Large Area Avalanche Photodiodes in X-rays and scintillation detection

Nuclear Instruments and Methods in Physics Research A 442 (2000) 230}237 Large Area Avalanche Photodiodes in X-rays and scintillation detection M. MoszynH ski *, M. Kapusta, M. Balcerzyk, M. Szawlowski, D. Wolski Soltan Institute for Nuclear Studies, PL 05-400 S! wierk-otwock, Poland Advanced Photonix, Inc., 1240 Avenida Acaso, Camarillo, CA 93012, USA Abstract The performance of 10 and 16 mm diameter beveled edge Large Area Avalanche Photodiodes (LAAPD) was studied in X-rays and scintillation detection. A good linearity of the LAAPD response to X-rays was observed up to energies of about 20 kev. The ratio of the APD gain for X-rays and light pulses remained constant (within 1%) for both the 5.9 and 16.6 kev photopeaks in the APD gain range up to 100. This allowed for use of soft X-rays as an accurate reference in APD scintillation detection study. The study covered measurements of the number of generated e}h pairs and energy resolution for several scintillators. Particularly, the energy resolution of 4.8$0.2% was measured with a small CsI(Tl) crystal for 662 kev γ-rays from a Cs source and 4.3$0.2% with YAP:Ce. The measured energy resolutions were comparable or better than those measured with a photomultiplier. 2000 Elsevier Science B.V. All rights reserved. Keywords: X-ray detectors; Scintillating detectors; Avalanche photodiodes 1. Introduction Growing interest in applications of avalanche photodiodes in nuclear physics [1] and medicine [2] stimulated the study of LAAPDs performance in the scintillation light readout. In the previous studies [3}6] we have shown the excellent properties of beveled edge LAAPDs in scintillation detection. A very good energy resolution, comparable or better than those measured with the XP2020Q photomultiplier, was reported. The measured high numbers of e}h pairs for di!erent scintillators con- "rmed a very good quantum e$ciency of LAAPDs and allowed for a quantitative discussion of the energy resolution [5,6]. The measurements with an LSO crystal resulted in a very good time resolution of 570 ps [3], obtained for γ-rays from a Co source. Recently, reported discrepancy in the APD gain for X-rays and light pulses for small area APDs [7,8], prompted us to verify if similar phenomenon would be observed for LAAPDs. Support for this work was provided by the Polish Committee for Scienti"c Research, Grants Nos. 8T 10C 005 15 and 8T 11B 037 13. * Corresponding author. Tel.: #48-22-718-0586; fax: #48-22-779-3481. E-mail address: marek@ipj.gov.pl (M. MoszynH ski) 2. Experimental details Our study was carried out with 10 and 16 mm diameter, windowless LAAPDs made by Advanced Photonix, Inc. (API). The main characteristics of 0168-9002/00/$ - see front matter 2000 Elsevier Science B.V. All rights reserved. PII: S 0 1 6 8-9 0 0 2 ( 9 9 ) 0 1 2 2 6-7

231 the tested diodes are listed in Table 1. These UVenhanced LAAPDs feature a high quantum e$ciency of about 77% at 400 nm, high gain, and dark current typically below 300 na for 16 mm devices and about 50 na for 10 mm devices at operating gain of 200. The X-ray and gamma ray detection was studied in the energy range up to 26 kev using Fe, Mo, and Am sources. The response for the light signal was measured with pulses from an XP22 light emitting diode [9]. The peak emission of the XP22 is at 620 nm and its emission spectrum spans between 550 and 800 nm [10]. The LED was driven by a voltage pulse with 50 ns decay time constant from a precise mercury relay pulser. The properties of LAAPDs in scintillation detection were studied for 16 mm devices using small NaI(Tl), CsI(Tl), LSO:Ce, and YAP:Ce crystals. The crystals were wrapped in several layers of white Te#on tape, except for the NaI(Tl), which had been assembled by the manufacturer. An overview of the crystals used is given in Table 2. In the experiments the signal from LAAPD was fed to a 142AH Ortec preampli"er and then to a TC244 Tennelec spectroscopy ampli"er. A PCbased multichannel analyzer was used to record energy spectra. 3. Results 3.1. Light and direct X-ray detection Fig. 1 shows the energy spectrum of Fe and Am sources measured with a 16 mm LAAPD at gain of 50 and shaping time constant in the ampli- "er of 0.25 μs. Note the obtained good energy resolution, especially taking into account the large area of the tested diode (200 mm ) and #ooded mode of illumination. Based on the measured Table 1 The main parameters of the tested LAAPDs Model 394-70-73-500 630-70-73-500 630-70-73-500 Enhancement UV Serial no. 121-07-05 121-1-7 113-03-01 Diameter 10 mm 16 mm Window None Q.E. at 400 nm 78% 77% 77% Gain 200 at 1780 V 200 at 1850 V 200 at 2468 V Dark current 54.3 na 367 na 226 na Capacitance 65 pf 130 pf 110 pf Rise time 10 ns 16 ns 11 ns Fig. 1. Energy spectrum of X-rays from Fe and Am sources. Table 2 Tested scintillators Crystal Size (mm) Mfg λ (nm) Decay (μs) Light output (ph/mev) NaI(Tl) H10 10 Amcrys-H 415 0.23 40 000$2000 CsI(Tl) H9 9 Bicron 530 0.8 61 000$3000 LSO 4 5 14.5 Russia 420 0.046 27 300$1400 YAP 3 3 10 Preciosa 365 0.025 21 400$1000 Ref. [12]. Ref. [13]. Ref. [14]. SECTION VI.

232 spectra, the linearity of the LAAPD response to X-rays was determined (Fig. 2). A good linearity for energies up to 26 kev was observed at LAAPD gains of 25, 50 and 100. A certain deviation of the linearity was observed for the 59.6 kev gamma line. Fig. 3 presents the LAAPD gain for X-rays normalized for the optical gain obtained with a 16 mm LAAPD for two di!erent energies of X-ray photopeaks. The measurements were carried out at simultaneous illumination by 5.9 and 16.6 kev X-rays from Fe and Mo sources, and the light pulses from the LED pulser. In the APD gain range up to 100 Fig. 2. The linearity of LAAPD signal versus energy of X-rays at gains of 25, 50, and 100. The error bars are within the con"nes of the points. the gain for X-rays followed that for light within 1%. For higher LAAPD gains (up to 250) a deviation down by about 5% was measured for 16.6 kev X-rays and asymmetry of the peak was observed. Obtained results showed di!erent performance of LAAPDs in X-ray detection as compared to the small area devices [7,8]. It has been reported [7] that the latter exhibit up to 50% gain reduction for 5.9 kev X-rays in optical gain range up to 100, making using X-rays as a reference in light measurements with these devices impractical. Nonlinear response of an avalanche diode at higher signal current density can be caused by a number of e!ects. The voltage drop across load resistance reducing the diode bias voltage (gain), space-charge e!ect decreasing electric "eld in the avalanche region, and local heating are the most probable causes. Detailed analysis needs further studies and is beyond the scope of this paper. We attribute much better linearity obtained with LAA- PDs to the di!erent diode structure, wider avalanche region, and di!erent electric "eld pro"le. The obtained results show that in the LAAPD gain range up to 100, both the 5.9 and 16.6 kev X-ray peaks are a good reference for measuring the number of e}h pairs generated by the detected light, particularly in scintillation light detection [3}6]. This "nding allows also for evaluation of the excess noise factor for tested LAAPDs based on the energy resolution of peaks produced by a light pulser. Fig. 4 presents an example of the energy Fig. 3. Ratio of the LAAPD gain for 16.6 and 5.9 kev X-rays and light pulses. Fig. 4. Energy spectrum of 5.9 kev X-rays from Fe source together with the peaks due to light pulser and charge pulser as measured for 10 mm LAAPD.

233 spectrum showing 5.9 kev X-rays, light pulser, and charge pulser peaks as observed with a 10 mm LAAPD. The obtained X-rays energy resolution is much better than for a 16 mm LAAPD. That result is mainly due to the better gain uniformity over diode active area. The charge pulse peak resolution represents LAAPD's dark noise and preampli"er noise contributions. The energy resolution obtainable in measurements with an avalanche photodiode is limited mainly by three factors: statistical limitations determined by the avalanche gain amplitude distribution, statistics of primary electron}hole pairs, Fano factor (in ionizing radiation detection), and photon noise (in light detection), gain non-uniformity in the diode detection volume, dark noise of the diode}preampli"er system. In the light detection the whole illuminated area contributes to the output signal, averaging local gains in points of photons interactions, so we can exclude gain non-uniformity contribution to peak broadening. Assuming Gaussian shape of the detected peak and negligible photon noise, the energy resolution E of the peak (expressed in kev) due to the light pulser, can be described by the following equation: E "(2.36) FEε#Δ (1) where F is the APD excess noise factor, E is the energy of the light peak in kev, ε is the energy per e}h pair creation (3.6 ev for silicon), and Δ is the noise contribution of the diode}preampli"er system (FWHM in kev). The quantities presented in Fig. 4 allow calculating the excess noise factor F using Eq. (1). Fig. 5 shows the dependence of the excess noise factor versus gain of a 16 mm LAAPD. To increase the accuracy, all the measurements were carried out at least three times and the average value was plotted. Up to the gain of 100 the excess noise factor is growing slowly, with the absolute value below 2. For higher gains it grows faster, re#ecting contribution of holes in the ampli"cation process. Fig. 5. The dependence of the excess noise factor versus gain for 16 mm LAAPD. Table 3 Number of e}h pairs and energy resolution for 662 kev γ-rays Crystal Number of e}h pairs (e}h/mev) Energy resolution (%) CsI(Tl), H9 9mm 37 000$1100 4.8$0.2 NaI(Tl), H10 10 mm 26 900$800 6.5$0.2 LSO:Ce, 4 5 14.5 mm 21 500$1000 10.3$0.3 YAP:Ce, 3 3 10 mm 15 500$700 4.3$0.2 Measured with the LAAPD gain of 50 and 3 μs shaping time constant. Measured with the LAAPD gain of 100 and the optimal shaping for each crystal. Obtained data are in good agreement with experimental results published earlier by API [11]. 3.2. Scintillation detection In the scintillation detection 16 mm LAAPDs were studied with small CsI(Tl), NaI(Tl), LSO, and YAP crystals attached to the sensitive area of the diodes. Table 3 presents the number of primary e}h pairs and energy resolution determined for 662 kev γ-rays from a Cs source for all the studied scintillators. The number of e}h pairs was measured by comparing the position of 662 kev γ-peak detected in the scintillator to that of 16.6 kev X-rays from a Mo source detected directly by the LAAPD. The measurements were carried out at the LAAPD SECTION VI.

234 gain of 50 and shaping time constant of the ampli- "er of 3 μs. Note high number of e}h pairs for CsI(Tl) and NaI(Tl) close to those reported in Refs. [12,13], as measured with pin photodiodes. Fig. 6a presents a pulse height spectrum of γ-rays from a Cs source obtained with the CsI(Tl) crystal at LAAPD gain of 100 and shaping time constant of 6 μs. Note the excellent energy resolution of 4.8$0.2%, possibly one of the best observed with scintillation detectors. The energy threshold is below the Ba X-ray peak of 32 kev re#ecting the large dynamic range of the measured energies. Such high-energy resolution was never observed before with CsI(Tl) crystals coupled to photomultipliers. Even better energy resolution of 4.3$0.2% was measured with YAP : Ce crystal having a peak emission at 365 nm (see Fig. 6b). In the last column of Table 3 the best energy resolution obtained with various crystals are collected. The measurements were carried out at LAAPD gain of 100 using the optimal shaping time constant for each crystal. Note a high-energy resolution for the studied crystals, except for LSO, for Fig. 6. Energy spectrum of 662 kev γ-rays from a Cs source measured with the YAP:Ce crystal and 0.25 μs shaping time constant (a). The same spectrum measured with the CsI(Tl) crystal at 6 μs shaping time constant (b). which the energy resolution, at present, is limited by the poor intrinsic resolution [4,5]. The energy resolution, E/E, of the full energy peak measured with a scintillator coupled to an APD can be written ( E/E) "(δ ) #( N/N } ) #(Δ /N } ) (2) where δ is the intrinsic resolution of the crystal, N/N represents the e}h pair statistical contribution, and Δ /N the APD noise contribu- } } tion. The intrinsic resolution of a crystal is connected with many e!ects such as inhomogeneities in the scintillator causing local variations of the light output, non-uniform re#ectivity of the re#ecting covering of the crystal, as well as the non-proportional response of the scintillator [15]. The statistical accuracy of the signal from an APD is a!ected by the excess noise factor F,re#ecting the statistical #uctuation of the APD gain. N/N } "2.36(F/N } ) (3) where N is a number of primary electron}hole } pairs. In good beveled edge APDs the excess noise factor of about 2 was observed depending on the APD gain, see Chapter 3.1. For electron injection alone the excess noise factor can be expressed as F"k M#(2!1/M)(1!k ) (4) where k is a weighted average ratio of the hole and electron ionization rates. The noise contribution is measured by using a test pulser to feed a known, small capacitance connected to the preampli"er input. In this study the charge injected by the pulser was calibrated in the number of e}h pairs using response of LAAPD to 5.9 kev X-rays from a Fe source, see Fig. 4. The absolute value expressed in the rms electrons was recalculated to the normalized value, as follows: Δ /N } "2.36Δ /N } (5) where Δ is noise expressed in rms electrons. The results of the analysis of the energy resolution measured with LAAPDs are summarized in Table 4. To evaluate the statistical contribution of e}h pairs, according to Eq. (4), an excess noise

235 Table 4 Energy resolution and its components for LAAPD light readout Crystal e}h number E/E (%) N/N } (%) Δ /N } (%) δ (%) LAAPD PMT CsI(Tl) 27 800$800 4.8$0.1 2.00$0.03 1.38$0.04 4.1$0.2 4.4$0.4 NaI(Tl) 17 150$500 6.5$0.2 2.55$0.04 0.94$0.03 5.9$0.2 5.8$0.3 LSO 13 900$400 10.6$0.3 2.90$0.87 0.87$0.03 10.2$0.5 9.1$0.4 YAP 10 300$460 4.3$0.2 3.3$0.1 1.15$0.04 2.5$0.4 1.3$0.5 For 662 kev, measured with optimal shaping time. For 662 kev and optimal shaping. See Ref. [5]. See Ref. [14]. factor F of 2 was assumed for the LAAPD gain of 100. Table 4 summarizes obtained results showing the contributions of di!erent factors to the energy resolution observed with LAAPDs. The contribution of e}h pair statistics is up to about a factor of 2 lower than those typically found with PMTs [5,14]. This is due to the high quantum e$ciency and the low excess noise factor of LAAPDs. The latter is a big advantage of beveled edge devices. The noise contribution is low and has weak in#uence on the measured energy resolution. In the last two columns the intrinsic resolution of the crystals calculated using Eq. (2) and those published for the measurements with photomultipliers are collected. Note a good agreement of these quantities con- "rming a good accuracy of the measurements and analysis. The study also demonstrates that the energy resolution is strongly limited by the intrinsic energy resolution for scintillators with a high light output, such as CsI(Tl), NaI(Tl), and LSO. The good energy resolution of the YAP crystal is due to the fact that it has one of the lowest contributions of intrinsic resolution [5,14,15]. The experiments presented above were carried out with 662 kev γ-rays from a Cs source. Consequently, the study characterizes scintillation detectors with LAAPD light readout mainly in the medium-energy γ-ray domain. Fig. 7 presents the pulse height spectra of 59.6 kev γ-rays from an Am source and 122 kev γ-rays from a Co source measured with a 10 mm Fig. 7. Energy spectra from Am and Co sources measured with NaI(Tl) crystal. in diameter by 10 mm high NaI(Tl) crystal coupled to a LAAPD. The best energy resolutions of 11.3$0.3% and 8.4$0.3%, respectively, were obtained at the LAAPD gain of 100 and 0.5 μs shaping time constant in the spectroscopy ampli"er. The energy threshold of the collected spectra is below 10 kev. These results are comparable to those measured with the best scintillators coupled to a photomultiplier. Note that the best energy resolution was obtained with a shorter time constant than that used previously, which allowed SECTION VI.

236 Fig. 8. The energy resolution of CsI(Tl) crystal versus energy of γ-rays measured at 2 μs shaping time constant. The thin continuous lines represent the contribution of noise of the LAAPD and the statistical accuracy of the signal. Thick continuous line is a guiding line for experimental data points. The error bars are within the con"nes of the points. for signi"cant reduction of the LAAPD noise contribution. Fig. 8 shows the energy resolution of CsI(Tl) crystal plotted versus energy of γ-rays in the range of 16.6 kev to 1.275 MeV. The thin continuous lines represent the contribution of noise of the LAAPD and the statistical accuracy of the signal. Thick continuous line is a guiding line for experimental data points (squares). All the curves are plotted for the 2 μs shaping time constant in the ampli"er. That shows clearly that the energy resolution for energies below 100 kev is mainly controlled by the noise. For higher energies the statistical accuracy and intrinsic resolution of the crystal determine the energy resolution. 4. Conclusions The presented study shows a good linearity of the beveled edge LAAPDs response in the direct X-rays detection up to the energy of 20 kev. A constant ratio of the LAAPD gain for X-rays and light pulses from an LED light pulser was observed at 5.9 and 16.6 kev X-rays and LAAPD gain up to 100. This fact allowed for determination of excess noise factor using the light peak calibrated against X-rays and measurements of the number of e}h pairs generated in scintillation detection. The excess noise factor measured versus LAAPD gain was below 2 for the gains up to 100. This study has shown also that new Large Area Avalanche Photodiodes recently produced by Advanced Photonix, Inc. demonstrate excellent performance when used in scintillation detection. This is re#ected by the signi"cantly better, or comparable, energy resolution obtained with LAAPDs for di!erent scintillators, to those observed with photomultipliers. The high number of e}h pairs obtained with various scintillators implies a better statistical accuracy of the detected light and an almost negligible contribution of noise for γ-rays with energies above 100 kev. The low noise level, particularly at 0.25 μs shaping, allows for a good energy resolution with the new fast Ce-doped scintillators having a lower light output than the classical CsI(Tl) and NaI(Tl) crystals. The analysis of the energy resolution measured with LAAPD readout allows the intrinsic resolution of the tested crystals to be calculated. The results are in a good agreement with the data reported from measurements with photomultipliers. For scintillators with a high light output such as CsI(Tl), NaI(Tl) and LSO the intrinsic resolution is the main limiting factor for the energy resolution. References [1] I. Holl, E. Lorenz, S. Natkaniec, D. Renker, C. Schmeltzer, B. Schwartz, IEEE Trans. Nucl. Sci. NS-42 (1995) 351. [2] C. Schmelz, S.M. Bradbury, I. Holl, E. Lorenz, D. Renker, S. Ziegler, IEEE Trans. Nucl. Sci. NS-42 (1995) 1080. [3] M. Moszynski, T. Ludziejewski, D. Wolski, W. Klamra, M. Szawlowski, M. Kapusta, IEEE Trans. Nucl. Sci. NS- 43 (1996) 1298. [4] M. MoszynH ski, M. Kapusta, D. Wolski, M. Szawlowski, W. Klamra, IEEE Trans. Nucl. Sci. NS-44 (1997) 436. [5] M. MoszynH ski, M. Kapusta, D. Wolski, M. Szawlowski, W. Klamra, IEEE Trans. Nucl. Sci. NS-45 (1998) 472. [6] M. MoszynH ski, M. Kapusta, J. Zalipska, M. Balcerzyk, D. Wolski, M. Szawlowski, W. Klamra, IEEE Trans. Nucl. Sci. 46 (1999) 243. [7] J.P. Pansart, Nucl. Instr. and Meth. 387 (1997) 186.

237 [8] C.P. Allier, H. Valk, J. Huizenga, V.R. Bom, R.W. Hollander, C.W.E. van Eijk, IEEE Trans. Nucl. Sci. NS-45 (1998) 576. [9] M. MoszynH ski, J. Vacher, Nucl. Instr. and Meth. 141 (1977) 319. [10] Ferranti data, Ferranti Ltd, Gem Mill, Chadderton, Oldham, Lancs, England, produced in the 1970s. [11] M. Szawlowski, S. Zhang, A. DeCecco, M. Madden, M. Lindberg, E. Gramsch, IEEE NSS 1992, Conference Record, p. 239. [12] I. Holl, E. Lorenz, G. Mergas, IEEE Trans. Nucl. Sci. NS-35 (1988) 105. [13] M. MoszynH ski, M. Kapusta, M. Mayhough, D. Wolski, S.O. Flyckt, IEEE Trans. Nucl. Sci. NS-44 (1997) 1052. [14] M. Kapusta, M. Balcerzyk, M. MoszynH ski, J. Pawelke, Nucl. Instr. and Meth. A 421 (1999) 610. [15] P. Dorenbos, J.T.M. de Hass, C.W.E. van Eijk, IEEE Trans. Nucl. Sci. NS-42 (1995) 2190. SECTION VI.