¹n_1MeV/cm²for a technical reason as explained in the text.

Date Facility Examined item Irradiated flux

Oct. 2014 Tandem facility at Kyushu Univ. S8664-55 2.2×10¹²n1MeV/cm² Jul. 2015 Tandem facility at Kobe Univ. S8664-55 2.5×10¹²n_1MeV/cm²

S8664-1010 2.5×10¹²n1MeV/cm²(*) Oct. 2015 ⁶⁰Co facility at Kyushu Univ. S8664-55 12 kGy

to 9 MeV and generates neutron beams with a ¹²C(d,n)¹³N reaction in a carbon fixed target. We prepared three samples of S8664-55 and exposed them to neutrons up to 2.2×10¹² n_1MeV/cm². To evaluate the performance, we created a test bench wherein an LED provided a constant light input to the tested APD, and a PMT monitored its magnitude and stability. The APD signals were amplified by amplifier electronics and measured by a waveform-sampling digitizer. We calculated their maximum wave height G_h and integrated chargeG_c by fitting them with the averaged waveform template; we evaluated their fluctuations σh and σc. Figure 4.4 shows changes in the S/N ratio G_h/σh and the resolutionσc/G_c as a function of the input light strength. The waveform became 1.3 times noisier, and the S/N ratio deteriorated. However, it is not fatal because the waveform fitting compensates it and preserve the resolution.

0 10000 20000 30000 40000 50000 Input Light Strength (a.u.) 0

50 100 150 200

S/N 250

Before Irradiation After Irradiation

(a)S/N ratio

0 10000 20000 30000 40000 50000 Input Light Strength (a.u.) 0

1 2 3 4

Resolution (%)

Before Irradiation After Irradiation

(b)Resolution

Figure 4.4: Changes in (a) the S/N ratio and (b) resolution of S8664-55 after neutron irradiation of 2.2× 10¹²n1MeV/cm². The horizontal axis represents the strength of the input LED light.

The neutron tolerance of S8664-1010 was evaluated with the tandem accelerator facility at Kobe University in 2015, where neutrons were produced with the ⁹Be(d,n)¹⁰ reaction from its 3 MeV deuteron beam. S8664-1010 showed a good tolerance against neutrons up to 2.4×10¹¹ n_1MeV/cm². Although the APD was exposed to 2.5×10¹²n_1MeV/cm²at most, the dark current increased consider-ably and the prepared electronics could not work safely to evaluate the performance. However, since S8664-1010 also hardly changed its resolution at 2.4×10¹¹ n_1MeV/cm² as well as S8664-55, it was assumed to possess the required tolerance.

Gamma-ray tolerance was studied using the⁶⁰Co facility at Kyushu University in 2015. S8664-55

4.2. Design was exposed to gamma rays of 12 kGy in total with energies of 1173 and 1332 keV from the⁶⁰Co beta decay. The resolution did not change considerably, and the dark-current was only 340 nA. We found that the neutron damage was more significant for the APD in COMET.

Finally, we performed the same evaluation procedure mentioned above to compare the nominal performance of S8664-55 and S8664-1010. Figure 4.5 shows the S/N ratio and resolution of both products. As expected, S8664-1010 has a distinct difference from S8664-55. The noise rises only by 30%, and the S/N ratio improves nearly four times, which is better than the cost that is 2.2 times as expensive as that of S8664-55.

0 10000 20000 30000 40000 50000 Input Light Strength (a.u.) 0

20 40 60 80 100 120 140

S/N 160

× 50 5 mm: Gain

× 200 5 mm: Gain

× 50 10 mm: Gain

× 200 10 mm: Gain

(a)S/N ratio

0 5000 10000 15000 20000

Input Light Strength (a.u.) 0

1 2 3 4 5

Resolution (%)

× 50 5 mm: Gain

× 200 5 mm: Gain

× 50 10 mm: Gain

× 200 10 mm: Gain

(b)Resolution

Figure 4.5:Comparison of the (a) S/N ratio and (b) resolution between S8664-55 (5×5 mm²) and S8664-1010 (10×10 mm²) at gains of 50 and 200. The horizontal axis represents the strength of the input LED light.