- Email: [email protected]
Neutron detection in high γ background using a micromegas detector
ARTICLE IN PRESS
Nuclear Instruments and Methods in Physics Research A 572 (2007) 859–865 www.elsevier.com/locate/nima
Neutron detection in high g background using a micromegas detector J. Pancina,, S. Aunea, E. Berthoumieuxa, S. Boyera, E. Delagnesa, V. Macarya, B. Poumaredeb, H. Safaa a
CEA Saclay-DSM/DAPNIA, Bat 534/34A, F-91191 Gif-Sur-Yvette, France b CEA-DRT/LIST Saclay, France
Received 20 April 2006; received in revised form 24 October 2006; accepted 7 December 2006 Available online 17 December 2006
Abstract The ability of Micromegas to detect neutrons over a wide energy range has already been demonstrated. However, in some nuclear experiments or applications, neutrons come with high photon flux disturbing the detectors. A new project for nuclear waste characterization using photonuclear reactions is under development at the CEA. One of the ideas is to detect the prompt neutrons produced by these kinds of reactions, which are accompanied by a strong g flash. The micromegas detector has been chosen to detect these neutrons since it is quite insensitive to g-rays under certain conditions. The configuration of this detector will be presented and its g insensitivity demonstrated. Results from simulations and experiments will be shown. r 2007 Elsevier B.V. All rights reserved. PACS: 25.85.Jg; 28.20.v; 07.85.Fv; 29.40.n Keywords: Micromegas; Photofission; Radioactive waste characterization; Neutron detection; g sensitivity
1. Introduction The SAPHIR facility at CEA Saclay is experienced in radioactive parcel characterization using delayed neutrons [1,2]. An intense g-ray beam is created by the slowing down of electrons in a copper or a tungsten target. The energy of the electrons is between 12 and 15 MeV with about 1012 electrons in one accelerator pulse of 2:5 ms. Photonuclear reactions are induced by the g-rays on the minor actinides and transuranium elements but also on the impurities of the parcel. The actinide and transuranium quantities are evaluated by measuring the delayed neutrons coming from fission fragments, after the accelerator pulse, using 3He counters placed all around the parcel. Detecting prompt neutrons during the accelerator pulse could be another means to characterize radioactive parcels. However, it would require detectors which are not blinded by the g-rays flash like 3He counters. A neutron detector based on a Micromegas concept [3] has been realized to detect these Corresponding author. Tel.: +33 1 69082648; fax: +33 1 69085669.
E-mail address: [email protected] (J. Pancin). 0168-9002/$ - see front matter r 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.nima.2006.12.010
neutrons in the huge g-rays flash. The principle of the detector is presented first. Some simulations have been performed to check the detector characteristics for neutron detection and evaluate its behavior in g-beams. Some experiments and results are finally reported, showing the detector characteristics. 2. The neutron detector 2.1. Technical description The Micromegas detector used here is quite similar to the one used for the n_TOF experiment [4]. The inner chamber, filled with a mixture of argon–isobutane or helium–isobutane at atmospheric pressure, has been reused. The detector, placed inside, is a double stage circular chamber with an active surface of 40 cm2 . It consists of a 9:5 mm drift gap and a 100 mm amplification gap. The two gaps are separated by a 5 mm thick nickel micromesh. The amplification occurs between the mesh and a copper printed circuit composed of eight strips of 9 mm wide spaced 100 mm apart. Some spacers insulate the
ARTICLE IN PRESS J. Pancin et al. / Nuclear Instruments and Methods in Physics Research A 572 (2007) 859–865
860
printed circuit from the micromesh and guarantee the uniformity of the amplification gap. The drift electrode, covering all the detection area, is a 12 mm thick polypropylene foil on which a 100 nm layer of aluminium has been deposited to ensure electrical conductivity. 2.2. Detector principle Electrons are produced between the drift electrode and the micromesh during the passing of charged particles through the gas. These electrons drift towards the micromesh under the action of the electric field applied in the drift gap ð1 kV=cmÞ. After the micromesh, they are multiplied in the narrow amplification gap (high field with 30 kV=cmÞ and this avalanche induces a current on the anode strips. However, a neutron to charged particle converter is necessary to detect the incident neutron. Since only fast neutrons are of interest here, neutron scattering on the nuclei of the polypropylene foil and of the gas is exploited (see Fig. 1). Above the detection threshold, the ionization created by the recoils is detected. For this application, the Micromegas detector is generally filled with He þ iC4H10 ð3:8%Þ (For safety reasons, the percentage of isobutane is sufficiently low to have a nonflammable gas). On the one hand, helium is chosen since, as a light gas, it is less sensitive to X-and g-ray photons. On the other hand, the lighter the nucleus, the larger is the energy transferred from the incident neutron. It is then the recoils of hydrogen coming from the foil and the gas, and the recoils of helium coming from the gas that are mostly detected. The hydrogen nucleus is of particular interest in fast neutron scattering since it can recoil with the entire neutron energy. In the case of helium, the reaction 4He(n,n)4He has a resonance around 1 MeV which increases the reaction rate.
fiers1 with a 2 ns rise time. These preamplifiers are linked to 1 GSample/s flash digitizer boards2 interfaced to a LabVIEW acquisition software by a VME-PCI optical link bridge module3 [5,6]. The whole acquisition system can run with a frequency of 250 Hz. This limitation comes from the PC data transfer rate for the digital representation of the strip current. However, the SAPHIR accelerator can deliver between 1 and 400 pulses per second and we have used it at only 12 Hz. 3. Simulations 3.1. Neutron efficiency The neutron detection efficiency is evaluated using a Monte-Carlo simulation and the SRIM code (Stopping Range of Ions in Matter) [7]. The different processes of detection (proton recoils from the plastic foil and helium or proton recoils from the gas) are considered. The reaction yield of each process is calculated as a function of the neutron energy using the neutron cross-section from the ENDF data base. The recoils are generated uniformly in depth in the plastic foil and in the drift gap taking into account the differential cross-sections for the angular distribution. The ionization deposited energy in the gas is then calculated using the SRIM code and the recoils depositing less than the threshold energy estimated to 30 keV are discarded. This threshold energy is in fact the threshold energy per strip, which corresponds to a 6 mV amplitude signal. The simulated efficiency as a function of energy is presented in Fig. 2. The carbon recoils from the gas and the plastic foil are not represented since their number is negligible. 3.2. g-rays sensitivity
2.3. Electronics and acquisition system The current from each strip is registered for each 2:5 ms pulse of the electrons accelerator. These accelerator pulses play the role of trigger for the acquisition system. The strips are connected individually to fast current preamplin - HV1
9.5 mm
n
Polypropylene foil
He + 3.8% Isobutane
- HV2 0.1
0 (not on scale) Recoil H, He (or C) Fig. 1. Micromegas principle for neutron detection.
One wants to be insensible to g-rays or at least be able to discriminate g-rays and neutrons [8]. Like neutrons, g-rays must be converted before being detected. Obviously, the conversion rate of g-rays in electrons through photoelectric or compton effects is more important in the surrounding material than in the gas of the detector. However, the deposited electron energy increases with its path inside the gas of the detector and more generally with the drift gap size. The electron and proton stopping powers are plotted in Fig. 3. Electrons deposit at least a factor 10 less than protons in helium. It should be easy to find an adequate gain for the detector permitting sensitivity to n-recoils but not to g- and X-rays. Nevertheless, a problem could arise at low energy recoils, close to the threshold energy of the detector, where it could be impossible to distinguish a low energy electron having a transversal path in the detector 1
ZFL-500LN from Minicircuit. Compatible with CAEN V1729 board. 3 V2718 from CAEN. 2
ARTICLE IN PRESS J. Pancin et al. / Nuclear Instruments and Methods in Physics Research A 572 (2007) 859–865
0.04
Efficiency [%]
Table 1 MCNP calculation results for 2 107 incident photons
H recoil in gas He recoil in gas H recoil in PP total
0.035 0.03
861
0.025 0.02 0.015
Deposited energy
Number of photons
0.01–1 keV 1–10 keV 10–20 keV 20–30 keV
2160743 548721 772 171
0.01 0.005 0 1
102 Energy [keV]
10
103
104
Fig. 2. Simulated neutron detection efficiency from 1 keV to 10 MeV.
0.1
electron proton
dE/dX [MeV/mm]
0.01
0.001 Fig. 4. Neutron detector response from spontaneous fission from
0.0001
1e-05 0.01
0.1
1
10
252
Cf.
depositing more than 1 keV in the detector. Since the pulse duration of the micromegas detector used here is over 40 ns, a baseline shift will appear in the strip signal of the detector.
Energy [MeV]
4. Experimental validation of the detector Fig. 3. Ionization energy losses of electrons and protons in helium as a function of energy up to 10 MeV.
from a low energy proton recoil. Therefore, the maximum electron energy deposit has to be calculated. MCNP (Monte Carlo N-Particle) [9] simulations were done to estimate the deposited energy distribution due to the g-rays flux of the SAPHIR facility. The simulation accounts for the exact geometry of the detector and the grays spectrum behind a radioactive parcel. The total number of g-rays passing through the detector is around 2 107 g=pulse (corresponding to 2 1011 g=cm2 =s). The results of the simulation are given in Table 1. The first column is the deposited energy over one strip while the second column is the number of photons leading to this energy deposit. One can see that most of the photons are depositing less than 10 keV and none over 30 keV. This means that in principle, with a threshold over 30 keV per strip, the detector is insensible to individual photon and only to recoils. It is not so simple in practice since the number of photons depositing small energies is still important during the short 2:5 ms accelerator pulse. Consequently, one has about one photon every 5 ns
4.1. Tests with radioactive sources Preliminary tests with neutron source (252Cf with 8 103 neutrons=s) and g-ray source (22Na with an activity of 720 kBq) have been made. The purpose was first, to validate the detector for neutron detection and second, to adapt the amplification gain for a good intrinsic g-ray rejection. Fig. 4 shows an example of the detector response to neutrons with high voltages of 340 V on the micromesh and 1540 V on the drift electrode, respectively. The detection efficiency has been reproduced within 20% of the simulation. The measured detection efficiency is 1:7ð0:2Þ 104 in agreement with 2ð0:2Þ 104 found by simulation. The error on the measurement is mainly given by geometrical aspects like the position of the source while the error on the simulation had been estimated by comparison with measurement of the n TOF flux. Systematic measurements have been realized in order to get the counting rate in neutrons and photons as a function of the micromesh high voltage. These tests have been performed with the whole detector set up (detector, electronics and data processing). The results are shown in
ARTICLE IN PRESS J. Pancin et al. / Nuclear Instruments and Methods in Physics Research A 572 (2007) 859–865
862
100
20
Cf source Na source
90
10
80 Amplitude [mV]
Counts/min
70 60 50 40 30 20
0 -10 -20 -30
Beam Window
10 0 -400
-40 -390
-380
-370 -360 -350 Vmesh [V]
-340
-330
-320
Fig. 5. Counting rate per minute as a function of the micromesh high voltage with 252Cf and 22Na sources.
Fig. 5. Over 360 V, the counting rate is low in both case. It is about 3 counts/min for the 22Na source, which is corresponding to the noise level. At this gain, the detector is not counting any photons and the counts coming from the 252Cf are then purely neutrons. The mean amplitude of these proton recoils is 12 mV in these conditions. Below 360 V, the counting rate is increasing rapidly and the detector is sensitive to photons for both sources. Consequently, one has to choose a micromesh high voltage superior to 360 V to have an intrinsic g-ray rejection. The value of 325 V (well over 360 V) has been chosen for the experiments in order to have a detection threshold (about 30 keV) sufficiently high to avoid individual g-ray detection with a high probability and sufficiently low to optimize the neutron detection efficiency. However, the detector has to be tested now in a high g-rays flux. 4.2. Test in high g rate environment Experiments were performed at the Gent University in Belgium to estimate the g-rays sensitivity of the micromegas detector [10]. The g-rays were produced by Bremsstrahlung on a 1 cm thick copper target. The beam was working at 1 Hz with 12:5 MeV electrons and a pulse width of 2 ms with a 15 nC charge/pulse. The detector was at 8 m from the target in the beam axis. The g-rays flux has been estimated by simulation to 107 g=pulse which corresponds to the conditions behind a parcel in SAPHIR. First, the detector behavior was normal when the accelerator was on and the beam off with a noise level of 2 mV peak to peak on each strip. It had been also verified that the preamplifiers and the cables in the g-rays flux did not perturb the detector signal. Twenty measurements of 600 accelerator pulses each have been registered. Fig. 6 shows the typical response of the detector to a high g-rays flux. The response is not saturated but the baseline shift is important.
8500
9000
9500
10000 Time [ns]
10500
11000
Fig. 6. Detector response to a 1:55 107 g-rays pulse.
The baseline shift is due to two phenomena. The first one is the simultaneous detection of many low energy photons, which contribute to a current creation in the detector. These photons create small signals which are summed together and provoke the baseline shift. The second phenomenon is linked to individual photon detection leading to higher energy deposit, which contributes to local variation of the baseline. One has to be more careful in this case with the data analysis. In high g-rays flux, the probability of an important energy deposit due to photons (in our case more than 30 keV) is not negligible anymore and recoils must not be mixed with electrons. The data analysis has to use more powerful means for the signal treatment like derivative peak detection or Fast Fourier Transform smoothing in order to detect individual pulses in a noisy baseline. A program using FFT has been written to analyze the SAPHIR data. It has been used here and no local variation over 6 mV has been found in the whole set of the Gent experiment data, which means that no individual photon leading to an energy deposit superior to 30 keV has been seen. Under these conditions, it should be possible to see neutrons in very high g-rays flux. 4.3. Experiment at SAPHIR facility The experiment took place in July 2005. The electron energy was about 13 MeV with a copper target to create the photon flux. The test parcel has been placed at 1, 2 and 3 m from the target and the Micromegas detector has always been placed just behind the parcel. It was running with helium/isobutane ð3:8%Þ and with 325 and 1400 V on the micromesh and the drift electrode, respectively. The accelerator was delivering some 2:5 ms time width pulses with 12 Hz frequency. The purpose was to see if one were able to detect some prompt fission neutrons (about 100 neutrons on the surface of the detector/accelerator pulse)
ARTICLE IN PRESS J. Pancin et al. / Nuclear Instruments and Methods in Physics Research A 572 (2007) 859–865
coming from the impurities of the test parcel in the high g-rays background of SAPHIR. The g-rays spectrum shape is shown in Fig. 7. 4.3.1. Preliminary results in images Fig. 8 shows the mean detector response on 50 accelerator pulses for the three different distances. One can observe the electromagnetic perturbation due to the accelerator (already observable in Gent) which corresponds to the first signals on run 21 and run 23 and which has nothing to do with the photon flux. The electron beam and then the photon noise really arrive on the detector at the relative time t ¼ 400 ns. For run 13, the acquisition has been started 200 ns after the beginning of the beam window in order to see the induced signal also after the accelerator pulse and eventually some neutrons that have been diffused
Number of photons
10000
1000
100
10 0.1
1 Energy [MeV]
10
Fig. 7. g-rays spectrum behind a parcel at 3 m from the target for 1 cm2 of detector and one pulse.
40 Run 21 : d = 90 cm Run 13 : d = 190 cm Run 23 : d = 290 cm
Signal (mV)
20 0 -20 -40 -60
-500
0
500 1000 Time (ns)
1500
Fig. 8. Mean detector signal (on 50 pulses) versus time for three different distances.
863
and slowed down in the parcel. The fact that the signal increases when the detector with the parcel are closer from the target is obvious. The structures in the detector response (run 21 for example) are due to the internal structures of the electron beam but also to the huge ionization in detector. For instance, the change of polarity at the end of the signal is probably due to an important ion transmission between the amplification gap and the drift gap. To conclude, the detector response is never saturated even for run 21 where the number of photons on the active surface of the detector has been estimated to more than 108 g=pulse (around 1012 g=cm2 =s). Fig. 9 shows two responses of the detector with a distance parcel-target of 2 m. Both contain one neutron signal, at t ¼ 0 ns for 9a and at t ¼ 2100 ns for 9b. The second one takes place after the end of the accelerator pulse, it is a neutron that has diffused in the parcel.
4.3.2. Data analysis and results Data analysis usually consists of looking for a threshold crossing to detect a neutron interaction. It is not possible to apply this method here because of the baseline shift and the noise implied by the photons. Two methods have been used: one using a FFT smoothed derivative peaked detector and the other based on substraction of the average baseline. The first method uses an FFT and an inverse FFT in order to cut off high frequencies superior to 50 MHz. The derivative of this smoother signal is then calculated with a threshold crossing to detect individual interactions (see Fig. 10). The efficiency of this new method has been evaluated as a function of its threshold by applying it to the calibration runs made with the neutron source. The threshold used on SAPHIR data is high enough to avoid noise perturbation and permits to detect 10% of the neutron pulses. The second method consists in calculating the mean strip signal using 50 pulses and subtracting it from each measured strip signal. It aims at reducing as much as possible the baseline shift (see Fig. 11). The analysis is then realized as usual by a threshold crossing. However, the noise is much more important and the threshold is put at 14 mV instead of 5 mV. The efficiency of this method has been estimated to 15%. Twenty runs of 10 000 accelerator pulses have been registered for the three different positions. Table 2 gives the results, efficiency of the methods included. It shows the number of detected neutrons for 10 000 accelerator pulses using the mean and the FFT methods. These results are in agreement with MCNP simulations realized to study the potential of this new method using prompt neutrons. The detector has detected prompt neutrons coming from photofission of impurities in the parcel despite the important g-rays background of 2 1011 g=cm2 =s. This background could even be higher as long as the strip signals are not saturated.
ARTICLE IN PRESS J. Pancin et al. / Nuclear Instruments and Methods in Physics Research A 572 (2007) 859–865
864
b
40
40
20
20
0
0 Signal [mV]
Signal [mV]
a
-20 -40 -60
-20 -40 -60
Beam Window
-80
Beam Window
-80
-100
-100 -500
0
500 1000 Time [ns]
1500
2000
0
500
1000 1500 Time [ns]
2000
2500
Fig. 9. Detector response to an accelerator pulse behind a parcel at 2 m with: (a) a neutron detected during the pulse, and (b) a neutron detected after the pulse.
Table 2 Results of the experiment at SAPHIR: number of detected neutrons for 10 000 accelerator pulses using the mean or the FFT method for three different distances
50 40 30 Signal [mv]
20
Distances (m)
Number of neutrons
Error
Mean
FFT
Mean
FFT
192.8 49.2 21.0
67.4 33.7
64.2 18.6 12.2
27.5 19.4
10 0 -10
1 2 3
-20 -30 -40
5. Conclusion
-50 0
500
1000 1500 Time [ns]
2000
2500
S- (mV)
Signal (mV)
Fig. 10. Detector signal (with a detected neutron) after an FFT and an inverse FFT, and its derivative (upper part).
20 10 0 -10 -20 -30 -40
It has been clearly demonstrated that it was possible to use a micromegas detector to detect neutrons in high g-rays flux of more than 1011 g=cm2 =s. The whole detector physics of this detector has been studied and confirmed by simulations and experiments. One drawback of this detector for the characterization of radioactive parcel using prompt neutrons is its low efficiency for neutron detection, which is degrading the sensibility limit of the method. However, this mean of detection could be useful in environments with high g-rays and neutron rate like fusion experiments or nuclear reactors. Acknowledgments
5 0 -5 -10 -15
We would like to thank L. Van Hoorebeke from Gent university for his welcome and F. Laine´ from SAPHIR for his technical support. References 0
500
1000 Time [ns]
1500
2000
Fig. 11. Four strip signals with beam: top, mean signal on 50 accelerator pulses, bottom, strip signal after the mean substraction.
[1] F. Jeanneau, et al., Active photonic interrogation and applications in a nuclear-waste inspection purpose, Proceedings of Ninth International Conference on Radioactive Waste Management and Environmental Remediation, September 2003.
ARTICLE IN PRESS J. Pancin et al. / Nuclear Instruments and Methods in Physics Research A 572 (2007) 859–865 [2] M. Gmar, et al., Assessment of actinide mass embedded in large concrete waste packages by photon interrogation and photofission, Appl. Radiat. Isotopes 63 (2005) 613–619. [3] Y. Giomataris, Ph. Rebourgeard, J.P. Robert, G. Charpak, Nucl. Intsr. and Meth. A 376 (1996) 29. [4] J. Pancin, et al., Nucl. Instr. and Meth. A 524 (2004) 102. [5] hhttp://www.ni.comi. [6] D. Breton, E. Delagnes, M. Houry, IEEE Trans. Nucl. Sci. NS-52-6 (2005) 2853.
865
[7] hhttp://www.srim.orgi. [8] M. Houry, et al., Nucl. Instr. and Meth. A 557-2 (2006) 648. [9] J.F. Briesmeister (Ed.), MCNP—A general Monte Carlo N-particle transport code, Version 4C, LA-13709-M (2000, version 4C3). [10] S. Boyer, et al., A fast neutron detector for nuclear waste characterization, Proceedings of Conference GLOBAL in Tsukuba, 2005.
Nuclear Instruments and Methods in Physics Research A 572 (2007) 859–865 www.elsevier.com/locate/nima
Neutron detection in high g background using a micromegas detector J. Pancina,, S. Aunea, E. Berthoumieuxa, S. Boyera, E. Delagnesa, V. Macarya, B. Poumaredeb, H. Safaa a
CEA Saclay-DSM/DAPNIA, Bat 534/34A, F-91191 Gif-Sur-Yvette, France b CEA-DRT/LIST Saclay, France
Received 20 April 2006; received in revised form 24 October 2006; accepted 7 December 2006 Available online 17 December 2006
Abstract The ability of Micromegas to detect neutrons over a wide energy range has already been demonstrated. However, in some nuclear experiments or applications, neutrons come with high photon flux disturbing the detectors. A new project for nuclear waste characterization using photonuclear reactions is under development at the CEA. One of the ideas is to detect the prompt neutrons produced by these kinds of reactions, which are accompanied by a strong g flash. The micromegas detector has been chosen to detect these neutrons since it is quite insensitive to g-rays under certain conditions. The configuration of this detector will be presented and its g insensitivity demonstrated. Results from simulations and experiments will be shown. r 2007 Elsevier B.V. All rights reserved. PACS: 25.85.Jg; 28.20.v; 07.85.Fv; 29.40.n Keywords: Micromegas; Photofission; Radioactive waste characterization; Neutron detection; g sensitivity
1. Introduction The SAPHIR facility at CEA Saclay is experienced in radioactive parcel characterization using delayed neutrons [1,2]. An intense g-ray beam is created by the slowing down of electrons in a copper or a tungsten target. The energy of the electrons is between 12 and 15 MeV with about 1012 electrons in one accelerator pulse of 2:5 ms. Photonuclear reactions are induced by the g-rays on the minor actinides and transuranium elements but also on the impurities of the parcel. The actinide and transuranium quantities are evaluated by measuring the delayed neutrons coming from fission fragments, after the accelerator pulse, using 3He counters placed all around the parcel. Detecting prompt neutrons during the accelerator pulse could be another means to characterize radioactive parcels. However, it would require detectors which are not blinded by the g-rays flash like 3He counters. A neutron detector based on a Micromegas concept [3] has been realized to detect these Corresponding author. Tel.: +33 1 69082648; fax: +33 1 69085669.
E-mail address: [email protected] (J. Pancin). 0168-9002/$ - see front matter r 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.nima.2006.12.010
neutrons in the huge g-rays flash. The principle of the detector is presented first. Some simulations have been performed to check the detector characteristics for neutron detection and evaluate its behavior in g-beams. Some experiments and results are finally reported, showing the detector characteristics. 2. The neutron detector 2.1. Technical description The Micromegas detector used here is quite similar to the one used for the n_TOF experiment [4]. The inner chamber, filled with a mixture of argon–isobutane or helium–isobutane at atmospheric pressure, has been reused. The detector, placed inside, is a double stage circular chamber with an active surface of 40 cm2 . It consists of a 9:5 mm drift gap and a 100 mm amplification gap. The two gaps are separated by a 5 mm thick nickel micromesh. The amplification occurs between the mesh and a copper printed circuit composed of eight strips of 9 mm wide spaced 100 mm apart. Some spacers insulate the
ARTICLE IN PRESS J. Pancin et al. / Nuclear Instruments and Methods in Physics Research A 572 (2007) 859–865
860
printed circuit from the micromesh and guarantee the uniformity of the amplification gap. The drift electrode, covering all the detection area, is a 12 mm thick polypropylene foil on which a 100 nm layer of aluminium has been deposited to ensure electrical conductivity. 2.2. Detector principle Electrons are produced between the drift electrode and the micromesh during the passing of charged particles through the gas. These electrons drift towards the micromesh under the action of the electric field applied in the drift gap ð1 kV=cmÞ. After the micromesh, they are multiplied in the narrow amplification gap (high field with 30 kV=cmÞ and this avalanche induces a current on the anode strips. However, a neutron to charged particle converter is necessary to detect the incident neutron. Since only fast neutrons are of interest here, neutron scattering on the nuclei of the polypropylene foil and of the gas is exploited (see Fig. 1). Above the detection threshold, the ionization created by the recoils is detected. For this application, the Micromegas detector is generally filled with He þ iC4H10 ð3:8%Þ (For safety reasons, the percentage of isobutane is sufficiently low to have a nonflammable gas). On the one hand, helium is chosen since, as a light gas, it is less sensitive to X-and g-ray photons. On the other hand, the lighter the nucleus, the larger is the energy transferred from the incident neutron. It is then the recoils of hydrogen coming from the foil and the gas, and the recoils of helium coming from the gas that are mostly detected. The hydrogen nucleus is of particular interest in fast neutron scattering since it can recoil with the entire neutron energy. In the case of helium, the reaction 4He(n,n)4He has a resonance around 1 MeV which increases the reaction rate.
fiers1 with a 2 ns rise time. These preamplifiers are linked to 1 GSample/s flash digitizer boards2 interfaced to a LabVIEW acquisition software by a VME-PCI optical link bridge module3 [5,6]. The whole acquisition system can run with a frequency of 250 Hz. This limitation comes from the PC data transfer rate for the digital representation of the strip current. However, the SAPHIR accelerator can deliver between 1 and 400 pulses per second and we have used it at only 12 Hz. 3. Simulations 3.1. Neutron efficiency The neutron detection efficiency is evaluated using a Monte-Carlo simulation and the SRIM code (Stopping Range of Ions in Matter) [7]. The different processes of detection (proton recoils from the plastic foil and helium or proton recoils from the gas) are considered. The reaction yield of each process is calculated as a function of the neutron energy using the neutron cross-section from the ENDF data base. The recoils are generated uniformly in depth in the plastic foil and in the drift gap taking into account the differential cross-sections for the angular distribution. The ionization deposited energy in the gas is then calculated using the SRIM code and the recoils depositing less than the threshold energy estimated to 30 keV are discarded. This threshold energy is in fact the threshold energy per strip, which corresponds to a 6 mV amplitude signal. The simulated efficiency as a function of energy is presented in Fig. 2. The carbon recoils from the gas and the plastic foil are not represented since their number is negligible. 3.2. g-rays sensitivity
2.3. Electronics and acquisition system The current from each strip is registered for each 2:5 ms pulse of the electrons accelerator. These accelerator pulses play the role of trigger for the acquisition system. The strips are connected individually to fast current preamplin - HV1
9.5 mm
n
Polypropylene foil
He + 3.8% Isobutane
- HV2 0.1
0 (not on scale) Recoil H, He (or C) Fig. 1. Micromegas principle for neutron detection.
One wants to be insensible to g-rays or at least be able to discriminate g-rays and neutrons [8]. Like neutrons, g-rays must be converted before being detected. Obviously, the conversion rate of g-rays in electrons through photoelectric or compton effects is more important in the surrounding material than in the gas of the detector. However, the deposited electron energy increases with its path inside the gas of the detector and more generally with the drift gap size. The electron and proton stopping powers are plotted in Fig. 3. Electrons deposit at least a factor 10 less than protons in helium. It should be easy to find an adequate gain for the detector permitting sensitivity to n-recoils but not to g- and X-rays. Nevertheless, a problem could arise at low energy recoils, close to the threshold energy of the detector, where it could be impossible to distinguish a low energy electron having a transversal path in the detector 1
ZFL-500LN from Minicircuit. Compatible with CAEN V1729 board. 3 V2718 from CAEN. 2
ARTICLE IN PRESS J. Pancin et al. / Nuclear Instruments and Methods in Physics Research A 572 (2007) 859–865
0.04
Efficiency [%]
Table 1 MCNP calculation results for 2 107 incident photons
H recoil in gas He recoil in gas H recoil in PP total
0.035 0.03
861
0.025 0.02 0.015
Deposited energy
Number of photons
0.01–1 keV 1–10 keV 10–20 keV 20–30 keV
2160743 548721 772 171
0.01 0.005 0 1
102 Energy [keV]
10
103
104
Fig. 2. Simulated neutron detection efficiency from 1 keV to 10 MeV.
0.1
electron proton
dE/dX [MeV/mm]
0.01
0.001 Fig. 4. Neutron detector response from spontaneous fission from
0.0001
1e-05 0.01
0.1
1
10
252
Cf.
depositing more than 1 keV in the detector. Since the pulse duration of the micromegas detector used here is over 40 ns, a baseline shift will appear in the strip signal of the detector.
Energy [MeV]
4. Experimental validation of the detector Fig. 3. Ionization energy losses of electrons and protons in helium as a function of energy up to 10 MeV.
from a low energy proton recoil. Therefore, the maximum electron energy deposit has to be calculated. MCNP (Monte Carlo N-Particle) [9] simulations were done to estimate the deposited energy distribution due to the g-rays flux of the SAPHIR facility. The simulation accounts for the exact geometry of the detector and the grays spectrum behind a radioactive parcel. The total number of g-rays passing through the detector is around 2 107 g=pulse (corresponding to 2 1011 g=cm2 =s). The results of the simulation are given in Table 1. The first column is the deposited energy over one strip while the second column is the number of photons leading to this energy deposit. One can see that most of the photons are depositing less than 10 keV and none over 30 keV. This means that in principle, with a threshold over 30 keV per strip, the detector is insensible to individual photon and only to recoils. It is not so simple in practice since the number of photons depositing small energies is still important during the short 2:5 ms accelerator pulse. Consequently, one has about one photon every 5 ns
4.1. Tests with radioactive sources Preliminary tests with neutron source (252Cf with 8 103 neutrons=s) and g-ray source (22Na with an activity of 720 kBq) have been made. The purpose was first, to validate the detector for neutron detection and second, to adapt the amplification gain for a good intrinsic g-ray rejection. Fig. 4 shows an example of the detector response to neutrons with high voltages of 340 V on the micromesh and 1540 V on the drift electrode, respectively. The detection efficiency has been reproduced within 20% of the simulation. The measured detection efficiency is 1:7ð0:2Þ 104 in agreement with 2ð0:2Þ 104 found by simulation. The error on the measurement is mainly given by geometrical aspects like the position of the source while the error on the simulation had been estimated by comparison with measurement of the n TOF flux. Systematic measurements have been realized in order to get the counting rate in neutrons and photons as a function of the micromesh high voltage. These tests have been performed with the whole detector set up (detector, electronics and data processing). The results are shown in
ARTICLE IN PRESS J. Pancin et al. / Nuclear Instruments and Methods in Physics Research A 572 (2007) 859–865
862
100
20
Cf source Na source
90
10
80 Amplitude [mV]
Counts/min
70 60 50 40 30 20
0 -10 -20 -30
Beam Window
10 0 -400
-40 -390
-380
-370 -360 -350 Vmesh [V]
-340
-330
-320
Fig. 5. Counting rate per minute as a function of the micromesh high voltage with 252Cf and 22Na sources.
Fig. 5. Over 360 V, the counting rate is low in both case. It is about 3 counts/min for the 22Na source, which is corresponding to the noise level. At this gain, the detector is not counting any photons and the counts coming from the 252Cf are then purely neutrons. The mean amplitude of these proton recoils is 12 mV in these conditions. Below 360 V, the counting rate is increasing rapidly and the detector is sensitive to photons for both sources. Consequently, one has to choose a micromesh high voltage superior to 360 V to have an intrinsic g-ray rejection. The value of 325 V (well over 360 V) has been chosen for the experiments in order to have a detection threshold (about 30 keV) sufficiently high to avoid individual g-ray detection with a high probability and sufficiently low to optimize the neutron detection efficiency. However, the detector has to be tested now in a high g-rays flux. 4.2. Test in high g rate environment Experiments were performed at the Gent University in Belgium to estimate the g-rays sensitivity of the micromegas detector [10]. The g-rays were produced by Bremsstrahlung on a 1 cm thick copper target. The beam was working at 1 Hz with 12:5 MeV electrons and a pulse width of 2 ms with a 15 nC charge/pulse. The detector was at 8 m from the target in the beam axis. The g-rays flux has been estimated by simulation to 107 g=pulse which corresponds to the conditions behind a parcel in SAPHIR. First, the detector behavior was normal when the accelerator was on and the beam off with a noise level of 2 mV peak to peak on each strip. It had been also verified that the preamplifiers and the cables in the g-rays flux did not perturb the detector signal. Twenty measurements of 600 accelerator pulses each have been registered. Fig. 6 shows the typical response of the detector to a high g-rays flux. The response is not saturated but the baseline shift is important.
8500
9000
9500
10000 Time [ns]
10500
11000
Fig. 6. Detector response to a 1:55 107 g-rays pulse.
The baseline shift is due to two phenomena. The first one is the simultaneous detection of many low energy photons, which contribute to a current creation in the detector. These photons create small signals which are summed together and provoke the baseline shift. The second phenomenon is linked to individual photon detection leading to higher energy deposit, which contributes to local variation of the baseline. One has to be more careful in this case with the data analysis. In high g-rays flux, the probability of an important energy deposit due to photons (in our case more than 30 keV) is not negligible anymore and recoils must not be mixed with electrons. The data analysis has to use more powerful means for the signal treatment like derivative peak detection or Fast Fourier Transform smoothing in order to detect individual pulses in a noisy baseline. A program using FFT has been written to analyze the SAPHIR data. It has been used here and no local variation over 6 mV has been found in the whole set of the Gent experiment data, which means that no individual photon leading to an energy deposit superior to 30 keV has been seen. Under these conditions, it should be possible to see neutrons in very high g-rays flux. 4.3. Experiment at SAPHIR facility The experiment took place in July 2005. The electron energy was about 13 MeV with a copper target to create the photon flux. The test parcel has been placed at 1, 2 and 3 m from the target and the Micromegas detector has always been placed just behind the parcel. It was running with helium/isobutane ð3:8%Þ and with 325 and 1400 V on the micromesh and the drift electrode, respectively. The accelerator was delivering some 2:5 ms time width pulses with 12 Hz frequency. The purpose was to see if one were able to detect some prompt fission neutrons (about 100 neutrons on the surface of the detector/accelerator pulse)
ARTICLE IN PRESS J. Pancin et al. / Nuclear Instruments and Methods in Physics Research A 572 (2007) 859–865
coming from the impurities of the test parcel in the high g-rays background of SAPHIR. The g-rays spectrum shape is shown in Fig. 7. 4.3.1. Preliminary results in images Fig. 8 shows the mean detector response on 50 accelerator pulses for the three different distances. One can observe the electromagnetic perturbation due to the accelerator (already observable in Gent) which corresponds to the first signals on run 21 and run 23 and which has nothing to do with the photon flux. The electron beam and then the photon noise really arrive on the detector at the relative time t ¼ 400 ns. For run 13, the acquisition has been started 200 ns after the beginning of the beam window in order to see the induced signal also after the accelerator pulse and eventually some neutrons that have been diffused
Number of photons
10000
1000
100
10 0.1
1 Energy [MeV]
10
Fig. 7. g-rays spectrum behind a parcel at 3 m from the target for 1 cm2 of detector and one pulse.
40 Run 21 : d = 90 cm Run 13 : d = 190 cm Run 23 : d = 290 cm
Signal (mV)
20 0 -20 -40 -60
-500
0
500 1000 Time (ns)
1500
Fig. 8. Mean detector signal (on 50 pulses) versus time for three different distances.
863
and slowed down in the parcel. The fact that the signal increases when the detector with the parcel are closer from the target is obvious. The structures in the detector response (run 21 for example) are due to the internal structures of the electron beam but also to the huge ionization in detector. For instance, the change of polarity at the end of the signal is probably due to an important ion transmission between the amplification gap and the drift gap. To conclude, the detector response is never saturated even for run 21 where the number of photons on the active surface of the detector has been estimated to more than 108 g=pulse (around 1012 g=cm2 =s). Fig. 9 shows two responses of the detector with a distance parcel-target of 2 m. Both contain one neutron signal, at t ¼ 0 ns for 9a and at t ¼ 2100 ns for 9b. The second one takes place after the end of the accelerator pulse, it is a neutron that has diffused in the parcel.
4.3.2. Data analysis and results Data analysis usually consists of looking for a threshold crossing to detect a neutron interaction. It is not possible to apply this method here because of the baseline shift and the noise implied by the photons. Two methods have been used: one using a FFT smoothed derivative peaked detector and the other based on substraction of the average baseline. The first method uses an FFT and an inverse FFT in order to cut off high frequencies superior to 50 MHz. The derivative of this smoother signal is then calculated with a threshold crossing to detect individual interactions (see Fig. 10). The efficiency of this new method has been evaluated as a function of its threshold by applying it to the calibration runs made with the neutron source. The threshold used on SAPHIR data is high enough to avoid noise perturbation and permits to detect 10% of the neutron pulses. The second method consists in calculating the mean strip signal using 50 pulses and subtracting it from each measured strip signal. It aims at reducing as much as possible the baseline shift (see Fig. 11). The analysis is then realized as usual by a threshold crossing. However, the noise is much more important and the threshold is put at 14 mV instead of 5 mV. The efficiency of this method has been estimated to 15%. Twenty runs of 10 000 accelerator pulses have been registered for the three different positions. Table 2 gives the results, efficiency of the methods included. It shows the number of detected neutrons for 10 000 accelerator pulses using the mean and the FFT methods. These results are in agreement with MCNP simulations realized to study the potential of this new method using prompt neutrons. The detector has detected prompt neutrons coming from photofission of impurities in the parcel despite the important g-rays background of 2 1011 g=cm2 =s. This background could even be higher as long as the strip signals are not saturated.
ARTICLE IN PRESS J. Pancin et al. / Nuclear Instruments and Methods in Physics Research A 572 (2007) 859–865
864
b
40
40
20
20
0
0 Signal [mV]
Signal [mV]
a
-20 -40 -60
-20 -40 -60
Beam Window
-80
Beam Window
-80
-100
-100 -500
0
500 1000 Time [ns]
1500
2000
0
500
1000 1500 Time [ns]
2000
2500
Fig. 9. Detector response to an accelerator pulse behind a parcel at 2 m with: (a) a neutron detected during the pulse, and (b) a neutron detected after the pulse.
Table 2 Results of the experiment at SAPHIR: number of detected neutrons for 10 000 accelerator pulses using the mean or the FFT method for three different distances
50 40 30 Signal [mv]
20
Distances (m)
Number of neutrons
Error
Mean
FFT
Mean
FFT
192.8 49.2 21.0
67.4 33.7
64.2 18.6 12.2
27.5 19.4
10 0 -10
1 2 3
-20 -30 -40
5. Conclusion
-50 0
500
1000 1500 Time [ns]
2000
2500
S-
Signal (mV)
Fig. 10. Detector signal (with a detected neutron) after an FFT and an inverse FFT, and its derivative (upper part).
20 10 0 -10 -20 -30 -40
It has been clearly demonstrated that it was possible to use a micromegas detector to detect neutrons in high g-rays flux of more than 1011 g=cm2 =s. The whole detector physics of this detector has been studied and confirmed by simulations and experiments. One drawback of this detector for the characterization of radioactive parcel using prompt neutrons is its low efficiency for neutron detection, which is degrading the sensibility limit of the method. However, this mean of detection could be useful in environments with high g-rays and neutron rate like fusion experiments or nuclear reactors. Acknowledgments
5 0 -5 -10 -15
We would like to thank L. Van Hoorebeke from Gent university for his welcome and F. Laine´ from SAPHIR for his technical support. References 0
500
1000 Time [ns]
1500
2000
Fig. 11. Four strip signals with beam: top, mean signal on 50 accelerator pulses, bottom, strip signal after the mean substraction.
[1] F. Jeanneau, et al., Active photonic interrogation and applications in a nuclear-waste inspection purpose, Proceedings of Ninth International Conference on Radioactive Waste Management and Environmental Remediation, September 2003.
ARTICLE IN PRESS J. Pancin et al. / Nuclear Instruments and Methods in Physics Research A 572 (2007) 859–865 [2] M. Gmar, et al., Assessment of actinide mass embedded in large concrete waste packages by photon interrogation and photofission, Appl. Radiat. Isotopes 63 (2005) 613–619. [3] Y. Giomataris, Ph. Rebourgeard, J.P. Robert, G. Charpak, Nucl. Intsr. and Meth. A 376 (1996) 29. [4] J. Pancin, et al., Nucl. Instr. and Meth. A 524 (2004) 102. [5] hhttp://www.ni.comi. [6] D. Breton, E. Delagnes, M. Houry, IEEE Trans. Nucl. Sci. NS-52-6 (2005) 2853.
865
[7] hhttp://www.srim.orgi. [8] M. Houry, et al., Nucl. Instr. and Meth. A 557-2 (2006) 648. [9] J.F. Briesmeister (Ed.), MCNP—A general Monte Carlo N-particle transport code, Version 4C, LA-13709-M (2000, version 4C3). [10] S. Boyer, et al., A fast neutron detector for nuclear waste characterization, Proceedings of Conference GLOBAL in Tsukuba, 2005.