We discuss the application of digital pulse shape algorithms to n/γ pulse shape discrimination (PSD) in boron-enriched liquid scintillator, where the use of digital ...
Digital pulse shape algorithms for scintillation-based neutron detectors S. D. Jastaniah and P. J. Sellin, Department of Physics, University of Surrey, Guildford GU2 7XH, UK. Abstract We discuss the application of digital pulse shape algorithms to n/γ pulse shape discrimination (PSD) in boron-enriched liquid scintillator, where the use of digital techniques has potential applications for neutron monitors. High-speed flash analogue-to-digital converters (ADCs) have opened up new possibilities for scintillator detector systems based upon digital event-by-event data acquisition. Using an 8-bit 1GS/s waveform digitiser, we sampled the scintillation detector signals with nanosecond time resolution to extract event parameters such as pulse height and rise time. Hence we have studied the variations in pulse shape and risetimes for gamma rays and neutrons and the quality of separation of these events. Even using a simple 10-90% risetime algorithm, we observe good gamma/neutron separation, and we present the PSD figure-of-merit obtained using this digital technique. Significant PSD between gamma events and the charged particles events produced from neutron capture was not observed.
1. Introduction In this experiment we address a digital PSD algorithm for n/γ-ray discrimination using boron loaded liquid scintillator BC523A1. In general, liquid scintillators show good sensitivity to fast neutrons and gamma rays, and can be adapted to detect thermal neutrons by 10B. For neutron detection applications, it can be advantageous to use PSD to discriminate the signal against the gamma background. Traditionally, PSD methods in liquid scintillators have used one of the following three analogue techniques; [1,2] firstly risetime inspection which determines the time at which the integrated light pulse reaches a certain fraction of its maximum amplitude. Secondly, the zero-crossing method, which allows an indirect evaluation of the risetime of the pulse performed by measurement of time for a bipolar signal to cross the baseline. Thirdly, charge comparison, which measures the integrated portions of the signal in the fast and slow components of the light pulse. In this work we investigated a simple digital algorithm that was not computer intensive and which could give comparable PSD performance to traditional analogue methods.
1
The boron loaded liquid scintillator cell used in this study was BC523A, loaded at 5% in weight with enriched boron (minimum of 90% 10B). The presence of a thermalization material (hydrogen) and a high concentration of 10B in BC523A produces a short mean neutron capture time of 0.5µs. This compares to a mean capture time of 2.2µs in the boron loaded plastic BC454 which contains only natural boron [3]. The possible interactions that take place when the BC523A detector is irradiated with fast neutrons can be described as follows:
The PSD properties of boron-loaded plastic scintillator BC454 at 5% in weight with natural boron have recently been characterised using conventional analogue techniques [4]. In general, BC454 shows worse PSD properties compared to liquid boron loaded scintillators. The boron loaded liquid scintillator BC523 has a higher light output and a larger H/C ratio than BC454 [2], hence is expected to have higher energy resolution. BC523 was used as part of a full-
www.bicron.com
0-7803-7324-3/02/$17.00 (C) 2002 IEEE
energy neutron spectroscopy system [5], and has provided high gamma ray background rejection without using a gamma ray shield. In order to investigate the differences in the shapes of the BC523A pulses by a simple risetime algorithm, we have used a high-speed 8-bit waveform digitiser with a 1 GS/s sampling rate2. The digitiser had a high bandwidth PCI bus connection to a PC to achieve event-by-event acquisition of complete detector waveforms at reasonable event rates (typically 1 kHz). The PC computer running dedicated data acquisition software performed real-time extraction of event parameters such as energy and risetime from the acquired waveforms. 2. Experimental considerations The liquid scintillator detector was irradiated with fast neutron and gamma radiation produced from an 18GBq Am-Be neutron source contained in a water tank (figure 1). A cylindrical BC523A liquid scintillator detector (measuring 52mm in diameter by 52mm in length) was positioned above the water tank. To produce fast neutrons a sealed air-filled plastic tube was placed into the water tank and positioned above the source.
detector operating bias voltage was 900V. A high performance PC computer running dedicated LabView data acquisition software captured the digitised waveforms from the detector on an event-by-event basis. The LabView software extracted the energy and rise-time from each pulse and performed histogramming of the data.
Figure 2: The calibration curve for high gain setting.
The response of the digital detection system to γ-rays, fast, and thermal neutrons was investigated. Two gain settings were used for the digitiser; a high gain setting to investigate the charged particle events due to thermal neutron capture, and a low gain setting to measure higher energy neutron and gamma ray interactions. The energy calibration curve for the high gain setting was determined using a variable target 241 Am X-ray source, with weighted mean energies taken for Ba, Tb targets, and a 59.5 keV 241Am gamma ray source as shown in figure 2.
Figure 1: The neutron tank and apparatus. Figure 3: The calibration curve for low gain setting.
Pulses from the photomultiplier tube (Electron Tubes type 9266B) were passed through an integrating preamplifier (Ortec type 113) to the waveform digitiser. The 2
For the higher energy events (scattered neutrons and gammas), a high energy was used, as shown in figure 3. For this calibration we used the Compton edge energies of four sources, as shown in Table
www.acqiris.com
0-7803-7324-3/02/$17.00 (C) 2002 IEEE
1. In each case, the calibration used the channel number 75% up the Compton edge [6]. Source 133 Ba 22 Na 137 Cs 60 Co 22 Na
spectra are shown in figure 5 and figure 6 respectively.
Compton Edge Energy 0.197 MeV 0.341 MeV 0.477 MeV 1.041 MeV 1.062 MeV
Table 1: Calibration energies used for the low gain calibration Figure 5: Energy spectra of Tb source using an 8bit digital pulse processing system.
Figure 4: The Am-Be energy spectra using the digital system.
3. Results
Figure 6: Energy spectra of Tb source using analogue pulse processing system with conventional 12-bit ADC.
3.1 Spectral Performance
3.2
After calibration, the high gain setting was used to acquire a pulse height spectrum from the Am-Be source. A simple maximum height algorithm was used to extract the amplitude of each pulse and no PSD was applied to the data at this stage. The resulting energy spectrum is shown in figure 4. The estimated energy of the neutron capture peak appeared at an electronequivalent energy (due to the reduced light yield of the liquid scintillator to charged particles [7]) of 60+/-15keV. To compare the performance of our 8-bit digital system to that of a conventional analogue spectroscopy system, we obtained X-ray photopeak spectra from Tb (44 keV) using both the digital system and a 12-bit multi channel analyser (MCA). These
Two typical waveforms observed from the output of the charge integrating preamplifier, produced by γ-ray and fast neutron interactions are shown in figure 7. The slower tail of the neutron pulse in the figure is clearly shown, and can be detected by the 10-90% rise-time algorithm. In these typical waveforms the measured risetimes are 10-20 ns for the γ-ray and 40-100 ns for the fast neutron. Digital filtering was used to remove the high frequency random noise components from the signal, using a fivepoint binomial smoothing filter [8]. The risetime algorithm first calculated the amplitude of the pulse from an average over the a number of points at the pulse maximum. The time difference between two amplitude limits, typically 10% and 90%, was then calculated as a measure of the
PSD algorithm
0-7803-7324-3/02/$17.00 (C) 2002 IEEE
pulse risetime. Saturated pulses larger than the full scale response of the digitiser were rejected.
Fn, Fγ are the full width half maximum values of the two peaks respectively.
Figure 8: Risetime vs. pulse height plot showing fast neutrons/γ-rays discrimination at low gain setting.
Figure 7: Digitised pulses from BC523A due to (upper) a gamma ray, and (lower) a fast neutron. The 10-90% risetime interval is indicated by the marker lines.
A 2D plot of pulse risetime versus pulse energy of different events at low and high gain setting shows the resulting n/γ discrimination. At low gain (figure 8), the algorithm provides n/γ discrimination between fast neutrons scatter in the scintillator, and γ−ray events (dominated by Compton scatter). At high gain (figure 9), the boron capture peak is clearly visible, and is not separated from the gamma ray events by pulse shape discrimination. By applying energy gates to the low gain 2D data shown in figure 8, the corresponding risetime distributions at low, medium, and high energies are obtained. Using the approach of reference [9] a figure of merit (FOM) was defined as S nγ FOM = Fn + Fγ where Snγ is the separation between the centroids of the neutron peak and the gamma peak in the risetime spectrum, and
Figure 9: Risetime vs. pulse height plot with poor capture neutrons/γ-rays discrimination at high gain setting.
The resulting risetime spectra are shown in figure 10. The FOM values deduced from these data are (a) 1.2, (b) 1.7, and (c) 1.3. The FOM for the lowest energy events (Figure 10(a)) is reduced due to the effect of noise on the low amplitude pulses which tends to broaden the peaks. The FOM values show a similar trend to conventional analogue PSD data [9], which typically shows a convergence of neutron and gamma events in two dimensional space at low energy. In our digital 10-90 risetime algorithm, the distribution of events in two dimensional risetime-energy space is different to analogue systems, but the under-
0-7803-7324-3/02/$17.00 (C) 2002 IEEE
1000
3000
6000
Counts
(c)
(b)
(a) 5000
2500
4000
2000
800
600 3000
1500
2000
1000
1000
500
400
200
0
0
0
0
25
50
75
100 125
150
0
25
50
75
100
125
150
0
25
50
75
100 125
Risetime (ns) Figure 10: Risetime spectra projected from the data in figure 8. The PSD FOM in each case is (a) 1.2 at low energy, (b) 1.7 at intermediate energy, and (c) 1.3 at high energy.
lying degradation of FOM at low energy is similar. For comparison, the same analysis was made using a slightly modified PSD algorithm in which the amplitude cuts were widened to 5 – 97%, instead of 10-90%. The corresponding projected risetime spectra are shown in figure 11. No significant variation in the FOM values were observed compared to the initial algorithm.
Figure 12: Co-60 energy spectrum using the digital pulse processing system connected to a HPGe detector.
Figure 13: Co-60 energy spectrum using an analogue pulse processing system connected to a HPGe detector.
It was of interest to compare the pulse height spectra that was derived from the digital technique with those obtained using a conventional 12-bit spectroscopy ADC, and to compare the energy resolution of both systems [10]. A 60Co spectrum was acquired using a high purity germanium (HPGe) detector, using both a conventional 12-bit analogue MCA system, and our 8-bit digital system. The pulse height spectra of both systems are shown in figures 12 and 13 respectively. Because of the difference in the ADC resolution for both systems, the digital system was of the order of three times worse (R=1.2%), than the analogue system (R=0.4%).
0-7803-7324-3/02/$17.00 (C) 2002 IEEE
150
1000
3000
6000
Counts
(c)
(b)
(a) 5000
2500
4000
2000
800
600 3000
1500
2000
1000
1000
500
400
200
0
0
0
0
25
50
75
100 125 150
0
25
50
75
100 125
150
0
25
50
75
100 125 150
Risetime (ns) Figure 11: Risetime spectra projected in the same manner as figure 10, using a 5% - 97% risetime algorithm. The PSD FOM in each case is very similar to that of figure 10.
4. Conclusion
6. References
We have used digital pulse processing to investigate the PSD properties of BC523A boron loaded liquid scintillator. First tests with an 8-bit 1 GS/s waveform digitiser have produced acceptable pulse height and rise time information from a scintillator detector. Using only simple pulse processing algorithms, clear pulse shape discrimination was achieved between fast neutron and gamma ray interactions in the detector. However, we observed poor pulse shape discrimination between gamma events and thermal neutrons [5]. We will further explore ways to enhance the technique, and investigate further digital PSD algorithms for use with organic scintillator detectors. These techniques have potential applications in neutron monitors where event-by-event n/γ discrimination is required.
[1] G. Ranucci, NIM A354 (1995) 389-399. [2] G. Knoll,”Radiation detection and measurements” 3rd. ed., 2000. [3] W. Feldman et al,. NIM A306 (1991) 350-365. [4] S. Normand et. al., NIM (Article in press). [5] T. Aoyama et. al., NIM A333 (1993) 492-501. [6] R. Cherubini et. al., Nucl. Instr. And Meth. A281 (1989) 349-352. [7] K. Geiger, et al., Nucl. Instr. And Meth. 131 (1975) 315-321. [8] P. Marchand et. al., Rev. Sci. Instrum.,Vol. 54, No. 8, August 1983. [9] S.C. Wang et al, Nucl. Instr. And Meth. A 432 (1999) 111-121. [10] J.M. Los Arcos, Nucl. Instr. And Meth. A 339 (1994) 99-101.
5. Acknowledgements Author S.J. gratefully acknowledges financial support from the Saudi Government.
0-7803-7324-3/02/$17.00 (C) 2002 IEEE
