1842 IEEE TRANSACTIONS ON NUCLEAR SCIENCE, VOL. 52, NO. 5, OCTOBER M. Marisaldi, C. Labanti, H. Soltau, C. Fiorini, A. Longoni, and F.

Transcription

1 1842 IEEE TRANSACTIONS ON NUCLEAR SCIENCE, VOL. 52, NO. 5, OCTOBER 2005 X- and Gamma-Ray Detection With a Silicon Drift Detector Coupled to a CsI(Tl) Scintillator Operated With Pulse Shape Discrimination Technique M. Marisaldi, C. Labanti, H. Soltau, C. Fiorini, A. Longoni, and F. Perotti Abstract A monolithic X- and gamma-ray detector based on a silicon drift detector coupled to a CsI(Tl) scintillator is presented. The detector is operated both as a direct X-ray detector for photons interacting in silicon and as an indirect detector for photons interacting in the scintillator. As interactions in silicon and in CsI yield different amount of charge per unit energy deposited, discrimination of the place of interaction is necessary to obtain the correct energy. Discrimination of the interaction type is carried out by means of pulse shape discrimination performed with two parallel processing chains with different shaping times. In this paper an extensive characterization of the detector with temperature is presented. It is shown that cooling the detector at 10 C allows to obtain a nearly 100% efficiency between 8 and 200 kev. Further cooling below 0 C allows pulse shape discrimination with 100% accuracy throughout the detector s energy range. The detector has also been tested with X-rays at various energies by means of a tunable X-ray facility. These tests allowed an investigation of the light yield nonproportionality in CsI(Tl) at low energies, necessary for a proper energy calibration of the detector. Index Terms Gamma-ray detectors, pulse shape discrimination, scintillation detectors, silicon drift detectors, X-ray detectors. I. INTRODUCTION AGAMMA-ray detector based on a CsI(Tl) scintillating crystal coupled to a silicon photodetector is also a direct X-ray detector, for radiation interacting in silicon. Charge pulses either from X-ray interaction in silicon or scintillation light collection are quite different in shape due to the different timing properties and can be discriminated easily. In fact, while the electron-hole pair creation from X-ray interaction in silicon originates a fast signal (about 10 ns rise time), the scintillation light collection is dominated by the fluorescence states de-excitation time [0.68 s (64%) and 3.34 s (36%) for CsI(Tl) at room temperature] and a few s shaping time is needed in this case to avoid significant ballistic deficit. Moreover, interactions in silicon and in CsI yield different amount of charge per unit energy deposited, so discrimination of the place of interaction is necessary to obtain the correct energy. Pulse shape discrimination (PSD) techniques allow discrimination between Manuscript received November 15, 2004; revised March 15, M. Marisaldi and C. Labanti are with the CNR-INAF/IASF Istituto di Astrofisica Spaziale e Fisica Cosmica, Bologna, Italy ( marisaldi@bo.iasf.cnr.it). H. Soltau is with PNSensor GmbH, München, Germany. C. Fiorini and A. Longoni are with Politecnico di Milano, Dipartimento di Elettronica e Informazione, Milano, Italy. F. Perotti is with the CNR-INAF/IASF Istituto di Astrofisica Spaziale e Fisica cosmica, Milano, Italy. Digital Object Identifier /TNS signals with different timing properties. Detectors based on a CsI(Tl) crystal coupled to a silicon photodetector operated with PSD have already been described: in [1] and [2] it was proposed as a detector for charged particle discrimination, while in [3] it was proposed as a X/gamma-ray instrument. However, in all these applications the silicon photodetector used was a p-i-n photodiode (PD). Since the energy threshold and the spectroscopic capabilities of such a detector are dominated by the electronic noise of the photodetector, it was decided to use a silicon drift detector (SDD) [4] instead of a traditional p-i-n PD, because of its much lower intrinsic electronic noise. The first experimental results obtained with this detector s concept have been published in a previous paper [5]. The energy threshold obtained was 1.5 kev for the direct X-ray detector and 16 kev for the scintillation detector, at room temperature. Energy resolution FWHM for X-ray interaction in silicon was 1.2% at kev, while energy resolution for gamma-rays interacting in CsI was 6.9% at 662 kev, with a corresponding light yield of 23.4 e-/kev. A nonlinearity in the light yield at low energy has been observed. This kind of detector could find application in space-based high energy astrophysics instruments. Simultaneous X and gamma-ray detection with a single, compact device could be a valuable solution to fulfill the strict weight and volume constraints of a space mission. For example, this detector could be employed as the detector unit (pixel) of a coded-mask based wide field X/gamma-ray transient monitor [6]. The novel contribution of this approach is the possibility to perform quite good spectroscopy in the kev region in the very early phases of a transient event, such as a gamma-ray burst (GRB), before the eventual re-pointing of a narrow field instrument. The detector s energy range extended up to a few hundreds kev would allow a prompt discrimination between GRB s and other softer X-ray transients and permit suitable observation strategies. II. EXPERIMENTAL SETUP A 10-mm active area SDD with integrated JFET supplied by PNSensor GmbH was used. Even if such detectors have been developed mainly as direct X-ray detectors [7], they have been also widely tested as photodetectors coupled to CsI(Tl) scintillators [8] [11]. The fully depleted active thickness of the device is 300 m. The chip was bonded on a custom made ceramic housing for easy handling and electrical connection. To allow bonding on the electronic side, the ceramic housing /$ IEEE

2 MARISALDI et al.: X- AND GAMMA-RAY DETECTION WITH A SILICON DRIFT DETECTOR 1843 III. RESULTS Fig. 1. Schematic view of the experimental setup. was designed with a hole corresponding to the integrated JFET. A CsI(Tl) cylindrical crystal, 10-mm high and 3.6 mm in diameter, was coupled to the detector s entrance window by means of a thin layer of previously cured siliconic resin. The crystal was wrapped with multi-layer polymeric film [12] and PTFE tape. A PTFE tool was used to keep the crystal in the proper position and to accurately press the crystal onto the transparent siliconic layer. Several methods are available to carry out pulse shape discrimination (see, for example, [13]). In this application the preamplified signal was sent to two different processing chains each of them comprising a spectroscopy amplifier with quasigaussian shaping and a 4096 channels ADC. Shaping times were set to 0.5 s in one processing chain (tailored to the detection of X-rays interaction in silicon, hereafter called the fast chain) and 3 s in the other (tailored to the detection of scintillation light, hereafter called the slow chain). The gain has been set to about the same value in the two processing chains. A threshold was set independently for each ADC by means of their low level discriminator setting. A coincidence signal was used to enable the digital conversion only when both discriminators have triggered. For each event, the two ADC values were saved on hard disk for subsequent analysis. A schematic view of the experimental setup is shown in Fig. 1. A detailed description of the electronics setup used can be found in [5]. The detector was irradiated with gamma-ray sources from the electronic side, in such a way that photons pass through the silicon detector first and then, if not absorbed, through the scintillator. The hole in the ceramic housing allows gamma-rays to reach the silicon detector without attenuation. For tests with radioactive sources, the detector has been placed in a climatic chamber and measurements have been performed at temperatures between C and C. Tests have also been carried out at room temperature by means of a tunable X-ray machine. The beam emitted by a commercial X-ray tube is collimated and selected in energy by means of Bragg diffraction through a LiF crystal. The crystal can be rotated with respect to the incoming photon beam direction, the rotation angle being a measure of the selected energy. A rotating platform allows positioning of the detector in front of the diffracted beam. A set of slits and collimators allows to obtain an X-ray beam with energies between 15 and 200 kev and a limited energy spread. The intrinsic energy spread of the beam was found to give rise only to a minor contribution to the detector s energy response, so, for our purposes, the X-ray beam produced can be considered monochromatic. A. Detector Performance The spectra obtained can be visualized as bidimensional histograms as shown in Fig. 2, where the two axes are the fast and slow pulse height, for measurements recorded at C. As expected, events distribute along two straight lines depending whether interaction takes place in silicon or in CsI. For each event, the key parameter for PSD is the ratio between the pulse height of the fast and slow processing chains, expressed in channels and corrected for ADC offset For X-rays interacting in silicon at C (it would be 1 if the gain of the two chains was exactly the same) and for gamma-rays interacting in CsI. and correspond roughly to the angular coefficients of the two straight lines shown in the bidimensional spectra of Fig. 2. The evident displaced peaks between ADC channel 800 and 1100, concerning the line corresponding to interaction in silicon in subfigure (a), are due to mixed interactions, i.e., events where part of the signal charge is provided by a gamma-ray interaction in the scintillator and part comes from interaction of a cesium or iodine characteristic X-ray interacting in silicon. The timing of such a pulse does not correspond to that of a pure interaction in CsI or in silicon, and the corresponding value lies in between and. A detailed description of this effect is reported in [5]. Table I summarizes some of the main detector s characteristics at different temperatures. The detector s equivalent noise charge (ENC) has been obtained fitting the Np fluorescence X-ray peak at kev, assuming that the energy resolution is given by the sum in quadrature of the contributions due to electronic noise and statistical spread. As expected, the slow chain exhibits a higher, strongly temperature-dependent, ENC than the fast chain, as the 3- s shaping time used is higher than the optimal one for ENC minimization, and so the parallel noise contribution is enhanced. Nevertheless, such a shaping time is required to avoid significant ballistic deficit in scintillation light collection because of the long CsI(Tl) scintillation decay times. The energy resolution at kev in silicon and at 662 kev in CsI have been calculated using the fast and slow chain spectra respectively. The light yield (LY) is defined as the charge collected per unit energy deposited in CsI. A temperature dependence of light yield is also evident in Table I, as expected from CsI(Tl) scintillation properties [14], [15]. Even if we used a3 s shaping time for the slow chain at all temperatures to allow comparison of the detector s PSD capabilities at different temperatures with a uniform setup, we are aware that at C a higher shaping time in the slow chain would allow a higher light yield at the price of a limited increase in the electronic noise [9]. A tradeoff could be performed to choose the more appropriate combination of operating temperature and slow chain shaping time. Table II shows the energy thresholds of the slow chain for interaction in silicon and in CsI at different temperatures. The energy threshold has been calculated

3 1844 IEEE TRANSACTIONS ON NUCLEAR SCIENCE, VOL. 52, NO. 5, OCTOBER 2005 Fig. 2. Bidimensional spectra obtained with Am (a) and Cs (b) sources at 020 C. Each image bin is square channels. Channel values have been corrected for ADC offset. Events in silicon in the Cs spectrum are due to the X-ray background and the Ba K X-rays. requiring the signal to be higher than five times the ENC level (see Table I). As the system requires both fast and slow chain signals above threshold for data acquisition and PSD, the whole detector threshold is limited by the slow chain s electronic noise, so the fast chain s energy thresholds have not been reported. has been calculated using the light yield value obtained with the 32-keV Ba fluorescence X-ray lines. Parameters and reported in Table II concern PSD capabilities and will be discussed in Section III-B. B. Pulse Shape Discrimination Capabilities The distribution of the parameter introduced in Section III-A can be fit with a Gaussian distribution, as shown in Fig. 3, where the distribution for a portion of a Am spectrum with the corresponding Gaussian fit is reported. The distribution s width depends on the pulse height interval considered in calculations. In fact, as pulse height approaches lower values, energy resolution worsens and the width of the distribution widens. Fig. 4 shows the centroid and the width of the distribution for interactions in silicon and in CsI at C as a function of the slow chain pulse height. Different figures of merit can be defined to establish the PSD capabilities of a detector system. In this paper we will consider the parameter defined as the ratio between and the sum of the FWHM and of the distributions [16] Provided the distributions are Gaussians, as in this case, events discrimination with 100% accuracy can be achieved if [17]. As the width of the distributions is dominated by the electronic noise, the parameter for a given energy range is improved with noise reduction, i.e., by cooling the device. Of course, also the energy threshold for 100% disentanglement is lowered by device cooling. Fig. 5 shows the parameter as a function of the slow chain pulse height, for different temperatures. As was expected, becomes higher as the temperature gets lower, for a given slow chain pulse height interval. As the temperature is reduced, also the threshold for (1)

4 MARISALDI et al.: X- AND GAMMA-RAY DETECTION WITH A SILICON DRIFT DETECTOR 1845 TABLE I MAIN DETECTOR S PROPERTIES AT DIFFERENT TEMPERATURES TABLE II ENERGY THRESHOLD IN Si AND CsI AT DIFFERENT TEMPERATURES AND PSD CAPABILITIES Fig. 4. r distribution centroids for the Si and CsI peaks versus the slow channel pulse height at 020 C. Each section is 20 channels wide (corresponding to 0.7 kev in silicon and 6.7 kev in CsI). Error bars are 61 FWHM. Fig. 3. r distribution for Am events with energies between kev for interactions in silicon, and kev for interactions in CsI at room temperature, from [5]. r is the ratio between fast and slow pulse height. 100% event discrimination gets lower. Table II shows the minimum energy required to have for interactions in silicon and in CsI at different temperatures. A decrease in the value can be observed in Fig. 5 between slow chain pulse height 200 and 250 ADC channels, for all temperatures but 10 C. This pulse height range corresponds to energies just above the kev peak in CsI. If a close look is taken to the Am bidimensional spectrum in this energy range, a group of events whose parameter deviates from the expected linear behavior can be observed. These events are possibly due to mixed interactions where a kev photon interacts in silicon by means of Compton scattering and the scattered photon Fig. 5. The figure of merit M versus the slow channel pulse height for different temperatures. The error on M is lower than the width of symbols. is absorbed in the scintillator. Thus, the signal charge is higher than that corresponding to complete absorption in CsI, as part of the energy is directly released in silicon. In this process only forward scattering is involved; it can be seen from Compton kinematics and simple geometrical considerations that only photons with energies from kev to 54.5 kev can be detected in CsI, so maximum 5 kev can be directly released in silicon and a continuum of events with energies between these limits should

5 1846 IEEE TRANSACTIONS ON NUCLEAR SCIENCE, VOL. 52, NO. 5, OCTOBER 2005 Fig. 6. r distributions for a 21 kev beam in the detector. In plot (b) higher order peaks can be seen. Such peaks are not clearly evident in plot (a) because of the limited resolution of the gray-scale. be observed. The deviation above the kev peak observed in the Am bidimensional spectrum is compatible with such a process. If we consider, for example, events at C between slow chain pulse height 200 and 220 ADC channels, there is an overlap between the distributions of the events in CsI and of this mixed interaction. The resulting distribution has a higher standard deviation and so the value is reduced. For higher ADC channel intervals the two distributions separate and the expected behavior of is restored. At 10 C this effect is not evident because of the higher light yield obtained at this temperature. In fact, the mixed events distribution are displaced in such a way that the two distributions can be easily discriminated for ADC channels higher than 200 and no superposition effect takes place. Other physical processes, such as Compton scattering in the scintillator or electron escape are not compatible with the observed experimental results. Anyway, it must be noted that this effect is not a concern for the detector s PSD cababilities, as in any case remains higher than 1.5. C. Light Yield Nonproportionality Light yield nonproportionality in CsI(Tl) crystals for photon energies lower than about 100 kev has been reported in literature [18], where a light yield maximum at about 10 kev has been observed, with a value about 15% higher than that obtained at 662 kev. A consequence of this behavior is the CsI(Tl) intrinsic energy resolution that is worse than the theoretical limit set by counting statistics [19]. As our detector is expected to work as a scintillation detector for energies as low as 10 kev, a characterization of the light yield nonproportionality is required for proper detector calibration. For this purpose, the detector has been tested with a tunable X-ray machine, as described in Section II. For energies lower than about 60 kev, the peak for direct interaction in silicon is evident, so the beam energy can be obtained using the electronic chains calibration previously performed with a Am source. The obtained values are in agreement with those calculated from the Bragg law within 1%. Fig. 6 shows the parameter distribution for a 21 kev beam, where both peaks regarding interaction in silicon and in CsI are evident. Several measurements at different energies have been performed at room temperature. For each energy, light yield in electrons/kev has been calculated. The obtained LY values are plot-ted in Fig. 7 normalized to unity at 662 kev. A light yield increase up to 20% with respect to that obtained at 662 kev is observed at 20 kev. Two maxima for energies about 50 kev and lower than 20 kev can be observed. This behavior is consistent with that reported in [18], and is ascribed to interactions of pho-

6 MARISALDI et al.: X- AND GAMMA-RAY DETECTION WITH A SILICON DRIFT DETECTOR 1847 Fig. 7. Light yield measured at different energies, at room temperature. Photon beams have been obtained with a tunable X-ray machine. Data are normalized to unity at 662 kev. The solid line represents only a linear interpolation of data. tons with CsI K- and L-shell electrons, leading to the production of photoelectrons with energies of about 10 kev, for which the light yield reaches a maximum. The obtained relative light yield is typically about 6% higher than that shown in [18]. It must be considered that literature values are related to measurements obtained with photomultiplier tubes (PMT), while in this case a solid state device with a sensitivity different from that of PMT s has been used. Further investigations concerning this issue are required. In order to address the possibility that the point of interaction in the scintillator may affect light yield, the detector was also tested as a typical scintillation detector, i.e., with photons incident directly on the scintillator, on the opposite side with respect to the photodetector. Low energy photons interact in the first layers of the crystal so, if either attenuation or inhomogeneity effects should be significant, different results should be obtained with standard and reversed configurations. Results were found to be consistent within 1% so the observed nonproportionality is reasonably an intrinsic characteristic of the scintillator, as expected. D. Energy Threshold and Detector Efficiency In Fig. 8 the detection efficiency for on-axis photons is shown. The energy thresholds of the slow chain for interaction in CsI at different temperatures are also shown. The efficiency for direct interaction in the SDD is limited at low energy by the dead layer thickness due to the integrated electronics and polarization rings. has been calculated by computing the absorption coefficients for all the dead layers involved (aluminum, silicon oxide, silicon nitride and nondepleted silicon) using the photon cross-sections provided by [20]. reaches a maximum at about 8 kev and then decreases due to the limited silicon thickness. The efficiency for the CsI crystal coupled to the SDD is limited at low energy by photon absorption in the SDD itself, and at high energy by the crystal thickness. About 25% efficiency at 1 MeV is obtained with a 1-cm thick crystal. A 500- m thick dead layer between SDD and the crystal has been considered in calculations to account for the optical coupling material. The detector total efficiency at energy will be Fig. 8. Expected detector efficiency for on-axis photons. Solid line: efficiency for interactions in silicon. Dashed line: efficiency for interactions in CsI. The K absorption edges for aluminum and silicon are evident. The energy thresholds of the slow chain for interactions in CsI at different temperatures are shown. for, and for.if is greater than about 8 kev, there is an efficiency gap between 8 kev and because of the decrease of. At room temperature, this efficiency gap extends between 8 and about 13 kev, while, by moderately cooling the device at 10 C the efficiency gap is eliminated, i.e., between 8 and about 200 kev. A further cooling does not improve the detector efficiency, despite other benefits in terms of energy resolution and PSD capabilities can be obtained. The low energy efficiency could be improved by a design of the SDD s electronics side tailored to this application, in order to reduce the dead layer to a minimum. Such an improvement would extend the detector s low energy range but would not affect neither its PSD capabilities, limited by noise considerations, nor the total efficiency for energies greater than about 8 kev. IV. CONCLUSION A X/gamma-ray detector based on a silicon drift detector coupled to a CsI(Tl) scintillator operated by means of pulse shape discrimination is described. The SDD works both as a direct X-ray detector for photons interacting in silicon, and as a photodetector for the scintillation light generated by photons interacting in the CsI crystal. Pulse shape discrimination is carried out by means of the comparison between signals provided by two parallel processing chains with different shaping time. The device has been succesfully tested at temperatures between C and 25 C. As expected, cooling improves both energy threshold, energy resolution for interaction in silicon, and PSD capabilities. A moderate cooling at 10 C lowers the threshold for interactions in CsI enough to fully exploit the efficiency of the CsI crystal and to fill the efficiency gap between silicon and the scintillator. In this way the detector s efficiency is nearly 100% between 8 and about 200 kev. These lower and upper limits are determined by the physical dimensions of the detector, namely by the SDD dead layer and by the crystal dimensions. In fact, it

7 1848 IEEE TRANSACTIONS ON NUCLEAR SCIENCE, VOL. 52, NO. 5, OCTOBER 2005 must be noted that the SDD used was not developed for this application, and an optimized design of the SDD electronic side in order to reduce the dead layers would result in higher efficiency in silicon at low energy. On the other hand, the use of a SDD with larger active area would allow the use of thicker crystals that, in turn, would result in higher efficiency at high energies. A further cooling does not improve the detector efficiency, despite other benefits in terms of energy resolution and PSD capabilities are obtained. In fact, cooling the device at 0 C allows PSD with 100% accuracy throughout about the detector s energy range. A further cooling at C evidences a decrease in the scintillation light yield, which in turn affects the energy resolution for interactions in the scintillator. CsI(Tl) scintillation light yield nonproportionality at low energies has been studied by means of a tunable X-ray facility, and quite a good agreement with previously published results has been found. As this instrument is expected to detect photons with energies of 15 kev and lower in the scintillator, this effect must be taken into account for proper detector calibration, so a thorough characterization must be performed at the chosen operating temperature. It must be noted that the efficiency gap filling and the good energy resolution and PSD performance can be obtained only by means of very low noise photodetectors such as SDD s. A traditional p-i-n PD with equivalent active area would show much worse performance due to the higher electronic noise. The proposed detector combines a low energy X-ray detector with good spectroscopic capabilities, with a gamma-ray detector with significant efficiency up to several hundreds kev in a single, compact device. Such a detector would be of great interest, especially for X- and gamma-ray astrophysics where space and weight optimization is a strong requirement. ACKNOWLEDGMENT The authors would like to to thank E. Rossi and A. Mauri at CNR-INAF/IASF, Bologna, for skillful electronics prototyping and help with crystal preparation. They also gratefully acknowledge P. Lechner at PNSensor GmbH, and L. Strüder at Max Planck Institut Halbleiterlabor, München, Germany. REFERENCES [1] G. Prete, G. Vedovato, G. Maron, D. Fabris, M. Poggi, G. Poggi, N. Taccetti, G. Pasquali, and F. Maletta, A logarithmic detection system for nuclear physics, Nucl. Instrum. Meth. A, vol. 315, pp , [2] J. Friese, A. Gillitzer, H. Korner, J. Reinhold, M. Peter, and M. Maier, The SISWICH, a detector telescope with intrinsic calibration, IEEE Trans. Nucl. Sci., pt. 1 2, vol. 40, no. 4, pp , Aug [3] A. Bird, T. Carter, Z. He, A. Dean, and D. Ramsden, Pulse shape analysis of signals from a CsI(Tl)/photodiode detector, Nucl. Instrum. Meth. A, vol. 353, pp , [4] E. Gatti and P. Rehak, Semiconductor drift chamber An application of a novel charge transport scheme, Nucl. Instrum. Meth., vol. 225, pp , [5] M. Marisaldi, C. Labanti, and H. Soltau, A pulse shape discrimination gamma-ray detector based on a silicon drift chamber coupled to a CsI(Tl) scintillator: prospects for a 1 kev-1 MeV monolithic detector, IEEE Trans. Nucl. Sci., pt. 2, vol. 51, no. 4, pp , Aug [6] M. Marisaldi and C. Labanti, Prospects for a 1 kev 1 MeV monolithic gamma-ray detector and possible application in X/gamma-ray astronomy, Mem. S. A. It. Suppl., vol. 5, pp , [7] P. Lechner, S. Eckbauer, R. Hartmann, S. Krisch, D. Hauff, R. Richter, H. Soltau, L. Strüder, C. Fiorini, E. Gatti, A. Longoni, and M. Sampietro, Silicon drift detectors for high resolution room temperature X-ray spectroscopy, Nucl. Instrum. Meth. A, vol. 377, pp , [8] C. Fiorini and F. Perotti, Scintillation detection using a silicon drift chamber with on-chip electronics, Nucl. Instrum. Meth. A, vol. 401, pp , [9] C. Fiorini, A. Longoni, F. Perotti, C. Labanti, P. Lechner, and L. Strüder, Gamma ray spectroscopy with CsI(Tl) scintillator coupled to silicon drift chamber, IEEE Trans. Nucl. Sci., pt. 3, vol. 44, no. 6, pp , [10] C. Fiorini, A. Longoni, F. Perotti, C. Labanti, P. Lechner, and L. Strüder, First experimental results of a new gamma-ray detector based on a silicon drift chamber coupled to a scintillator, Nucl. Instrum. Meth. A, vol. 409, pp , [11] C. Fiorini, A. Longoni, F. Perotti, C. Labanti, E. Rossi, P. Lechner, and L. Strüder, Detectors for high-resolution gamma-ray imaging based on a single CsI(Tl) scintillator coupled to an array of silicon drift detectors, IEEE Trans. Nucl. Sci., pt. 1, vol. 48, no. 3, pp , Jun [12] M. Weber, C. Stover, L. Gilbert, T. Nevitt, and A. Ouderkirk, Giant birefringent optics in multilayer polymer mirrors, Science, vol. 287, pp , [13] F. Frontera, D. Dal Fiume, G. Landini, E. Artina, M. Biserni, V. Chiaverini, F. Monzani, E. Costa, and R. C. Butler, Performances of the pulse shape electronics of the high energy experiment PDS on board the X-ray astronomy satellite SAX, IEEE Trans. Nucl. Sci., pt. 1 2, vol. 40, no. 4, pp , Aug [14] J. Valentine, W. Moses, S. Derenzo, D. Wehe, and G. Knoll, Temperature dependence of CsI(Tl) gamma-ray scintillation decay time constants and emission spectrum, Proc. SPIE Gamma-Ray Detectors, vol. 1734, [15] J. Valentine, D. Wehe, G. Knoll, and C. Moss, Temperature dependence of absolute CsI(Tl) scintillation yield, in Conf. Rec. IEEE Nuclear Science Symp. Medical Imaging Conf., 1991, pp [16] R. Winyard, J. Lutkin, and G. McBeth, Pulse shape discrimination in inorganic and organic scintillators. I, Nucl. Instrum. Meth., vol. 95, pp , [17] R. Winyard and G. McBeth, Pulse shape discrimination in inorganic and organic scintillators, II, Nucl. Instrum. Meth. A, vol. 98, pp , [18] W. Mengesha, T. Taulbee, B. Rooney, and J. Valentine, Light yield nonproportionality of CsI(Tl), CsI(Na), and YAP, IEEE Trans. Nucl. Sci., pt. 1, vol. 45, no. 3, pp , Jun [19] W. Moses, Current trends in scintillator detectors and materials, Nucl. Instrum. Meth. A, vol. 487, pp , [20] M. Berger, J. Hubbell, S. Seltzer, J. Coursey, and D. Zucker. (1999) XCOM: Photon Cross Section Database (Version 1.2) [Online]. Available: