Gamma-ray Spectroscopy with LaBr 3 :Ce Scintillator Readout by a Silicon Drift Detector C. Fiorini, member, IEEE, A. Gola, M. Zanchi, A. Longoni, P. Lechner, H. Soltau, L. Strüder Abstract In this work we propose a gamma-ray spectrometer based on a LaBr 3 :Ce scintillator coupled to a Silicon Drift Detector (SDD). The SDD is a photodetector characterized by a very low noise thanks to the low value of output capacitance independent from the active area. With respect to a PMT, the SDD offers a higher quantum efficiency which reduces the spread associated to the statistic of photoelectrons generation. Also with respect to an APD, the SDD offers a lower photoelectrons statistic contribution, which, in the APD, is worsened by the excess noise factor with respect to pure Poisson statistics. Moreover, the SDD has a stable behaviour, less sensitive to temperature and bias drift. In the past years, good energy resolutions were measured using a SDD coupled to a CsI:Tl crystal. However, the long shaping time, to be used with this scintillator to prevent ballistic deficit, was too far to exploit the best noise performances achievable with a SDD obtained at shaping times in the order of 1us. On the contrary, this optimum shaping time is fully compatible with the short decay time of the LaBr 3 :Ce crystal (about 25ns). The results of the experimental characterization of the LaBr 3 :Ce-SDD gamma-ray spectrometer are presented in this work and are compared with the performances achieved with the same crystal coupled to a PMT and to a CsI(Tl) crystal coupled to the same SDD. The SDD has an active area of 30mm 2. Antireflective coatings have been implemented. Good energy resolutions were measured at room temperature, thanks to the low leakage current of the detector: 2.7% at the 137 Cs 661.7 KeV line and 6.1% at the 57 Co 122 KeV line. A resolution of 5.7% at 122 KeV line was measured at 0 C. T I. INTRODUCTION he LaBr 3 :Ce scintillator has been recently introduced as a very valuable alternative to the most conventional scintillators [1-4]. This crystal, in fact, is characterized by a 5.3 g/cm 3 density, a very high light output (> 60000 photons/mev), of the same order of the CsI(Tl), and at the same time by a very short decay time (< 25ns), at the same level of LSO and much better than NaI(Tl). LaBr3:Ce crystals have emission peaks at 360 nm and 380 nm, very close to the wavelength of peak detection efficiency of PMTs and still in the range of detection efficiency of most solid state photodetectors. The light output is essentially constant in a wide range of temperatures. Saint Gobain has recently measured a very little change of light yield (less than 1%) with its LaBr3:Ce samples in the range of -10 C to 50 C All these features make this scintillator very attractive for both space Manuscript received November 11, 2005. C. Fiorini, A. Gola, M. Zanchi and A. Longoni are with the Politecnico di Milano Dipartimento di Elettronica e Informazione, Milano, Italy and with the INFN, Sezione di Milano, Milano, Italy. P. Lechner and H. Soltau are with PNSensor GmbH, Munich, Germany. L. Strüder is with MPI Semiconductor Laboratory, Munich, Germany. applications (potentially also as an alternative to CZT detectors) [5] and for medical imaging [6]. To fully exploit the best performances of a good crystal as the LaBr3:Ce, also the choice of the best photodetector for the scintillation light is of relevant importance. For this reason let us consider the following well-known equation, which gives the expression of the energy resolution, for a generic γ-ray spectroscopy system composed by a scintillator and a photodetector: ΔE E = 2.35 2 ENC 2 M ( ηn ) ph 2 α + ηn ph ΔE + E 2 np, inh where ENC is the equivalent noise charge of the detector, η is its quantum efficiency, M the internal gain (if present), α is a worsening factor due to the statistics of the multiplication mechanism, N ph is the number of photons generated in the scintillator. Unitary light collection efficiency has been considered. The first member is the electronic noise contribution (R el ), the second is the statistical contribution (R st ) and the last is the intrinsic contribution due to the scintillator and the assembly non-idealities (R intr ). Typical values for the parameters of (1) for the different photodetectors that are considered here are given in Table 1. Photomultiplier tubes (PMTs) are still the most used devices for scintillator readout. The high gain of the PMT makes the contribution of the electronics noise R el practically negligible, however the statistical contribution R st plays an important role. In fact the disadvantage of the PMT is represented by its low quantum efficiency η (in the order of 25%) which results in a photoelectrons generation spread R st not optimal with respect to the intrinsic capabilities of the scintillator. Moreover, the mentioned contribution is worsened from pure Poisson statistics by the statistics of the multiplication itself (α ~1.25). A possible alternative to the PMTs is represented by the Avalanche Photodiode (APD) [7]. It combines the high η of a Silicon photodiode, which affects both the electronic and statistical contributions, with the benefits of avalanche multiplication which helps to keep R el low. However, also in the case of APDs, the statistical component to the resolution R st is affected by the statistics of the multiplication itself: α = F, where F is the excess noise factor (>2). Moreover, as a practical drawback, APDs show a high sensitivity of the gain and η to temperature and bias variations as well as the need to be operated at high voltages. Another approach is to use the Silicon Drift Detector (SDD), which has a low electronic noise, thanks to the low (1)
value of output capacitance. In this case R el can be significantly low, without using an internal multiplication mechanism. The statistical contribution is kept close to the pure Poisson limit thanks to both the high quantum efficiency of the photodetector, and by the absence of the multiplication, so that α = 1 (see Table 1). If we plot the energy resolution vs. γ-ray energy, as shown in Fig.1, there is a low-energy region where the electronic contribution is expected to be dominant, if present, like in APDs and SDDs; then we find a region dominated by the statistical contribution, and in this case the SDD is expected to have better energy resolution performances with respect to the other photodetectors. Finally there is a region where the intrinsic contribution of the scintillator is dominant. In this case, the choice of the photodetector has no significant effect. II. THE SDD DETECTOR The Silicon Drift Detector is a silicon detector, characterized by a particular charge collection mechanism, such as the electrons generated in the active area of the device drift towards the anode, which has a small size and a very small capacitance, independent from the total active area of the device. Equation 2 expresses the ENC of the detector, as a function of the shaping time used for signal processing. 2 2 A 1 ENC = C T a + A2 2 πa f + A3τb τ (2) s In this equation C T is the value of the total anode capacitance (detector + preamplifier + parasitic), τ s is the shaping time, a is the white series noise, a f is the 1/f noise coefficient and b is the white parallel noise due to the leakage current, which can be reduced by cooling. From (2) it can be seen that a small value of C T reduces the electronic noise without the need of a multiplication mechanism. Moreover, as the ENC is dependent on the shaping time, there is an optimum value for this parameter, which is usually of the order of 1 μs for the SDDs (this value is dependent on the operating temperature). SDDs used for scintillation readout have already demonstrated to achieve state-of-the-art energy resolution in γ-ray spectroscopy using a CsI:Tl scintillator [8] as well as sub-millimetre position resolution in γ-ray imaging [9]. However, the use of a slow scintillator like CsI:Tl (decay time ~1 μs) did not allow to fully exploit the best noise performances offered by the SDD, which are obtained using shaping times of the order of 1 μs. Longer shaping times were needed with CsI:Tl for not being excessively affected by ballistic deficit. In this case, the electronic noise of the SDD was dominated by the contribution of the leakage current (white parallel noise, see (2)) which had to be reduced by cooling. The recently introduced LaBr 3: Ce scintillator has a much shorter decay time, of the order of 25 ns, which does not generate ballistic deficit, even when used at shaping times of 1 μs or less. This benefit has however to be considered together with the collection mechanism of the scintillation-generated charge inside the SDD. In fact, the drift time necessary for the charge to reach the anode of the photodetector prevents the use of a too short shaping time. For the device presented in the next section, the drift time from the outermost region of the active area is estimated to be about 0.7μs. III. DETECTOR CHARACTERIZATION The detector we have used for the measurements is a circular SDD, with an active area of 30mm 2, produced at the Semiconductor Laboratory of Max Planck Institut in Munich. The light enters from the bottom side of the detector, which is made by an homogeneous entrance window. The input JFET of the readout electronics has been integrated in the detector itself, in order to further reduce the total anode capacitance, and consequently the electronic noise at short shaping times. Recent technological improvements in the fabrication process of the SDD include a reduced leakage current, of about 200pA/cm 2 at 25 C, which makes the device very attractive for measurements near room temperature. Another feature is the presence of the bonding pads outside the active area; a layout of the chip is shown in Fig. 2. The principle schematic of the readout electronics used in the measurements with the SDD is shown in Fig.3. The JFET integrated on the detector is operated in a source follower configuration with an external current source. The voltage signal is obtained by the conversion of the charge signal into voltage on the total capacitance C d connected at the detector output. The signal is amplified by means of a voltage amplifier. This is realized by means of a coupling capacitor followed by a charge preamplifer. The voltage gain on the signal step is given by C K /C F while the decay time of the preamplifier waveform is equals to R F C F. The signal is further processed by a Tennelec TC244 shaping amplifier. The electronic noise of the device have been characterized by directly irradiating the device with an X-ray source of 55 Fe (5.9 KeV Mn Kα), at different temperatures, ranging from - 10 C to 23 C. The optimum energy resolution achieved was 145 ev (9.8 e - rms), obtained at -10 C and with a shaping time of 1 μs. At room temperature we obtained 237 ev at 0.25 μs. The corresponding spectra are shown in Fig. 4. If we plot the measured electronics noise with respect to the shaping time we find a behavior, in accordance to (2), which is shown in Fig. 5. It is also worth mentioning that the shaping times compatible with LaBr 3: Ce correspond roughly to the minimum of the ENC curve, while the CsI:Tl lays in a region with higher noise, especially at room temperature. The entrance window of the detector was also provided with custom anti-reflective coatings (ARCs). The presence of these coatings is important, because they improve the Quantum Efficiency η at the wavelengths of interest, increasing the collected signal charge, and, from (1), this has positive effects both on the electronic contribution R el and on the statistical contribution R st. A plot of the measured η vs.
incident wavelength for these ARCs and for standard coatings is shown in Fig. 6. The ARC coatings were originally designed to have a maximum efficiency at λ = 565 nm (CsI:Tl), where we have measured η = 90%, but still have η = 80% at λ = 400nm (the lower end of our QE measurement sensitivity). IV. γ-ray SPECTROSCOPY SET-UP AND MEASUREMENTS For γ-ray measurements the SDD was optically coupled to a LaBr 3 :Ce Brillance 380 crystal module, produced by Saint Gobain. The crystal has a cylindrical shape with 5mm diameter and 5mm thickness. No specifications were available on the Ce doping of the crystal and on the light output of the assembled module. The crystal was already mounted with a light reflector and sealed in an aluminium housing. A picture of the scintillator coupled to the SDD photodetector is shown in Fig. 7. The back side of the detector was biased at -86 V, while the last ring, providing the charge drift mechanism, was biased at about -110V. The first measurement was done to evaluate the conversion gain between the energy of the γ-photon and the number of photoelectrons. Using the data from the X-ray measurements, we obtained a conversion gain of about 28 e - /KeV. Assuming, from the data sheets, a light yield of the scintillator of 60 ph/kev and extrapolating from the curve shown in Fig. 6 a value for η of about 70% at λ = 370 nm (the central wavelength of the emission spectrum of LaBr 3 :Ce), we can evaluate a light collection efficiency of about 70%. This means that there is still room for improvements, both in the assembly and in η, which should be tuned to match λ = 370 nm. With the system described, we carried out a set of spectroscopic measurements at room temperature, with different gamma sources. Fig. 8 shows the spectrum obtained irradiating the scintillator with a 137 Cs source. The energy resolution is 2.7% and the 662 kev escape peak, as shown in the expanded region plot on the right side, is clearly distinguishable. As far as the electronic noise contribution is already known from the X-ray spectroscopic measurements, and the statistical contribution can be obtained from the conversion gain, it is possible to evaluate from (1) the intrinsic contribution, due to the crystal and to the assembly, which is 2% in this case. This value is quite high, if compared to the one measured by Shah and co-workers [7], who estimated the same contribution to be of 0.9%, using a LaBr 3 :Ce scintillator by RMD, coupled to a 8mm 2 APD cooled at 250K. It is worth mentioning that the same scintillator was used with a PMT by the manufacturer, and an energy resolution of 3% have been measured. This is actually not among the best values measured with a LaBr 3 :Ce scintillator. The same manufacturer reports resolutions as good as 2.8% with other samples. We could estimate the potential improvement in resolution using the SDD photodetector with a LaBr 3 :Ce scintillator characterized by a better intrinsic contribution. For instance, considering the intrinsic resolution of 0.9% quoted by RMD and using the R el and R st contributions for the SDD quoted in the first column of Table II, a resolution of 2.0% at 662keV can be estimated. In Fig. 9, the spectrum of a 57 Co source measured at room temperature is shown. The energy resolution () at the 122keV peak is of 6.1%. In this spectrum, the peak at 14 kev is well distinguished from the noise threshold. The same measurement, carried out with the PMT by the manufacturer, has given a value of 6.5%. In this case the electronic noise is no longer negligible; to improve the results we repeated the measurements with a moderate cooling, operating the system at 0 C. The corresponding spectrum, reported in Fig. 10, shows an energy resolution of 5.7%. The energy resolutions measured with the SDD and PMT are listed in Table 2, together with the estimated contributions. Combining these measurements with others, made with a 241 Am source, it is possible to draw a plot of the energy resolution vs. energy, shown in Fig. 11, which fits the experimental data. It is also interesting to compare the results obtained at 122 kev with LaBr 3 :Ce with the ones obtained with a CsI:Tl crystal, using the same SDD as a photodetector. The comparison is shown in Fig. 12. The main difference between the two scintillators is that CsI:Tl has a much longer decay time which forces to use the SDD at longer shaping times and with a higher electronic noise, as anticipated in section II. The energy resolution of the LaBr 3 :Ce crystal readout by the PMT is also reported. V. CONCLUSIONS In conclusion the measurements have shown that the SDD coupled to a LaBr 3 :Ce scintillator is a good photodetector for γ-ray spectroscopy, and can obtain better results than a PMT. The performances achieved are due to the low electronic noise of the SDD and to an improved photodetection efficiency, but there is still room for improvements, especially in the optimization of the detection module. Applications in medical imaging and γ-ray spectrometry are foreseen in the near future. VI. ACKNOWLEDGMENTS The authors would like to thank Mike Mayhugh (Saint Gobain) for his support. REFERENCES [1] E.V.D.van Loef, P.Dorenbos, C.W.E. van Eijk, K.Krämer, H.U.Güdel, Appl.Phys.Lett., vol.79, p. 1573, 2001. [2] P. Dorenbos, Nucl. Instr. Meth., vol. A 486, pp. 208-213, 2002. [3] E. V. D. van Loef, et al., Nucl. Instr. Meth., vol. A 486, pp. 254-258, 2002. [4] P. Dorenbos, J. T. M. de Haas, C. W. E. Van Eijk, IEEE Trans. Nucl. Sci., vol.51, n.3, p.1289, 2004. [5] M.L.McConnel, et al., SPIE 5488, 2004. [6] W.Moses, K.S.Shah, Nucl. Instr. Meth., vol. A 537, pp. 317-320, 2005. [7] K.S.Shah, et al., IEEE Trans. Nucl. Sci., vol.51, n.5, p.2395, 2004.
[8] C. Fiorini, et al., IEEE Trans. Nucl. Sci., vol. 44, no. 6, pp. 2553-2560, 1997. [9] C.Fiorini, F.Perotti, Review of Scientific Instruments, 76, 044303, 2005.
TABLE I PHOTODETECTOR TYPICAL PARAMETERS Parameter PMT APD SDD η ~25% >80% >80% α 1+υ (~1.25) F (>2) 1 M ~ 10 6 ~ 10 3 1 TABLE II DIFFERENT CONTRIBUTIONS TO THE ENERGY RESOLUTION SDD 137 Cs T amb PMT 137 Cs T amb SDD 57 Co T amb PMT 57 Co T amb SDD 57 Co T = 0 C R el 0.5% ~0% 2.9% ~0% 1.9% R st 1.7% 2.2% 4.0% 5.4% 4.0% R intr 2% 2% 3.6% 3.6% 3.6% R tot 2.7% 3% 6.1% 6.5% 5.7% Energy resolution (%) 10 electronics noise total statistics scintillator non idealities 1 10 100 1000 Energy (kev) Fig. 1. Plot of the energy resolution vs. incoming γ-ray energy, for a system composed of a scintillator crystal coupled to a photodetector. It is possible to distinguish three regions, each one dominated by a particular contribution to the resolution. In the region on the right side of the graph, above few hundreds of kev, scintillator non idealities dominates with respect to the other contributions. Fig. 2. Layout of the Silicon Drift Detector used for the γ-ray measurement. On the left side of the figure one can see the bonding pads, which are external to the active area and are connected to the central region of the device.
Fig. 3. Principle schematic of the readout electronics used in the measurements. 4000 55 Fe spectrum 5.9 kev 5000 55 Fe spectrum 5.9 kev Counts 3000 2000 237eV Counts 4000 3000 2000 145eV 1000 1000 4000 5000 6000 7000 Energy [ev] 4000 5000 6000 7000 Energy [ev] Fig. 4. Energy spectra of a 55 Fe source, measured at room temperature (left) with 0.25μs shaping time and at -10 C (right) with 1μs shaping time. The longer shaping time used at -10 C is used thanks to the reduced leakage current (see also Fig. 4).
Fig. 5. ENC curve of the 30 mm 2 SDD measured at different temperatures. In this figure the shaping time range to be used with two different scintillators, due to their decay times, are also shown. Fig. 6. Graph of the Quantum Efficiency η measured with the 30 mm 2 SDD detector, with custom and standard anti-reflection coatings. Fig. 7. Picture of the circular SDD coupled to the LaBr 3:Ce module. The detector is placed on a ceramic substrate for biasing and signal extraction. 2x10 6 137 Cs spectrum 661.7 kev 6x10 4 661.7 kev 1x10 6 4x10 4 2.7% escape 5x10 5 2x10 4 32 kev Ba X-rays 0 200 400 600 E [k V] 600 650 700 Energy [kev]
Fig. 8. Spectrum of a 137 Cs source, measured at room temperature (23 C). In the left side of the figure, the same spectrum is shown in two different vertical scales to better show both 32keV X-ray lines and 662keV line. The escape peak of the 661.7 KeV line (shown in the expanded region on the right side of the figure) is clearly distinguishable and the noise threshold is well below the 32 KeV X-ray line. 2000 57 Co spectrum 122 kev Counts 1500 1000 6.1% 500 0 14 kev Escape 136.4 kev 0 40 80 120 160 Energy [kev] Fig. 9. Spectrum of a 57 Co source, measured at room temperature (23 C). The 14 KeV line is clearly distinguished from noise. Laescape peak is visible in figure. 2000 57 Co spectrum 122 kev Counts 1500 1000 5.7% 500 0 14 kev Escape 136.4 kev 0 40 80 120 160 Energy [kev] Fig. 10. Same measurements of Fig. 8, but made at a lower temperature, in order to reduce the electronic noise contribution. The energy resolution has improved to 5.7%.
Fig. 11. Plot of the energy resolution vs. energy for the 30 mm 2 SDD used for γ-ray measurements. Fig. 12. Comparison between the energy resolution obtained at different shaping times for LaBr 3 :Ce and CsI:Tl. Also the measurement with the LaBr 3:Ce crystal using a PMT is reported.