Characterization of Scintillators by Modern ... - IEEE Xplore

2886

IEEE TRANSACTIONS ON NUCLEAR SCIENCE, VOL. 57, NO. 5, OCTOBER 2010

Characterization of Scintillators by Modern Photomultipliers—A New Source of Errors Marek Moszyn´ski, Fellow, IEEE, Tomasz Szcz˛esńiak, Member, IEEE, Maciej Kapusta, Marek Szawlowski, Joanna Iwanowska, Member, IEEE, Michał Gierlik, Agnieszka Syntfeld-Ka˙zuch, Member, IEEE, Łukasz S´widerski, Member, IEEE, Chuck L. Melcher, Senior Member, IEEE, Lars A. Eriksson, Fellow, IEEE, and Jarek Glodo, Member, IEEE

Abstract—The observed discrepancy in the light output, measured for a number of LSO, LYSO and BGO scintillators by different photomultipliers (PMTs), triggered studies to understand the problem. For that purpose the photoelectron number was measured by two different methods: the direct one based on a comparison of the full energy peak to that of the single photoelectron and by a method based on the pulse height resolution of the peak due to the light pulser. In this study, a significant number of different PMTs from Photonis and Hamamatsu were used. We concluded that the number of photoelectrons measured by means of the direct method was higher than the number of photoelectrons calculated from the pulse height resolution of the light pulser peak for all of the PMTs but XP2020Q. It leads to a large dispersion in the estimated light output for a given scintillator. In detail, the light output of BGO and LSO determined with the R6231 and R2059 PMTs is comparable to those measured with XP2020Q PMT and the S3590-18 pin photodiode, when photoelectron number is calculated from the pulse height resolution. Further in-depth studies of the photoelectron number at different bias voltages suggested that the effect is related to the space charge created in the dynode structure of the PMTs. Operation of PMTs at lower bias/gain minimizes this effect; thus, low noise electronics are recommended to determine the single photoelectron peak under these conditions. Moreover, the absolute light output of scintillators is affected by differences in the quantum efficiency calibrations by Photonis and Hamamatsu. Index Terms—BGO, light output of scintillators, LSO, LYSO, photoelectron number, photomultipliers, single photoelectron peak, space charge in photomultipliers.

I. INTRODUCTION

T

HE determination of the light output of scintillators in terms of emitted photons per unit of deposited gamma-ray energy (e.g, ph/MeV) requires the correction of the measured number of photoelectrons (phe) or electron—hole (e-h) pairs Manuscript received October 21, 2009; revised April 11, 2010; accepted June 08, 2010. Date of publication August 30, 2010; date of current version October 15, 2010. This work was supported in part by EU Structural Funds Project POIG.01.01.02-14-012/08-00 and by the International Atomic Energy Agency, Research Contract 14360. M. Moszyn´ski, T. Szcz˛esńiak, M. Szawlowski, J. Iwanowska, M. Gierlik, A. Syntfeld-Ka˙zuch, and Ł. S´widerski are with the Soltan Institute for Nuclear Studies, PL 05-400 S´wierk-Otwock, Poland (e-mail: [email protected]). M. Kapusta is with Photonis, F-19106 Brive La Gaillarde Cedex, France (e-mail: [email protected]). C. L. Melcher is with the Science and Engineering Facility, University of Tennessee,, Knoxville, TN 37996-2000 USA (e-mail: [email protected]). L. A. Eriksson is with Siemens Medical Solutions, Knoxville, TN 37932 USA (e-mail: [email protected]). J. Glodo is with Radiation Monitoring Devices, Watertown, MA 02472 USA (e-mail: [email protected]). Digital Object Identifier 10.1109/TNS.2010.2054111

produced by a scintillation light pulse for the quantum efficiency of the applied photodetector [1], [2]. In the case of photomultipliers (PMT), the phe number is measured directly, comparing the position of the full energy peak of known energy with that of the single photoelectron peak, which determines the gain of the PMT [2]. In the case of Si photodiodes, the e-h number is measured comparing the scintillator response to that of low energy gamma rays or X-rays detected directly in the photodiode. This simple approach becomes complex and difficult when one is looking for the absolute light output of scintillating materials. The efficiency of light collection from the scintillators, the photoelectron collection in PMTs, the photocathode reflectivity and light trapping in crystals are some processes affecting the measured light output [1]–[6]. The present studies were triggered by a comparative test of the light output of some selected LSO crystals carried out at the University of Tennessee (UT) and in our laboratory by the direct method [2]. Three different calibrated PMTs were used, Hamamatsu R2059 at UT and R6231MOD and Photonis XP2020Q in our laboratory. A large discrepancy in the measured numbers of photons seen by the PMTs was observed for the classical XP2020Q and those of the Hamamatsu PMTs. The XP2020Q measurements lead to light output values somewhat above 30,000 ph/MeV, while the Hamamatsu PMTs showed numbers close to about 40,000 ph/MeV. The first part of the investigation includes systematic studies to clarify the observed disagreement [7]. In the course of the work, tests with several other LSO, LYSO, and BGO crystals were added. Measurements employing a Hamamatsu S3590-18 Si photodiode were done for verification purposes in cases where the crystal size was compatible with the photodiode. The observed discrepancies of the light output measured with the different PMTs were correlated with the recently noticed unclear excess of the photoelectron number measured by the direct method in two samples of the R9420 PMT [8]. The R9420, no CF9214, showed the photoelectron number of 11500 300 phe/MeV for the LSO crystal, while the expected number reblue, was about 7000 lated to its blue sensitivity of 12.6 phe/MeV, see [9]. Moreover, the high phe number leads to a false interpretation of the measured time resolution. To clear up the problem, the phe number was crosschecked in [8] by the indirect method based on the calculation of the phe number from the pulse height resolution of a peak due to a light pulser [10]–[12]. It leads to the photoelectron number of 6900 200 phe/MeV, as expected for the quoted blue sensitivity of the PMT.

0018-9499/$26.00 © 2010 IEEE

´ SKI et al.: CHARACTERIZATION OF SCINTILLATORS BY MODERN PHOTOMULTIPLIERS MOSZYN

TABLE I THE SCINTILLATORS UNDER THE TESTS

Thus this method was first applied to the XP2020Q and R6231MOD. Moreover, the number of Photonis PMTs type: XP2020Q, XP20D0, XP5212, XP5200, XP5301, XP5302 and Hamamatsu PMTs type: R6231, R5320, R3998, R9420 and R2059 were tested for the same LSO and BGO crystals. The choice of PMTs was motivated by the new features introduced by major manufacturers over last decade as enhanced quantum efficiency of photocathodes and high gain dynodes. The main goal of the second part of the study [13], was to understand the origin of the observed discrepancies by means of analyzing the results of our in-depth studies and to specify a consistent method and technique to measure the photoelectrons number of scintillation crystals. The effect of applied HV on calculated number of phe for R6231 and XP5200, in comparison to XP2020Q is discussed in details, suggesting that an influence of the space charge effect in the PMT dynode structure leads to an incorrect PMT gain.

2887

TABLE II TESTED PHOTODETECTORS

a) Tested at the University of Tennessee

II. EXPERIMENTAL DETAILS In the first part of the study, related to the light output measurements, a number of scintillators, listed in Table I was tested. to assure a full Most scintillators have a size of photoelectron collection in PMTs, when coupled in the center of 51 mm diameter tubes, see [2]. Then in the study of the PMTs response, the BGO crystal of size and the selected LSO 2005 of size were used. Table II, in turn, lists the PMTs and photodiode used in the studies of the light output and some other PMTs of Photonis and Hamamatsu applied for comparative tests of photoelectron numbers. The first three calibrated PMTs were used in the study of the light output. Their quantum efficiency characteristics are presented in Fig. 1, following manufacturer data. In the last row of Table II the S3590-18 Si photodiode is added, with its quantum efficiency for BGO and LSO crystals

Fig. 1. Quantum efficiency characteristics of the PMTs used in the light output measurements, following manufacturer data.

quoted by Hamamatsu. This new photodiode is characterized by an enhanced quantum efficiency characteristic presented in Fig. 2, mainly due to the improved antireflection coating. In all measurements, the tested crystals, coated with Teflon tape, were coupled to the photodetectors by Down Corning DC 200 silicone grease. The tested PMTs were fitted with tapered voltage dividers to increase the linearity of the output signals. In most of the experiments, the signal from the PMT anode was fed to a Canberra 2005 scintillation preamplifier and in the

2888


Fig. 2. A typical quantum efficiency characteristic of S3590-18 photodiode, following the manufacturer data.

case of Si photodiode to a Polon 1001 charge sensitive preamplifier. The preamplifier signal was sent to a Tennelec TC244 spectroscopy amplifier operating with 3 bipolar shaping time constant for PMTs and unipolar shaping for the photodiode. The PC-based multichannel analyzer (Tukan 8k) [14] was used to record energy spectra. Gaussian functions were fitted to the full energy peaks by using procedures in the analyzer to determine their position and energy resolution. It included also the analysis of complex double peaks allowing for a correction of the peaks distorted by an escape peak. Afterwards, in the measurements with a reduced HV at PMTs, the Cremat 112 low noise preamplifier [15] was used too. Its low noise and high gain allowed recording single photoelectron spectra at the reduced PMT gain by a factor of about 10. This was of the importance to detect a possible space charge effect in the PMTs. III. RESULTS A. Light Output Measurements All measurements used the 661.6 keV -rays from a source, following method of [2]. In the case of PMTs, the number of photoelectrons per energy unit was measured by the direct method of Bertolaccini et al. [16] and used further in [2], [4], [5] and [9], for example. In this method the number of photoelectrons is measured directly by comparing the position of the 661.6 keV full-energy peak detected in the scintillator with that of the single photoelectron peak, which determines the gain of the PMT. The dark noise peak of single photoelectrons was adopted as the reference, following a commonly used method. In the case of Si photodiode, the number of electron-hole pairs per energy unit was measured by comparing the position of the 661.6 keV full-energy peak from the scintillator to that of the source detected directly in the 59.6 keV -rays from a photodiode [1], [2]. To get a number of photons, the integral quantum efficiency of the photodetectors corresponding to the emission spectrum of tested crystals has to be applied. It is done in Table III, presenting both phe (e-h) numbers and the light outputs. In the

footnote of Table III, the integral quantum efficiencies of the photodetectors, calculated using the typical emission spectra of LSO and BGO, are listed. It was assumed that the LSO and LYSO have the same emission spectra. The quantum efficiency was not corrected for the reflectivity of PMTs and S3590-18 photodiode. In spite of the improved antireflection coating in the latter, in comparison to the Si photodiodes used in [1] and [2], the presence of certain systematic errors resulting from different reflectivity of PMT and Si photodiodes cannot be excluded. Table III shows a systematic effect of the larger light output measured with the Hamamatsu PMTs compared to those determined using the XP2020Q PMT and S3590-18 Si photodiode. In Table IV the results of an independent experiment carried out at the University of Tennessee for BGO and LSO crystals, which used the calibrated XP2020Q and R2059 PMTs, are collected. It was an important crosscheck and confirmation of the above tests done in an independent laboratory using different photodetectors and finally done by the other scientists. In the experiment a precise evaluation of the integral quantum efficiency of the PMTs for both the crystals was done, correcting the measured emission spectra for wavelength response of the radioluminescence set-up. This was accomplished by recording the response of the set-up to a “standard” lamp that is calibrated by NIST. The generated correction file was used further to correct the LSO and BGO spectra. Data collected in Table IV confirms the earlier conclusion of a different light output measured with the Photonis and Hamamatsu PMTs, particularly important in that the precisely evaluated integral quantum efficiency of PMTs for the tested crystals was used. But which numbers, given in Tables III and IV, reflect the true light output? To solve the confusion about true (real) value of the BGO light output, the present data were compared in Table V with those from [1] and [2], considered by authors as the most precise data. A good agreement of the measured light output with the XP2020Q and S3590-18 photodiode to those of the earlier measurements, presented in Table V, suggests strongly that the use of the new Hamamatsu PMTs leads to the false excess of the photon number. Thus, the next question appears, is it the effect of a Hamamatsu technology or is it the effect of the modern technologies applied to photomultipliers with the enhanced quantum efficiency and higher gain dynodes, independent of the manufacturer? B. Study of PMTs Response In the studies of different photomultipliers, the number of photoelectrons measured by the direct comparison to the single photoelectron peak, due to the PMT noise, was compared to that of the number of photoelectrons calculated from the pulse height resolution (PHR) of the peak due to a light pulser [8], [10]–[12]. In each case, the measurements were repeated at least three times, in independent experiments. This was of particular importance for the R6231 PMT. A precise analysis of the light pulser peak resolution requires knowing the contribution of the PMT itself to the measured


2889

TABLE III LIGHT OUTPUT OF TESTED CRYSTALS AS MEASURED WITH DIFFERENT PHOTODETECTORS

QE = 15%, QE = 21 7%, QE = 87 8%, QE = 17 8%, LSO/LYSO: QE = 21 9%, QE = 29 1%, QE = 82 6%, QE = 24 1%, Measured for the other samples of 10 2 10 2 2 mm see [17]. BGO:

:

:

:

:

:

:

:

TABLE IV THE LIGHT OUTPUT OF BGO AND LSO CRYSTALS MEASURED WITH THE XP2020Q AND R2059 PMTS AT THE UNIVERSITY OF TENNESSEE

The knowledge of the PHR and the excess noise factor (ENF) of the PMT allows calculation of the number of photoelectrons in the peak (N). If the light pulser peak position corresponds to the gamma-peak position of a given energy, the number of photoelectrons per MeV can be easily defined. The excess noise factor, ENF, was calculated from the pulse of the single photoelectron peak fitted height resolution by a Gaussian function, as follows: (2)

QE = 12 23%, QE = 19 85%, QE = 15 85%, QE = 23 13% :

:

:

:

quantity. Thus, the shape and pulse height resolution were further measured for the single photoelectron peak produced by the light pulser under single photon illumination. The determination of the photoelectron number from the pulse height resolution is based on the following consideration. The statistical contribution of the photoelectron number to the PHR measured with the photomultiplier is given by [18]: (1)

The measurements were done with a 360 nm LED light pulser in relation to the LSO 2005 crystal, see Table I. The LED was excited by a pulse from a precision BNC PB-5 generator. The number of photons was adjusted by varying the amplitude of the pulse triggering the LED. It allows setting the number of photoelectrons to correspond to that in the 662 keV source. gamma ray peak from a gamma-rays In Fig. 3, spectra of single photoelectrons, and the 360 nm LED light source measured with the XP2020Q PMT are presented. It allows comparing directly the number of photoelectrons measured by the direct method and that determined from the pulse height resolution of the light pulser peak.

2890


TABLE V A COMPARISON OF THE MEASURED BGO LIGHT OUTPUT [ph/MeV] AND THOSE REPORTED IN REFS [1] AND [2]

Mean of 21 samples,

Mean of 4 samples tested using 4 different photodetectors. TABLE VI ANALYSIS OF PULSE HEIGHT RESOLUTION OF THE SINGLE PHOTOELECTRON AND LIGHT PEAKS FOR SOME SELECTED PMTS

Fig. 3. Spectra recorded with Photonis XP2020Q: single photoelectron, Cs gamma spectrum measured with LSO 2005 crystal and LED 360 nm light source.

The photoelectron number measured for the LSO crystal by the direct method was equal to 6150 150 phe/MeV; also see Table III, which lists the results of the earlier independent measurements done in 2006. The pulse height resolution of the light 200 phe/MeV, pulser peak of 3.82% corresponds to 6300 taking an ENF of 1.09, calculated from the pulse height resolution of the single photoelectron peak, see Table VI below. A good agreement of both the numbers confirms a good accuracy of the PHR method. In addition to the tests of the XP2020Q and R6231, similar measurements were carried out for a number of other Photonis and Hamamatsu PMTs, listed in Table II. Particularly, it includes a test of the XP5200, similar to the R6231, recommended by crystals. Photonis for gamma spectrometry with The analysis of the PHR of the light peak requires a precise measurement of the single photoelectron peak to determine accurately the excess noise factor of the PMTs. It was done by observing the dark noise and then the peak under single photon illumination by the light pulser. Fig. 4 presents a comparison of the single photoelectron peaks due to dark noise and the light pulser for the XP2020Q, R6231MOD, XP5200 and XP5301 PMTs. In the test with the light pulser, the LED was triggered by the BNC PB-5 output signal reshaped further in the measurements of single photoelectron response by an Ortec 579 Fast Filter Amplifier.

Measured under single photon illumination, The peak position corresponds to that of 662 keV -peak detected in the LSO-2005 crystal and the photoelectron number corresponds to that of LSO.

The Tukan 8k MCA was gated by the coincidence between the trigger pulse of the BNC pulser and the output pulse of the PMT under study, selected at the output of the spectroscopy amplifier by a single channel analyzer with the threshold set well below of the single phe peak. The intensity of the signal from the light pulser was adjusted to record less than 5% of pulses that triggered LED. It assures mainly single photon illumination of the PMT. The single phe peak, corrected for the exponential background, was fitted with a Gaussian function and its centroid was adopted for the single phe peak position. The spectrum displayed for the XP2020Q follows those known for more than twenty years. The single photoelectron peak due to the light pulser follows that of the dark noise. Its shape is more precise since it is not affected by the last dynode dark noise component dominating below the single photoelectron peak. Note a good pulse height resolution of 71%, which leads to an ENF of 1.09. In the case of the R6231MOD, the dark noise peak representing single photoelectrons is much poorer since it is strongly overloaded by the dynode noise. The measurement with the light pulser results in a cleaner single phe peak with a pulse height resolution of 77%. The positions of both peaks are the same, however, with a lower accuracy because of a poorer peak definition of the dark noise.


PHOTOELECTRON NUMBER OF LSO CRYSTAL

2891

TABLE VII MEASURED BY DIFFERENT METHODS WITH DIFFERENT PMTS

2 2

LSO: 10 10 5 mm , selected 2005, the ratio of the phe number measured by the standard single phe method, in relation to the noise peak, to that measured by the pulse height resolution method, reduced HV

Fig. 4. A comparison of the single photoelectron spectra due to dark noise of PMTs and due to single photons from the light pulser (marked by the FWHM) for the XP2020Q, R6231MOD, XP5200 and XP5301 PMTs.

In contrast, for the case of the XP5301, an unexpected shift up of the light pulser peak in relation to that of the dark noise is clearly seen, in spite of the fact that both peaks are very well

defined with the pulse height resolution below 70%. This is an unexplained effect, which could be responsible for a false photoelectron number measured in relation to the dark noise single phe peak, see also Table VII below presenting the numerical data on similar shifts for some other PMTs. It is important also to note a recent observation reported to us by Hamamatsu that a similar effect is observed for the R9420 [19]. The R9420 studied by us in [8] was sent back to the manufacturer for a more precise evaluation. Table VI presents the results of the analysis of the single photoelectron peak and the light pulser peak resolutions leading to the number of photoelectrons determined by the pulse height resolution method for the selected PMTs. The quoted numbers are the result of the fit of Gaussian functions to the measured single photoelectron peaks and light peaks. The quoted errors on the pulse height resolution for the single photoelectron and light pulser peaks reflect a variation of the numbers between successive independent measurements, which includes systematic errors related to the experiments themselves. The statistical errors of the fits are well below the adopted errors. Table VII summarizes test results of a number of photomultipliers obtained with HV/gain values sufficiently high to allow recording of single photoelectron peaks using standard electronics. Of the 12 tested PMTs, four of them, the XP2020Q, XP20D0, XP5212 and R3998, showed the same PHE number measured by both methods. The remaining PMTs exhibit a significant excess of photoelectrons measured by the direct method in relation to that determined by the PHR method, see the last column of Table VII. The XP5301 and R9420 PMT also tested earlier in [8] have shown a large excess of more than 1.4. These discrepancies, for most of the PMTs, could be explained by a larger gain for multi-photoelectron pulses than that for the single photoelectrons. For the remaining PMTs like

2892

XP5301, XP5302 and R9420, another effect, related to the shift of the single photoelectron peak position measured for thermal emission of the photocathode and with the light pulser affects the number of photoelectrons, see Table VII column 3 and 4. In columns 5 and 6 of the Table VII, the respective photoelectron numbers measured with the LSO crystal in relation to the single photoelectron peaks are listed. Column 7 summarizes the number of photoelectrons measured by the PHR method, and column 8 summarizes the ratio of the numbers measured by the direct and PHR methods, called later on as the PHE ratio. The data presented in Table VII lead to the conclusion that two independent effects are observed in modern photomultipliers. The first one suggests a different gain of the photomultiplier for the multi-photoelectron light pulses than that of the single photoelectron peak, and the second one is related to the shift of the single photoelectron peak. The accuracy of the obtained results is defined by the following factors: (1) gain stability and noise of PMT and associated electronics, (2) precision of the pulse height resolution of the light pulser peak, additionally limited by the stability of the light pulse and statistics of the LED emission. Any accidental degradation of the resolution will be reflected in the reduced number of photoelectrons. Thus, the high precision BNC PB-5 resolution and 10 ppm jitter of the pulse pulser with a 155 height was used to trigger the LED. Its high temperature stability and the short time of the spectra acquisition of 5 ppm per suggest that the influence of temperature on the measurements is negligible. The good precision of the PHR method applied to the measurement of the photoelectron numbers, is confirmed by the good agreement of the measured numbers by two methods for the XP2020Q and XP20D0, see Table VII. Finally, the light output of the LSO crystal determined with the R6231MOD and based on the photoelectron number of 8500 300 phe/MeV, measured by the PHR method, is equal to 29200 1500 ph/MeV, very close to that measured with the XP2020Q of 28800 1500 ph/MeV. It further confirms the excess of the PMT gain for the R6231MOD. But what is the origin of the observed effect?


Fig. 5. Photoelectron number of LSO at R6231 (upper panel) and XP5200 (bottom panel) versus HV. Square and open points measured by the direct method of comparison to the single photoelectron peak. Triangle points correspond to the PHE number calculated from pulse height resolution of light pulser peak.

C. Photoelectron Number and High Voltage at PMTs Despite the high precision of the measurements themselves, there are large discrepancies between results listed in Table VII. The discrepancies for different PMTs are probably related to the shift of the PMT gain occurring in some tubes due to intense light pulses. An improvement of energy resolution measured crystals coupled to the XP5500 which was operated with at reduced HV suggests that the origin of the problem could be the high gain of the PMTs [20]. The data from Table VII for the XP5301, where a lower photoelectron ratio was measured at a reduced HV, support this conclusion. To verify this with the other PMTs operating at reduced gain the single photoelectron peak position needs to be measured with sufficient accuracy. Here, the application of a low noise and high gain Cremat 112 preamplifier was particularly important. Measurements over a broad range of HV, corresponding to a dynamic range of PMT gain of about 10, were done for the R6231, XP5200 and XP2020Q PMTs using both the Canberra

Fig. 6. The same as in Fig. 5 for BGO coupled to R6231.

2005 and Cremat preamplifiers. The covered gain variation of the R6231 and XP5200 PMTs is the same with an accuracy of 10%, in spite of the large range of HV. Fig. 5 presents the photoelectron number versus HV at the PMTs, measured by the both methods for the LSO crystal coupled to the R6231 and XP5200 PMTs, respectively. Fig. 6 presents, in turn, a similar plot for the BGO crystal coupled to the R6231 PMT. Both figures show that the discrepancy in the number of photoelectrons measured by two methods continuously decreases with a reduced HV applied to PMTs and is very small at the lowest HV values. The results are summarized in


2893

Fig. 8. Photoelectron number measured directly versus the number determined from the PHR for the R6231, approximated by two straight lines. The error bars are within the size of the points.

Fig. 7. The ratio of PHE number in R6231 (upper panel) and XP5200 (lower panel), measured by the direct method to that calculated from pulse height resolution.

Fig. 7 that presents the photoelectron ratio versus HV, measured by two methods for the LSO and BGO crystals. It is worth noting that the phe number measured by the direct method decreases with the reduction of HV bias, while that calculated from the pulse height resolution, increases. It suggests that the mechanism in the gain process in PMTs affects both the gain for multi-photoelectron signals, and the pulse height resolution. Moreover, as shown Fig. 7, the photoelectron ratio is lower for the BGO which indicates that the observed degradation in the R6231 and XP5200 PMTs is correlated with the intensity of the multi-photoelectron signals. It is further confirmed by the plot of the photoelectron number measured directly versus that determined from the PHR for the R6231, see Fig. 8. The dependency, presented there, is approximated by two straight lines with the slopes corresponding to the photoelectron ratio measured for the LSO at large phe numbers of 1.26 and that close to 1.0 at lower phe numbers. In contrast to the PMTs discussed above, typically used in the gamma spectrometry, the photoelectron number obtained with the XP2020Q is the same for both methods of measurements and is practically independent of the applied high voltage in the tested range, as shown in Fig. 9. IV. DISCUSSION In the first part of the study, presented in Sections III-A and III-B, it was concluded that the application of modern PMTs leads to discrepancies in the estimated number of photons from different scintillators when the direct single photoelectron method is used. An unexpected gain shift

Fig. 9. PHE number of LSO at XP2020Q. Square and open points measured by the direct method of comparison to the single photoelectron peak. Triangle points correspond to the PHE number calculated from pulse height resolution of light pulser peak.

of PMTs for multi-photoelectron pulses as compared with the single photoelectron events was observed. To get a true number of photoelectrons and then the number of photons, the pulse height resolution method was proposed, following [8], [10]–[12]. Since the accuracy of the PHR method is affected by the variance of the PMT gain, the experiment was optimised to get well defined single photoelectron peaks, as presented in Fig. 4. First, several preliminary tests were done to find the origin of the observed effect, including: • Test of the influence of the colour of light from the light pulser. Tests done with a red laser showed the same excess of photoelectrons as that measured with the blue LED [8]. • Drift of the peak position of the PMT at very low counting rates, below 1000 c/s, observed in the past in our laboratory for some PMTs and related to a possible charging of insulators, showed no effect for the tested PMTs. • Afterpulses in the modern PMTs with enhanced quantum efficiency. Comparative tests done with the XP5301, R6231 and XP2020Q showed a negligible contribution to

2894


TABLE VIII COMPARISON OF THE PHOTOELECTRON NUMBERS AND LIGHT OUTPUT OF BGO AND LSO

QE = 15% QE = 87 8% QE = 21 7%, d) QE = 15 3%, LSO: e) QE = 21 9%, f) QE = 29 1%, g) QE = 20 8%, h) calculated using QE = 24 3%, j) QE = 17 1%, LSO: k) QE = 31 3%, l) QE = 22 4%—Estimated.

BGO: a) , b) , c) : QE characteristic measured by Photonis, BGO: i)

:

:

:

the total number of photoelectrons which was below 1% for the tested PMTs. • A downward shift of the single photoelectron peak due to electrons released not only from photocathode but also from other electrodes with a more positive voltage. This was excluded for the R9420 because of the excellent time jitter measured in [8]. However, it could be responsible for the shift down of the single photoelectron peak due to dark noise in relation to that induced by the light pulser. The obtained results of the photoelectron numbers versus the PMTs gain showed that a high PMT gain influences the photoelectron numbers obtained in the measurements by both methods. It activates the mechanism affecting PMT gain for multi-photoelectron signals, simultaneously also degrading the pulse height resolution. Moreover, its contribution is larger for intense fast light pulses, as seen by comparing the results of the LSO with those of the BGO crystals, see Fig. 7. It was further confirmed by a test of the photoelectron number measured directly versus the number determined from the PHR for the R6231 and approximated by two straight lines of different slopes, see Fig. 8. The measurement of the photoelectron number based on the comparison method usually requires operation of a PMT at high gain to record a single photoelectron peak. This in turn provides intense electron beams in the dynode structure for the light pulses from scintillators. Thus, we propose to explain these observations by the space charge created in photomultipliers by more intense signals. This space charge could affect the linear response of the PMTs and is manifested by the over-linearity of the intense light pulses [21]. This hypothesis is supported by the fact that the largest effect was observed for the R6231, XP5200, XP5301 and XP5302, i.e., the PMTs used in gamma spectrometry. These PMTs are equipped with modern dynodes of enhanced gain. The structures of the multipliers consist of two box-and-grid dynodes followed by a number of linear focusing stages of a reduced size. The effect of over-linearity caused by the space charge effect is weaker in the R2059 timing PMT than in the spectrometry

:

:

:

:

:

:

PMTs. Well recognized fast timing PMTs, such as the XP2020Q and XP20D0, with a high linearity of the current pulse up to 100 mA, did not show variations correlated with applied HV. The degradation of the pulse height resolution by the space charge effect is less evident. A serious influence of the space charge for fast timing with fast plastic scintillators, like NE111, was reported in [22] and [23] and discussed in [24]. Thus, we cannot exclude a similar effect on the pulse height resolution in the PMT. The degradation of the multi-photoelectron response was particularly pronounced for a C31024 PMT consisting of 4 stages based on high gain GaP dynodes, and was reflected in a large broadening of the output pulses for more intense anode pulses [25]. It suggests that this type of PMT is more sensitive to the space charge effect than classical PMTs with a much larger number of stages. The space charge influence on the measured phe number of the tested PMTs was practically eliminated by operation of PMT at lower HV/gain. Recording of the single photoelectron spectrum with a well pronounced peak at lower PMT gains required low noise input electronics. It was feasible by implementation of the low noise Cremat 112 preamplifier, characterized by a two times lower dark noise and a higher gain by about factor of four, instead of the commonly used Canberra 2005. Shown in Figs. 5 and 6, data at reduced HV were obtained using the Cremat 112 preamplifier. In the first part of the study, carried out at high gain/HV with classical electronics, a good agreement of the light output measured with the R6231 and XP2020Q was obtained using the photoelectron number for the R6231 measured by the pulse height resolution method. However, in the light of the study discussed in Section III-C, this quantity can be underestimated because of a deterioration of the pulse height resolution, see Fig. 5. The comparison of the light output results measured by the R6231 at different HV to those obtained with the XP2020Q, R2059 and S3590-18 photodiode is shown in Table VIII. For the R6231 PMT, three numbers of photoelectrons and photons are given in Table VIII, corresponding to those mea-


Fig. 10. A comparison of the quantum efficiency characteristics of R6231 measured by Hamamatsu and Photonis.

sured directly and then evaluated from the pulse height resolution for two HV voltages on the PMT. It reflects, as discussed above, an excellent agreement of the photon numbers measured at 1700 V by the PHR method to those determined with the use of the XP2020Q and the S3590-18 Si photodiode. The same quantities measured for 1000 V on the PMT gave a certain spread of the photon number. The same effect is observed for the R2059. The PMT integral quantum efficiency for the BGO and LSO was evaluated, in this case, using QE data of the calibrated R2059 (see Fig. 1) and corrected for the blue sensitivity ratio. To clear up further the problem, the quantum efficiency characteristic of the Hamamatsu R6231 was re-measured at the Photonis test bench. The results are showed in Fig. 10. Although the shapes of the curves are similar, there is a significant difference of about 10% in the wavelength range of 400–550 nm, which is critical for the LSO and BGO. The quantum efficiency, according to Photonis, is higher, leading to a lower light output of the tested scintillators compared to the calibration by Hamamatsu. It is shown in Table VI, by the additional numbers of photons calculated for the R6231 at 1000 V, and the R2059, assuming the Photonis calibration of the quantum efficiency. These numbers agree well with those measured by means of the XP2020Q and S3590-18 photodiode. The above analysis pointed out the next serious source of error in the measurement of the light output. This is the accuracy of the quantum efficiency calibration of PMTs by different manufacturers. In most of the earlier studies, the quantum efficiency characteristics given by PMT manufacturers were considered as accurate data and were not discussed in the error analysis. V. CONCLUSIONS This study showed that the evaluation of the photoelectron number by the method of direct comparison to the single photoelectron peak done with modern PMTs, operated at high gain with commonly used electronics, leads to an overestimation of the phe number. To minimize this effect, operation of PMTs at low bias/gain and implementation of low noise electronics, which allow single photoelectron peak detection, is required.

2895

The pulse height resolution method is much less affected by the observed effects; thus, it is preferred in measurements with the tested PMTs. Here, we must underline the successful application of the direct method based on the comparison of the full energy peak to that of single photoelectron used in our measurements with the XP2020Q PMT in [2] as well as in the present experiments. The observed effect is probably the result of the space charge build-up in the dynode structure leading to the over-linearity of the PMT and a degradation of the pulse height resolution. It is particularly pronounced for the PMTs used in a gamma spectrometry and it is further supported by a larger effect for intense fast light pulses as seen by comparing results of the LSO with those of the BGO crystals. Timing PMTs, able to sustain high signal currents for short intervals and PMTs with a large number of stages, based on dynodes with a moderate gain are much less sensitive to this effect. In the measurements of the light output of scintillators, the accuracy of the quantum efficiency calibration directly influences the results and has to be taken into account. Comparison of the calibration of the R6231 done at Photonis and at Hamamatsu showed about 10% difference, reflected afterwards in the measured light output of scintillators. The implications of the observed effects in the modern photomultipliers are of the great importance for the characterization of scintillators. Besides the incorrect light output, as determined by the classical method of the measured photoelectron number, it affects also results of the analysis of the energy resolution based on the measured number of photoelectrons. ACKNOWLEDGMENT The authors are very indebted to Hamamatsu Photonics K. K., for the R2059 PMT put at our disposal. REFERENCES [1] I. Holl, E. Lorenz, and G. Mageras, “A measurements of light yield of common inorganic scintillators,” IEEE Trans. Nucl. Sci., vol. 35, no. 1, pp. 105–109, Feb. 1988. [2] M. Moszyn´ski, M. Kapusta, M. Mayhugh, and S. O. Flyckt, “Absolute light output of scintillators,” IEEE Trans. Nucl. Sci., vol. 44, no. 3, pp. 1052–1061, Jun. 1997. [3] J. T. M. de Haas and P. Dorenbos, “Advances in yield calibration of scintillators,” IEEE Trans. Nucl. Sci., vol. 55, no. 3, pp. 1086–1092, Jun. 2008. [4] J. T. M. de Haas, P. Dorenbos, and C. W. E. van Eijk, “Measuring the absolute light yield of scintillators,” Nucl. Instrum. Methods Phys. Res. A, vol. A537, no. 1–2, pp. 97–100, Jan. 2005. [5] M. Gierlik, M. Moszyn´ski, A. Nassalski, A. Syntfeld-Ka˙zuch, T. Szcz˛esńiak, and Ł. S´widerski, “Investigation of absolute light output measurement techniques,” IEEE Trans. Nucl. Sci., vol. 54, no. 4, pp. 1367–1371, Aug. 2005. [6] G. F. Knoll, Radiation Detection and Measurement, 3rd ed. New York: Wiley, 1999, p. 249. [7] M. Moszyn´ski, T. Szcz˛esńiak, M. Kapusta, M. Szawlowski, M. Gierlik, J. Iwanowska, A. Syntfeld-Ka˙zuch, Ł. S´widerski, C. L. Melcher, and L. Eriksson, “Light output of BGO, LSO and LYSO measured with different photomultipliers and silicon pin photodiode,” presented at the SCINT Conf., Jeju, Korea, Jun. 2009. [8] T. Szcz˛esńiak, M. Moszyn´ski, Ł. S´widerski, A. Nassalski, A. SyntfeldKa˙zuch, A. G. Dehaine, and M. Kapusta, “A comparative study of fast photomultipliers for timing experiments and TOF PET,” IEEE Trans. Nucl. Sci., vol. 56, no. 3, pp. 1017–1023, Jun. 2009.

2896

[9] T. Szcz˛esńiak, M. Moszyn´ski, Ł. S´widerski, A. Nassalski, P. Lavoute, and M. Kapusta, “Fast photomultipliers for TOF PET,” IEEE Trans. Nucl. Sci., vol. 56, no. 1, pp. 173–181, Feb. 2009. [10] G. C. Kelley, P. R. Bell, K. C. Davis, and N. H. Lazara, “Intrinsic energy resolution,” IRE Trans. Nucl. Sci., vol. NS-3, pp. 57–58, 1956. [11] F. J. Lynch, “New liquid scintillators with higher speed and efficiency,” IEEE Trans. Nucl. Sci., vol. NS-15, no. 3, pp. 102–116, Jun. 1968. [12] M. Moszyn´ski, J. Vacher, and R. Odru, “Timing study with R1294U microchannel plate photomultiplier,” Nucl. Instrum. Methods, vol. 204, no. 2–3, pp. 471–478, Jan. 1983. [13] M. Moszyn´ski, T. Szcz˛esńiak, M. Kapusta, M. Szawłowski, J. Iwanowska, M. Gierlik, A. Syntfeld-Ka˙zuch, Ł. S´widerski, C. L. Melcher, L. Eriksson, and J. Glodo, “Characterization of scintillators by modern photomultipliers—A new source of errors,” in Proc. IEEE Nuclear Science Symp. Conf. Rec., Orlando, FL, Oct. 2009. [14] Z. Guzik, S. Borsuk, K. Traczyk, and M. Plominski, “Enhanced 8K pulse height analyzer and multi-channel scaler (TUKAN) with PCI or USB interfaces,” IEEE Trans. Nucl. Sci., vol. 53, no. 1, pp. 231–235, Feb. 2006. [15] Model 112 preamplifier purchased from Cremat Inc, Watertown, MA [Online]. Available: http://cremat.com [16] M. Bertolaccini, S. Cova, and C. Bussolati, “A technique for absolute measurements of the effective photoelectron yield in scintillation counters,” in Proc. Nucl. Electronics Symp., Versailles, France, 1968.


[17] M. A. Spurrier, P. Supryczynski, K. Yang, A. A. Carey, and C. L. Melcher, “Effects of Ca Co-doping on the scintillation properties of LSO:Ce,” IEEE Trans. Nucl. Sci., vol. 55, no. 3, pp. 1178–1182, Jun. 2008. [18] M. Moszyn´ski, “Inorganic scintillators in -spectrometry,” Nucl. Instrum. Methods Phys. Res. A, vol. A505, no. 1–2, pp. 101–110, Jun. 2003. [19] J. Baszak, Hamamatsu Photonics K. K., private communication. [20] P. Lavoute, Photonis Inc. to M. Moszynski, private communication. [21] Photonis, Photomultipliers Tubes, Principles and Applications S.-O. Flyckt and C. Marmonier, Eds., 2002. [22] B. Bengtson and M. Moszyn´ski, “Timing improved by the use of dynode signals studied with different scintillators and photomultipliers,” Nucl. Instrum. Methods, vol. 204, no. 1, pp. 129–140, Dec. 1982. [23] M. Moszyn´ski, “Prospects for new fast photomultipliers,” Nucl. Instrum. Methods Phys. Res. A, vol. A337, no. 1, pp. 154–164, Dec. 1993. [24] A. J. Person, Photomultiplier Space-Charge Effects and Transit-Time Spread ET Application Note R/P064, ET Enterprises Ltd., Uxbridge, U.K. [25] R. E. McHose, Time Characteristics of Photomultipliers—Some General Observations, Burle TP-114 Application Note, Burle Industries Inc., Lancaster, PA.