A multipixel silicon APD with ultralow dark count rate at liquid nitrogen temperature

Transcription

1 A multipixel silicon APD with ultralow dark count rate at liquid nitrogen temperature M. Akiba 1, K. Tsujino 1, K. Sato 2, and M. Sasaki 1 1 National Institute of Information and Communications Technology, 4-2-1, Nukuikitamachi, Koganei-city,Tokyo , Japan 2 Hamamatsu Photonics K.K., , Higashi-ku, Hamamatsu-city, Shizuoka , Japan Corresponding author: akiba@nict.go.jp Abstract: The dark count rate of a multipixel photon counter (MPPC) was reduced to less than 0.2 cps by cooling the counter to 78.5 K. Characteristics of the MPPC other than the dark count rate were also determined at 78.5 K and 295 K. The photodetection efficiency and the timing jitter at 78.5 K were comparable to those at 295 K. The MPPC had a count rate of up to 50 MHz at 78.5 K and above 100 MHz at 295 K. Detailed measurements were performed for determining the afterpulse probability. References and Links 1. R. Charvin, Photo-multiplicateur eclaire par des impulsions subnanosecondes, Opt. Acta, 28, (1981). 2. R. S. Bondurant, P. Kumar, J. H. Shapiro, and M. M. Salour, Photon-counting statistics of pulsed light sources, Opt. Lett. 7, (1982). 3. B. E. Kardynal, Z. L. Yuan and A. J. Shields, An avalanche-photodiode-based photon-number-resolving detector, Nature photonics 2, (2008). 4. K. Banaszek and I. A. Walmsley, Photon counting with a loop detector, Opt. Lett. 28, (2003). 5. D. Achilles, C. Silberhorn, C. Sliwa, K. Banaszek, and I.A. Walmsley, Fiber-assisted detection with photon number resolution, Opt. Lett. 28, (2003). 6. M. J. Fitch, B. C. Jacobs, T. B. Pittman, and J. D. Franson, Photon-number resolution using timemultiplexed single-photon detectors, Phys. Rev. A 68, (2003) 7. M. Moszyliski, T. Ludziejewski, D. Wolski, W. Klamra, M. Szawlowski, M. Kapusta, Subnanosecond Timing with Large Area Avalanche Photodiodes and LSO Scintillator, IEEE Trans. Nucl. Sci. 43, (1996). 8. V. N. Solovov, F. Neves, V. Chepel, M. I. Lopes, R. F. Marques, and A. J. P. L. Policarpo, Low temperature performance of a large area avalanche photodiode, J. Mod. Opt. 51, (2004). 9. N. G. Woodard, E. G. Hufstedler, and G. P. Lafyatis, Photon counting using a large area avalanche photodiode cooled to 100 K, Appl. Phys. Lett. 64, (1994). 10. J. J. Fox, N. Woodard, and G. P. Lafyatis, Characterization of cooled large-area silicon avalanche photodiodes, Rev. Sci. Inst. 70, (1999). 11. P. Buzhan, B. Dolgoshein, L. Filatov, A. Ilyin, V. Kantserov, V. Kaplin, A. Karakash, F. Kayumov, S. Klemin, E. Popova, and S. Smirnov, Silicon photomultiplier and its possible application, Nucl. Instr. and Meth. A 504, (2003). 12. S. Castelletto, I. P. Degiovanni, V. Schettini, and A. Migdall, Reduced deadtime and higher rate photoncounting detection using multiplexed detector array, J. Mod. Opt. 54, (2007). 13. P. Eraerds, M. Legre, A. Rochas, H. Zbinden1, and N. Gisin, SiPM for fast photon-counting and multiphoton detection, Opt. Express 15, (2007). 14. D. L. Robinson and B. D. Metscher, Photon detection with cooled avalanche photodiodes, Appl. Phys. Lett. 51, (1987). 15. J. G. Rarity, T. E. Wall, K. D. Ridley, P. C. M. Owens, and P. R. Tapster, Single-photon counting for the nm range by use of Peltier-cooled and passively quenched InGaAs avalanche photodiodes, Appl. Opt. 39, (2000).

2 1. Introduction Single-photon detection technology is used in various fields such as quantum information, medicine, and high-energy physics. Photomultiplier tubes (PMTs) and Geiger-mode silicon avalanche photodiodes (Si APDs) are widely used as single-photon counters in the visible and near-infrared wavelengths because they are commercially available and convenient to operate. However, further improvements are required for some applications, and much effort has been spent to improve their performances. Recent advances in quantum information and communications technologies require high photodetection efficiencies, high count rates, and photon-number-resolving abilities. The detection of low-intensity light from scintillation (in high-energy physics and positron emission tomography (PET), for instance) requires large detection areas [1,2]. The low photodetection efficiencies of PMTs (<20%) are a serious limitation in many low-light-level experiments. Hallensleben et al. demonstrated that the photodetection efficiencies of PMTs can be improved by the use of multiple total internal reflection, but at the cost of the detection area and field of view [3]. In contrast, Geiger-mode Si APDs have low count rates (<20 MHz), no photon-number-resolving ability, and small detection areas. The low count rate is responsible for the long detector-deadtime. The deadtime can be reduced by 1/N by employing an array of N detectors and a 1-by-N optical switch [4]. The self-differencing mode makes it possible for Si APDs to resolve photon numbers, but the resolution is still poor [5]. Time-multiplexed detectors can correctly measure the number of simultaneously incident photons by using conventional photon counters. The basic idea behind time-multiplexed detectors is that an incident light pulse is divided into several pulses that are separated by a time interval, allowing each pulse to be measured subsequently by the photon counters [6,7,8]. Thus, time-multiplexed detectors reduce the count rates rather than those of photon counters used. Furthermore, although large-area APDs have been used to determine the single-photon count in the linear mode at cryogenic temperatures [9,10], their count rates are only about 1 MHz [9]. A multipixel Si APD is a novel photon counting device that comprises several Geigermode APDs in parallel [11]. This design overcomes the drawbacks of Si APDs [11,12,13]. Each Geiger-mode APD produces a pulse of almost the same level regardless of the number of incident photons, and the pulses generated in different pixels can be superposed. Therefore, when all the photons are injected into different pixels, the output pulse height is proportional to the number of incident photons. The photon distribution over the pixels is effective in reducing the deadtime, because even when some pixels are not active, others continue to be active. Multipixel Si APDs also have the advantage of having a large area because of the parallel combination of Si APDs. The dark count rates of Multipixel Si APDs, however, are much higher than those of single APDs due to the parallel combination of APDs [11,13]. A direct method to reduce the dark count rate is to reduce the temperature [14]. On the other hand, afterpulses generally increase in number as the temperature is reduced [10]. Hence, characteristics other than the dark count must be determined at low temperatures in order to allow the use of multipixel Si APDs in various applications. In this study, we demonstrate that the dark count rate of a multipixel Si APD can be dramatically reduced by cooling to liquid nitrogen temperature and present the characteristics of the detector at this temperature. 2. Experimental setup and time response The multipixel APD used in this experiment is a multipixel photon counter (MPPC S U) supplied by Hamamatsu Photonics. The pixel size and the number of pixels are 100 µm 100 µm and 100, respectively, and the effective area of the APD is 1 mm 2. The experimental setup is shown in Fig. 1. The detector was placed in a liquid nitrogen cryostat and was electrically connected to a room-temperature circuit through a CuNi SMA cable,

3 Fig. 1. Schematic diagram of apparatus used to measure the characteristics of the MPPC. The blue parts in the diagram are cooled to 78.5 K. The light is coupled to the MPPC through a single-mode fiber and a collimator lens. The intensity of the collimated beam is almost uniform within a diameter of 5 mm. For the measurement of the time response of the MPPC, the high-pass filter is not used. The oscilloscope is triggered by the timing signal from the LD controller for the jitter measurements. Fig. 2. Output signals from the MPPC when no high-pass filter is used (a), and output signals from the high-pass filter when two photons happened to be incident successively (b). which was used for thermal isolation. The temperature of the detector was 78.5 K. The light from a UV pulsed laser diode (LD, wavelength: 407 nm, pulse width: 75 ps (FWHM), Hamamatsu Photonics PLP10-040C) or light- emitting diode (LED, wavelength: 450 nm) was passed through neutral density filters and was then focused on a single-mode fiber by using focus lenses. The light signal from the single mode fiber was collimated through a collimator lens. The intensity of the collimated beam was nearly uniform over a diameter of 5 mm and was varied by 2% in order to reduce the effect of the position of the detector on the intensity of the light injected into the detector. The signal from the detector was acquired by using a high-speed oscilloscope (Lecroy SDA3000A). All the following results were obtained on analyzing the signal. We first measured the time response of the detector without the high-pass filter (Fig. 2a). The slope of the exponential tail of the response was found to increase from 25 ns at 295 K to 210 ns at 78.5 K. When successive photons are incident, the output signal of the multipixel APD is a superposition of the signals from the individual APDs. Thus, since levels of tails of the signals from the individual APDs are also superposed, the output signal level varies as a whole with

4 the signal levels of the tails if the time interval between the signals from the individual APDs is smaller than the time scale of the tail. This problem is solved by using a high-pass filter [13]. Fig. 2b illustrates signal outputs from the 300-MHz RC high-pass filter when two photons were incident successively. As can be seen, the two pulses separated by 2 ns can be distinguished at 78.5 K. All the following measurements were carried out with a high-pass filter. 3. Dark count rate, afterpulse probability, and photodetection efficiency During the measurement of the dark count rates, the optical window of the cryostat was closed. Fig. 3 shows the dark count rates that include afterpulses as a function of the bias voltage. The dark count rates increased rapidly when the bias voltages exceeded 60 V for 78.5 K and 70 V for 295 K due to an increase in the afterpulses. At 78.5 K, the dark count rate of the MPPC was found to be extremely low compared with that of the silicon photomultiplier (a few kcps) described in [11]. Fig. 3. Dark count rate as a function of the bias voltage at temperatures of 78.5 K (a) and 295 K (b). Fig. 4. Schematic of the method for the measurement of the afterpulse probability. The light pulses are so weak that hardly more than two signal pulses are generated. The time interval of the light pulses is set to 10 µs, and after this interval, the afterpulses are completely suppressed.

5 A 450-nm LED was used for the determination of photodetection efficiency and afterpulse probability because the intensity of the LED was more stable as compared to that of the LD. The afterpulse probability was determined by using two methods by using a correlation function, which is generally used [15], and by using a very weak light pulse with a low repetition rate. After a signal pulse generated by the light pulse was detected, the number of output pulses was counted (see Fig. 4). A repetition rate of 100 kcps was set in order to ensure that there was sufficient time for the afterpulses to be subdued. The intensity of light pulses (0.01 counts/pulse) was set so that there was negligible probability of more than two signal pulses being generated. This method has the merit of measuring the afterpulse probability precisely, but has a demerit of being unable to measure the afterpulse probability when the number of incident photons is large or when the dark count rate is high. The pulse height distributions for signal pulses generated by single photons and afterpulses at the bias voltage of 59.5 V at 78.5 K are shown in Fig. 4. The number of afterpulses increased with a decrease in pulse height because many afterpulses were generated at the tail of the signal pulse. However, this feature was not seen at 295 K. The difference in the distributions of the signal pulse and the afterpulse indicates that it is important to set an adequate threshold level in order to differentiate signal pulses from afterpulses and noise at 78.5 K, depending on the application. In the following measurements, each threshold level was set at the minimum of the valley (see Fig. 4). Figure 5 shows afterpulse probabilities as a function of the bias voltage. At 295 K, the afterpulse could not be measured by the method using a very weak light pulse with a low repetition rate because of the high dark count rate. There is a clear difference between the afterpulse probabilities obtained from the two methods at 78.5 K. The most notable difference between the two methods is in the count rate of incident photons the count rate when the correlation function was used was 30 Mcps, and the count rate when the very weak light pulse with a low repetition rate was used was 100 kcps. The dependence of the afterpulse probability on the count rate of incident photons was examined (Fig. 6), and it was observed that the afterpulse probability below 3 Mcps was consistent with that determined at 78.5 K by using the weak light pulse. It was observed that the afterpulse probability decreased with an increase in the count rate of the incident photons above 30 Mcps at 295 K. The photodetection efficiencies at 78.5 K and 295 K are presented in Fig. 7. The photodetection efficiencies are the net values not including the afterpulses and are measured at an incident photon count rate of 30 Mcps. When compared to the photodetection efficiency, the afterpulse probability and dark count rate of the MPPC measured at 78.5 K decrease rapidly with a decrease in the bias voltage. Therefore, lowering the bias voltage may reduce the afterpulse probability at the cost of a small reduction in the photodetection efficiency. Fig. 5. Afterpulse probability as a function of the bias voltage. Open circle: measurement using a very weak and low-repetition-rate light pulse. Filled circle: measurement using the correlation function. At 295 K, the measurements using very weak and low repetition rate light pulse can not be carried out due to the high dark count rates.

6 Fig. 6. Afterpulse probability as a function of the input photon number Fig. 7. Photodetection efficiency as a function of the bias voltage at temperatures of 78.5 K (a) and 295 K (b). 4. Timing jitter and fast counting ability The parameters discussed in this section were measured by using the UV pulsed LD. The timing jitter between the timing signal from the laser controller and the laser pulses (electronics timing jitter) was measured to be 75 ps. Fig. 8a plots the timing jitter for single photoelectrons, including the laser pulse width and the electronics timing jitter. The timing jitter of the MPPC was found to be 158 ps at 78.5 K and 181 ps at 295 K after subtracting the laser and electronics contributions. The jitter showed a slight improvement at 78.5 K and in the time response. Fig. 8. (a): Timing jitters for single photoelectrons. (b): Detection rate to high-repetition-rate light pulses. The average number of photoelectrons, which are determined by the data below 10 MHz, is 1.6 pulse per incident light pulse. The slope value of 0.8 represents the photodetection probability per incident light pulse.

7 In order to investigate the fast-counting ability of the detector, the responses of the detector to high repetition rate light pulses were determined. The responses are affected by the number of photons contained in the light pulses because the detection rate decreases with the number of active pixels when the repetition interval is smaller than the time scale of the tail of the time response. For the measurements, the average number of photoelectrons produced in different pixels was adjusted to 1.6 per incident light pulse. From this value, the probability of generation of an output pulse by a light pulse is obtained to be of 0.8. The number of photon was measured at a repetition interval of 0.1 ms, which is sufficiently large compared to the tail of the time response, which is 210 ns at 78.5K. Fig. 8b shows the response of the detector. While the photodetection efficiency at 78.5 K decreases at a high repetition rate of 100 MHz, that at 295 K does not decrease. 5. Photon number resolution and cross-talk probability By combining single APDs in parallel, multipixel APDs acquire photon-number-resolving ability. However, a peculiar problem of cross-talk is encountered. The photon-numberresolving ability and cross-talk probability were determined from the output pulse height distributions obtained when several photons were simultaneously incident on the MPPC. The output pulse height distributions measured at 78.5 and 295 K and theoretical curves of the distribution are shown in Fig. 9. The repetition rate of the light pulse from the first UV pulsed LD was 100 khz. The theoretical curve D(V) as a function of the output voltage V was calculated from the following expressions: 2 1 ( V V ) n D( V ) = p( n) exp 2 n 2πσ 2σ, (1) σ = σ + nσ, c p where V n is the output voltage corresponding to the number of detections n, σ c is the circuit noise, and σ p is the standard deviation of the pulse height distribution when a single photon is detected. The probability that n photons are simultaneously detected, p(n), is expressed as follows: pth( n, m) + ( n 1) pth( n 1, m) pct p( n) =, 1+ np (2) ct p(0) = pth(0, m), where p th (n,m) is a Poissonian distribution with a mean value m and p ct is the cross-talk probability. The expression of p(n) is the same as that in [13] except that p(n) is replaced by p th (n,m) in the second term of the numerator. This means that a pulse generated by cross-talk is assumed not to generate another pulse. The expression for p(n) used in [13] was not able to fit the measured distributions when reasonable values of parameters were used. The values of Fig. 9. Pulse height distributions for short light pulses at 78.5 K (a) and 295 K (b). The parameters used for the determination of the distribution are listed in Table 1.

8 Table 1. Parameter values used for the calculation for the pulse height distributions. Parameters 78.5 K 295 K m V n (mv) σ p (mv) V n /σ p σ n (mv) p ct the parameters used in the fitting calculations are listed in Table 1. The signal-to-noise ratiofor a pulse height V n /σ p is also listed in the table to represent the photon-numberresolving ability. 6. Discussion The drawbacks of the MPPC at 78.5 K are the increase in the time scale of the tail of the time response, the decrease in the detection rate at high repetition rates, and the increase in the afterpulse probability. We discuss here the first two drawbacks. To overcome these drawbacks, it is effective to use a detector with a smaller pixel size. The tail of the signal is derived from the current that compensates for the depleted charge in a pixel. The charge-up time is shorter for a small-sized pixel because the amount of charge that can be stored in it is small. Moreover, since a detector with smaller pixels will have a larger number of them for the same effective area, a small-pixel-size detector is thought to have a high count rate because of its large number of active APDs. We tentatively measured the time response and detection rate for an MPPC with a small pixel size of 25 µm 25 µm (S U). Figure 10 shows the time responses. The time scale of the tail decreases considerably. The detection rate is shown in Fig. 11. The average number of photoelectrons per incident light pulse was 2.8, and the probability that an output pulse is generated by a light pulse was Despite a large number of photoelecrons generated per incident light pulse, there was no reduction in the detection rate at a repetition rate of 100 MHz. The photodetection efficiency of the MPPC was reduced to 13% because of a small filling factor. This reduction in efficiency, however, will be solved by using a microlens array for some applications. Further study on a MPPC with smaller-sized pixels will be published elsewhere. MPPCs with an effective area of 9 mm 2 and a pixel size of 100 µm 100 µm are now available. The MPPCs are expected to have a high count rate of 500 MHz without a reduction in the photodetection efficiency due to the increased number of pixels. Fig. 10. Output signals from the 25 µm 25 µm MPPC when no high-pass filter is used (a), and when the high-pass filter is used (b).

9 Fig. 11. Detection rate of the 25 µm 25 µm MPPC to high-repetitionrate light pulses. The average number of photoelectrons is 2.8 per incident light pulse. The detection rate is consistent up to 100 MHz. Table 2. Comparison of parameter values between the MPPC and the SiPM. Parameters MPPC SiPM (id Quantique) Temperature 78.5 K 295 K 7 C Pixel size (µm ) Fill factor (%) Number of pixels Photodetection efficiency (%) Dark count rate (cps) ,000 30,000 Count rate (MHz) 50 > Cross-talk probability* Afterpulse probability < Timing jitter (ps) The values for the MPPC corresponding to bias voltage of 59.5 V for 78.5 K and 69.7 V for 295 K, except for the timing jitters. The values for the SiPM are obtained from [13]. *Note that the definition of the crosstalk probability in Eq. 2 is different from that in [13]. 7. Conclusion The values of the parameters measured at 78.5 K and 295 K are shown in Table 2. The values for a detector of the same type manufactured by id Quantique are also shown for comparison. By cooling the multipixel Si APD to liquid nitrogen temperature, an ultralow dark count rate (less than 0.2 cps) was obtained. The dark count rate, photodetection efficiency, and afterpulse probability were determined to be decreasing functions of the bias voltage, each having different decreasing scales. Furthermore, we found that the afterpulse probabilities decreased when the incident photon count rate exceeded 3 Mcps at 78.5 K and 30 Mcps at 295 K. The timing jitter and temporal resolution for two successively incident pulses at 78.5 K showed a slight improvement and had values of 158 ps and 2 ns, respectively. On the other hand, the count rate decreased from more than 100 MHz at 295 K to 50 MHz at 78.5 K. The

10 count rate, however, improved to be more than 100 MHz when an MPPC with smaller-sized pixels or a larger number of pixels was used. The theoretically obtained pulse height distributions for simultaneously incident photons were in good agreement with the observed distributions. The parameter values used for calculation indicate that at 78.5 K, the photon number resolution somewhat deteriorates and that cross-talk probability decreases considerably.