- Email: [email protected]
The gain, photon detection efficiency and excess noise factor of multi-pixel Geiger-mode avalanche photodiodes
ARTICLE IN PRESS
Nuclear Instruments and Methods in Physics Research A 567 (2006) 57–61 www.elsevier.com/locate/nima
The gain, photon detection efficiency and excess noise factor of multi-pixel Geiger-mode avalanche photodiodes Y. Musienko,1, S. Reucroft, J. Swain Department of Physics, Northeastern University, Boston, MA 02115, USA Available online 27 June 2006
Abstract In this work we propose a method to characterize multi-pixel Geiger-mode (MPGM) avalanche photodiodes (APDs) and present systematic studies of new devices from three different manufacturers. The basic properties of MGPM APDs such as gain, photon detection efficiency, excess noise factor and noise, as well as their dependence on operating voltage have been measured. Spectral response was measured in the range 350–800 nm. It was shown that despite very good pixel-to-pixel gain uniformity, the excess noise factor of these APDs can be significantly greater than 1. r 2006 Elsevier B.V. All rights reserved. PACS: 29.40Wk Keywords: Silicon avalanche photodiodes; Light detection; Pixellated detectors
1. Introduction Recently developed multi-pixel Geiger-mode avalanche photodiodes (MPGM APDs [1–3]) are very promising candidates for many HEP, astrophysical and medical applications [4]. They have many advantages over conventional photosensors (such as photomultiplier tubes (PMTs), APDs and normal photodiodes) because of their compact size, low power consumption, high gain, reasonably high QE in the green and red regions of the optical spectrum, nonsensitivity to magnetic field, etc. However, the measurement techniques for characterization of these devices have not been well developed. In this work we propose a method to characterize MPGM APDs and present systematic studies of new devices from three different manufacturers. 2. MPGM APDs studied For this paper we have studied the performance of three MPGM APDs which were developed by three different groups. Corresponding author. CERN, CH-1211, Geneva 23, Switzerland.
Tel.: +41 767 16 31; fax: +41 767 89 40. E-mail address: [email protected] (Y. Musienko). 1 On leave from Institute for Nuclear Research, Moscow. 0168-9002/$ - see front matter r 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.nima.2006.05.214
The APDs from MEPhI/PULSAR and Center of Perspective Technologies and Apparatus (CPTA) are both based on the n+-p-p+ structure. They have 1 mm2 sensitive area, which is subdivided into 576 (MEPhI/PULSAR) and 556 (CPTA) separate pixels, respectively. Each of these pixels is quenched by an individual resistor. The structure of these APDs was optimised for green-red light detection. In the case of the CPTA devices, pixels are separated by grooves, which are filled with an optically nontransparent material. This is a new development based on the structure described in Refs. [2] and [5]. The third device (from Dubna) is made on an n-type silicon substrate. It has a 0.5 mm2 sensitive area, which is subdivided into 10,000 separate avalanche regions-micro-wells (for more details see Ref. [6]). The structure of this APD was optimised for the blue-UV part of the light spectrum. 3. Dependence of gain and photon detection efficiency on the bias voltage One can assume that the MPGM APD gain is equal to the charge (Q1) stored in a pixel capacitance when a single pixel is fired [2,4]: Q1 ¼ C pix ðV V B Þ
(1)
ARTICLE IN PRESS Y. Musienko et al. / Nuclear Instruments and Methods in Physics Research A 567 (2006) 57–61
58
where Cpix is the equivalent pixel capacitance, V is the APD operating voltage and VB is its breakdown voltage. However, it was found [7] that light emitted during the pixel breakdown penetrated adjacent pixels due to optical cross-talk and fired these pixels. Thus the average number of pixels fired by a primary photoelectron is typically more than 1. Then the real APD gain (MAPD) is equal to the charge Q1 multiplied by the average number of fired pixels (Npix): M APD ¼ N pix Q1 .
(2)
Fig. 1 shows a single electron spectrum (SES) of the MEPhI/PULSAR APD measured at reduced temperature (with the MPGM APD placed inside a commercial freezer). The ‘‘pedestal’’ events were carefully subtracted from this plot. The green LED (l ¼ 515 nm) pulse amplitude was reduced to the level of 51 photon/pulse using neutral density filters.
From Fig. 1 one can see that there is a significant probability that 2 or more pixels are fired by an absorbed photon. From the measurements of the SES at different voltages one can calculate the MPGM APD gains, corresponding to these voltages. Fig. 2 shows the MEPhI/PULSAR APD gain and Q1 charge as a functions of the bias voltage measured at T ¼ 28 C. One can see that at U ¼ 61 V the average number of fired pixels is more than 2. The gain of the MPGM APD can also be measured in a different way. The charge Q measured by the ADC in response to the LED pulse is related to the MPGM APD gain by: Q ¼ N pe G amp M APD
(3)
where Npe is the number of photoelectrons and Gamp is the gain of the amplifier. If the APD noise is small, the measured low light LED amplitude spectra (see, for example, Fig. 1 from Ref. [8]) can be compared with the Poisson distribution and the mean, Npe, can be calculated using the well-known property of this distribution: N pe ¼ lnðPð0ÞÞ.
(4)
Fig. 1. Single electron spectrum of the MEPhI/PULSAR APD.
Fig. 2. Gain and Q1 of the MEPhI/PULSAR APD as a function of voltage measured at T ¼ 28 1C.
Fig. 3. Gain (top) and PDE (bottom) as a function of voltage for the MEPhI/PULSAR APD measured at two temperatures.
ARTICLE IN PRESS Y. Musienko et al. / Nuclear Instruments and Methods in Physics Research A 567 (2006) 57–61
Fig. 4. Gain (top) and PDE (bottom) as a function of voltage for the CPTA APD measured at two temperatures.
Here, P(0) is the probability to observe the ‘‘pedestal’’ events. It should be mentioned that Npe depends on the APD voltage and should be calculated for each measurement point. Using this method, the gain of the MPGM APDs was measured using blue (470 nm) and green (515 nm) LEDs operated in pulsed mode. The signals from the MPGM APDs were amplified with a fast transimpedance amplifier (gain ¼ 60) and digitised with a LeCroy 2249 W ADC. The APDs were illuminated with small LED pulses via a 0.5 mm collimator. The number of photons in the LED pulse was measured using a calibrated XP2020 photomultiplier. This allowed the calculation of the MPGM APD gain and PDE as a function of voltage using the same measurement points. Figs. 3 and 4 show the MEPhI/ PULSAR and CPTA APDs gain and PDE as a function of voltage measured at two temperatures with a 515 nm LED light. Higher gains and PDE can be reached at reduced temperature. This is due to the fact that in silicon the dark carrier generation rate is significantly reduced when the temperature is lowered. The dark current and power dissipation inside the MPGM APDs are lower than they are at room temperature. As a result, the APDs can operate
59
Fig. 5. Gain, PDE (top figure) and excess noise factor (bottom figure) as a function of voltage for the Dubna APD.
at higher values of ‘‘overvoltage’’ (V–VB), and higher values of the gain and PDE become possible. The gain and PDE as a function of voltage of the Dubna APD are shown in Fig. 5 (top figure). The gain of this APD is lower in comparison to that of the other devices due to smaller pixel capacitance.
4. Multiplication noise of the MPGM APDs As was discussed in Section 2, the optical cross-talk between the pixels causes an increase of the average MPGM APD gain (see Fig. 2), but with the number of pixels which are fired by a primary photoelectron fluctuating from event to event (see the SES in Fig. 1). As a result, the multiplication noise of the MPGM APD is not negligible as one would expect from the very small spread of the charge Q1 for all the MPGM APD pixels (this spread is small due to small variation of the breakdown voltage over the 1 mm2 APD area). This multiplication noise can be numerically expressed in a similar way as is usually done for PMTs and APDs operated below breakdown in terms of an ‘‘excess noise factor’’, F. The excess noise factor (F)
ARTICLE IN PRESS 60
Y. Musienko et al. / Nuclear Instruments and Methods in Physics Research A 567 (2006) 57–61
of the MPGM APD can be calculated from the single electron distribution using the equation: s2 F ¼ 1 þ 2M M APD
(5)
where MAPD is the MPGM APD gain and s2M is its variance. The excess noise factor F as a function of voltage for the Dubna APD is plotted in Fig. 5 (bottom). Fig. 6 (top) and Fig. 6 (bottom) show F as function of gain for MEPhI/ PULSAR and CPTA APDs, respectively. These results show that at high gains the F is significantly different from unity for all three devices. From Fig. 6 one can also see that the dependence of F on the gain is the same even at temperatures differing by 50 C. It is important to notice that comparison between the numbers of photoelectrons calculated from the width of the LED spectrum and calculated from Eq. (2) gives another method for the determination of F (see Ref. [8]). It was found that both methods give very similar results for all the devices studied in this work.
5. Spectral response of the MPGM APDs For the spectral response measurements, an ‘‘Optometrics’’ SDMC1-03 spectrophotometer was used. The spectrophotometer light intensity was significantly reduced using neutral density filters to the level when the maximum current measured with the APD was only 30% higher than its dark current. This was done to avoid non-linearity effects caused by high pixel illumination. The photocurrent measured with the APD was compared with the photocurrent measured with the PIN photodiode. In addition, measurements with the LED pulsed light were used for absolute spectral response calibration (at least 2 different LED measurements were done for each APD). Fig. 7 shows the spectral response of the three MPGM APDs measured at room temperature. The lower PDE of the MEPhI/PULSAR APD in comparison to that of the CPTA APD is due to the smaller ratio of the active pixel area to the total APD area (the so-called geometric factor). In the case of the MEPhI/ PULSAR APD the geometric factor is 25%. Due to an optimized design, the geometric factor of the CPTA APD is 70%. A high value of PDE (45% at 630 nm) was measured for this APD at T ¼ 28 1C [8]. The structure of the Dubna APD was optimized for blue/UV light detection [6]. The geometric factor of this APD is 100%. This device cannot operate at high values of the ‘‘overvoltage’’ ((V–VB)/VBo3%)) due to the rapid increase of the APD noise. Because of this, the probability of the Geiger discharge being initiated by a primary photoelectron is small [9] and the PDE of this APD does not reach the highest possible values. 6. Conclusion In this work we propose a method to characterize MPGM APDs and we present systematic studies of new devices from three different manufacturers. The basic properties of MGPM APDs such as gain, PDE and excess
Fig. 6. Excess noise factor as a function of gain for MEPhI/PULSAR (top) and CPTA (bottom) APDs.
Fig. 7. Photon detection efficiency as a function of the wavelength of light for three MPGM APDs measured at room temperature.
ARTICLE IN PRESS Y. Musienko et al. / Nuclear Instruments and Methods in Physics Research A 567 (2006) 57–61
noise factor, as well as their dependence on operating voltage have been measured. Acknowledgments The authors would like to acknowledge Boris Dolgoshein of MEPhI, Zair Sadygov of JINR, Victor Golovin of CPTA and David McNally of Photonique [10] for providing us with the samples of MPGM APDs, and the National Science Foundation for its continued support.
[2] [3] [4] [5]
[6] [7] [8] [9] [10]
References [1] A.G. Gasanov, V.M. Golovin, Z.Y. Sadygov, N.Y. Yusipov, Lett. J. Technol. Phys. 16 (1990) 14 (in Russian).
61
G. Bondarenko, et al., Nucl. Instr. and Meth. A 442 (2000) 187. Z. Sadygov, et al., Nucl. Instr. and Meth. A 504 (2003) 301. P. Buzhan, et al., Nucl. Instr. and Meth. A 504 (2003) 48. A. Akindinov, et al., Proceedings of the Eighth ICATPP Conference on Advanced Technology and Particle Physics, Villa Erba, Como, 6–10 October 2003, World Scientific, Singapore, 2004. Z. Sadygov, et al., these proceeding. A. Lacaita, et al., IEEE Trans. Electron Devices, ED ED-40 (1993) 577. E. Guschin, et al., these proceedings. W.G. Oldham, R.R. Samuelson, P. Antognetti, IEEE Trans. Electron Devices, ED ED-19 (1972) 1056. SA. Photonique, Geneva (www.photonique.ch).
Nuclear Instruments and Methods in Physics Research A 567 (2006) 57–61 www.elsevier.com/locate/nima
The gain, photon detection efficiency and excess noise factor of multi-pixel Geiger-mode avalanche photodiodes Y. Musienko,1, S. Reucroft, J. Swain Department of Physics, Northeastern University, Boston, MA 02115, USA Available online 27 June 2006
Abstract In this work we propose a method to characterize multi-pixel Geiger-mode (MPGM) avalanche photodiodes (APDs) and present systematic studies of new devices from three different manufacturers. The basic properties of MGPM APDs such as gain, photon detection efficiency, excess noise factor and noise, as well as their dependence on operating voltage have been measured. Spectral response was measured in the range 350–800 nm. It was shown that despite very good pixel-to-pixel gain uniformity, the excess noise factor of these APDs can be significantly greater than 1. r 2006 Elsevier B.V. All rights reserved. PACS: 29.40Wk Keywords: Silicon avalanche photodiodes; Light detection; Pixellated detectors
1. Introduction Recently developed multi-pixel Geiger-mode avalanche photodiodes (MPGM APDs [1–3]) are very promising candidates for many HEP, astrophysical and medical applications [4]. They have many advantages over conventional photosensors (such as photomultiplier tubes (PMTs), APDs and normal photodiodes) because of their compact size, low power consumption, high gain, reasonably high QE in the green and red regions of the optical spectrum, nonsensitivity to magnetic field, etc. However, the measurement techniques for characterization of these devices have not been well developed. In this work we propose a method to characterize MPGM APDs and present systematic studies of new devices from three different manufacturers. 2. MPGM APDs studied For this paper we have studied the performance of three MPGM APDs which were developed by three different groups. Corresponding author. CERN, CH-1211, Geneva 23, Switzerland.
Tel.: +41 767 16 31; fax: +41 767 89 40. E-mail address: [email protected] (Y. Musienko). 1 On leave from Institute for Nuclear Research, Moscow. 0168-9002/$ - see front matter r 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.nima.2006.05.214
The APDs from MEPhI/PULSAR and Center of Perspective Technologies and Apparatus (CPTA) are both based on the n+-p-p+ structure. They have 1 mm2 sensitive area, which is subdivided into 576 (MEPhI/PULSAR) and 556 (CPTA) separate pixels, respectively. Each of these pixels is quenched by an individual resistor. The structure of these APDs was optimised for green-red light detection. In the case of the CPTA devices, pixels are separated by grooves, which are filled with an optically nontransparent material. This is a new development based on the structure described in Refs. [2] and [5]. The third device (from Dubna) is made on an n-type silicon substrate. It has a 0.5 mm2 sensitive area, which is subdivided into 10,000 separate avalanche regions-micro-wells (for more details see Ref. [6]). The structure of this APD was optimised for the blue-UV part of the light spectrum. 3. Dependence of gain and photon detection efficiency on the bias voltage One can assume that the MPGM APD gain is equal to the charge (Q1) stored in a pixel capacitance when a single pixel is fired [2,4]: Q1 ¼ C pix ðV V B Þ
(1)
ARTICLE IN PRESS Y. Musienko et al. / Nuclear Instruments and Methods in Physics Research A 567 (2006) 57–61
58
where Cpix is the equivalent pixel capacitance, V is the APD operating voltage and VB is its breakdown voltage. However, it was found [7] that light emitted during the pixel breakdown penetrated adjacent pixels due to optical cross-talk and fired these pixels. Thus the average number of pixels fired by a primary photoelectron is typically more than 1. Then the real APD gain (MAPD) is equal to the charge Q1 multiplied by the average number of fired pixels (Npix): M APD ¼ N pix Q1 .
(2)
Fig. 1 shows a single electron spectrum (SES) of the MEPhI/PULSAR APD measured at reduced temperature (with the MPGM APD placed inside a commercial freezer). The ‘‘pedestal’’ events were carefully subtracted from this plot. The green LED (l ¼ 515 nm) pulse amplitude was reduced to the level of 51 photon/pulse using neutral density filters.
From Fig. 1 one can see that there is a significant probability that 2 or more pixels are fired by an absorbed photon. From the measurements of the SES at different voltages one can calculate the MPGM APD gains, corresponding to these voltages. Fig. 2 shows the MEPhI/PULSAR APD gain and Q1 charge as a functions of the bias voltage measured at T ¼ 28 C. One can see that at U ¼ 61 V the average number of fired pixels is more than 2. The gain of the MPGM APD can also be measured in a different way. The charge Q measured by the ADC in response to the LED pulse is related to the MPGM APD gain by: Q ¼ N pe G amp M APD
(3)
where Npe is the number of photoelectrons and Gamp is the gain of the amplifier. If the APD noise is small, the measured low light LED amplitude spectra (see, for example, Fig. 1 from Ref. [8]) can be compared with the Poisson distribution and the mean, Npe, can be calculated using the well-known property of this distribution: N pe ¼ lnðPð0ÞÞ.
(4)
Fig. 1. Single electron spectrum of the MEPhI/PULSAR APD.
Fig. 2. Gain and Q1 of the MEPhI/PULSAR APD as a function of voltage measured at T ¼ 28 1C.
Fig. 3. Gain (top) and PDE (bottom) as a function of voltage for the MEPhI/PULSAR APD measured at two temperatures.
ARTICLE IN PRESS Y. Musienko et al. / Nuclear Instruments and Methods in Physics Research A 567 (2006) 57–61
Fig. 4. Gain (top) and PDE (bottom) as a function of voltage for the CPTA APD measured at two temperatures.
Here, P(0) is the probability to observe the ‘‘pedestal’’ events. It should be mentioned that Npe depends on the APD voltage and should be calculated for each measurement point. Using this method, the gain of the MPGM APDs was measured using blue (470 nm) and green (515 nm) LEDs operated in pulsed mode. The signals from the MPGM APDs were amplified with a fast transimpedance amplifier (gain ¼ 60) and digitised with a LeCroy 2249 W ADC. The APDs were illuminated with small LED pulses via a 0.5 mm collimator. The number of photons in the LED pulse was measured using a calibrated XP2020 photomultiplier. This allowed the calculation of the MPGM APD gain and PDE as a function of voltage using the same measurement points. Figs. 3 and 4 show the MEPhI/ PULSAR and CPTA APDs gain and PDE as a function of voltage measured at two temperatures with a 515 nm LED light. Higher gains and PDE can be reached at reduced temperature. This is due to the fact that in silicon the dark carrier generation rate is significantly reduced when the temperature is lowered. The dark current and power dissipation inside the MPGM APDs are lower than they are at room temperature. As a result, the APDs can operate
59
Fig. 5. Gain, PDE (top figure) and excess noise factor (bottom figure) as a function of voltage for the Dubna APD.
at higher values of ‘‘overvoltage’’ (V–VB), and higher values of the gain and PDE become possible. The gain and PDE as a function of voltage of the Dubna APD are shown in Fig. 5 (top figure). The gain of this APD is lower in comparison to that of the other devices due to smaller pixel capacitance.
4. Multiplication noise of the MPGM APDs As was discussed in Section 2, the optical cross-talk between the pixels causes an increase of the average MPGM APD gain (see Fig. 2), but with the number of pixels which are fired by a primary photoelectron fluctuating from event to event (see the SES in Fig. 1). As a result, the multiplication noise of the MPGM APD is not negligible as one would expect from the very small spread of the charge Q1 for all the MPGM APD pixels (this spread is small due to small variation of the breakdown voltage over the 1 mm2 APD area). This multiplication noise can be numerically expressed in a similar way as is usually done for PMTs and APDs operated below breakdown in terms of an ‘‘excess noise factor’’, F. The excess noise factor (F)
ARTICLE IN PRESS 60
Y. Musienko et al. / Nuclear Instruments and Methods in Physics Research A 567 (2006) 57–61
of the MPGM APD can be calculated from the single electron distribution using the equation: s2 F ¼ 1 þ 2M M APD
(5)
where MAPD is the MPGM APD gain and s2M is its variance. The excess noise factor F as a function of voltage for the Dubna APD is plotted in Fig. 5 (bottom). Fig. 6 (top) and Fig. 6 (bottom) show F as function of gain for MEPhI/ PULSAR and CPTA APDs, respectively. These results show that at high gains the F is significantly different from unity for all three devices. From Fig. 6 one can also see that the dependence of F on the gain is the same even at temperatures differing by 50 C. It is important to notice that comparison between the numbers of photoelectrons calculated from the width of the LED spectrum and calculated from Eq. (2) gives another method for the determination of F (see Ref. [8]). It was found that both methods give very similar results for all the devices studied in this work.
5. Spectral response of the MPGM APDs For the spectral response measurements, an ‘‘Optometrics’’ SDMC1-03 spectrophotometer was used. The spectrophotometer light intensity was significantly reduced using neutral density filters to the level when the maximum current measured with the APD was only 30% higher than its dark current. This was done to avoid non-linearity effects caused by high pixel illumination. The photocurrent measured with the APD was compared with the photocurrent measured with the PIN photodiode. In addition, measurements with the LED pulsed light were used for absolute spectral response calibration (at least 2 different LED measurements were done for each APD). Fig. 7 shows the spectral response of the three MPGM APDs measured at room temperature. The lower PDE of the MEPhI/PULSAR APD in comparison to that of the CPTA APD is due to the smaller ratio of the active pixel area to the total APD area (the so-called geometric factor). In the case of the MEPhI/ PULSAR APD the geometric factor is 25%. Due to an optimized design, the geometric factor of the CPTA APD is 70%. A high value of PDE (45% at 630 nm) was measured for this APD at T ¼ 28 1C [8]. The structure of the Dubna APD was optimized for blue/UV light detection [6]. The geometric factor of this APD is 100%. This device cannot operate at high values of the ‘‘overvoltage’’ ((V–VB)/VBo3%)) due to the rapid increase of the APD noise. Because of this, the probability of the Geiger discharge being initiated by a primary photoelectron is small [9] and the PDE of this APD does not reach the highest possible values. 6. Conclusion In this work we propose a method to characterize MPGM APDs and we present systematic studies of new devices from three different manufacturers. The basic properties of MGPM APDs such as gain, PDE and excess
Fig. 6. Excess noise factor as a function of gain for MEPhI/PULSAR (top) and CPTA (bottom) APDs.
Fig. 7. Photon detection efficiency as a function of the wavelength of light for three MPGM APDs measured at room temperature.
ARTICLE IN PRESS Y. Musienko et al. / Nuclear Instruments and Methods in Physics Research A 567 (2006) 57–61
noise factor, as well as their dependence on operating voltage have been measured. Acknowledgments The authors would like to acknowledge Boris Dolgoshein of MEPhI, Zair Sadygov of JINR, Victor Golovin of CPTA and David McNally of Photonique [10] for providing us with the samples of MPGM APDs, and the National Science Foundation for its continued support.
[2] [3] [4] [5]
[6] [7] [8] [9] [10]
References [1] A.G. Gasanov, V.M. Golovin, Z.Y. Sadygov, N.Y. Yusipov, Lett. J. Technol. Phys. 16 (1990) 14 (in Russian).
61
G. Bondarenko, et al., Nucl. Instr. and Meth. A 442 (2000) 187. Z. Sadygov, et al., Nucl. Instr. and Meth. A 504 (2003) 301. P. Buzhan, et al., Nucl. Instr. and Meth. A 504 (2003) 48. A. Akindinov, et al., Proceedings of the Eighth ICATPP Conference on Advanced Technology and Particle Physics, Villa Erba, Como, 6–10 October 2003, World Scientific, Singapore, 2004. Z. Sadygov, et al., these proceeding. A. Lacaita, et al., IEEE Trans. Electron Devices, ED ED-40 (1993) 577. E. Guschin, et al., these proceedings. W.G. Oldham, R.R. Samuelson, P. Antognetti, IEEE Trans. Electron Devices, ED ED-19 (1972) 1056. SA. Photonique, Geneva (www.photonique.ch).