- Email: [email protected]
High resolution gamma-ray spectroscopy at high count rates with a prototype High Purity Germanium detector
Accepted Manuscript High resolution gamma-ray spectroscopy at high count rates with a prototype High Purity Germanium detector R.J. Cooper, M. Amman, K. Vetter
PII: DOI: Reference:
S0168-9002(17)31459-6 https://doi.org/10.1016/j.nima.2017.12.053 NIMA 60394
To appear in:
Nuclear Inst. and Methods in Physics Research, A
Received date : 3 April 2017 Revised date : 8 December 2017 Accepted date : 15 December 2017 Please cite this article as: R.J. Cooper, M. Amman, K. Vetter, High resolution gamma-ray spectroscopy at high count rates with a prototype High Purity Germanium detector, Nuclear Inst. and Methods in Physics Research, A (2017), https://doi.org/10.1016/j.nima.2017.12.053 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
*Manuscript Click here to view linked References
1
High Resolution Gamma-Ray Spectroscopy at High Count Rates with a Prototype
2
High Purity Germanium Detector
3 R.J. Coopera*, M. Ammana, K. Vettera,b
4 5 6 7
a
Lawrence Berkeley National Laboratory, 1 Cyclotron Road, Berkeley, CA 94720, USA b
Department of Nuclear Engineering, University of California, Berkeley, CA 94720,
8
USA
9 10
Abstract
11
High-resolution gamma-ray spectrometers are required for applications in nuclear
12
safeguards, emergency response, and fundamental nuclear physics. To overcome one of
13
the shortcomings of conventional High Purity Germanium (HPGe) detectors, we have
14
developed a prototype device capable of achieving high event throughput and high
15
energy resolution at very high count rates. This device, the design of which we have
16
previously reported on, features a planar HPGe crystal with a reduced-capacitance strip
17
electrode geometry. This design is intended to provide good energy resolution at the
18
short shaping or digital filter times that are required for high rate operation and which are
19
enabled by the fast charge collection afforded by the planar geometry crystal. In this
20
work, we report on the initial performance of the system at count rates up to and
21
including two million counts per second.
22
23
Keywords: Gamma-ray detectors, gamma-ray spectroscopy, high-purity germanium, high
24
count rate.
25 26
Corresponding author: Tel.: +1 510 486 7296
27
Email address: [email protected] (R.J. Cooper)
28 29 30
1.0 Introduction
31
While High Purity Germanium (HPGe) detectors are typically limited to use in count rate
32
regimes up to around 10 kcps, a number of applications exist where high count rate
33
operation is desirable. These applications range from basic nuclear physics [1] to the
34
non-destructive assay of spent nuclear fuel [2]. For such applications, the ability to
35
maintain both fine energy resolution and high event throughput at count rates exceeding 1
36
mega counts per second (Mcps) would allow the analysis of relatively weak features in
37
complex gamma-ray spectra to be performed with increased sensitivity. Unfortunately,
38
however, conventional HPGe spectrometers typically sacrifice energy resolution for
39
higher throughput by employing short pulse shaping times in order to reduce deadtime.
40
This fundamental trade-off defines the challenge of achieving both high throughput and
41
fine energy resolution at high count rates.
42 43
In Ref. [3], results from the operation of a coaxial HPGe detector adapted for high rate
44
performance were reported. The system was able to achieve 8 keV FWHM at 662 keV
45
and 39% throughput when operating at an incident count rate of 1.03 Mcps. To our
46
knowledge, this represents the best reported result for high resolution gamma-ray
47
spectroscopy at high count rate. In Ref. [4], a coaxial HPGe detector was operated using
48
specially developed electronics and demonstrated less than 2.4 keV FWHM at 662 keV at
49
an incident rate 1 Mcps. However, the associated dead time at this count rate was 98%.
50 51
To overcome these resolution and throughput limitations, we have developed a novel
52
HPGe detector concept and have demonstrated its performance at millions of counts per
53
second. In Ref. [5], the design and basic electrical and spectroscopic performance of the
54
first prototype device were presented.
55
performance of a second, slightly modified prototype.
In this work, we report on the high-rate
56 57 58
2.0 Prototype Detector
59
The prototype detector discussed in Ref. [5] is based on a 100 mm diameter x 16 mm
60
thick planar, p-type HPGe crystal and features a single-sided strip electrode configuration
61
of ten 71 mm long, 2.85 mm wide charge collecting strips, each surrounded by 0.95 mm
62
wide steering grid strips. The collecting strips are separated by 4.75 mm and the steering
63
grid is employed in order to maintain good charge collection efficiency in the regions
64
between collecting strips. All contacts were produced using amorphous germanium [6].
65
The device we report on here features the same basic design but with the full-area contact
66
(i.e. the contact on the opposite face to the strips) produced using lithium diffusion. This
67
modification was made in order to improve the leakage current stability of the detector at
68
high count rates. This modification was made after a significant increase in leakage
69
current was observed when two prototype detectors that featured full-area contacts
70
fabricated using amorphous germanium were operated under high rate irradiation. A
71
series of small test devices were fabricated in order to investigate this effect but the
72
behavior was not consistently reproducible across all devices. The underlying cause of
73
the increased leakage current is not currently well understood. However, as it has not
74
been observed in any device in which the full-area contact is fabricated using lithium
75
diffusion, it is possible that it is related to barrier lowering at the amorphous germanium
76
contact as a result of charge build up during high-rate irradiation.
77 78
The detector is operated with a positive bias of 1000 V applied to the full-area contact,
79
the collecting strips connected to virtual ground through charge sensitive preamplifiers,
80
and a positive bias of 200 V applied to the steering grid. Figure 1 shows a photograph of
81
the detector, taken with the segmented electrode on display, along with a schematic
82
diagram of the planar geometry HPGe crystal and the electrode configuration.
83
Employing a planar geometry allows a strong, uniform electric field to be established
84
throughout the crystal and ensures fast charge collection. This, coupled to the short drift
85
distance between electrodes, results in signal rise times which are short and exhibit
86
minimal variation. Fig. 2 shows the distribution of signal risetimes recorded from a
87
typical strip while the detector was uniformly illuminated with a
88
risetime was calculated on a signal-by-signal basis as the time taken for the signal to rise
89
from 10% to 90% of its maximum value. The distribution in Fig. 2 was generated using
90
five thousand events from within the 662 keV photopeak and shows that the risetime
91
varies from a few tens of nanoseconds to just over 250 ns, consistent with the simulated
137
Cs source. The
92
values presented in Ref. [5]. When using digital trapezoidal filtering, this small variation
93
in rise time allows short gap times (e.g. 350 ns or shorter) to be employed in digital
94
trapezoidal filtering without degrading the energy resolution. This offers a significant
95
advantage over coaxial HPGe geometries for high count rate applications. However,
96
employing a planar geometry crystal does result in reduced detection efficiency relative
97
to larger volume coaxial geometries, particularly at high energy. Employing thinner
98
planar crystals would offer even smaller variations in the rise time and therefore the
99
potential for increased count rate performance. However, they would also result in even
100
greater reductions in efficiency. The relative efficiency [7] of the prototype detector has
101
been measured to be approximately 6% when operated in singles mode (i.e. when each
102
strip is read out individually and the resulting spectra summed). In any real application,
103
however, the detector would be operated in add-back mode [8] where the full energy
104
deposited by Compton scatted gamma rays is recovered by summing the energies
105
associated with simultaneous interactions in multiple strips. Monte Carlo simulations
106
suggest that the relative efficiency would be approximately 20% in this case. The add-
107
back technique may also be applied to multiple detectors. For the spectroscopy of high
108
energy gamma rays it is envisioned that multiple detectors could be employed in a stack
109
configuration to increase the total detection efficiency. This approach only meaningfully
110
increases the detection efficiency in systems where each individual detector is able to
111
maintain low dead time losses (i.e. high event throughput). Given the planar geometry, a
112
single device is optimally suited to the spectroscopy of lower energy gamma rays where
113
the detection efficiency is greater. For example, the intrinsic efficiency at 81 keV, as
114
measured using a 133Ba source, is 81%.
115 116 117
The specific strip electrode geometry was designed to reduce the capacitance associated
118
with each readout electrode relative to conventional HPGe strip detectors while
119
maintaining a relatively modest channel count. This is achieved by employing wide gaps
120
between the collecting strips. At 4.75 mm, the separation between collecting strips in the
121
device is approximately an order of magnitude larger than that of typical, conventional
122
strip detectors. By reducing the inter-strip capacitance, the series noise is reduced and
123
this in turn allows for improved energy resolution at short peaking times. By enabling
124
the use of both short peaking and gap times without significant degradation of the energy
125
resolution, the device is designed to allow high energy resolution and high throughput to
126
be maintained at count rates of hundreds of kcps per strip (i.e. millions of counts per
127
second across the entire detector).
128 129
At an intermediate count rate of approximately 2 kcps per strip, the average energy
130
resolution of the collecting strips is 2.4 keV FWHM at 662 keV, as measured using the
131
digital nanoMCA [9] with a peaking time of 1 s and a gap time of 350 ns applied in
132
online trapezoidal filtering. The average width of a pulser peak is 1.8 keV FWHM.
133
Figure 3 shows an example of a gamma-ray spectrum from a single, central strip (strip 6)
134
recorded in the presence of a
135
the 662 keV peak is 2.3 keV FWHM. In addition to the 662 keV 137Cs photopeak and the
136
peak associated with the pulser, the K and K X-rays associated with 137mBa can also be
137
observed just above the low energy threshold. As shown in the inset, these X-rays, which
137
Cs source and simultaneous pulser input. The width of
138
have characteristic energies of 32 keV and 36 keV, can easily be resolved. The energy
139
resolution of the 36 keV peak is 1.8 keV FWHM.
140 141
Fig. 4 shows the energy resolution of each of the ten collecting strips, as measured with
142
the nanoMCA with a peaking time of 1 s and a gap time of 350 ns. It can be seen that
143
the energy resolution of the edge strips (strips 1 and 10) at 662 keV is slightly degraded
144
relative to that of the others. As the width of the pulser peak remains generally constant
145
across all strips, this suggests that this is a result of reduced charge collection
146
performance rather than increased electronic noise.
147 148
3.0 Performance at High Count Rates
149
The performance of the prototype detector was assessed at a range of count rates using an
150
uncollimated 1.8 mCi
151
each source-to-detector distance, the ten collecting strips were read out using charge
152
sensitive preamplifiers and spectra were acquired by sequentially connecting individual
153
strips to the single-channel nanoMCA. The energy was calculated using the MCA’s
154
onboard trapezoidal filter algorithm with a peaking time of 500 ns and a gap time of 350
155
ns. In each case, the measured count rate was defined as the rate at which counts were
156
recorded to the energy spectrum. By performing fits to the full energy peaks, values of
157
energy resolution were extracted from each spectrum. For each measurement, the per-
158
strip event throughput was then calculated using the average dead time fraction reported
159
by the nanoMCA.
160
correcting the measured count rate according to this reported dead time, effectively
137
Cs source positioned at various distances from the device. At
The incident count rate was calculated on a per-strip basis by
161
dividing the number of recorded counts by the live time of the measurement. The
162
uncertainty associated with the incident count rate was calculated by assuming that the
163
live time is subject to uncertainties in both the determination of the dead fraction and the
164
real time. An uncertainty of 1% in the determination of the dead fraction was assumed
165
and a 10 ms uncertainty in the real time was assigned based on the specifications of the
166
nanoMCA. These uncertainties, along with the statistical uncertainty associated with the
167
number of counts recorded, were then propagated through the calculation of the incident
168
count rate. Similarly, the assumed uncertainty in the dead fraction was propagated in
169
order to assign uncertainties to the throughput. The incident count rates were verified
170
using the results of corresponding Monte Carlo simulations performed using MCNP6
171
[10]. At all source-to-detector distances, good agreement between the simulated and
172
measured count rates was observed. At the shortest source-to-detector distance, for
173
example, the discrepancy between the simulated count rate of strip 6 and that estimated
174
from the data was approximately 2% of the measured value.
175 176
Figure 5 shows the spectrum acquired from strip 6 during a measurement in which the
177
incident single-strip count rate was calculated to be 178.5(+/-1.0) kcps. The continuum
178
of counts above the 662 keV photopeak is a result of piled-up events. The inset image
179
shows the region of the spectrum below 100 keV, where the X-rays from 137mBa can once
180
again be observed. The additional peaks at 59 keV and 69 keV result from fluorescence
181
X-rays emitted from the Tungsten shielding surrounding the
182
amount of symmetric tailing is observed below the Full Width at Tenth Maximum
183
(FWTM) points on both the high and low energy sides of the 622 keV peak. This tailing
137
Cs source. A small
184
is not observed at low and intermediate count rates and the magnitude of the observed
185
tailing increases with count rate. This is not an unexpected result and may be caused by
186
uncertainty in determining the signal baseline when the count rate is large.
187 188
Figure 6 shows how the energy resolution of the 662 keV photopeak varies as a function
189
of the incident count rate for two central strips (strip 5 and 6) as well as the two edge
190
strips (strips 1 and 10). Figure 7 shows how the event throughout varies as a function of
191
incident count rate for the same strips. A linear fit was applied to the energy resolution
192
data and a double exponential fit to the throughput data. The results of these fits are
193
plotted as dashed lines on each figure panel and serve to allow interpolation between the
194
measured data points. The double exponential model was used to fit the throughput as a
195
function of incident count rate as the data is not well described by a single exponential
196
function. This implies that the behavior of the detector and signal processing chain is not
197
well described by a simple, paralyzable dead-time model which assumes a Poisson
198
distribution of pulse arrival times. This may be due, in part, to the contribution from
199
image charge signals and the effect of signals arising from charge sharing interactions but
200
is likely also a result of the triggering architecture employed by the nanoMCA system.
201 202
The results presented in Figs. 6 and 7 show that both the representative central strips and
203
the edge strips perform well across a range of count rates. The fitted curves suggest that
204
strip 5, the performance of which is generally representative of all non-edge strips,
205
maintains an energy resolution of 3.3 keV FWHM and a throughput of 78% when
206
operating at an incident count rate of 100 kcps. At 100 kcps, the throughput of the edge
207
strips is similar to that of the central strips but the energy resolution is slightly worse at
208
3.5 keV FHWM.
209 210
At the maximum count rate studied, the incident count rates associated with the central
211
strips were 211.5(+/-1.3) kcps (strip 5) and 212.5(+/-1.4) kcps (strip 6). At these count
212
rates, the throughput values associated with strips 5 and 6 were 61.4(+/-0.4)% and
213
61.2(+/-0.4)%, respectively. The energy resolution of each strip was 4.1 keV FWHM at
214
662 keV. Due to the small distance between the source and the detector during these
215
measurements, the incident count rate associated with the edge strips was slightly lower
216
than that of the central strips at 170.2(+/-0.8) kcps for strip 1 and 173.0(+/-0.9) kcps for
217
strip 10. The edge strips exhibited throughput values of 67.2(+/-0.3)% (strip 1) and
218
67.0(+/-0.3)% (strip 10). These values are consistent with the throughput of the central
219
strips at the same incident count rates. The energy resolution of each edge strip was 4.0
220
keV FWHM at 662 keV. During these measurements, the sum of the incident count rates
221
of all ten collecting strips was 2.0 Mcps. Figure 8 shows the incident count rate (top),
222
event throughput (middle), and energy resolution (bottom) associated with all collecting
223
strips.
224
throughput is 62.8(+/-0.8)%.
The performance of all ten collecting strips is similar and the total event
225 226 227
4.0 Summary
228
We have demonstrated the performance of a novel, prototype HPGe detector for high-
229
rate, high-resolution gamma-ray spectroscopy. The detector design features a reduced-
230
capacitance, single sided strip electrode configuration with ten collecting strips and a grid
231
of narrow steering electrodes to maintain charge collection performance. The ability of
232
all ten strips to maintain high event throughput and fine energy resolution has been
233
demonstrated during measurements in which the incident count rate across the entire
234
detector was 2.0 Mcps. During these measurements, a single centrally located strip
235
operating at 212.5(+/-1.4) kcps, exhibited event throughput of 61.2(+/-0.4)% and energy
236
resolution of 4.1 keV FWHM at 662 keV.
237 238
This performance represents a significant improvement over previously demonstrated
239
systems for high-resolution gamma-ray spectroscopy at high count rate. The use of a
240
segmented electrode geometry allows the incident count rate to be distributed over
241
multiple channels, the short charge collection times offered by the planar geometry
242
allows the use of short gap times without loss of energy resolution, and the reduced
243
capacitance strip-electrode geometry reduces the degradation of energy resolution at short
244
peaking times. Due to the reduced capacitance of the electrode design, the device also
245
exhibits low electronic noise and good energy resolution at low energy. This may be
246
particularly advantageous for applications such as the non-destructive assay of spent
247
nuclear fuel where spectroscopy of the low energy features associated with Plutonium
248
and Uranium isotopes would allow direct measurements of fissile content to be made.
249 250
At a count rate of 1 Mcps, the small coaxial HPGe system presented in Ref. [3] was able
251
to achieve 8 keV FWHM at 662 keV and 39% throughput. The results presented here
252
represent improvement by a factor of greater than two in both energy resolution and
253
throughput at the same incident count rate. Extrapolations based on the performance of
254
this detector (Figs. 6 and 7), suggest that in its current implementation, an incident count
255
rate of 500 kcps per strip (equivalent to 5 Mcps total) could be reached before the
256
throughput dropped to 39%. At this count rate, the corresponding energy resolution
257
would be approximately 6.2 keV FWHM at 662 keV.
258 259
With the use of front-end electronics and a digital data acquisition system specifically
260
optimized for high rate operation, increases in both throughput and energy resolution may
261
be achieved. For example, reducing the capacitance of the front end electronics could
262
further improve the energy resolution at short peaking times while the application of
263
digital pile-up recovery algorithms could be used to improve the throughput.
264 265 266
Acknowledgements
267
This work was performed under the auspices of the US Department of Energy by
268
Lawrence Berkeley National Laboratory under Contract DE-AC02-05CH11231. The
269
project was funded by the US Department of Energy, National Nuclear Security
270
Administration, Office of Defense Nuclear Nonproliferation Research and Development
271
(DNN R&D).
272 273 274
References
275
[1] B. Loher et al., Nuclear Instruments and Methods in Physics Research A 686 (2012)
276
1-6.
277
[2] A.J. Gilbert et al., Proceedings of IEEE Nuclear Science Symposium and Medical
278
Imaging Conference, 2015.
279
[3] B.A. VanDevender et al., IEEE Transactions on Nuclear Science, 61 5 (2014) 2619-
280
2627.
281
[4] J. Plagnard et al., Applied Radiation and Isotopes 60 (2004) 179-183.
282
[5] R.J. Cooper et al., Nuclear Instruments and Methods in Physics Research A 795
283
(2015) 167-173.
284
[6] M. Amman and P.N. Luke, Nuclear Instruments and Methods in Physics Research A
285
452 (2000) 155.
286
[7] IEEE Std. 325 1996
287
[8] M. Schumaker and C.E. Svensson, Nuclear Instruments and Methods in Physics
288
Research A 575 (2007) 421-432.
289
[9] http://www.labzy.com/index.html
290 291
[10] T. Goorley et al. Annals of Nuclear Energy 87 (2016): 772-783.
292 293 294 295 296 297
298 299
Figure 1.
300
segmented electrode on display, along with a schematic diagram of the planar geometry
301
HPGe crystal (top) and electrode configuration.
302 303 304 305
A photograph of the prototype HPGe detector (bottom), taken with the
700 600
Counts
500 400 300 200 100 0 306
0
100
200
300
Signal Risetime (ns)
307
Figure 2. Distribution of signal risetimes in the prototype detector. The risetime is
308
defined as the time taken for the signal to rise from 10% to 90% of its maximum. The
309
distribution was generated using the signals associated with five thousand events from
310
within the 662 keV photpeak of 137Cs.
311 312 313 314 315 316 317 318 319
320 137
321
Figure 3. A gamma-ray spectrum from a single strip recorded in the presence of a
322
source and simultaneous pulser input at a count rate of around 2 kcps. The width of the
323
137
324
shows the low energy portion of the spectrum where the 32 keV and 36 keV X-rays
325
associated with 137mBa can also be observed.
326 327 328 329 330 331 332 333 334
Cs
Cs and pulser peaks are 2.4 keV FWHM and 1.8 keV FWHM, respectively. The inset
4 Pulser 662 keV Gamma Peak
Peak Width, FWHM (keV)
3.5 3 2.5 2 1.5 1 0.5 0
0
2
335
4 6 8 Strip Number
10
336
Figure 4. The width of the 662 keV peak from 137Cs and an electronic pulser for each of
337
the ten collecting strips.
338 339 340 341 342 343 344
345 346
Figure 5. A gamma-ray spectrum from a single, central strip (strip 6) recorded in the
347
presence of a
348
kcps. The inset shows the 32 keV and 36 keV X-ray peaks associated with
349
the 59 keV and 69 keV peaks that result from fluorescence X-rays emitted from Tungsten
350
shielding surrounding the 137Cs source.
351 352 353 354 355 356 357 358 359
137
Cs source and acquired at an incident single-strip count rate of 178.5 137m
Ba and
360 361
Figure 6. The energy resolution of the 662 keV photopeak as a function of the incident
362
count rate for two central strips and the two edge strips.
363 364 365 366 367 368 369 370
371 372
Figure 7. The event throughput as a function of the incident count rate for two central
373
strips and the two edge strips.
374 375 376 377 378 379 380 381
382 383
Figure 8. The incident count rate (top), event throughput (middle), and energy resolution
384
at 662 keV (bottom) of all collecting strips at the highest total count rate studied.
PII: DOI: Reference:
S0168-9002(17)31459-6 https://doi.org/10.1016/j.nima.2017.12.053 NIMA 60394
To appear in:
Nuclear Inst. and Methods in Physics Research, A
Received date : 3 April 2017 Revised date : 8 December 2017 Accepted date : 15 December 2017 Please cite this article as: R.J. Cooper, M. Amman, K. Vetter, High resolution gamma-ray spectroscopy at high count rates with a prototype High Purity Germanium detector, Nuclear Inst. and Methods in Physics Research, A (2017), https://doi.org/10.1016/j.nima.2017.12.053 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
*Manuscript Click here to view linked References
1
High Resolution Gamma-Ray Spectroscopy at High Count Rates with a Prototype
2
High Purity Germanium Detector
3 R.J. Coopera*, M. Ammana, K. Vettera,b
4 5 6 7
a
Lawrence Berkeley National Laboratory, 1 Cyclotron Road, Berkeley, CA 94720, USA b
Department of Nuclear Engineering, University of California, Berkeley, CA 94720,
8
USA
9 10
Abstract
11
High-resolution gamma-ray spectrometers are required for applications in nuclear
12
safeguards, emergency response, and fundamental nuclear physics. To overcome one of
13
the shortcomings of conventional High Purity Germanium (HPGe) detectors, we have
14
developed a prototype device capable of achieving high event throughput and high
15
energy resolution at very high count rates. This device, the design of which we have
16
previously reported on, features a planar HPGe crystal with a reduced-capacitance strip
17
electrode geometry. This design is intended to provide good energy resolution at the
18
short shaping or digital filter times that are required for high rate operation and which are
19
enabled by the fast charge collection afforded by the planar geometry crystal. In this
20
work, we report on the initial performance of the system at count rates up to and
21
including two million counts per second.
22
23
Keywords: Gamma-ray detectors, gamma-ray spectroscopy, high-purity germanium, high
24
count rate.
25 26
Corresponding author: Tel.: +1 510 486 7296
27
Email address: [email protected] (R.J. Cooper)
28 29 30
1.0 Introduction
31
While High Purity Germanium (HPGe) detectors are typically limited to use in count rate
32
regimes up to around 10 kcps, a number of applications exist where high count rate
33
operation is desirable. These applications range from basic nuclear physics [1] to the
34
non-destructive assay of spent nuclear fuel [2]. For such applications, the ability to
35
maintain both fine energy resolution and high event throughput at count rates exceeding 1
36
mega counts per second (Mcps) would allow the analysis of relatively weak features in
37
complex gamma-ray spectra to be performed with increased sensitivity. Unfortunately,
38
however, conventional HPGe spectrometers typically sacrifice energy resolution for
39
higher throughput by employing short pulse shaping times in order to reduce deadtime.
40
This fundamental trade-off defines the challenge of achieving both high throughput and
41
fine energy resolution at high count rates.
42 43
In Ref. [3], results from the operation of a coaxial HPGe detector adapted for high rate
44
performance were reported. The system was able to achieve 8 keV FWHM at 662 keV
45
and 39% throughput when operating at an incident count rate of 1.03 Mcps. To our
46
knowledge, this represents the best reported result for high resolution gamma-ray
47
spectroscopy at high count rate. In Ref. [4], a coaxial HPGe detector was operated using
48
specially developed electronics and demonstrated less than 2.4 keV FWHM at 662 keV at
49
an incident rate 1 Mcps. However, the associated dead time at this count rate was 98%.
50 51
To overcome these resolution and throughput limitations, we have developed a novel
52
HPGe detector concept and have demonstrated its performance at millions of counts per
53
second. In Ref. [5], the design and basic electrical and spectroscopic performance of the
54
first prototype device were presented.
55
performance of a second, slightly modified prototype.
In this work, we report on the high-rate
56 57 58
2.0 Prototype Detector
59
The prototype detector discussed in Ref. [5] is based on a 100 mm diameter x 16 mm
60
thick planar, p-type HPGe crystal and features a single-sided strip electrode configuration
61
of ten 71 mm long, 2.85 mm wide charge collecting strips, each surrounded by 0.95 mm
62
wide steering grid strips. The collecting strips are separated by 4.75 mm and the steering
63
grid is employed in order to maintain good charge collection efficiency in the regions
64
between collecting strips. All contacts were produced using amorphous germanium [6].
65
The device we report on here features the same basic design but with the full-area contact
66
(i.e. the contact on the opposite face to the strips) produced using lithium diffusion. This
67
modification was made in order to improve the leakage current stability of the detector at
68
high count rates. This modification was made after a significant increase in leakage
69
current was observed when two prototype detectors that featured full-area contacts
70
fabricated using amorphous germanium were operated under high rate irradiation. A
71
series of small test devices were fabricated in order to investigate this effect but the
72
behavior was not consistently reproducible across all devices. The underlying cause of
73
the increased leakage current is not currently well understood. However, as it has not
74
been observed in any device in which the full-area contact is fabricated using lithium
75
diffusion, it is possible that it is related to barrier lowering at the amorphous germanium
76
contact as a result of charge build up during high-rate irradiation.
77 78
The detector is operated with a positive bias of 1000 V applied to the full-area contact,
79
the collecting strips connected to virtual ground through charge sensitive preamplifiers,
80
and a positive bias of 200 V applied to the steering grid. Figure 1 shows a photograph of
81
the detector, taken with the segmented electrode on display, along with a schematic
82
diagram of the planar geometry HPGe crystal and the electrode configuration.
83
Employing a planar geometry allows a strong, uniform electric field to be established
84
throughout the crystal and ensures fast charge collection. This, coupled to the short drift
85
distance between electrodes, results in signal rise times which are short and exhibit
86
minimal variation. Fig. 2 shows the distribution of signal risetimes recorded from a
87
typical strip while the detector was uniformly illuminated with a
88
risetime was calculated on a signal-by-signal basis as the time taken for the signal to rise
89
from 10% to 90% of its maximum value. The distribution in Fig. 2 was generated using
90
five thousand events from within the 662 keV photopeak and shows that the risetime
91
varies from a few tens of nanoseconds to just over 250 ns, consistent with the simulated
137
Cs source. The
92
values presented in Ref. [5]. When using digital trapezoidal filtering, this small variation
93
in rise time allows short gap times (e.g. 350 ns or shorter) to be employed in digital
94
trapezoidal filtering without degrading the energy resolution. This offers a significant
95
advantage over coaxial HPGe geometries for high count rate applications. However,
96
employing a planar geometry crystal does result in reduced detection efficiency relative
97
to larger volume coaxial geometries, particularly at high energy. Employing thinner
98
planar crystals would offer even smaller variations in the rise time and therefore the
99
potential for increased count rate performance. However, they would also result in even
100
greater reductions in efficiency. The relative efficiency [7] of the prototype detector has
101
been measured to be approximately 6% when operated in singles mode (i.e. when each
102
strip is read out individually and the resulting spectra summed). In any real application,
103
however, the detector would be operated in add-back mode [8] where the full energy
104
deposited by Compton scatted gamma rays is recovered by summing the energies
105
associated with simultaneous interactions in multiple strips. Monte Carlo simulations
106
suggest that the relative efficiency would be approximately 20% in this case. The add-
107
back technique may also be applied to multiple detectors. For the spectroscopy of high
108
energy gamma rays it is envisioned that multiple detectors could be employed in a stack
109
configuration to increase the total detection efficiency. This approach only meaningfully
110
increases the detection efficiency in systems where each individual detector is able to
111
maintain low dead time losses (i.e. high event throughput). Given the planar geometry, a
112
single device is optimally suited to the spectroscopy of lower energy gamma rays where
113
the detection efficiency is greater. For example, the intrinsic efficiency at 81 keV, as
114
measured using a 133Ba source, is 81%.
115 116 117
The specific strip electrode geometry was designed to reduce the capacitance associated
118
with each readout electrode relative to conventional HPGe strip detectors while
119
maintaining a relatively modest channel count. This is achieved by employing wide gaps
120
between the collecting strips. At 4.75 mm, the separation between collecting strips in the
121
device is approximately an order of magnitude larger than that of typical, conventional
122
strip detectors. By reducing the inter-strip capacitance, the series noise is reduced and
123
this in turn allows for improved energy resolution at short peaking times. By enabling
124
the use of both short peaking and gap times without significant degradation of the energy
125
resolution, the device is designed to allow high energy resolution and high throughput to
126
be maintained at count rates of hundreds of kcps per strip (i.e. millions of counts per
127
second across the entire detector).
128 129
At an intermediate count rate of approximately 2 kcps per strip, the average energy
130
resolution of the collecting strips is 2.4 keV FWHM at 662 keV, as measured using the
131
digital nanoMCA [9] with a peaking time of 1 s and a gap time of 350 ns applied in
132
online trapezoidal filtering. The average width of a pulser peak is 1.8 keV FWHM.
133
Figure 3 shows an example of a gamma-ray spectrum from a single, central strip (strip 6)
134
recorded in the presence of a
135
the 662 keV peak is 2.3 keV FWHM. In addition to the 662 keV 137Cs photopeak and the
136
peak associated with the pulser, the K and K X-rays associated with 137mBa can also be
137
observed just above the low energy threshold. As shown in the inset, these X-rays, which
137
Cs source and simultaneous pulser input. The width of
138
have characteristic energies of 32 keV and 36 keV, can easily be resolved. The energy
139
resolution of the 36 keV peak is 1.8 keV FWHM.
140 141
Fig. 4 shows the energy resolution of each of the ten collecting strips, as measured with
142
the nanoMCA with a peaking time of 1 s and a gap time of 350 ns. It can be seen that
143
the energy resolution of the edge strips (strips 1 and 10) at 662 keV is slightly degraded
144
relative to that of the others. As the width of the pulser peak remains generally constant
145
across all strips, this suggests that this is a result of reduced charge collection
146
performance rather than increased electronic noise.
147 148
3.0 Performance at High Count Rates
149
The performance of the prototype detector was assessed at a range of count rates using an
150
uncollimated 1.8 mCi
151
each source-to-detector distance, the ten collecting strips were read out using charge
152
sensitive preamplifiers and spectra were acquired by sequentially connecting individual
153
strips to the single-channel nanoMCA. The energy was calculated using the MCA’s
154
onboard trapezoidal filter algorithm with a peaking time of 500 ns and a gap time of 350
155
ns. In each case, the measured count rate was defined as the rate at which counts were
156
recorded to the energy spectrum. By performing fits to the full energy peaks, values of
157
energy resolution were extracted from each spectrum. For each measurement, the per-
158
strip event throughput was then calculated using the average dead time fraction reported
159
by the nanoMCA.
160
correcting the measured count rate according to this reported dead time, effectively
137
Cs source positioned at various distances from the device. At
The incident count rate was calculated on a per-strip basis by
161
dividing the number of recorded counts by the live time of the measurement. The
162
uncertainty associated with the incident count rate was calculated by assuming that the
163
live time is subject to uncertainties in both the determination of the dead fraction and the
164
real time. An uncertainty of 1% in the determination of the dead fraction was assumed
165
and a 10 ms uncertainty in the real time was assigned based on the specifications of the
166
nanoMCA. These uncertainties, along with the statistical uncertainty associated with the
167
number of counts recorded, were then propagated through the calculation of the incident
168
count rate. Similarly, the assumed uncertainty in the dead fraction was propagated in
169
order to assign uncertainties to the throughput. The incident count rates were verified
170
using the results of corresponding Monte Carlo simulations performed using MCNP6
171
[10]. At all source-to-detector distances, good agreement between the simulated and
172
measured count rates was observed. At the shortest source-to-detector distance, for
173
example, the discrepancy between the simulated count rate of strip 6 and that estimated
174
from the data was approximately 2% of the measured value.
175 176
Figure 5 shows the spectrum acquired from strip 6 during a measurement in which the
177
incident single-strip count rate was calculated to be 178.5(+/-1.0) kcps. The continuum
178
of counts above the 662 keV photopeak is a result of piled-up events. The inset image
179
shows the region of the spectrum below 100 keV, where the X-rays from 137mBa can once
180
again be observed. The additional peaks at 59 keV and 69 keV result from fluorescence
181
X-rays emitted from the Tungsten shielding surrounding the
182
amount of symmetric tailing is observed below the Full Width at Tenth Maximum
183
(FWTM) points on both the high and low energy sides of the 622 keV peak. This tailing
137
Cs source. A small
184
is not observed at low and intermediate count rates and the magnitude of the observed
185
tailing increases with count rate. This is not an unexpected result and may be caused by
186
uncertainty in determining the signal baseline when the count rate is large.
187 188
Figure 6 shows how the energy resolution of the 662 keV photopeak varies as a function
189
of the incident count rate for two central strips (strip 5 and 6) as well as the two edge
190
strips (strips 1 and 10). Figure 7 shows how the event throughout varies as a function of
191
incident count rate for the same strips. A linear fit was applied to the energy resolution
192
data and a double exponential fit to the throughput data. The results of these fits are
193
plotted as dashed lines on each figure panel and serve to allow interpolation between the
194
measured data points. The double exponential model was used to fit the throughput as a
195
function of incident count rate as the data is not well described by a single exponential
196
function. This implies that the behavior of the detector and signal processing chain is not
197
well described by a simple, paralyzable dead-time model which assumes a Poisson
198
distribution of pulse arrival times. This may be due, in part, to the contribution from
199
image charge signals and the effect of signals arising from charge sharing interactions but
200
is likely also a result of the triggering architecture employed by the nanoMCA system.
201 202
The results presented in Figs. 6 and 7 show that both the representative central strips and
203
the edge strips perform well across a range of count rates. The fitted curves suggest that
204
strip 5, the performance of which is generally representative of all non-edge strips,
205
maintains an energy resolution of 3.3 keV FWHM and a throughput of 78% when
206
operating at an incident count rate of 100 kcps. At 100 kcps, the throughput of the edge
207
strips is similar to that of the central strips but the energy resolution is slightly worse at
208
3.5 keV FHWM.
209 210
At the maximum count rate studied, the incident count rates associated with the central
211
strips were 211.5(+/-1.3) kcps (strip 5) and 212.5(+/-1.4) kcps (strip 6). At these count
212
rates, the throughput values associated with strips 5 and 6 were 61.4(+/-0.4)% and
213
61.2(+/-0.4)%, respectively. The energy resolution of each strip was 4.1 keV FWHM at
214
662 keV. Due to the small distance between the source and the detector during these
215
measurements, the incident count rate associated with the edge strips was slightly lower
216
than that of the central strips at 170.2(+/-0.8) kcps for strip 1 and 173.0(+/-0.9) kcps for
217
strip 10. The edge strips exhibited throughput values of 67.2(+/-0.3)% (strip 1) and
218
67.0(+/-0.3)% (strip 10). These values are consistent with the throughput of the central
219
strips at the same incident count rates. The energy resolution of each edge strip was 4.0
220
keV FWHM at 662 keV. During these measurements, the sum of the incident count rates
221
of all ten collecting strips was 2.0 Mcps. Figure 8 shows the incident count rate (top),
222
event throughput (middle), and energy resolution (bottom) associated with all collecting
223
strips.
224
throughput is 62.8(+/-0.8)%.
The performance of all ten collecting strips is similar and the total event
225 226 227
4.0 Summary
228
We have demonstrated the performance of a novel, prototype HPGe detector for high-
229
rate, high-resolution gamma-ray spectroscopy. The detector design features a reduced-
230
capacitance, single sided strip electrode configuration with ten collecting strips and a grid
231
of narrow steering electrodes to maintain charge collection performance. The ability of
232
all ten strips to maintain high event throughput and fine energy resolution has been
233
demonstrated during measurements in which the incident count rate across the entire
234
detector was 2.0 Mcps. During these measurements, a single centrally located strip
235
operating at 212.5(+/-1.4) kcps, exhibited event throughput of 61.2(+/-0.4)% and energy
236
resolution of 4.1 keV FWHM at 662 keV.
237 238
This performance represents a significant improvement over previously demonstrated
239
systems for high-resolution gamma-ray spectroscopy at high count rate. The use of a
240
segmented electrode geometry allows the incident count rate to be distributed over
241
multiple channels, the short charge collection times offered by the planar geometry
242
allows the use of short gap times without loss of energy resolution, and the reduced
243
capacitance strip-electrode geometry reduces the degradation of energy resolution at short
244
peaking times. Due to the reduced capacitance of the electrode design, the device also
245
exhibits low electronic noise and good energy resolution at low energy. This may be
246
particularly advantageous for applications such as the non-destructive assay of spent
247
nuclear fuel where spectroscopy of the low energy features associated with Plutonium
248
and Uranium isotopes would allow direct measurements of fissile content to be made.
249 250
At a count rate of 1 Mcps, the small coaxial HPGe system presented in Ref. [3] was able
251
to achieve 8 keV FWHM at 662 keV and 39% throughput. The results presented here
252
represent improvement by a factor of greater than two in both energy resolution and
253
throughput at the same incident count rate. Extrapolations based on the performance of
254
this detector (Figs. 6 and 7), suggest that in its current implementation, an incident count
255
rate of 500 kcps per strip (equivalent to 5 Mcps total) could be reached before the
256
throughput dropped to 39%. At this count rate, the corresponding energy resolution
257
would be approximately 6.2 keV FWHM at 662 keV.
258 259
With the use of front-end electronics and a digital data acquisition system specifically
260
optimized for high rate operation, increases in both throughput and energy resolution may
261
be achieved. For example, reducing the capacitance of the front end electronics could
262
further improve the energy resolution at short peaking times while the application of
263
digital pile-up recovery algorithms could be used to improve the throughput.
264 265 266
Acknowledgements
267
This work was performed under the auspices of the US Department of Energy by
268
Lawrence Berkeley National Laboratory under Contract DE-AC02-05CH11231. The
269
project was funded by the US Department of Energy, National Nuclear Security
270
Administration, Office of Defense Nuclear Nonproliferation Research and Development
271
(DNN R&D).
272 273 274
References
275
[1] B. Loher et al., Nuclear Instruments and Methods in Physics Research A 686 (2012)
276
1-6.
277
[2] A.J. Gilbert et al., Proceedings of IEEE Nuclear Science Symposium and Medical
278
Imaging Conference, 2015.
279
[3] B.A. VanDevender et al., IEEE Transactions on Nuclear Science, 61 5 (2014) 2619-
280
2627.
281
[4] J. Plagnard et al., Applied Radiation and Isotopes 60 (2004) 179-183.
282
[5] R.J. Cooper et al., Nuclear Instruments and Methods in Physics Research A 795
283
(2015) 167-173.
284
[6] M. Amman and P.N. Luke, Nuclear Instruments and Methods in Physics Research A
285
452 (2000) 155.
286
[7] IEEE Std. 325 1996
287
[8] M. Schumaker and C.E. Svensson, Nuclear Instruments and Methods in Physics
288
Research A 575 (2007) 421-432.
289
[9] http://www.labzy.com/index.html
290 291
[10] T. Goorley et al. Annals of Nuclear Energy 87 (2016): 772-783.
292 293 294 295 296 297
298 299
Figure 1.
300
segmented electrode on display, along with a schematic diagram of the planar geometry
301
HPGe crystal (top) and electrode configuration.
302 303 304 305
A photograph of the prototype HPGe detector (bottom), taken with the
700 600
Counts
500 400 300 200 100 0 306
0
100
200
300
Signal Risetime (ns)
307
Figure 2. Distribution of signal risetimes in the prototype detector. The risetime is
308
defined as the time taken for the signal to rise from 10% to 90% of its maximum. The
309
distribution was generated using the signals associated with five thousand events from
310
within the 662 keV photpeak of 137Cs.
311 312 313 314 315 316 317 318 319
320 137
321
Figure 3. A gamma-ray spectrum from a single strip recorded in the presence of a
322
source and simultaneous pulser input at a count rate of around 2 kcps. The width of the
323
137
324
shows the low energy portion of the spectrum where the 32 keV and 36 keV X-rays
325
associated with 137mBa can also be observed.
326 327 328 329 330 331 332 333 334
Cs
Cs and pulser peaks are 2.4 keV FWHM and 1.8 keV FWHM, respectively. The inset
4 Pulser 662 keV Gamma Peak
Peak Width, FWHM (keV)
3.5 3 2.5 2 1.5 1 0.5 0
0
2
335
4 6 8 Strip Number
10
336
Figure 4. The width of the 662 keV peak from 137Cs and an electronic pulser for each of
337
the ten collecting strips.
338 339 340 341 342 343 344
345 346
Figure 5. A gamma-ray spectrum from a single, central strip (strip 6) recorded in the
347
presence of a
348
kcps. The inset shows the 32 keV and 36 keV X-ray peaks associated with
349
the 59 keV and 69 keV peaks that result from fluorescence X-rays emitted from Tungsten
350
shielding surrounding the 137Cs source.
351 352 353 354 355 356 357 358 359
137
Cs source and acquired at an incident single-strip count rate of 178.5 137m
Ba and
360 361
Figure 6. The energy resolution of the 662 keV photopeak as a function of the incident
362
count rate for two central strips and the two edge strips.
363 364 365 366 367 368 369 370
371 372
Figure 7. The event throughput as a function of the incident count rate for two central
373
strips and the two edge strips.
374 375 376 377 378 379 380 381
382 383
Figure 8. The incident count rate (top), event throughput (middle), and energy resolution
384
at 662 keV (bottom) of all collecting strips at the highest total count rate studied.