Suitability of high-pressure xenon as scintillator for gamma ray spectroscopy
F. Resnati, U. Gendotti, R. Chandra, A. Curioni, G. Davatz, H. Frederich, A. Gendotti, L. Goeltl, R. Jebali, D. Murer, A. Rubbia
SSuitability of high-pressure xenon as scintillator forgamma ray spectroscopy
F. Resnati a , U. Gendotti b , R. Chandra b , A. Curioni a , G. Davatz b , H.Frederich b , A. Gendotti a , L. Goeltl b , R. Jebali b , D. Murer b , A. Rubbia a a ETH Zurich, Institute for Particle Physics, Zurich, Switzerland b Arktis Radiation Detectors Ltd
Abstract
In this paper we report the experimental study of high-pressure xenon usedas a scintillator, in the context of developing a gamma ray detector. Wemeasure a light yield near 2 photoelectrons per keV for xenon at 40 bar.Together with the light yield, we also measured an energy resolution of ∼ Keywords: high-pressure xenon, scintillation, gamma ray spectroscopy
1. Introduction
Xenon detectors have been used and are in use for a number of applica-tions, in particular for the detection of gamma rays (for a partial list see [1]and references therein, [2, 3, 4, 5]. For exhaustive review work see [6] and [7]).Xenon detectors are also preeminent in the field of Dark Matter searches andneutrinoless double beta decay (e.g. the XENON experiment [8], the LUXexperiment [9], and the EXO experiment [10]). Xenon is an attractive mate-rial for gamma ray detection, in particular thanks to its high atomic number(Z = 54) and, as a consequence, large cross-section for photoelectric absorp-tion. Xenon is also a rather dense material (the density of liquid xenon is3.1 g/mL, while the density of 50 bar xenon at 293 K is 0.4 g/mL), whichallows to build compact and efficient detectors. The scintillation light fromXe is at 175 nm, in the VUV, and it is therefore rather difficult to detectefficiently. While a number of works presenting high-pressure Xe ionizationchambers can be found in the literature, very few are available that discuss
Preprint submitted to Nuclear instruments and Methods A October 31, 2018 a r X i v : . [ phy s i c s . i n s - d e t ] D ec he properties of high-pressure Xe as a scintillator (known to these authors:[11]). Compared to a liquid Xe detector, a high-pressure Xe detector doesnot require cryogenics, and this simplification may be important for someapplications. It has to be noted that the energy resolution achieved in Xegas proportional chambers is significantly better than the one in liquid Xe[12]. It is therefore very interesting to compare the energy resolution ob-tained from primary scintillation in Xe gas to the one measured for liquidXe.In this work we have performed an experimental study of the propertiesof high-pressure Xe as a scintillator, in the context of developing a gammaray detector for the detection of Special Nuclear Materials . Our first goalhas been to study experimentally the light yield and energy resolution froma high-pressure Xe tube, studying the dependence on thermodynamic condi-tions as well. A short description of the apparatus is given in Sec. 2, and theresults are discussed in Sec. 3.
2. Experimental apparatus
We have used a high-pressure tube (designed to withstand pressures up to200 bar) filled with high purity xenon [13]. The tube has a length of 200 mmand a diameter of 44 mm, with UV transparent windows at both ends. Theinner surface of the tube is lined with a reflector combined with a wavelengthshifter, that shifts the VUV light from xenon scintillation into visible light .The isochoric curves of xenon are shown in Fig. 1. The critical point forxenon is at 289.77 K and 58.41 bar, therefore at room temperature (293 K),xenon is in supercritical phase (for pressures above 58.41 bar), while fortemperatures below 290 K it becomes liquid. The detector was operated inthe pressure range between 30 and 60 bar, i.e. with xenon as a dense vapor(density changing between - roughly - 0.2 and 0.7 g/mL).The scintillation light from the high-pressure tube was read out usingtwo Hamamatsu R580 (quartz window) photomultiplier tubes (PMTs). Aschematic diagram of the experimental setup is shown in Fig. 2. The datawere taken using a CAEN V1751 digitizer (1 GS/s, 10 bit). A waveform p r e ss u r e ( ba r) temperature ( o C)400 g/L800 g/L1200 g/L1600 g/L2000 g/L
Figure 1: Isochoric curves for xenon. Data from NIST(http://webbook.nist.gov/chemistry/fluid/). is shown in Fig. 3. The two PMTs were digitized independently, and thetrigger, directly programmed on the digitizer, required the coincidence of thetwo PMTs.We took data with various gamma ray sources, most of the time with thesource collimated with lead bricks, having a 3 mm wide slit illuminating thedetector.
Our analysis procedure consists of four steps:1.
Single photoelectron (p.e.) calibration:
The single p.e. is taken fromthe digitized events. A peak finding algorithm is applied to the digi-tized waveform (Fig. 3) and the area for each peak (in units of ADCcounts × ns) is determined integrating the signal over the time windowin which the waveform is above threshold; the threshold is 2 times thenoise (rms of the baseline), and a minimum of 3 consecutive samplesabove threshold is required. The time windows for each peak are shownin Fig. 3. The interesting region for the single p.e. is easily determined,producing a spectrum as the one shown in Fig. 4. The superimposed3 b PbPMT PMTpressure vesselADC HV windows
44 mm 20 cm
Figure 2: Schematic diagram of the experimental setup. See text for explanations. function, fitted to the spectrum, is A g (0 , s ) + n =3 (cid:88) n =1 A n g ( n × m, √ n × σ ) , where A g (0 , s ) is a Gaussian with amplitude A , mean zero and sigma s to account for the background, and the sum of the Gaussian functionsmodels the photoelectron peaks.2. Each waveform is analyzed:
The procedure is similar to the one usedfor the single p.e. calibration, but this time the amplitudes of the peaksin the event are summed up, to define the number of p.e. per PMT forthe given event. This integration procedure is performed over a timeinterval of 400 ns following the first threshold crossing.3. At this point it is possible to apply basic selections to the reconstructedevents. We have three different selections: first of all, the time differ-ence between the first identified peaks of the two PMT signals mustbe less than 40 ns (this selection rejects less than 0.1% of the events).Second, the difference between the integrals of the two PMT signals is4
940 950 960 970 980 990 1000 0 200 400 600 800 1000 a m p li t ude ( A DC c oun t s ) time (ns) Figure 3: Digitized waveform from one the PMTs. The sampling rate is 1 GS/s, with aresolution of 10 bit. Also shown are the integration windows for the signal (see text forexplanation). required to be less than 200 p.e.. Third, the integrals of the two PMTsignals are required to differ less than 30%. The last two selections onthe asymmetry between the two PMTs tend to improve the peak-to-total ratio in the energy spectrum and are especially effective for datawith the source collimated in the center of the detector, because eventsshowing largely asymmetric signals between the two PMTs tend to beonly partially contained. One example showing the impact of these se-lections is given in Fig. 5, for a collimated
Cs source: after applyingthe selections described here, the peak-to-total ratio goes from 3.8%to 7%.4. The event integrals for each PMT are summed up, giving the amountof light for each event (in units of p.e.) and the final spectrum. Here for the “total” we integrate over the entire spectrum, including the X-ray lines. it integral (ADC counts x sample rate)0 50 100 150 200 250 300 c oun t s / b i n / Figure 4: The single p.e. obtained from the digitized events. The superimposed function,fitted to the spectrum, includes a Gaussian to account for the background, and a sum ofGaussian functions to model the peaks of the n p.e. ( n = 1 , , ... ).
3. Results
In the development of a gamma ray detector, our main goal was thedetermination of the light yield of high-pressure xenon, when exposed togamma rays.The energy spectra from exposure to a
Ba and a
Cs are shown inFig. 6 (left and right, respectively). In both cases the source was collimatedat the center of the detector by lead bricks.
Cs gives a line at 662 keV(and 32 keV X-ray line), while
Ba gives several lines: the most prominentare at 31 keV (X-ray), 80 keV, 302 keV and 356 keV, with additional weakerlines at 276 keV and 384 keV, which, with the available energy resolution,are partially merged with the 302 keV and 356 keV lines, and hard to detectindividually. The backscattered peak is large for this particular detectorgeometry, where the mass of the stainless steel tube largely exceeds the massof the xenon inside. Data were taken also with a Na source, which giveslines at 511 and 1275 keV. From the data, we derive a light yield of 1 . ± .
05 p.e./keV at 662 keV.The energy dependence of the light yield (per unit energy) is shown in6 ollected light (p.e.)0 200 400 600 800 1000 1200 1400 1600 c oun t s / b i n Figure 5: Energy spectrum of a collimated
Cs before (continuous red line) and after(dotted black line) the selections described in the text.
Fig. 7, where we present the “normalized light yield”, i.e. normalized to thelight yield per keV at 662 keV (
Cs). There is a systematic trend, showing alarger light yield at lower energies. We interpret this as due, at least in part,to non-containment of the electron from gamma conversion. Fig. 8 showsthe fraction of energy deposited in a pressurized xenon tube as a functionof electron energy, for different gas pressures. Here monochromatic electrons(1000 for each energy) are generated in the center of a tube with a diameterof 40 mm, using the GEANT4 simulation package . At low pressures anddensities the fraction of energy deposited drops rapidly with energy, becauseMeV electrons are, on average, not contained. At higher pressures the elec-trons stop inside the tube, but energetic Bremsstrahlung photons still escapefrom the tube. Considering the CSDA range for electrons, we see that itexceeds 20 mm in 40 bar xenon when the energy of the electron is larger than1.1 MeV: given a diameter of the detector of 44 mm, the probability of fullcontainment of electrons with a range exceeding 20 mm is rather small. http://geant4.cern.ch/ Continuous Slowing Down Approximation, which gives an upper limit to the rangeof the electron. It can be easily calculated using the ESTAR program from NIST,http://physics.nist.gov/PhysRefData/Star/Text/ESTAR.html. ollected light (p.e.)0 200 400 600 800 1000 1200 1400 1600 no r m a li z ed c oun t s / b i n no r m a li z ed c oun t s / b i n
356 keV302 keV31 keV80 keV 662 keV
Figure 6:
Left : Ba spectrum.
Ba gives lines at 31 keV (X-ray), 80 keV, 302 keVand 356 keV. At ∼
30 keV several X-ray lines from iron and lead are also possible.
Right : Cs spectrum, no off-line selections applied. Both sources were collimated in the centerof the detector.
Again in Fig. 7, the measured energy resolution and the expected energyresolution (for the measured light yield, and assuming that the spread can bemodeled with a Poisson distribution), are shown. The measured energy res-olution is consistently worse than the one expected from simple photostatis-tics. Again, the discrepancy increases with energy. The non-containment ofthe electron is one reason for this systematic effect. Another reason is thatCompton scattering becomes more and more important when the energy ofthe gamma ray increases, and therefore the event is less localized and theenergy is deposited within a larger fraction of the active volume. This leadsto a degradation of the energy resolution, as can be seen in Fig. 9 and inFig. 10, which show that in our setup the light yield has a significant (ofthe order of 15% over the full volume) dependence on the interaction posi-tion inside the detector. In Fig. 9 we show the number of photoelectronscollected on one PMT vs. the one from the other PMT, for an uncollimated
Cs source at a distance of 2 m. It is clear that the band correspondingto the full energy peak has a “banana shape”, i.e. when the sharing of the8 no r m a li z ed li gh t y i e l d pe r un i t ene r g y F W H M ( % ) energy (keV)light yieldresolutionstatistical resolution Figure 7: Normalized light yield, energy resolution, and limit to the energy resolutionfrom photostatistics, vs gamma ray energy. The normalized light yield is defined as theratio between the light yield at a given energy divided by the light yield at 662 keV ( Csline). The energy resolution is derived from fitting the peaks in the spectrum with aGaussian plus a second order polynomial for modeling the background. The “statisticalresolution” is calculated from the number of photoelectrons, assuming that the spread inthe peak can be modeled with a Poisson distribution. light is almost equal between the two PMTs, the total collected light is lessthan when the light distribution between the two PMTs is very asymmetric.The data shown in Fig. 10 were taken in different conditions, with a colli-mated
Cs source, scanning the length of the tube up to 7 cm from thetube center. The dependence of the number of photoelectrons with respectto the position of the source is quite clear. In fact, even when summing upboth PMTs, we notice that the light yield for an event in the center of thedetector is about 15% lower than for an event happening next to one of thetwo PMTs. To summarize, the lower the energy of the gamma ray, the betterthe localization, because1. the range of the secondary electron is shorter;2. the gamma ray is more likely to deposit all the energy in one singleinteraction (photoelectric absorption, to be contrasted with multipleCompton interactions).A better localization immediately translates in a reduced spread on the light9 E dep / E t o t energy (MeV)10 bar20 bar30 bar40 bar50 bar Figure 8: Fraction of energy deposited in a pressurized xenon cylinder of 20 mm radius forelectrons starting at the center of the cylinder as a function of the electron initial energy,for five different pressures. yield, and on the energy resolution .With the help of a Monte Carlo simulation we estimate the light collec-tion efficiency of 17%, not including the quantum efficiency of the PMTs,with large systematic errors due to the lack of independent knowledge of theoptical properties of the various materials. Given the geometry of the tubeand an aspect ratio of almost 5, only a small fraction of the scintillation lightis directly detected by the PMTs, without being converted after interactionon the detector walls. The light collection efficiency takes into account thepropagation of the light in the tube, i.e. the wavelength shifting process onthe detector walls (for VUV photons: assumed to be 100% as estimated fromavailable data), the reflection from the detector walls (for visible photons),the transmission/reflection from the windows (separately for VUV and visi-ble photons), and the possibility of absorption. The poor knowledge of thereflectivity of the material lining the tube introduces the largest uncertaintyin our calculation. If we keep the reflectivity as the only free parameter inthe calculation, we can determine it from the data, in particular from the A partial correction for the position dependence of the light yield can be introducedoffline, allowing to recover the energy resolution. ± +5 − % Thequantum efficiency of the PMTs is ∼ s ) of 26 +7 − eVto produce one scintillation photon in high-pressure xenon at 662 keV. Thiscan be compared to a value of 23.7 eV for liquid xenon [14].For completeness, we notice that our value for W s is significantly smallerthan the one of 76 ±
12 eV from [15] for 15 bar Xe at 60 keV, and 72 ± s of 111 ±
16 eV was reported in [17] for 1 bar pressure at 5.9keV. When comparing these conflicting results one should also consider thedifferent thermodynamic conditions of the xenon. collected light - left - (p.e.)0 200 400 600 800 1000 1200 1400 c o ll e c t ed li gh t - r i gh t - ( p . e . ) c oun t s / b i n Figure 9: PMT-left vs PMT-right. Detector exposed to an uncollimated
Cs source.
We have also studied the light yield as a function of the xenon pressure:results are shown in Fig. 11. From the point of view of an application, thisis important because an efficient detector should be operated at the largestpossible density. We clearly see that the energy resolution deteriorates whenthe xenon gas gets closer to the supercritical phase, essentially because manyscattering centers within the fluid hamper the uniform propagation of light.The abrupt change in the optical properties of the Xe is easily observable bysimply watching the Xe cell while changing the pressure: the high pressure Xegoes from a perfectly transparent medium to a turbid one when the pressure11 no r m a li z ed c o ll e c t ed li gh t distance (cm)leftrightsum Figure 10: Collected light (normalized) as a function of the source position. Detectorexposed to a collimated
Cs source, moved along the detector. exceeds 45 bars. In fact, while the light yield does not change enough tomodify much the Poisson term in the resolution, the effect is particularlyvisible on the measured energy resolution, pointing to a systematic smearingwhich applies to the light production or to the light propagation, or to both.It is interesting to compare our results in terms of energy resolution (9%FWHM at 662 keV, with a light yield of 1.9 p.e./keV) to the results ob-tained using liquid Xe detectors: Akerib et al. [9] report a light yield of 8p.e./keV and an energy resolution of 11.7% FWHM at 662 keV, while Ni etal. [18] report 6 p.e./ keV and 14.5% (FWHM at 662 keV). Therefore wesee that, while the energy resolution in high-pressure xenon is dominated byphotostatistics, this is not the case for liquid xenon.
4. Conclusion
In this paper we have presented experimental results that give an initialcharacterization of the properties of high-pressure Xe gas as a scintillator,in particular in terms of the light yield. Using a detector which is not fullyoptimized for efficient light collection, we measure a light yield of 1 . ± .
05 p.e./keV at 662 keV at 40 bar pressure. The corresponding energyresolution is 9% FWHM. We find this result encouraging for applications ofhigh-pressure Xe as a scintillator for gamma ray spectroscopy.12 F W H M ( % ) c o ll e c t ed li gh t ( p . e . ) pressure (bar)light yieldresolution Figure 11: Light yield and energy resolution as a function of the xenon pressure. Detectorexposed to a collimated
Cs source.
Some of the shortcomings of the present setup are well understood interms of non-uniformity of the light yield and can be readily fixed in anoptimized setup. A more uniform light collection across the active volumeshould push the energy resolution down to the limit given by photostatistics,which right now is reached only at the lowest energies. We have also studiedthe behavior of xenon as a function of the pressure, and we find that, whenthe gas gets close to its supercritical phase, the energy resolution degradessignificantly, well beyond the effect of the slightly reduced light yield. Thissystematic effect is due to the presence of many scattering centers that affectthe propagation of light in xenon, and determines an upper limit to thedensity of xenon used as a scintillator.
Acknowledgments
This work is supported by the European Union through the MODES SNMproject (Call FP7-SEC-2011-1, ERC grant agreement n ◦ References [1] V.V. Dmitrenko, S.E. Ulin, V.M. Grachev, K.F. Vlasik, Z.M. Uteshev,I.V. Chernyseva, K.V. Krivova, and A.G. Dukhvalov. Perspectives of13igh Pressure Xenon Gamma Ray Spectrometers to Detect and IdentifyRadioactive and Fissile Materials. S. Apikyan et al. (eds.), Prevention,Detection and Response to Nuclear and Radiological Threats, 155175,2008 Springer. (doi: )[2] A. Bolozdynya, V. Egorov, A. Kuchenkov, G. Safronov, G. Smirnov,S. Medved and V. Morgunov, Nucl. Instrum. Meth. A (1997) 225.[3] A. Curioni, E. Aprile, T. Doke, K. L. Giboni, M. Kobayashiand U. G. Oberlack, Nucl. Instrum. Meth. A (2007) 350[physics/0702078 [PHYSICS]].[4] E. Aprile, A. Curioni, K. L. Giboni, M. Kobayashi, U. G. Oberlack andS. Zhang, Nucl. Instrum. Meth. A , 414 (2008) [arXiv:0805.0290[physics.ins-det]].[5] G. Carugno, E. Conti, A. T. Meneguzzo, R. Onofrio, U. Beriotto, S. DeBiasia, M. Nicoletto and R. Pedrotta et al. , Nucl. Instrum. Meth. A (1996) 149.[6] E. Aprile and T. Doke, Rev. Mod. Phys. (2010) 2053 [arXiv:0910.4956[physics.ins-det]].[7] Aprile, Elena; Bolotnikov, Aleksey E.; Doke, Tadayoshi (2006). NobleGas Detectors. Wiley-VCH. pp. 89. ISBN 3-527-60963-6.[8] E. Aprile et al. [XENON100 Collaboration], Phys. Rev. Lett. (2011)131302 [arXiv:1104.2549 [astro-ph.CO]].[9] D. S. Akerib et al. [LUX Collaboration], arXiv:1210.4569 [astro-ph.IM].[10] M. Auger et al. [EXO Collaboration], Phys. Rev. Lett. (2012)032505 [arXiv:1205.5608 [hep-ex]].[11] A. Bolotnikov, B. Ramsey, Nucl. Instrum. Meth. A , 391 (1999)[12] A. Bolotnikov, B. Ramsey, Nucl. Instrum. Meth. A , 360 (1997)[13] R. Chandra et al., JINST 7 C03035 (2012) doi:10.1088/1748-0221/7/03/C03035[14] T. Doke and K. Masuda, Nucl. Instrum. Meth. A (1999) 62.1415] A. Parsons et al., IEEE Trans. Nucl. Sci. 37 (1990) 541.[16] L. M. P. Fernandes, E. D. C. Freitas, M. Ball, J. J. Gomez-Cadenas,C. M. B. Monteiro, N. Yahlali, D. Nygren and J. M. F. d. Santos, JINST (2010) P09006 [Erratum-ibid. (2010) A12001] [arXiv:1009.2719[astro-ph.IM]].[17] S. J. C. do Carmo et al., JINST (2010) P07004 doi:10.1088/1748-0221/3/07/P07004[18] K. Ni, E. Aprile, K. L. Giboni, P. Majewski and M. Yamashita, JINST1